aboutsummaryrefslogtreecommitdiffstats
path: root/arch/powerpc/kernel
diff options
context:
space:
mode:
authorGlenn Elliott <gelliott@cs.unc.edu>2012-03-04 19:47:13 -0500
committerGlenn Elliott <gelliott@cs.unc.edu>2012-03-04 19:47:13 -0500
commitc71c03bda1e86c9d5198c5d83f712e695c4f2a1e (patch)
treeecb166cb3e2b7e2adb3b5e292245fefd23381ac8 /arch/powerpc/kernel
parentea53c912f8a86a8567697115b6a0d8152beee5c8 (diff)
parent6a00f206debf8a5c8899055726ad127dbeeed098 (diff)
Merge branch 'mpi-master' into wip-k-fmlpwip-k-fmlp
Conflicts: litmus/sched_cedf.c
Diffstat (limited to 'arch/powerpc/kernel')
-rw-r--r--arch/powerpc/kernel/Makefile22
-rw-r--r--arch/powerpc/kernel/align.c4
-rw-r--r--arch/powerpc/kernel/asm-offsets.c40
-rw-r--r--arch/powerpc/kernel/btext.c2
-rw-r--r--arch/powerpc/kernel/cpu_setup_44x.S1
-rw-r--r--arch/powerpc/kernel/cpu_setup_6xx.S40
-rw-r--r--arch/powerpc/kernel/cpu_setup_a2.S114
-rw-r--r--arch/powerpc/kernel/cpu_setup_fsl_booke.S24
-rw-r--r--arch/powerpc/kernel/cpu_setup_power7.S91
-rw-r--r--arch/powerpc/kernel/cputable.c174
-rw-r--r--arch/powerpc/kernel/crash.c164
-rw-r--r--arch/powerpc/kernel/crash_dump.c50
-rw-r--r--arch/powerpc/kernel/dbell.c65
-rw-r--r--arch/powerpc/kernel/dma-iommu.c23
-rw-r--r--arch/powerpc/kernel/dma.c38
-rw-r--r--arch/powerpc/kernel/e500-pmu.c2
-rw-r--r--arch/powerpc/kernel/entry_32.S12
-rw-r--r--arch/powerpc/kernel/entry_64.S67
-rw-r--r--arch/powerpc/kernel/exceptions-64e.S204
-rw-r--r--arch/powerpc/kernel/exceptions-64s.S257
-rw-r--r--arch/powerpc/kernel/fpu.S11
-rw-r--r--arch/powerpc/kernel/ftrace.c8
-rw-r--r--arch/powerpc/kernel/head_32.S13
-rw-r--r--arch/powerpc/kernel/head_40x.S9
-rw-r--r--arch/powerpc/kernel/head_44x.S3
-rw-r--r--arch/powerpc/kernel/head_64.S71
-rw-r--r--arch/powerpc/kernel/head_8xx.S1
-rw-r--r--arch/powerpc/kernel/head_fsl_booke.S13
-rw-r--r--arch/powerpc/kernel/ibmebus.c415
-rw-r--r--arch/powerpc/kernel/idle_power4.S21
-rw-r--r--arch/powerpc/kernel/idle_power7.S97
-rw-r--r--arch/powerpc/kernel/io-workarounds.c188
-rw-r--r--arch/powerpc/kernel/iommu.c14
-rw-r--r--arch/powerpc/kernel/irq.c333
-rw-r--r--arch/powerpc/kernel/kgdb.c192
-rw-r--r--arch/powerpc/kernel/kvm.c596
-rw-r--r--arch/powerpc/kernel/kvm_emul.S302
-rw-r--r--arch/powerpc/kernel/l2cr_6xx.S2
-rw-r--r--arch/powerpc/kernel/legacy_serial.c30
-rw-r--r--arch/powerpc/kernel/lparcfg.c70
-rw-r--r--arch/powerpc/kernel/machine_kexec.c41
-rw-r--r--arch/powerpc/kernel/machine_kexec_32.c4
-rw-r--r--arch/powerpc/kernel/misc.S5
-rw-r--r--arch/powerpc/kernel/misc_32.S12
-rw-r--r--arch/powerpc/kernel/misc_64.S14
-rw-r--r--arch/powerpc/kernel/mpc7450-pmu.c2
-rw-r--r--arch/powerpc/kernel/nvram_64.c504
-rw-r--r--arch/powerpc/kernel/of_platform.c9
-rw-r--r--arch/powerpc/kernel/paca.c88
-rw-r--r--arch/powerpc/kernel/pci-common.c18
-rw-r--r--arch/powerpc/kernel/pci_32.c2
-rw-r--r--arch/powerpc/kernel/pci_64.c9
-rw-r--r--arch/powerpc/kernel/pci_dn.c17
-rw-r--r--arch/powerpc/kernel/pci_of_scan.c4
-rw-r--r--arch/powerpc/kernel/perf_callchain.c86
-rw-r--r--arch/powerpc/kernel/perf_event.c230
-rw-r--r--arch/powerpc/kernel/perf_event_fsl_emb.c149
-rw-r--r--arch/powerpc/kernel/power4-pmu.c2
-rw-r--r--arch/powerpc/kernel/power5+-pmu.c2
-rw-r--r--arch/powerpc/kernel/power5-pmu.c2
-rw-r--r--arch/powerpc/kernel/power6-pmu.c2
-rw-r--r--arch/powerpc/kernel/power7-pmu.c2
-rw-r--r--arch/powerpc/kernel/ppc970-pmu.c4
-rw-r--r--arch/powerpc/kernel/ppc_ksyms.c12
-rw-r--r--arch/powerpc/kernel/ppc_save_regs.S3
-rw-r--r--arch/powerpc/kernel/process.c69
-rw-r--r--arch/powerpc/kernel/prom.c118
-rw-r--r--arch/powerpc/kernel/prom_init.c30
-rw-r--r--arch/powerpc/kernel/prom_parse.c122
-rw-r--r--arch/powerpc/kernel/ptrace.c129
-rw-r--r--arch/powerpc/kernel/ptrace32.c7
-rw-r--r--arch/powerpc/kernel/rtas-rtc.c29
-rw-r--r--arch/powerpc/kernel/rtas.c13
-rw-r--r--arch/powerpc/kernel/rtas_flash.c56
-rw-r--r--arch/powerpc/kernel/rtasd.c8
-rw-r--r--arch/powerpc/kernel/setup-common.c23
-rw-r--r--arch/powerpc/kernel/setup_32.c5
-rw-r--r--arch/powerpc/kernel/setup_64.c53
-rw-r--r--arch/powerpc/kernel/signal_32.c57
-rw-r--r--arch/powerpc/kernel/signal_64.c17
-rw-r--r--arch/powerpc/kernel/smp.c318
-rw-r--r--arch/powerpc/kernel/swsusp.c1
-rw-r--r--arch/powerpc/kernel/swsusp_32.S2
-rw-r--r--arch/powerpc/kernel/sys_ppc32.c1
-rw-r--r--arch/powerpc/kernel/sysfs.c38
-rw-r--r--arch/powerpc/kernel/time.c350
-rw-r--r--arch/powerpc/kernel/traps.c81
-rw-r--r--arch/powerpc/kernel/udbg.c2
-rw-r--r--arch/powerpc/kernel/udbg_16550.c53
-rw-r--r--arch/powerpc/kernel/vdso.c12
-rw-r--r--arch/powerpc/kernel/vdso32/Makefile6
-rw-r--r--arch/powerpc/kernel/vdso32/sigtramp.S2
-rw-r--r--arch/powerpc/kernel/vdso64/Makefile6
-rw-r--r--arch/powerpc/kernel/vdso64/sigtramp.S2
-rw-r--r--arch/powerpc/kernel/vector.S3
-rw-r--r--arch/powerpc/kernel/vio.c29
-rw-r--r--arch/powerpc/kernel/vmlinux.lds.S2
97 files changed, 4570 insertions, 2080 deletions
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index 1dda70129141..e8b981897d44 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -29,18 +29,23 @@ endif
29obj-y := cputable.o ptrace.o syscalls.o \ 29obj-y := cputable.o ptrace.o syscalls.o \
30 irq.o align.o signal_32.o pmc.o vdso.o \ 30 irq.o align.o signal_32.o pmc.o vdso.o \
31 init_task.o process.o systbl.o idle.o \ 31 init_task.o process.o systbl.o idle.o \
32 signal.o sysfs.o cacheinfo.o 32 signal.o sysfs.o cacheinfo.o time.o \
33obj-y += vdso32/ 33 prom.o traps.o setup-common.o \
34 udbg.o misc.o io.o dma.o \
35 misc_$(CONFIG_WORD_SIZE).o vdso32/
34obj-$(CONFIG_PPC64) += setup_64.o sys_ppc32.o \ 36obj-$(CONFIG_PPC64) += setup_64.o sys_ppc32.o \
35 signal_64.o ptrace32.o \ 37 signal_64.o ptrace32.o \
36 paca.o nvram_64.o firmware.o 38 paca.o nvram_64.o firmware.o
37obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o 39obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
38obj-$(CONFIG_PPC_BOOK3S_64) += cpu_setup_ppc970.o cpu_setup_pa6t.o 40obj-$(CONFIG_PPC_BOOK3S_64) += cpu_setup_ppc970.o cpu_setup_pa6t.o
41obj-$(CONFIG_PPC_BOOK3S_64) += cpu_setup_power7.o
39obj64-$(CONFIG_RELOCATABLE) += reloc_64.o 42obj64-$(CONFIG_RELOCATABLE) += reloc_64.o
40obj-$(CONFIG_PPC_BOOK3E_64) += exceptions-64e.o idle_book3e.o 43obj-$(CONFIG_PPC_BOOK3E_64) += exceptions-64e.o idle_book3e.o
44obj-$(CONFIG_PPC_A2) += cpu_setup_a2.o
41obj-$(CONFIG_PPC64) += vdso64/ 45obj-$(CONFIG_PPC64) += vdso64/
42obj-$(CONFIG_ALTIVEC) += vecemu.o 46obj-$(CONFIG_ALTIVEC) += vecemu.o
43obj-$(CONFIG_PPC_970_NAP) += idle_power4.o 47obj-$(CONFIG_PPC_970_NAP) += idle_power4.o
48obj-$(CONFIG_PPC_P7_NAP) += idle_power7.o
44obj-$(CONFIG_PPC_OF) += of_platform.o prom_parse.o 49obj-$(CONFIG_PPC_OF) += of_platform.o prom_parse.o
45obj-$(CONFIG_PPC_CLOCK) += clock.o 50obj-$(CONFIG_PPC_CLOCK) += clock.o
46procfs-y := proc_powerpc.o 51procfs-y := proc_powerpc.o
@@ -55,7 +60,9 @@ obj-$(CONFIG_IBMVIO) += vio.o
55obj-$(CONFIG_IBMEBUS) += ibmebus.o 60obj-$(CONFIG_IBMEBUS) += ibmebus.o
56obj-$(CONFIG_GENERIC_TBSYNC) += smp-tbsync.o 61obj-$(CONFIG_GENERIC_TBSYNC) += smp-tbsync.o
57obj-$(CONFIG_CRASH_DUMP) += crash_dump.o 62obj-$(CONFIG_CRASH_DUMP) += crash_dump.o
63ifeq ($(CONFIG_PPC32),y)
58obj-$(CONFIG_E500) += idle_e500.o 64obj-$(CONFIG_E500) += idle_e500.o
65endif
59obj-$(CONFIG_6xx) += idle_6xx.o l2cr_6xx.o cpu_setup_6xx.o 66obj-$(CONFIG_6xx) += idle_6xx.o l2cr_6xx.o cpu_setup_6xx.o
60obj-$(CONFIG_TAU) += tau_6xx.o 67obj-$(CONFIG_TAU) += tau_6xx.o
61obj-$(CONFIG_HIBERNATION) += swsusp.o suspend.o 68obj-$(CONFIG_HIBERNATION) += swsusp.o suspend.o
@@ -67,20 +74,16 @@ endif
67obj64-$(CONFIG_HIBERNATION) += swsusp_asm64.o 74obj64-$(CONFIG_HIBERNATION) += swsusp_asm64.o
68obj-$(CONFIG_MODULES) += module.o module_$(CONFIG_WORD_SIZE).o 75obj-$(CONFIG_MODULES) += module.o module_$(CONFIG_WORD_SIZE).o
69obj-$(CONFIG_44x) += cpu_setup_44x.o 76obj-$(CONFIG_44x) += cpu_setup_44x.o
70obj-$(CONFIG_FSL_BOOKE) += cpu_setup_fsl_booke.o dbell.o 77obj-$(CONFIG_PPC_FSL_BOOK3E) += cpu_setup_fsl_booke.o dbell.o
71obj-$(CONFIG_PPC_BOOK3E_64) += dbell.o 78obj-$(CONFIG_PPC_BOOK3E_64) += dbell.o
72 79
73extra-y := head_$(CONFIG_WORD_SIZE).o 80extra-y := head_$(CONFIG_WORD_SIZE).o
74extra-$(CONFIG_PPC_BOOK3E_32) := head_new_booke.o
75extra-$(CONFIG_40x) := head_40x.o 81extra-$(CONFIG_40x) := head_40x.o
76extra-$(CONFIG_44x) := head_44x.o 82extra-$(CONFIG_44x) := head_44x.o
77extra-$(CONFIG_FSL_BOOKE) := head_fsl_booke.o 83extra-$(CONFIG_FSL_BOOKE) := head_fsl_booke.o
78extra-$(CONFIG_8xx) := head_8xx.o 84extra-$(CONFIG_8xx) := head_8xx.o
79extra-y += vmlinux.lds 85extra-y += vmlinux.lds
80 86
81obj-y += time.o prom.o traps.o setup-common.o \
82 udbg.o misc.o io.o dma.o \
83 misc_$(CONFIG_WORD_SIZE).o
84obj-$(CONFIG_PPC32) += entry_32.o setup_32.o 87obj-$(CONFIG_PPC32) += entry_32.o setup_32.o
85obj-$(CONFIG_PPC64) += dma-iommu.o iommu.o 88obj-$(CONFIG_PPC64) += dma-iommu.o iommu.o
86obj-$(CONFIG_KGDB) += kgdb.o 89obj-$(CONFIG_KGDB) += kgdb.o
@@ -102,8 +105,11 @@ obj-$(CONFIG_KEXEC) += machine_kexec.o crash.o \
102obj-$(CONFIG_AUDIT) += audit.o 105obj-$(CONFIG_AUDIT) += audit.o
103obj64-$(CONFIG_AUDIT) += compat_audit.o 106obj64-$(CONFIG_AUDIT) += compat_audit.o
104 107
108obj-$(CONFIG_PPC_IO_WORKAROUNDS) += io-workarounds.o
109
105obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o 110obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o
106obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o 111obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o
112obj-$(CONFIG_FTRACE_SYSCALLS) += ftrace.o
107obj-$(CONFIG_PERF_EVENTS) += perf_callchain.o 113obj-$(CONFIG_PERF_EVENTS) += perf_callchain.o
108 114
109obj-$(CONFIG_PPC_PERF_CTRS) += perf_event.o 115obj-$(CONFIG_PPC_PERF_CTRS) += perf_event.o
@@ -127,6 +133,8 @@ ifneq ($(CONFIG_XMON)$(CONFIG_KEXEC),)
127obj-y += ppc_save_regs.o 133obj-y += ppc_save_regs.o
128endif 134endif
129 135
136obj-$(CONFIG_KVM_GUEST) += kvm.o kvm_emul.o
137
130# Disable GCOV in odd or sensitive code 138# Disable GCOV in odd or sensitive code
131GCOV_PROFILE_prom_init.o := n 139GCOV_PROFILE_prom_init.o := n
132GCOV_PROFILE_ftrace.o := n 140GCOV_PROFILE_ftrace.o := n
diff --git a/arch/powerpc/kernel/align.c b/arch/powerpc/kernel/align.c
index b876e989220b..8184ee97e484 100644
--- a/arch/powerpc/kernel/align.c
+++ b/arch/powerpc/kernel/align.c
@@ -889,7 +889,7 @@ int fix_alignment(struct pt_regs *regs)
889#ifdef CONFIG_PPC_FPU 889#ifdef CONFIG_PPC_FPU
890 preempt_disable(); 890 preempt_disable();
891 enable_kernel_fp(); 891 enable_kernel_fp();
892 cvt_df(&data.dd, (float *)&data.v[4], &current->thread); 892 cvt_df(&data.dd, (float *)&data.v[4]);
893 preempt_enable(); 893 preempt_enable();
894#else 894#else
895 return 0; 895 return 0;
@@ -933,7 +933,7 @@ int fix_alignment(struct pt_regs *regs)
933#ifdef CONFIG_PPC_FPU 933#ifdef CONFIG_PPC_FPU
934 preempt_disable(); 934 preempt_disable();
935 enable_kernel_fp(); 935 enable_kernel_fp();
936 cvt_fd((float *)&data.v[4], &data.dd, &current->thread); 936 cvt_fd((float *)&data.v[4], &data.dd);
937 preempt_enable(); 937 preempt_enable();
938#else 938#else
939 return 0; 939 return 0;
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 1c0607ddccc0..36e1c8a29be8 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -48,11 +48,11 @@
48#ifdef CONFIG_PPC_ISERIES 48#ifdef CONFIG_PPC_ISERIES
49#include <asm/iseries/alpaca.h> 49#include <asm/iseries/alpaca.h>
50#endif 50#endif
51#ifdef CONFIG_KVM 51#if defined(CONFIG_KVM) || defined(CONFIG_KVM_GUEST)
52#include <linux/kvm_host.h> 52#include <linux/kvm_host.h>
53#ifndef CONFIG_BOOKE
54#include <asm/kvm_book3s.h>
55#endif 53#endif
54#if defined(CONFIG_KVM) && defined(CONFIG_PPC_BOOK3S)
55#include <asm/kvm_book3s.h>
56#endif 56#endif
57 57
58#ifdef CONFIG_PPC32 58#ifdef CONFIG_PPC32
@@ -61,7 +61,7 @@
61#endif 61#endif
62#endif 62#endif
63 63
64#if defined(CONFIG_FSL_BOOKE) 64#if defined(CONFIG_PPC_FSL_BOOK3E)
65#include "../mm/mmu_decl.h" 65#include "../mm/mmu_decl.h"
66#endif 66#endif
67 67
@@ -74,6 +74,7 @@ int main(void)
74 DEFINE(AUDITCONTEXT, offsetof(struct task_struct, audit_context)); 74 DEFINE(AUDITCONTEXT, offsetof(struct task_struct, audit_context));
75 DEFINE(SIGSEGV, SIGSEGV); 75 DEFINE(SIGSEGV, SIGSEGV);
76 DEFINE(NMI_MASK, NMI_MASK); 76 DEFINE(NMI_MASK, NMI_MASK);
77 DEFINE(THREAD_DSCR, offsetof(struct thread_struct, dscr));
77#else 78#else
78 DEFINE(THREAD_INFO, offsetof(struct task_struct, stack)); 79 DEFINE(THREAD_INFO, offsetof(struct task_struct, stack));
79#endif /* CONFIG_PPC64 */ 80#endif /* CONFIG_PPC64 */
@@ -181,17 +182,19 @@ int main(void)
181 offsetof(struct slb_shadow, save_area[SLB_NUM_BOLTED - 1].vsid)); 182 offsetof(struct slb_shadow, save_area[SLB_NUM_BOLTED - 1].vsid));
182 DEFINE(SLBSHADOW_STACKESID, 183 DEFINE(SLBSHADOW_STACKESID,
183 offsetof(struct slb_shadow, save_area[SLB_NUM_BOLTED - 1].esid)); 184 offsetof(struct slb_shadow, save_area[SLB_NUM_BOLTED - 1].esid));
185 DEFINE(SLBSHADOW_SAVEAREA, offsetof(struct slb_shadow, save_area));
184 DEFINE(LPPACASRR0, offsetof(struct lppaca, saved_srr0)); 186 DEFINE(LPPACASRR0, offsetof(struct lppaca, saved_srr0));
185 DEFINE(LPPACASRR1, offsetof(struct lppaca, saved_srr1)); 187 DEFINE(LPPACASRR1, offsetof(struct lppaca, saved_srr1));
186 DEFINE(LPPACAANYINT, offsetof(struct lppaca, int_dword.any_int)); 188 DEFINE(LPPACAANYINT, offsetof(struct lppaca, int_dword.any_int));
187 DEFINE(LPPACADECRINT, offsetof(struct lppaca, int_dword.fields.decr_int)); 189 DEFINE(LPPACADECRINT, offsetof(struct lppaca, int_dword.fields.decr_int));
188 DEFINE(SLBSHADOW_SAVEAREA, offsetof(struct slb_shadow, save_area)); 190 DEFINE(LPPACA_DTLIDX, offsetof(struct lppaca, dtl_idx));
191 DEFINE(PACA_DTL_RIDX, offsetof(struct paca_struct, dtl_ridx));
189#endif /* CONFIG_PPC_STD_MMU_64 */ 192#endif /* CONFIG_PPC_STD_MMU_64 */
190 DEFINE(PACAEMERGSP, offsetof(struct paca_struct, emergency_sp)); 193 DEFINE(PACAEMERGSP, offsetof(struct paca_struct, emergency_sp));
191 DEFINE(PACAHWCPUID, offsetof(struct paca_struct, hw_cpu_id)); 194 DEFINE(PACAHWCPUID, offsetof(struct paca_struct, hw_cpu_id));
192 DEFINE(PACAKEXECSTATE, offsetof(struct paca_struct, kexec_state)); 195 DEFINE(PACAKEXECSTATE, offsetof(struct paca_struct, kexec_state));
193 DEFINE(PACA_STARTPURR, offsetof(struct paca_struct, startpurr)); 196 DEFINE(PACA_STARTTIME, offsetof(struct paca_struct, starttime));
194 DEFINE(PACA_STARTSPURR, offsetof(struct paca_struct, startspurr)); 197 DEFINE(PACA_STARTTIME_USER, offsetof(struct paca_struct, starttime_user));
195 DEFINE(PACA_USER_TIME, offsetof(struct paca_struct, user_time)); 198 DEFINE(PACA_USER_TIME, offsetof(struct paca_struct, user_time));
196 DEFINE(PACA_SYSTEM_TIME, offsetof(struct paca_struct, system_time)); 199 DEFINE(PACA_SYSTEM_TIME, offsetof(struct paca_struct, system_time));
197 DEFINE(PACA_TRAP_SAVE, offsetof(struct paca_struct, trap_save)); 200 DEFINE(PACA_TRAP_SAVE, offsetof(struct paca_struct, trap_save));
@@ -207,7 +210,6 @@ int main(void)
207 DEFINE(RTASENTRY, offsetof(struct rtas_t, entry)); 210 DEFINE(RTASENTRY, offsetof(struct rtas_t, entry));
208 211
209 /* Interrupt register frame */ 212 /* Interrupt register frame */
210 DEFINE(STACK_FRAME_OVERHEAD, STACK_FRAME_OVERHEAD);
211 DEFINE(INT_FRAME_SIZE, STACK_INT_FRAME_SIZE); 213 DEFINE(INT_FRAME_SIZE, STACK_INT_FRAME_SIZE);
212 DEFINE(SWITCH_FRAME_SIZE, STACK_FRAME_OVERHEAD + sizeof(struct pt_regs)); 214 DEFINE(SWITCH_FRAME_SIZE, STACK_FRAME_OVERHEAD + sizeof(struct pt_regs));
213#ifdef CONFIG_PPC64 215#ifdef CONFIG_PPC64
@@ -394,12 +396,14 @@ int main(void)
394 DEFINE(VCPU_HOST_STACK, offsetof(struct kvm_vcpu, arch.host_stack)); 396 DEFINE(VCPU_HOST_STACK, offsetof(struct kvm_vcpu, arch.host_stack));
395 DEFINE(VCPU_HOST_PID, offsetof(struct kvm_vcpu, arch.host_pid)); 397 DEFINE(VCPU_HOST_PID, offsetof(struct kvm_vcpu, arch.host_pid));
396 DEFINE(VCPU_GPRS, offsetof(struct kvm_vcpu, arch.gpr)); 398 DEFINE(VCPU_GPRS, offsetof(struct kvm_vcpu, arch.gpr));
397 DEFINE(VCPU_MSR, offsetof(struct kvm_vcpu, arch.msr)); 399 DEFINE(VCPU_VRSAVE, offsetof(struct kvm_vcpu, arch.vrsave));
398 DEFINE(VCPU_SPRG4, offsetof(struct kvm_vcpu, arch.sprg4)); 400 DEFINE(VCPU_SPRG4, offsetof(struct kvm_vcpu, arch.sprg4));
399 DEFINE(VCPU_SPRG5, offsetof(struct kvm_vcpu, arch.sprg5)); 401 DEFINE(VCPU_SPRG5, offsetof(struct kvm_vcpu, arch.sprg5));
400 DEFINE(VCPU_SPRG6, offsetof(struct kvm_vcpu, arch.sprg6)); 402 DEFINE(VCPU_SPRG6, offsetof(struct kvm_vcpu, arch.sprg6));
401 DEFINE(VCPU_SPRG7, offsetof(struct kvm_vcpu, arch.sprg7)); 403 DEFINE(VCPU_SPRG7, offsetof(struct kvm_vcpu, arch.sprg7));
402 DEFINE(VCPU_SHADOW_PID, offsetof(struct kvm_vcpu, arch.shadow_pid)); 404 DEFINE(VCPU_SHADOW_PID, offsetof(struct kvm_vcpu, arch.shadow_pid));
405 DEFINE(VCPU_SHARED, offsetof(struct kvm_vcpu, arch.shared));
406 DEFINE(VCPU_SHARED_MSR, offsetof(struct kvm_vcpu_arch_shared, msr));
403 407
404 /* book3s */ 408 /* book3s */
405#ifdef CONFIG_PPC_BOOK3S 409#ifdef CONFIG_PPC_BOOK3S
@@ -464,11 +468,27 @@ int main(void)
464 DEFINE(VCPU_FAULT_ESR, offsetof(struct kvm_vcpu, arch.fault_esr)); 468 DEFINE(VCPU_FAULT_ESR, offsetof(struct kvm_vcpu, arch.fault_esr));
465#endif /* CONFIG_PPC_BOOK3S */ 469#endif /* CONFIG_PPC_BOOK3S */
466#endif 470#endif
471
472#ifdef CONFIG_KVM_GUEST
473 DEFINE(KVM_MAGIC_SCRATCH1, offsetof(struct kvm_vcpu_arch_shared,
474 scratch1));
475 DEFINE(KVM_MAGIC_SCRATCH2, offsetof(struct kvm_vcpu_arch_shared,
476 scratch2));
477 DEFINE(KVM_MAGIC_SCRATCH3, offsetof(struct kvm_vcpu_arch_shared,
478 scratch3));
479 DEFINE(KVM_MAGIC_INT, offsetof(struct kvm_vcpu_arch_shared,
480 int_pending));
481 DEFINE(KVM_MAGIC_MSR, offsetof(struct kvm_vcpu_arch_shared, msr));
482 DEFINE(KVM_MAGIC_CRITICAL, offsetof(struct kvm_vcpu_arch_shared,
483 critical));
484 DEFINE(KVM_MAGIC_SR, offsetof(struct kvm_vcpu_arch_shared, sr));
485#endif
486
467#ifdef CONFIG_44x 487#ifdef CONFIG_44x
468 DEFINE(PGD_T_LOG2, PGD_T_LOG2); 488 DEFINE(PGD_T_LOG2, PGD_T_LOG2);
469 DEFINE(PTE_T_LOG2, PTE_T_LOG2); 489 DEFINE(PTE_T_LOG2, PTE_T_LOG2);
470#endif 490#endif
471#ifdef CONFIG_FSL_BOOKE 491#ifdef CONFIG_PPC_FSL_BOOK3E
472 DEFINE(TLBCAM_SIZE, sizeof(struct tlbcam)); 492 DEFINE(TLBCAM_SIZE, sizeof(struct tlbcam));
473 DEFINE(TLBCAM_MAS0, offsetof(struct tlbcam, MAS0)); 493 DEFINE(TLBCAM_MAS0, offsetof(struct tlbcam, MAS0));
474 DEFINE(TLBCAM_MAS1, offsetof(struct tlbcam, MAS1)); 494 DEFINE(TLBCAM_MAS1, offsetof(struct tlbcam, MAS1));
diff --git a/arch/powerpc/kernel/btext.c b/arch/powerpc/kernel/btext.c
index 625942ae5585..60b3e377b1e4 100644
--- a/arch/powerpc/kernel/btext.c
+++ b/arch/powerpc/kernel/btext.c
@@ -99,7 +99,7 @@ void __init btext_prepare_BAT(void)
99 99
100/* This function can be used to enable the early boot text when doing 100/* This function can be used to enable the early boot text when doing
101 * OF booting or within bootx init. It must be followed by a btext_unmap() 101 * OF booting or within bootx init. It must be followed by a btext_unmap()
102 * call before the logical address becomes unuseable 102 * call before the logical address becomes unusable
103 */ 103 */
104void __init btext_setup_display(int width, int height, int depth, int pitch, 104void __init btext_setup_display(int width, int height, int depth, int pitch,
105 unsigned long address) 105 unsigned long address)
diff --git a/arch/powerpc/kernel/cpu_setup_44x.S b/arch/powerpc/kernel/cpu_setup_44x.S
index 7d606f89a839..e32b4a9a2c22 100644
--- a/arch/powerpc/kernel/cpu_setup_44x.S
+++ b/arch/powerpc/kernel/cpu_setup_44x.S
@@ -35,6 +35,7 @@ _GLOBAL(__setup_cpu_440grx)
35_GLOBAL(__setup_cpu_460ex) 35_GLOBAL(__setup_cpu_460ex)
36_GLOBAL(__setup_cpu_460gt) 36_GLOBAL(__setup_cpu_460gt)
37_GLOBAL(__setup_cpu_460sx) 37_GLOBAL(__setup_cpu_460sx)
38_GLOBAL(__setup_cpu_apm821xx)
38 mflr r4 39 mflr r4
39 bl __init_fpu_44x 40 bl __init_fpu_44x
40 bl __fixup_440A_mcheck 41 bl __fixup_440A_mcheck
diff --git a/arch/powerpc/kernel/cpu_setup_6xx.S b/arch/powerpc/kernel/cpu_setup_6xx.S
index 55cba4a8a959..f8cd9fba4d35 100644
--- a/arch/powerpc/kernel/cpu_setup_6xx.S
+++ b/arch/powerpc/kernel/cpu_setup_6xx.S
@@ -18,7 +18,7 @@
18#include <asm/mmu.h> 18#include <asm/mmu.h>
19 19
20_GLOBAL(__setup_cpu_603) 20_GLOBAL(__setup_cpu_603)
21 mflr r4 21 mflr r5
22BEGIN_MMU_FTR_SECTION 22BEGIN_MMU_FTR_SECTION
23 li r10,0 23 li r10,0
24 mtspr SPRN_SPRG_603_LRU,r10 /* init SW LRU tracking */ 24 mtspr SPRN_SPRG_603_LRU,r10 /* init SW LRU tracking */
@@ -27,60 +27,60 @@ BEGIN_FTR_SECTION
27 bl __init_fpu_registers 27 bl __init_fpu_registers
28END_FTR_SECTION_IFCLR(CPU_FTR_FPU_UNAVAILABLE) 28END_FTR_SECTION_IFCLR(CPU_FTR_FPU_UNAVAILABLE)
29 bl setup_common_caches 29 bl setup_common_caches
30 mtlr r4 30 mtlr r5
31 blr 31 blr
32_GLOBAL(__setup_cpu_604) 32_GLOBAL(__setup_cpu_604)
33 mflr r4 33 mflr r5
34 bl setup_common_caches 34 bl setup_common_caches
35 bl setup_604_hid0 35 bl setup_604_hid0
36 mtlr r4 36 mtlr r5
37 blr 37 blr
38_GLOBAL(__setup_cpu_750) 38_GLOBAL(__setup_cpu_750)
39 mflr r4 39 mflr r5
40 bl __init_fpu_registers 40 bl __init_fpu_registers
41 bl setup_common_caches 41 bl setup_common_caches
42 bl setup_750_7400_hid0 42 bl setup_750_7400_hid0
43 mtlr r4 43 mtlr r5
44 blr 44 blr
45_GLOBAL(__setup_cpu_750cx) 45_GLOBAL(__setup_cpu_750cx)
46 mflr r4 46 mflr r5
47 bl __init_fpu_registers 47 bl __init_fpu_registers
48 bl setup_common_caches 48 bl setup_common_caches
49 bl setup_750_7400_hid0 49 bl setup_750_7400_hid0
50 bl setup_750cx 50 bl setup_750cx
51 mtlr r4 51 mtlr r5
52 blr 52 blr
53_GLOBAL(__setup_cpu_750fx) 53_GLOBAL(__setup_cpu_750fx)
54 mflr r4 54 mflr r5
55 bl __init_fpu_registers 55 bl __init_fpu_registers
56 bl setup_common_caches 56 bl setup_common_caches
57 bl setup_750_7400_hid0 57 bl setup_750_7400_hid0
58 bl setup_750fx 58 bl setup_750fx
59 mtlr r4 59 mtlr r5
60 blr 60 blr
61_GLOBAL(__setup_cpu_7400) 61_GLOBAL(__setup_cpu_7400)
62 mflr r4 62 mflr r5
63 bl __init_fpu_registers 63 bl __init_fpu_registers
64 bl setup_7400_workarounds 64 bl setup_7400_workarounds
65 bl setup_common_caches 65 bl setup_common_caches
66 bl setup_750_7400_hid0 66 bl setup_750_7400_hid0
67 mtlr r4 67 mtlr r5
68 blr 68 blr
69_GLOBAL(__setup_cpu_7410) 69_GLOBAL(__setup_cpu_7410)
70 mflr r4 70 mflr r5
71 bl __init_fpu_registers 71 bl __init_fpu_registers
72 bl setup_7410_workarounds 72 bl setup_7410_workarounds
73 bl setup_common_caches 73 bl setup_common_caches
74 bl setup_750_7400_hid0 74 bl setup_750_7400_hid0
75 li r3,0 75 li r3,0
76 mtspr SPRN_L2CR2,r3 76 mtspr SPRN_L2CR2,r3
77 mtlr r4 77 mtlr r5
78 blr 78 blr
79_GLOBAL(__setup_cpu_745x) 79_GLOBAL(__setup_cpu_745x)
80 mflr r4 80 mflr r5
81 bl setup_common_caches 81 bl setup_common_caches
82 bl setup_745x_specifics 82 bl setup_745x_specifics
83 mtlr r4 83 mtlr r5
84 blr 84 blr
85 85
86/* Enable caches for 603's, 604, 750 & 7400 */ 86/* Enable caches for 603's, 604, 750 & 7400 */
@@ -194,10 +194,10 @@ setup_750cx:
194 cror 4*cr0+eq,4*cr0+eq,4*cr1+eq 194 cror 4*cr0+eq,4*cr0+eq,4*cr1+eq
195 cror 4*cr0+eq,4*cr0+eq,4*cr2+eq 195 cror 4*cr0+eq,4*cr0+eq,4*cr2+eq
196 bnelr 196 bnelr
197 lwz r6,CPU_SPEC_FEATURES(r5) 197 lwz r6,CPU_SPEC_FEATURES(r4)
198 li r7,CPU_FTR_CAN_NAP 198 li r7,CPU_FTR_CAN_NAP
199 andc r6,r6,r7 199 andc r6,r6,r7
200 stw r6,CPU_SPEC_FEATURES(r5) 200 stw r6,CPU_SPEC_FEATURES(r4)
201 blr 201 blr
202 202
203/* 750fx specific 203/* 750fx specific
@@ -225,12 +225,12 @@ BEGIN_FTR_SECTION
225 andis. r11,r11,L3CR_L3E@h 225 andis. r11,r11,L3CR_L3E@h
226 beq 1f 226 beq 1f
227END_FTR_SECTION_IFSET(CPU_FTR_L3CR) 227END_FTR_SECTION_IFSET(CPU_FTR_L3CR)
228 lwz r6,CPU_SPEC_FEATURES(r5) 228 lwz r6,CPU_SPEC_FEATURES(r4)
229 andi. r0,r6,CPU_FTR_L3_DISABLE_NAP 229 andi. r0,r6,CPU_FTR_L3_DISABLE_NAP
230 beq 1f 230 beq 1f
231 li r7,CPU_FTR_CAN_NAP 231 li r7,CPU_FTR_CAN_NAP
232 andc r6,r6,r7 232 andc r6,r6,r7
233 stw r6,CPU_SPEC_FEATURES(r5) 233 stw r6,CPU_SPEC_FEATURES(r4)
2341: 2341:
235 mfspr r11,SPRN_HID0 235 mfspr r11,SPRN_HID0
236 236
diff --git a/arch/powerpc/kernel/cpu_setup_a2.S b/arch/powerpc/kernel/cpu_setup_a2.S
new file mode 100644
index 000000000000..7f818feaa7a5
--- /dev/null
+++ b/arch/powerpc/kernel/cpu_setup_a2.S
@@ -0,0 +1,114 @@
1/*
2 * A2 specific assembly support code
3 *
4 * Copyright 2009 Ben Herrenschmidt, IBM Corp.
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11
12#include <asm/asm-offsets.h>
13#include <asm/ppc_asm.h>
14#include <asm/ppc-opcode.h>
15#include <asm/processor.h>
16#include <asm/reg_a2.h>
17#include <asm/reg.h>
18#include <asm/thread_info.h>
19
20/*
21 * Disable thdid and class fields in ERATs to bump PID to full 14 bits capacity.
22 * This also prevents external LPID accesses but that isn't a problem when not a
23 * guest. Under PV, this setting will be ignored and MMUCR will return the right
24 * number of PID bits we can use.
25 */
26#define MMUCR1_EXTEND_PID \
27 (MMUCR1_ICTID | MMUCR1_ITTID | MMUCR1_DCTID | \
28 MMUCR1_DTTID | MMUCR1_DCCD)
29
30/*
31 * Use extended PIDs if enabled.
32 * Don't clear the ERATs on context sync events and enable I & D LRU.
33 * Enable ERAT back invalidate when tlbwe overwrites an entry.
34 */
35#define INITIAL_MMUCR1 \
36 (MMUCR1_EXTEND_PID | MMUCR1_CSINV_NEVER | MMUCR1_IRRE | \
37 MMUCR1_DRRE | MMUCR1_TLBWE_BINV)
38
39_GLOBAL(__setup_cpu_a2)
40 /* Some of these are actually thread local and some are
41 * core local but doing it always won't hurt
42 */
43
44#ifdef CONFIG_PPC_WSP_COPRO
45 /* Make sure ACOP starts out as zero */
46 li r3,0
47 mtspr SPRN_ACOP,r3
48
49 /* Enable icswx instruction */
50 mfspr r3,SPRN_A2_CCR2
51 ori r3,r3,A2_CCR2_ENABLE_ICSWX
52 mtspr SPRN_A2_CCR2,r3
53
54 /* Unmask all CTs in HACOP */
55 li r3,-1
56 mtspr SPRN_HACOP,r3
57#endif /* CONFIG_PPC_WSP_COPRO */
58
59 /* Enable doorbell */
60 mfspr r3,SPRN_A2_CCR2
61 oris r3,r3,A2_CCR2_ENABLE_PC@h
62 mtspr SPRN_A2_CCR2,r3
63 isync
64
65 /* Setup CCR0 to disable power saving for now as it's busted
66 * in the current implementations. Setup CCR1 to wake on
67 * interrupts normally (we write the default value but who
68 * knows what FW may have clobbered...)
69 */
70 li r3,0
71 mtspr SPRN_A2_CCR0, r3
72 LOAD_REG_IMMEDIATE(r3,0x0f0f0f0f)
73 mtspr SPRN_A2_CCR1, r3
74
75 /* Initialise MMUCR1 */
76 lis r3,INITIAL_MMUCR1@h
77 ori r3,r3,INITIAL_MMUCR1@l
78 mtspr SPRN_MMUCR1,r3
79
80 /* Set MMUCR2 to enable 4K, 64K, 1M, 16M and 1G pages */
81 LOAD_REG_IMMEDIATE(r3, 0x000a7531)
82 mtspr SPRN_MMUCR2,r3
83
84 /* Set MMUCR3 to write all thids bit to the TLB */
85 LOAD_REG_IMMEDIATE(r3, 0x0000000f)
86 mtspr SPRN_MMUCR3,r3
87
88 /* Don't do ERAT stuff if running guest mode */
89 mfmsr r3
90 andis. r0,r3,MSR_GS@h
91 bne 1f
92
93 /* Now set the I-ERAT watermark to 15 */
94 lis r4,(MMUCR0_TLBSEL_I|MMUCR0_ECL)@h
95 mtspr SPRN_MMUCR0, r4
96 li r4,A2_IERAT_SIZE-1
97 PPC_ERATWE(r4,r4,3)
98
99 /* Now set the D-ERAT watermark to 31 */
100 lis r4,(MMUCR0_TLBSEL_D|MMUCR0_ECL)@h
101 mtspr SPRN_MMUCR0, r4
102 li r4,A2_DERAT_SIZE-1
103 PPC_ERATWE(r4,r4,3)
104
105 /* And invalidate the beast just in case. That won't get rid of
106 * a bolted entry though it will be in LRU and so will go away eventually
107 * but let's not bother for now
108 */
109 PPC_ERATILX(0,0,0)
1101:
111 blr
112
113_GLOBAL(__restore_cpu_a2)
114 b __setup_cpu_a2
diff --git a/arch/powerpc/kernel/cpu_setup_fsl_booke.S b/arch/powerpc/kernel/cpu_setup_fsl_booke.S
index 0adb50ad8031..8053db02b85e 100644
--- a/arch/powerpc/kernel/cpu_setup_fsl_booke.S
+++ b/arch/powerpc/kernel/cpu_setup_fsl_booke.S
@@ -51,6 +51,7 @@ _GLOBAL(__e500_dcache_setup)
51 isync 51 isync
52 blr 52 blr
53 53
54#ifdef CONFIG_PPC32
54_GLOBAL(__setup_cpu_e200) 55_GLOBAL(__setup_cpu_e200)
55 /* enable dedicated debug exception handling resources (Debug APU) */ 56 /* enable dedicated debug exception handling resources (Debug APU) */
56 mfspr r3,SPRN_HID0 57 mfspr r3,SPRN_HID0
@@ -63,6 +64,12 @@ _GLOBAL(__setup_cpu_e500v2)
63 bl __e500_icache_setup 64 bl __e500_icache_setup
64 bl __e500_dcache_setup 65 bl __e500_dcache_setup
65 bl __setup_e500_ivors 66 bl __setup_e500_ivors
67#ifdef CONFIG_FSL_RIO
68 /* Ensure that RFXE is set */
69 mfspr r3,SPRN_HID1
70 oris r3,r3,HID1_RFXE@h
71 mtspr SPRN_HID1,r3
72#endif
66 mtlr r4 73 mtlr r4
67 blr 74 blr
68_GLOBAL(__setup_cpu_e500mc) 75_GLOBAL(__setup_cpu_e500mc)
@@ -72,3 +79,20 @@ _GLOBAL(__setup_cpu_e500mc)
72 bl __setup_e500mc_ivors 79 bl __setup_e500mc_ivors
73 mtlr r4 80 mtlr r4
74 blr 81 blr
82#endif
83/* Right now, restore and setup are the same thing */
84_GLOBAL(__restore_cpu_e5500)
85_GLOBAL(__setup_cpu_e5500)
86 mflr r4
87 bl __e500_icache_setup
88 bl __e500_dcache_setup
89#ifdef CONFIG_PPC_BOOK3E_64
90 bl .__setup_base_ivors
91 bl .setup_perfmon_ivor
92 bl .setup_doorbell_ivors
93 bl .setup_ehv_ivors
94#else
95 bl __setup_e500mc_ivors
96#endif
97 mtlr r4
98 blr
diff --git a/arch/powerpc/kernel/cpu_setup_power7.S b/arch/powerpc/kernel/cpu_setup_power7.S
new file mode 100644
index 000000000000..4f9a93fcfe07
--- /dev/null
+++ b/arch/powerpc/kernel/cpu_setup_power7.S
@@ -0,0 +1,91 @@
1/*
2 * This file contains low level CPU setup functions.
3 * Copyright (C) 2003 Benjamin Herrenschmidt (benh@kernel.crashing.org)
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License
7 * as published by the Free Software Foundation; either version
8 * 2 of the License, or (at your option) any later version.
9 *
10 */
11
12#include <asm/processor.h>
13#include <asm/page.h>
14#include <asm/cputable.h>
15#include <asm/ppc_asm.h>
16#include <asm/asm-offsets.h>
17#include <asm/cache.h>
18
19/* Entry: r3 = crap, r4 = ptr to cputable entry
20 *
21 * Note that we can be called twice for pseudo-PVRs
22 */
23_GLOBAL(__setup_cpu_power7)
24 mflr r11
25 bl __init_hvmode_206
26 mtlr r11
27 beqlr
28 li r0,0
29 mtspr SPRN_LPID,r0
30 bl __init_LPCR
31 bl __init_TLB
32 mtlr r11
33 blr
34
35_GLOBAL(__restore_cpu_power7)
36 mflr r11
37 mfmsr r3
38 rldicl. r0,r3,4,63
39 beqlr
40 li r0,0
41 mtspr SPRN_LPID,r0
42 bl __init_LPCR
43 bl __init_TLB
44 mtlr r11
45 blr
46
47__init_hvmode_206:
48 /* Disable CPU_FTR_HVMODE_206 and exit if MSR:HV is not set */
49 mfmsr r3
50 rldicl. r0,r3,4,63
51 bnelr
52 ld r5,CPU_SPEC_FEATURES(r4)
53 LOAD_REG_IMMEDIATE(r6,CPU_FTR_HVMODE_206)
54 xor r5,r5,r6
55 std r5,CPU_SPEC_FEATURES(r4)
56 blr
57
58__init_LPCR:
59 /* Setup a sane LPCR:
60 *
61 * LPES = 0b01 (HSRR0/1 used for 0x500)
62 * PECE = 0b111
63 * DPFD = 4
64 *
65 * Other bits untouched for now
66 */
67 mfspr r3,SPRN_LPCR
68 ori r3,r3,(LPCR_LPES0|LPCR_LPES1)
69 xori r3,r3, LPCR_LPES0
70 ori r3,r3,(LPCR_PECE0|LPCR_PECE1|LPCR_PECE2)
71 li r5,7
72 sldi r5,r5,LPCR_DPFD_SH
73 andc r3,r3,r5
74 li r5,4
75 sldi r5,r5,LPCR_DPFD_SH
76 or r3,r3,r5
77 mtspr SPRN_LPCR,r3
78 isync
79 blr
80
81__init_TLB:
82 /* Clear the TLB */
83 li r6,128
84 mtctr r6
85 li r7,0xc00 /* IS field = 0b11 */
86 ptesync
872: tlbiel r7
88 addi r7,r7,0x1000
89 bdnz 2b
90 ptesync
911: blr
diff --git a/arch/powerpc/kernel/cputable.c b/arch/powerpc/kernel/cputable.c
index 1f9123f412ec..9fb933248ab6 100644
--- a/arch/powerpc/kernel/cputable.c
+++ b/arch/powerpc/kernel/cputable.c
@@ -48,6 +48,7 @@ extern void __setup_cpu_440x5(unsigned long offset, struct cpu_spec* spec);
48extern void __setup_cpu_460ex(unsigned long offset, struct cpu_spec* spec); 48extern void __setup_cpu_460ex(unsigned long offset, struct cpu_spec* spec);
49extern void __setup_cpu_460gt(unsigned long offset, struct cpu_spec* spec); 49extern void __setup_cpu_460gt(unsigned long offset, struct cpu_spec* spec);
50extern void __setup_cpu_460sx(unsigned long offset, struct cpu_spec *spec); 50extern void __setup_cpu_460sx(unsigned long offset, struct cpu_spec *spec);
51extern void __setup_cpu_apm821xx(unsigned long offset, struct cpu_spec *spec);
51extern void __setup_cpu_603(unsigned long offset, struct cpu_spec* spec); 52extern void __setup_cpu_603(unsigned long offset, struct cpu_spec* spec);
52extern void __setup_cpu_604(unsigned long offset, struct cpu_spec* spec); 53extern void __setup_cpu_604(unsigned long offset, struct cpu_spec* spec);
53extern void __setup_cpu_750(unsigned long offset, struct cpu_spec* spec); 54extern void __setup_cpu_750(unsigned long offset, struct cpu_spec* spec);
@@ -61,11 +62,17 @@ extern void __setup_cpu_745x(unsigned long offset, struct cpu_spec* spec);
61extern void __setup_cpu_ppc970(unsigned long offset, struct cpu_spec* spec); 62extern void __setup_cpu_ppc970(unsigned long offset, struct cpu_spec* spec);
62extern void __setup_cpu_ppc970MP(unsigned long offset, struct cpu_spec* spec); 63extern void __setup_cpu_ppc970MP(unsigned long offset, struct cpu_spec* spec);
63extern void __setup_cpu_pa6t(unsigned long offset, struct cpu_spec* spec); 64extern void __setup_cpu_pa6t(unsigned long offset, struct cpu_spec* spec);
65extern void __setup_cpu_a2(unsigned long offset, struct cpu_spec* spec);
64extern void __restore_cpu_pa6t(void); 66extern void __restore_cpu_pa6t(void);
65extern void __restore_cpu_ppc970(void); 67extern void __restore_cpu_ppc970(void);
66extern void __setup_cpu_power7(unsigned long offset, struct cpu_spec* spec); 68extern void __setup_cpu_power7(unsigned long offset, struct cpu_spec* spec);
67extern void __restore_cpu_power7(void); 69extern void __restore_cpu_power7(void);
70extern void __restore_cpu_a2(void);
68#endif /* CONFIG_PPC64 */ 71#endif /* CONFIG_PPC64 */
72#if defined(CONFIG_E500)
73extern void __setup_cpu_e5500(unsigned long offset, struct cpu_spec* spec);
74extern void __restore_cpu_e5500(void);
75#endif /* CONFIG_E500 */
69 76
70/* This table only contains "desktop" CPUs, it need to be filled with embedded 77/* This table only contains "desktop" CPUs, it need to be filled with embedded
71 * ones as well... 78 * ones as well...
@@ -111,7 +118,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
111 .pmc_type = PPC_PMC_IBM, 118 .pmc_type = PPC_PMC_IBM,
112 .oprofile_cpu_type = "ppc64/power3", 119 .oprofile_cpu_type = "ppc64/power3",
113 .oprofile_type = PPC_OPROFILE_RS64, 120 .oprofile_type = PPC_OPROFILE_RS64,
114 .machine_check = machine_check_generic,
115 .platform = "power3", 121 .platform = "power3",
116 }, 122 },
117 { /* Power3+ */ 123 { /* Power3+ */
@@ -127,7 +133,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
127 .pmc_type = PPC_PMC_IBM, 133 .pmc_type = PPC_PMC_IBM,
128 .oprofile_cpu_type = "ppc64/power3", 134 .oprofile_cpu_type = "ppc64/power3",
129 .oprofile_type = PPC_OPROFILE_RS64, 135 .oprofile_type = PPC_OPROFILE_RS64,
130 .machine_check = machine_check_generic,
131 .platform = "power3", 136 .platform = "power3",
132 }, 137 },
133 { /* Northstar */ 138 { /* Northstar */
@@ -143,7 +148,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
143 .pmc_type = PPC_PMC_IBM, 148 .pmc_type = PPC_PMC_IBM,
144 .oprofile_cpu_type = "ppc64/rs64", 149 .oprofile_cpu_type = "ppc64/rs64",
145 .oprofile_type = PPC_OPROFILE_RS64, 150 .oprofile_type = PPC_OPROFILE_RS64,
146 .machine_check = machine_check_generic,
147 .platform = "rs64", 151 .platform = "rs64",
148 }, 152 },
149 { /* Pulsar */ 153 { /* Pulsar */
@@ -159,7 +163,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
159 .pmc_type = PPC_PMC_IBM, 163 .pmc_type = PPC_PMC_IBM,
160 .oprofile_cpu_type = "ppc64/rs64", 164 .oprofile_cpu_type = "ppc64/rs64",
161 .oprofile_type = PPC_OPROFILE_RS64, 165 .oprofile_type = PPC_OPROFILE_RS64,
162 .machine_check = machine_check_generic,
163 .platform = "rs64", 166 .platform = "rs64",
164 }, 167 },
165 { /* I-star */ 168 { /* I-star */
@@ -175,7 +178,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
175 .pmc_type = PPC_PMC_IBM, 178 .pmc_type = PPC_PMC_IBM,
176 .oprofile_cpu_type = "ppc64/rs64", 179 .oprofile_cpu_type = "ppc64/rs64",
177 .oprofile_type = PPC_OPROFILE_RS64, 180 .oprofile_type = PPC_OPROFILE_RS64,
178 .machine_check = machine_check_generic,
179 .platform = "rs64", 181 .platform = "rs64",
180 }, 182 },
181 { /* S-star */ 183 { /* S-star */
@@ -191,7 +193,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
191 .pmc_type = PPC_PMC_IBM, 193 .pmc_type = PPC_PMC_IBM,
192 .oprofile_cpu_type = "ppc64/rs64", 194 .oprofile_cpu_type = "ppc64/rs64",
193 .oprofile_type = PPC_OPROFILE_RS64, 195 .oprofile_type = PPC_OPROFILE_RS64,
194 .machine_check = machine_check_generic,
195 .platform = "rs64", 196 .platform = "rs64",
196 }, 197 },
197 { /* Power4 */ 198 { /* Power4 */
@@ -200,14 +201,13 @@ static struct cpu_spec __initdata cpu_specs[] = {
200 .cpu_name = "POWER4 (gp)", 201 .cpu_name = "POWER4 (gp)",
201 .cpu_features = CPU_FTRS_POWER4, 202 .cpu_features = CPU_FTRS_POWER4,
202 .cpu_user_features = COMMON_USER_POWER4, 203 .cpu_user_features = COMMON_USER_POWER4,
203 .mmu_features = MMU_FTR_HPTE_TABLE, 204 .mmu_features = MMU_FTRS_POWER4,
204 .icache_bsize = 128, 205 .icache_bsize = 128,
205 .dcache_bsize = 128, 206 .dcache_bsize = 128,
206 .num_pmcs = 8, 207 .num_pmcs = 8,
207 .pmc_type = PPC_PMC_IBM, 208 .pmc_type = PPC_PMC_IBM,
208 .oprofile_cpu_type = "ppc64/power4", 209 .oprofile_cpu_type = "ppc64/power4",
209 .oprofile_type = PPC_OPROFILE_POWER4, 210 .oprofile_type = PPC_OPROFILE_POWER4,
210 .machine_check = machine_check_generic,
211 .platform = "power4", 211 .platform = "power4",
212 }, 212 },
213 { /* Power4+ */ 213 { /* Power4+ */
@@ -216,14 +216,13 @@ static struct cpu_spec __initdata cpu_specs[] = {
216 .cpu_name = "POWER4+ (gq)", 216 .cpu_name = "POWER4+ (gq)",
217 .cpu_features = CPU_FTRS_POWER4, 217 .cpu_features = CPU_FTRS_POWER4,
218 .cpu_user_features = COMMON_USER_POWER4, 218 .cpu_user_features = COMMON_USER_POWER4,
219 .mmu_features = MMU_FTR_HPTE_TABLE, 219 .mmu_features = MMU_FTRS_POWER4,
220 .icache_bsize = 128, 220 .icache_bsize = 128,
221 .dcache_bsize = 128, 221 .dcache_bsize = 128,
222 .num_pmcs = 8, 222 .num_pmcs = 8,
223 .pmc_type = PPC_PMC_IBM, 223 .pmc_type = PPC_PMC_IBM,
224 .oprofile_cpu_type = "ppc64/power4", 224 .oprofile_cpu_type = "ppc64/power4",
225 .oprofile_type = PPC_OPROFILE_POWER4, 225 .oprofile_type = PPC_OPROFILE_POWER4,
226 .machine_check = machine_check_generic,
227 .platform = "power4", 226 .platform = "power4",
228 }, 227 },
229 { /* PPC970 */ 228 { /* PPC970 */
@@ -233,7 +232,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
233 .cpu_features = CPU_FTRS_PPC970, 232 .cpu_features = CPU_FTRS_PPC970,
234 .cpu_user_features = COMMON_USER_POWER4 | 233 .cpu_user_features = COMMON_USER_POWER4 |
235 PPC_FEATURE_HAS_ALTIVEC_COMP, 234 PPC_FEATURE_HAS_ALTIVEC_COMP,
236 .mmu_features = MMU_FTR_HPTE_TABLE, 235 .mmu_features = MMU_FTRS_PPC970,
237 .icache_bsize = 128, 236 .icache_bsize = 128,
238 .dcache_bsize = 128, 237 .dcache_bsize = 128,
239 .num_pmcs = 8, 238 .num_pmcs = 8,
@@ -242,7 +241,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
242 .cpu_restore = __restore_cpu_ppc970, 241 .cpu_restore = __restore_cpu_ppc970,
243 .oprofile_cpu_type = "ppc64/970", 242 .oprofile_cpu_type = "ppc64/970",
244 .oprofile_type = PPC_OPROFILE_POWER4, 243 .oprofile_type = PPC_OPROFILE_POWER4,
245 .machine_check = machine_check_generic,
246 .platform = "ppc970", 244 .platform = "ppc970",
247 }, 245 },
248 { /* PPC970FX */ 246 { /* PPC970FX */
@@ -252,7 +250,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
252 .cpu_features = CPU_FTRS_PPC970, 250 .cpu_features = CPU_FTRS_PPC970,
253 .cpu_user_features = COMMON_USER_POWER4 | 251 .cpu_user_features = COMMON_USER_POWER4 |
254 PPC_FEATURE_HAS_ALTIVEC_COMP, 252 PPC_FEATURE_HAS_ALTIVEC_COMP,
255 .mmu_features = MMU_FTR_HPTE_TABLE, 253 .mmu_features = MMU_FTRS_PPC970,
256 .icache_bsize = 128, 254 .icache_bsize = 128,
257 .dcache_bsize = 128, 255 .dcache_bsize = 128,
258 .num_pmcs = 8, 256 .num_pmcs = 8,
@@ -261,7 +259,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
261 .cpu_restore = __restore_cpu_ppc970, 259 .cpu_restore = __restore_cpu_ppc970,
262 .oprofile_cpu_type = "ppc64/970", 260 .oprofile_cpu_type = "ppc64/970",
263 .oprofile_type = PPC_OPROFILE_POWER4, 261 .oprofile_type = PPC_OPROFILE_POWER4,
264 .machine_check = machine_check_generic,
265 .platform = "ppc970", 262 .platform = "ppc970",
266 }, 263 },
267 { /* PPC970MP DD1.0 - no DEEPNAP, use regular 970 init */ 264 { /* PPC970MP DD1.0 - no DEEPNAP, use regular 970 init */
@@ -280,7 +277,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
280 .cpu_restore = __restore_cpu_ppc970, 277 .cpu_restore = __restore_cpu_ppc970,
281 .oprofile_cpu_type = "ppc64/970MP", 278 .oprofile_cpu_type = "ppc64/970MP",
282 .oprofile_type = PPC_OPROFILE_POWER4, 279 .oprofile_type = PPC_OPROFILE_POWER4,
283 .machine_check = machine_check_generic,
284 .platform = "ppc970", 280 .platform = "ppc970",
285 }, 281 },
286 { /* PPC970MP */ 282 { /* PPC970MP */
@@ -290,7 +286,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
290 .cpu_features = CPU_FTRS_PPC970, 286 .cpu_features = CPU_FTRS_PPC970,
291 .cpu_user_features = COMMON_USER_POWER4 | 287 .cpu_user_features = COMMON_USER_POWER4 |
292 PPC_FEATURE_HAS_ALTIVEC_COMP, 288 PPC_FEATURE_HAS_ALTIVEC_COMP,
293 .mmu_features = MMU_FTR_HPTE_TABLE, 289 .mmu_features = MMU_FTRS_PPC970,
294 .icache_bsize = 128, 290 .icache_bsize = 128,
295 .dcache_bsize = 128, 291 .dcache_bsize = 128,
296 .num_pmcs = 8, 292 .num_pmcs = 8,
@@ -299,7 +295,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
299 .cpu_restore = __restore_cpu_ppc970, 295 .cpu_restore = __restore_cpu_ppc970,
300 .oprofile_cpu_type = "ppc64/970MP", 296 .oprofile_cpu_type = "ppc64/970MP",
301 .oprofile_type = PPC_OPROFILE_POWER4, 297 .oprofile_type = PPC_OPROFILE_POWER4,
302 .machine_check = machine_check_generic,
303 .platform = "ppc970", 298 .platform = "ppc970",
304 }, 299 },
305 { /* PPC970GX */ 300 { /* PPC970GX */
@@ -309,7 +304,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
309 .cpu_features = CPU_FTRS_PPC970, 304 .cpu_features = CPU_FTRS_PPC970,
310 .cpu_user_features = COMMON_USER_POWER4 | 305 .cpu_user_features = COMMON_USER_POWER4 |
311 PPC_FEATURE_HAS_ALTIVEC_COMP, 306 PPC_FEATURE_HAS_ALTIVEC_COMP,
312 .mmu_features = MMU_FTR_HPTE_TABLE, 307 .mmu_features = MMU_FTRS_PPC970,
313 .icache_bsize = 128, 308 .icache_bsize = 128,
314 .dcache_bsize = 128, 309 .dcache_bsize = 128,
315 .num_pmcs = 8, 310 .num_pmcs = 8,
@@ -317,7 +312,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
317 .cpu_setup = __setup_cpu_ppc970, 312 .cpu_setup = __setup_cpu_ppc970,
318 .oprofile_cpu_type = "ppc64/970", 313 .oprofile_cpu_type = "ppc64/970",
319 .oprofile_type = PPC_OPROFILE_POWER4, 314 .oprofile_type = PPC_OPROFILE_POWER4,
320 .machine_check = machine_check_generic,
321 .platform = "ppc970", 315 .platform = "ppc970",
322 }, 316 },
323 { /* Power5 GR */ 317 { /* Power5 GR */
@@ -326,7 +320,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
326 .cpu_name = "POWER5 (gr)", 320 .cpu_name = "POWER5 (gr)",
327 .cpu_features = CPU_FTRS_POWER5, 321 .cpu_features = CPU_FTRS_POWER5,
328 .cpu_user_features = COMMON_USER_POWER5, 322 .cpu_user_features = COMMON_USER_POWER5,
329 .mmu_features = MMU_FTR_HPTE_TABLE, 323 .mmu_features = MMU_FTRS_POWER5,
330 .icache_bsize = 128, 324 .icache_bsize = 128,
331 .dcache_bsize = 128, 325 .dcache_bsize = 128,
332 .num_pmcs = 6, 326 .num_pmcs = 6,
@@ -338,7 +332,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
338 */ 332 */
339 .oprofile_mmcra_sihv = MMCRA_SIHV, 333 .oprofile_mmcra_sihv = MMCRA_SIHV,
340 .oprofile_mmcra_sipr = MMCRA_SIPR, 334 .oprofile_mmcra_sipr = MMCRA_SIPR,
341 .machine_check = machine_check_generic,
342 .platform = "power5", 335 .platform = "power5",
343 }, 336 },
344 { /* Power5++ */ 337 { /* Power5++ */
@@ -347,7 +340,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
347 .cpu_name = "POWER5+ (gs)", 340 .cpu_name = "POWER5+ (gs)",
348 .cpu_features = CPU_FTRS_POWER5, 341 .cpu_features = CPU_FTRS_POWER5,
349 .cpu_user_features = COMMON_USER_POWER5_PLUS, 342 .cpu_user_features = COMMON_USER_POWER5_PLUS,
350 .mmu_features = MMU_FTR_HPTE_TABLE, 343 .mmu_features = MMU_FTRS_POWER5,
351 .icache_bsize = 128, 344 .icache_bsize = 128,
352 .dcache_bsize = 128, 345 .dcache_bsize = 128,
353 .num_pmcs = 6, 346 .num_pmcs = 6,
@@ -355,7 +348,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
355 .oprofile_type = PPC_OPROFILE_POWER4, 348 .oprofile_type = PPC_OPROFILE_POWER4,
356 .oprofile_mmcra_sihv = MMCRA_SIHV, 349 .oprofile_mmcra_sihv = MMCRA_SIHV,
357 .oprofile_mmcra_sipr = MMCRA_SIPR, 350 .oprofile_mmcra_sipr = MMCRA_SIPR,
358 .machine_check = machine_check_generic,
359 .platform = "power5+", 351 .platform = "power5+",
360 }, 352 },
361 { /* Power5 GS */ 353 { /* Power5 GS */
@@ -364,7 +356,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
364 .cpu_name = "POWER5+ (gs)", 356 .cpu_name = "POWER5+ (gs)",
365 .cpu_features = CPU_FTRS_POWER5, 357 .cpu_features = CPU_FTRS_POWER5,
366 .cpu_user_features = COMMON_USER_POWER5_PLUS, 358 .cpu_user_features = COMMON_USER_POWER5_PLUS,
367 .mmu_features = MMU_FTR_HPTE_TABLE, 359 .mmu_features = MMU_FTRS_POWER5,
368 .icache_bsize = 128, 360 .icache_bsize = 128,
369 .dcache_bsize = 128, 361 .dcache_bsize = 128,
370 .num_pmcs = 6, 362 .num_pmcs = 6,
@@ -373,7 +365,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
373 .oprofile_type = PPC_OPROFILE_POWER4, 365 .oprofile_type = PPC_OPROFILE_POWER4,
374 .oprofile_mmcra_sihv = MMCRA_SIHV, 366 .oprofile_mmcra_sihv = MMCRA_SIHV,
375 .oprofile_mmcra_sipr = MMCRA_SIPR, 367 .oprofile_mmcra_sipr = MMCRA_SIPR,
376 .machine_check = machine_check_generic,
377 .platform = "power5+", 368 .platform = "power5+",
378 }, 369 },
379 { /* POWER6 in P5+ mode; 2.04-compliant processor */ 370 { /* POWER6 in P5+ mode; 2.04-compliant processor */
@@ -382,10 +373,9 @@ static struct cpu_spec __initdata cpu_specs[] = {
382 .cpu_name = "POWER5+", 373 .cpu_name = "POWER5+",
383 .cpu_features = CPU_FTRS_POWER5, 374 .cpu_features = CPU_FTRS_POWER5,
384 .cpu_user_features = COMMON_USER_POWER5_PLUS, 375 .cpu_user_features = COMMON_USER_POWER5_PLUS,
385 .mmu_features = MMU_FTR_HPTE_TABLE, 376 .mmu_features = MMU_FTRS_POWER5,
386 .icache_bsize = 128, 377 .icache_bsize = 128,
387 .dcache_bsize = 128, 378 .dcache_bsize = 128,
388 .machine_check = machine_check_generic,
389 .oprofile_cpu_type = "ppc64/ibm-compat-v1", 379 .oprofile_cpu_type = "ppc64/ibm-compat-v1",
390 .oprofile_type = PPC_OPROFILE_POWER4, 380 .oprofile_type = PPC_OPROFILE_POWER4,
391 .platform = "power5+", 381 .platform = "power5+",
@@ -397,7 +387,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
397 .cpu_features = CPU_FTRS_POWER6, 387 .cpu_features = CPU_FTRS_POWER6,
398 .cpu_user_features = COMMON_USER_POWER6 | 388 .cpu_user_features = COMMON_USER_POWER6 |
399 PPC_FEATURE_POWER6_EXT, 389 PPC_FEATURE_POWER6_EXT,
400 .mmu_features = MMU_FTR_HPTE_TABLE, 390 .mmu_features = MMU_FTRS_POWER6,
401 .icache_bsize = 128, 391 .icache_bsize = 128,
402 .dcache_bsize = 128, 392 .dcache_bsize = 128,
403 .num_pmcs = 6, 393 .num_pmcs = 6,
@@ -408,7 +398,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
408 .oprofile_mmcra_sipr = POWER6_MMCRA_SIPR, 398 .oprofile_mmcra_sipr = POWER6_MMCRA_SIPR,
409 .oprofile_mmcra_clear = POWER6_MMCRA_THRM | 399 .oprofile_mmcra_clear = POWER6_MMCRA_THRM |
410 POWER6_MMCRA_OTHER, 400 POWER6_MMCRA_OTHER,
411 .machine_check = machine_check_generic,
412 .platform = "power6x", 401 .platform = "power6x",
413 }, 402 },
414 { /* 2.05-compliant processor, i.e. Power6 "architected" mode */ 403 { /* 2.05-compliant processor, i.e. Power6 "architected" mode */
@@ -417,10 +406,9 @@ static struct cpu_spec __initdata cpu_specs[] = {
417 .cpu_name = "POWER6 (architected)", 406 .cpu_name = "POWER6 (architected)",
418 .cpu_features = CPU_FTRS_POWER6, 407 .cpu_features = CPU_FTRS_POWER6,
419 .cpu_user_features = COMMON_USER_POWER6, 408 .cpu_user_features = COMMON_USER_POWER6,
420 .mmu_features = MMU_FTR_HPTE_TABLE, 409 .mmu_features = MMU_FTRS_POWER6,
421 .icache_bsize = 128, 410 .icache_bsize = 128,
422 .dcache_bsize = 128, 411 .dcache_bsize = 128,
423 .machine_check = machine_check_generic,
424 .oprofile_cpu_type = "ppc64/ibm-compat-v1", 412 .oprofile_cpu_type = "ppc64/ibm-compat-v1",
425 .oprofile_type = PPC_OPROFILE_POWER4, 413 .oprofile_type = PPC_OPROFILE_POWER4,
426 .platform = "power6", 414 .platform = "power6",
@@ -431,13 +419,13 @@ static struct cpu_spec __initdata cpu_specs[] = {
431 .cpu_name = "POWER7 (architected)", 419 .cpu_name = "POWER7 (architected)",
432 .cpu_features = CPU_FTRS_POWER7, 420 .cpu_features = CPU_FTRS_POWER7,
433 .cpu_user_features = COMMON_USER_POWER7, 421 .cpu_user_features = COMMON_USER_POWER7,
434 .mmu_features = MMU_FTR_HPTE_TABLE | 422 .mmu_features = MMU_FTRS_POWER7,
435 MMU_FTR_TLBIE_206,
436 .icache_bsize = 128, 423 .icache_bsize = 128,
437 .dcache_bsize = 128, 424 .dcache_bsize = 128,
438 .machine_check = machine_check_generic,
439 .oprofile_type = PPC_OPROFILE_POWER4, 425 .oprofile_type = PPC_OPROFILE_POWER4,
440 .oprofile_cpu_type = "ppc64/ibm-compat-v1", 426 .oprofile_cpu_type = "ppc64/ibm-compat-v1",
427 .cpu_setup = __setup_cpu_power7,
428 .cpu_restore = __restore_cpu_power7,
441 .platform = "power7", 429 .platform = "power7",
442 }, 430 },
443 { /* Power7 */ 431 { /* Power7 */
@@ -446,21 +434,33 @@ static struct cpu_spec __initdata cpu_specs[] = {
446 .cpu_name = "POWER7 (raw)", 434 .cpu_name = "POWER7 (raw)",
447 .cpu_features = CPU_FTRS_POWER7, 435 .cpu_features = CPU_FTRS_POWER7,
448 .cpu_user_features = COMMON_USER_POWER7, 436 .cpu_user_features = COMMON_USER_POWER7,
449 .mmu_features = MMU_FTR_HPTE_TABLE | 437 .mmu_features = MMU_FTRS_POWER7,
450 MMU_FTR_TLBIE_206,
451 .icache_bsize = 128, 438 .icache_bsize = 128,
452 .dcache_bsize = 128, 439 .dcache_bsize = 128,
453 .num_pmcs = 6, 440 .num_pmcs = 6,
454 .pmc_type = PPC_PMC_IBM, 441 .pmc_type = PPC_PMC_IBM,
442 .oprofile_cpu_type = "ppc64/power7",
443 .oprofile_type = PPC_OPROFILE_POWER4,
455 .cpu_setup = __setup_cpu_power7, 444 .cpu_setup = __setup_cpu_power7,
456 .cpu_restore = __restore_cpu_power7, 445 .cpu_restore = __restore_cpu_power7,
446 .platform = "power7",
447 },
448 { /* Power7+ */
449 .pvr_mask = 0xffff0000,
450 .pvr_value = 0x004A0000,
451 .cpu_name = "POWER7+ (raw)",
452 .cpu_features = CPU_FTRS_POWER7,
453 .cpu_user_features = COMMON_USER_POWER7,
454 .mmu_features = MMU_FTRS_POWER7,
455 .icache_bsize = 128,
456 .dcache_bsize = 128,
457 .num_pmcs = 6,
458 .pmc_type = PPC_PMC_IBM,
457 .oprofile_cpu_type = "ppc64/power7", 459 .oprofile_cpu_type = "ppc64/power7",
458 .oprofile_type = PPC_OPROFILE_POWER4, 460 .oprofile_type = PPC_OPROFILE_POWER4,
459 .oprofile_mmcra_sihv = POWER6_MMCRA_SIHV, 461 .cpu_setup = __setup_cpu_power7,
460 .oprofile_mmcra_sipr = POWER6_MMCRA_SIPR, 462 .cpu_restore = __restore_cpu_power7,
461 .oprofile_mmcra_clear = POWER6_MMCRA_THRM | 463 .platform = "power7+",
462 POWER6_MMCRA_OTHER,
463 .platform = "power7",
464 }, 464 },
465 { /* Cell Broadband Engine */ 465 { /* Cell Broadband Engine */
466 .pvr_mask = 0xffff0000, 466 .pvr_mask = 0xffff0000,
@@ -470,14 +470,13 @@ static struct cpu_spec __initdata cpu_specs[] = {
470 .cpu_user_features = COMMON_USER_PPC64 | 470 .cpu_user_features = COMMON_USER_PPC64 |
471 PPC_FEATURE_CELL | PPC_FEATURE_HAS_ALTIVEC_COMP | 471 PPC_FEATURE_CELL | PPC_FEATURE_HAS_ALTIVEC_COMP |
472 PPC_FEATURE_SMT, 472 PPC_FEATURE_SMT,
473 .mmu_features = MMU_FTR_HPTE_TABLE, 473 .mmu_features = MMU_FTRS_CELL,
474 .icache_bsize = 128, 474 .icache_bsize = 128,
475 .dcache_bsize = 128, 475 .dcache_bsize = 128,
476 .num_pmcs = 4, 476 .num_pmcs = 4,
477 .pmc_type = PPC_PMC_IBM, 477 .pmc_type = PPC_PMC_IBM,
478 .oprofile_cpu_type = "ppc64/cell-be", 478 .oprofile_cpu_type = "ppc64/cell-be",
479 .oprofile_type = PPC_OPROFILE_CELL, 479 .oprofile_type = PPC_OPROFILE_CELL,
480 .machine_check = machine_check_generic,
481 .platform = "ppc-cell-be", 480 .platform = "ppc-cell-be",
482 }, 481 },
483 { /* PA Semi PA6T */ 482 { /* PA Semi PA6T */
@@ -486,7 +485,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
486 .cpu_name = "PA6T", 485 .cpu_name = "PA6T",
487 .cpu_features = CPU_FTRS_PA6T, 486 .cpu_features = CPU_FTRS_PA6T,
488 .cpu_user_features = COMMON_USER_PA6T, 487 .cpu_user_features = COMMON_USER_PA6T,
489 .mmu_features = MMU_FTR_HPTE_TABLE, 488 .mmu_features = MMU_FTRS_PA6T,
490 .icache_bsize = 64, 489 .icache_bsize = 64,
491 .dcache_bsize = 64, 490 .dcache_bsize = 64,
492 .num_pmcs = 6, 491 .num_pmcs = 6,
@@ -495,7 +494,6 @@ static struct cpu_spec __initdata cpu_specs[] = {
495 .cpu_restore = __restore_cpu_pa6t, 494 .cpu_restore = __restore_cpu_pa6t,
496 .oprofile_cpu_type = "ppc64/pa6t", 495 .oprofile_cpu_type = "ppc64/pa6t",
497 .oprofile_type = PPC_OPROFILE_PA6T, 496 .oprofile_type = PPC_OPROFILE_PA6T,
498 .machine_check = machine_check_generic,
499 .platform = "pa6t", 497 .platform = "pa6t",
500 }, 498 },
501 { /* default match */ 499 { /* default match */
@@ -504,12 +502,11 @@ static struct cpu_spec __initdata cpu_specs[] = {
504 .cpu_name = "POWER4 (compatible)", 502 .cpu_name = "POWER4 (compatible)",
505 .cpu_features = CPU_FTRS_COMPATIBLE, 503 .cpu_features = CPU_FTRS_COMPATIBLE,
506 .cpu_user_features = COMMON_USER_PPC64, 504 .cpu_user_features = COMMON_USER_PPC64,
507 .mmu_features = MMU_FTR_HPTE_TABLE, 505 .mmu_features = MMU_FTRS_DEFAULT_HPTE_ARCH_V2,
508 .icache_bsize = 128, 506 .icache_bsize = 128,
509 .dcache_bsize = 128, 507 .dcache_bsize = 128,
510 .num_pmcs = 6, 508 .num_pmcs = 6,
511 .pmc_type = PPC_PMC_IBM, 509 .pmc_type = PPC_PMC_IBM,
512 .machine_check = machine_check_generic,
513 .platform = "power4", 510 .platform = "power4",
514 } 511 }
515#endif /* CONFIG_PPC_BOOK3S_64 */ 512#endif /* CONFIG_PPC_BOOK3S_64 */
@@ -1805,11 +1802,25 @@ static struct cpu_spec __initdata cpu_specs[] = {
1805 .machine_check = machine_check_440A, 1802 .machine_check = machine_check_440A,
1806 .platform = "ppc440", 1803 .platform = "ppc440",
1807 }, 1804 },
1808 { /* 476 core */ 1805 { /* 464 in APM821xx */
1809 .pvr_mask = 0xffff0000, 1806 .pvr_mask = 0xffffff00,
1810 .pvr_value = 0x11a50000, 1807 .pvr_value = 0x12C41C80,
1808 .cpu_name = "APM821XX",
1809 .cpu_features = CPU_FTRS_44X,
1810 .cpu_user_features = COMMON_USER_BOOKE |
1811 PPC_FEATURE_HAS_FPU,
1812 .mmu_features = MMU_FTR_TYPE_44x,
1813 .icache_bsize = 32,
1814 .dcache_bsize = 32,
1815 .cpu_setup = __setup_cpu_apm821xx,
1816 .machine_check = machine_check_440A,
1817 .platform = "ppc440",
1818 },
1819 { /* 476 DD2 core */
1820 .pvr_mask = 0xffffffff,
1821 .pvr_value = 0x11a52080,
1811 .cpu_name = "476", 1822 .cpu_name = "476",
1812 .cpu_features = CPU_FTRS_47X, 1823 .cpu_features = CPU_FTRS_47X | CPU_FTR_476_DD2,
1813 .cpu_user_features = COMMON_USER_BOOKE | 1824 .cpu_user_features = COMMON_USER_BOOKE |
1814 PPC_FEATURE_HAS_FPU, 1825 PPC_FEATURE_HAS_FPU,
1815 .mmu_features = MMU_FTR_TYPE_47x | 1826 .mmu_features = MMU_FTR_TYPE_47x |
@@ -1833,6 +1844,20 @@ static struct cpu_spec __initdata cpu_specs[] = {
1833 .machine_check = machine_check_47x, 1844 .machine_check = machine_check_47x,
1834 .platform = "ppc470", 1845 .platform = "ppc470",
1835 }, 1846 },
1847 { /* 476 others */
1848 .pvr_mask = 0xffff0000,
1849 .pvr_value = 0x11a50000,
1850 .cpu_name = "476",
1851 .cpu_features = CPU_FTRS_47X,
1852 .cpu_user_features = COMMON_USER_BOOKE |
1853 PPC_FEATURE_HAS_FPU,
1854 .mmu_features = MMU_FTR_TYPE_47x |
1855 MMU_FTR_USE_TLBIVAX_BCAST | MMU_FTR_LOCK_BCAST_INVAL,
1856 .icache_bsize = 32,
1857 .dcache_bsize = 128,
1858 .machine_check = machine_check_47x,
1859 .platform = "ppc470",
1860 },
1836 { /* default match */ 1861 { /* default match */
1837 .pvr_mask = 0x00000000, 1862 .pvr_mask = 0x00000000,
1838 .pvr_value = 0x00000000, 1863 .pvr_value = 0x00000000,
@@ -1891,7 +1916,9 @@ static struct cpu_spec __initdata cpu_specs[] = {
1891 .platform = "ppc5554", 1916 .platform = "ppc5554",
1892 } 1917 }
1893#endif /* CONFIG_E200 */ 1918#endif /* CONFIG_E200 */
1919#endif /* CONFIG_PPC32 */
1894#ifdef CONFIG_E500 1920#ifdef CONFIG_E500
1921#ifdef CONFIG_PPC32
1895 { /* e500 */ 1922 { /* e500 */
1896 .pvr_mask = 0xffff0000, 1923 .pvr_mask = 0xffff0000,
1897 .pvr_value = 0x80200000, 1924 .pvr_value = 0x80200000,
@@ -1946,6 +1973,26 @@ static struct cpu_spec __initdata cpu_specs[] = {
1946 .machine_check = machine_check_e500mc, 1973 .machine_check = machine_check_e500mc,
1947 .platform = "ppce500mc", 1974 .platform = "ppce500mc",
1948 }, 1975 },
1976#endif /* CONFIG_PPC32 */
1977 { /* e5500 */
1978 .pvr_mask = 0xffff0000,
1979 .pvr_value = 0x80240000,
1980 .cpu_name = "e5500",
1981 .cpu_features = CPU_FTRS_E5500,
1982 .cpu_user_features = COMMON_USER_BOOKE | PPC_FEATURE_HAS_FPU,
1983 .mmu_features = MMU_FTR_TYPE_FSL_E | MMU_FTR_BIG_PHYS |
1984 MMU_FTR_USE_TLBILX,
1985 .icache_bsize = 64,
1986 .dcache_bsize = 64,
1987 .num_pmcs = 4,
1988 .oprofile_cpu_type = "ppc/e500mc",
1989 .oprofile_type = PPC_OPROFILE_FSL_EMB,
1990 .cpu_setup = __setup_cpu_e5500,
1991 .cpu_restore = __restore_cpu_e5500,
1992 .machine_check = machine_check_e500mc,
1993 .platform = "ppce5500",
1994 },
1995#ifdef CONFIG_PPC32
1949 { /* default match */ 1996 { /* default match */
1950 .pvr_mask = 0x00000000, 1997 .pvr_mask = 0x00000000,
1951 .pvr_value = 0x00000000, 1998 .pvr_value = 0x00000000,
@@ -1960,10 +2007,25 @@ static struct cpu_spec __initdata cpu_specs[] = {
1960 .machine_check = machine_check_e500, 2007 .machine_check = machine_check_e500,
1961 .platform = "powerpc", 2008 .platform = "powerpc",
1962 } 2009 }
1963#endif /* CONFIG_E500 */
1964#endif /* CONFIG_PPC32 */ 2010#endif /* CONFIG_PPC32 */
2011#endif /* CONFIG_E500 */
1965 2012
1966#ifdef CONFIG_PPC_BOOK3E_64 2013#ifdef CONFIG_PPC_A2
2014 { /* Standard A2 (>= DD2) + FPU core */
2015 .pvr_mask = 0xffff0000,
2016 .pvr_value = 0x00480000,
2017 .cpu_name = "A2 (>= DD2)",
2018 .cpu_features = CPU_FTRS_A2,
2019 .cpu_user_features = COMMON_USER_PPC64,
2020 .mmu_features = MMU_FTRS_A2,
2021 .icache_bsize = 64,
2022 .dcache_bsize = 64,
2023 .num_pmcs = 0,
2024 .cpu_setup = __setup_cpu_a2,
2025 .cpu_restore = __restore_cpu_a2,
2026 .machine_check = machine_check_generic,
2027 .platform = "ppca2",
2028 },
1967 { /* This is a default entry to get going, to be replaced by 2029 { /* This is a default entry to get going, to be replaced by
1968 * a real one at some stage 2030 * a real one at some stage
1969 */ 2031 */
@@ -1984,7 +2046,7 @@ static struct cpu_spec __initdata cpu_specs[] = {
1984 .machine_check = machine_check_generic, 2046 .machine_check = machine_check_generic,
1985 .platform = "power6", 2047 .platform = "power6",
1986 }, 2048 },
1987#endif 2049#endif /* CONFIG_PPC_A2 */
1988}; 2050};
1989 2051
1990static struct cpu_spec the_cpu_spec; 2052static struct cpu_spec the_cpu_spec;
@@ -2048,8 +2110,8 @@ static void __init setup_cpu_spec(unsigned long offset, struct cpu_spec *s)
2048 * pointer on ppc64 and booke as we are running at 0 in real mode 2110 * pointer on ppc64 and booke as we are running at 0 in real mode
2049 * on ppc64 and reloc_offset is always 0 on booke. 2111 * on ppc64 and reloc_offset is always 0 on booke.
2050 */ 2112 */
2051 if (s->cpu_setup) { 2113 if (t->cpu_setup) {
2052 s->cpu_setup(offset, s); 2114 t->cpu_setup(offset, t);
2053 } 2115 }
2054#endif /* CONFIG_PPC64 || CONFIG_BOOKE */ 2116#endif /* CONFIG_PPC64 || CONFIG_BOOKE */
2055} 2117}
diff --git a/arch/powerpc/kernel/crash.c b/arch/powerpc/kernel/crash.c
index 4457382f8667..4e6ee944495a 100644
--- a/arch/powerpc/kernel/crash.c
+++ b/arch/powerpc/kernel/crash.c
@@ -48,7 +48,7 @@ int crashing_cpu = -1;
48static cpumask_t cpus_in_crash = CPU_MASK_NONE; 48static cpumask_t cpus_in_crash = CPU_MASK_NONE;
49cpumask_t cpus_in_sr = CPU_MASK_NONE; 49cpumask_t cpus_in_sr = CPU_MASK_NONE;
50 50
51#define CRASH_HANDLER_MAX 2 51#define CRASH_HANDLER_MAX 3
52/* NULL terminated list of shutdown handles */ 52/* NULL terminated list of shutdown handles */
53static crash_shutdown_t crash_shutdown_handles[CRASH_HANDLER_MAX+1]; 53static crash_shutdown_t crash_shutdown_handles[CRASH_HANDLER_MAX+1];
54static DEFINE_SPINLOCK(crash_handlers_lock); 54static DEFINE_SPINLOCK(crash_handlers_lock);
@@ -64,9 +64,9 @@ void crash_ipi_callback(struct pt_regs *regs)
64 return; 64 return;
65 65
66 hard_irq_disable(); 66 hard_irq_disable();
67 if (!cpu_isset(cpu, cpus_in_crash)) 67 if (!cpumask_test_cpu(cpu, &cpus_in_crash))
68 crash_save_cpu(regs, cpu); 68 crash_save_cpu(regs, cpu);
69 cpu_set(cpu, cpus_in_crash); 69 cpumask_set_cpu(cpu, &cpus_in_crash);
70 70
71 /* 71 /*
72 * Entered via soft-reset - could be the kdump 72 * Entered via soft-reset - could be the kdump
@@ -77,8 +77,8 @@ void crash_ipi_callback(struct pt_regs *regs)
77 * Tell the kexec CPU that entered via soft-reset and ready 77 * Tell the kexec CPU that entered via soft-reset and ready
78 * to go down. 78 * to go down.
79 */ 79 */
80 if (cpu_isset(cpu, cpus_in_sr)) { 80 if (cpumask_test_cpu(cpu, &cpus_in_sr)) {
81 cpu_clear(cpu, cpus_in_sr); 81 cpumask_clear_cpu(cpu, &cpus_in_sr);
82 atomic_inc(&enter_on_soft_reset); 82 atomic_inc(&enter_on_soft_reset);
83 } 83 }
84 84
@@ -87,7 +87,7 @@ void crash_ipi_callback(struct pt_regs *regs)
87 * This barrier is needed to make sure that all CPUs are stopped. 87 * This barrier is needed to make sure that all CPUs are stopped.
88 * If not, soft-reset will be invoked to bring other CPUs. 88 * If not, soft-reset will be invoked to bring other CPUs.
89 */ 89 */
90 while (!cpu_isset(crashing_cpu, cpus_in_crash)) 90 while (!cpumask_test_cpu(crashing_cpu, &cpus_in_crash))
91 cpu_relax(); 91 cpu_relax();
92 92
93 if (ppc_md.kexec_cpu_down) 93 if (ppc_md.kexec_cpu_down)
@@ -109,7 +109,7 @@ static void crash_soft_reset_check(int cpu)
109{ 109{
110 unsigned int ncpus = num_online_cpus() - 1;/* Excluding the panic cpu */ 110 unsigned int ncpus = num_online_cpus() - 1;/* Excluding the panic cpu */
111 111
112 cpu_clear(cpu, cpus_in_sr); 112 cpumask_clear_cpu(cpu, &cpus_in_sr);
113 while (atomic_read(&enter_on_soft_reset) != ncpus) 113 while (atomic_read(&enter_on_soft_reset) != ncpus)
114 cpu_relax(); 114 cpu_relax();
115} 115}
@@ -125,14 +125,14 @@ static void crash_kexec_prepare_cpus(int cpu)
125 smp_wmb(); 125 smp_wmb();
126 126
127 /* 127 /*
128 * FIXME: Until we will have the way to stop other CPUSs reliabally, 128 * FIXME: Until we will have the way to stop other CPUs reliably,
129 * the crash CPU will send an IPI and wait for other CPUs to 129 * the crash CPU will send an IPI and wait for other CPUs to
130 * respond. 130 * respond.
131 * Delay of at least 10 seconds. 131 * Delay of at least 10 seconds.
132 */ 132 */
133 printk(KERN_EMERG "Sending IPI to other cpus...\n"); 133 printk(KERN_EMERG "Sending IPI to other cpus...\n");
134 msecs = 10000; 134 msecs = 10000;
135 while ((cpus_weight(cpus_in_crash) < ncpus) && (--msecs > 0)) { 135 while ((cpumask_weight(&cpus_in_crash) < ncpus) && (--msecs > 0)) {
136 cpu_relax(); 136 cpu_relax();
137 mdelay(1); 137 mdelay(1);
138 } 138 }
@@ -144,52 +144,24 @@ static void crash_kexec_prepare_cpus(int cpu)
144 * user to do soft reset such that we get all. 144 * user to do soft reset such that we get all.
145 * Soft-reset will be used until better mechanism is implemented. 145 * Soft-reset will be used until better mechanism is implemented.
146 */ 146 */
147 if (cpus_weight(cpus_in_crash) < ncpus) { 147 if (cpumask_weight(&cpus_in_crash) < ncpus) {
148 printk(KERN_EMERG "done waiting: %d cpu(s) not responding\n", 148 printk(KERN_EMERG "done waiting: %d cpu(s) not responding\n",
149 ncpus - cpus_weight(cpus_in_crash)); 149 ncpus - cpumask_weight(&cpus_in_crash));
150 printk(KERN_EMERG "Activate soft-reset to stop other cpu(s)\n"); 150 printk(KERN_EMERG "Activate soft-reset to stop other cpu(s)\n");
151 cpus_in_sr = CPU_MASK_NONE; 151 cpumask_clear(&cpus_in_sr);
152 atomic_set(&enter_on_soft_reset, 0); 152 atomic_set(&enter_on_soft_reset, 0);
153 while (cpus_weight(cpus_in_crash) < ncpus) 153 while (cpumask_weight(&cpus_in_crash) < ncpus)
154 cpu_relax(); 154 cpu_relax();
155 } 155 }
156 /* 156 /*
157 * Make sure all CPUs are entered via soft-reset if the kdump is 157 * Make sure all CPUs are entered via soft-reset if the kdump is
158 * invoked using soft-reset. 158 * invoked using soft-reset.
159 */ 159 */
160 if (cpu_isset(cpu, cpus_in_sr)) 160 if (cpumask_test_cpu(cpu, &cpus_in_sr))
161 crash_soft_reset_check(cpu); 161 crash_soft_reset_check(cpu);
162 /* Leave the IPI callback set */ 162 /* Leave the IPI callback set */
163} 163}
164 164
165/* wait for all the CPUs to hit real mode but timeout if they don't come in */
166#ifdef CONFIG_PPC_STD_MMU_64
167static void crash_kexec_wait_realmode(int cpu)
168{
169 unsigned int msecs;
170 int i;
171
172 msecs = 10000;
173 for (i=0; i < NR_CPUS && msecs > 0; i++) {
174 if (i == cpu)
175 continue;
176
177 while (paca[i].kexec_state < KEXEC_STATE_REAL_MODE) {
178 barrier();
179 if (!cpu_possible(i)) {
180 break;
181 }
182 if (!cpu_online(i)) {
183 break;
184 }
185 msecs--;
186 mdelay(1);
187 }
188 }
189 mb();
190}
191#endif
192
193/* 165/*
194 * This function will be called by secondary cpus or by kexec cpu 166 * This function will be called by secondary cpus or by kexec cpu
195 * if soft-reset is activated to stop some CPUs. 167 * if soft-reset is activated to stop some CPUs.
@@ -210,7 +182,7 @@ void crash_kexec_secondary(struct pt_regs *regs)
210 * exited using 'x'(exit and recover) or 182 * exited using 'x'(exit and recover) or
211 * kexec_should_crash() failed for all running tasks. 183 * kexec_should_crash() failed for all running tasks.
212 */ 184 */
213 cpu_clear(cpu, cpus_in_sr); 185 cpumask_clear_cpu(cpu, &cpus_in_sr);
214 local_irq_restore(flags); 186 local_irq_restore(flags);
215 return; 187 return;
216 } 188 }
@@ -224,7 +196,7 @@ void crash_kexec_secondary(struct pt_regs *regs)
224 * then start kexec boot. 196 * then start kexec boot.
225 */ 197 */
226 crash_soft_reset_check(cpu); 198 crash_soft_reset_check(cpu);
227 cpu_set(crashing_cpu, cpus_in_crash); 199 cpumask_set_cpu(crashing_cpu, &cpus_in_crash);
228 if (ppc_md.kexec_cpu_down) 200 if (ppc_md.kexec_cpu_down)
229 ppc_md.kexec_cpu_down(1, 0); 201 ppc_md.kexec_cpu_down(1, 0);
230 machine_kexec(kexec_crash_image); 202 machine_kexec(kexec_crash_image);
@@ -233,7 +205,8 @@ void crash_kexec_secondary(struct pt_regs *regs)
233 crash_ipi_callback(regs); 205 crash_ipi_callback(regs);
234} 206}
235 207
236#else 208#else /* ! CONFIG_SMP */
209
237static void crash_kexec_prepare_cpus(int cpu) 210static void crash_kexec_prepare_cpus(int cpu)
238{ 211{
239 /* 212 /*
@@ -251,75 +224,39 @@ static void crash_kexec_prepare_cpus(int cpu)
251 224
252void crash_kexec_secondary(struct pt_regs *regs) 225void crash_kexec_secondary(struct pt_regs *regs)
253{ 226{
254 cpus_in_sr = CPU_MASK_NONE; 227 cpumask_clear(&cpus_in_sr);
255} 228}
256#endif 229#endif /* CONFIG_SMP */
257#ifdef CONFIG_SPU_BASE
258
259#include <asm/spu.h>
260#include <asm/spu_priv1.h>
261
262struct crash_spu_info {
263 struct spu *spu;
264 u32 saved_spu_runcntl_RW;
265 u32 saved_spu_status_R;
266 u32 saved_spu_npc_RW;
267 u64 saved_mfc_sr1_RW;
268 u64 saved_mfc_dar;
269 u64 saved_mfc_dsisr;
270};
271 230
272#define CRASH_NUM_SPUS 16 /* Enough for current hardware */ 231/* wait for all the CPUs to hit real mode but timeout if they don't come in */
273static struct crash_spu_info crash_spu_info[CRASH_NUM_SPUS]; 232#if defined(CONFIG_SMP) && defined(CONFIG_PPC_STD_MMU_64)
274 233static void crash_kexec_wait_realmode(int cpu)
275static void crash_kexec_stop_spus(void)
276{ 234{
277 struct spu *spu; 235 unsigned int msecs;
278 int i; 236 int i;
279 u64 tmp;
280 237
281 for (i = 0; i < CRASH_NUM_SPUS; i++) { 238 msecs = 10000;
282 if (!crash_spu_info[i].spu) 239 for (i=0; i < nr_cpu_ids && msecs > 0; i++) {
283 continue; 240 if (i == cpu)
284
285 spu = crash_spu_info[i].spu;
286
287 crash_spu_info[i].saved_spu_runcntl_RW =
288 in_be32(&spu->problem->spu_runcntl_RW);
289 crash_spu_info[i].saved_spu_status_R =
290 in_be32(&spu->problem->spu_status_R);
291 crash_spu_info[i].saved_spu_npc_RW =
292 in_be32(&spu->problem->spu_npc_RW);
293
294 crash_spu_info[i].saved_mfc_dar = spu_mfc_dar_get(spu);
295 crash_spu_info[i].saved_mfc_dsisr = spu_mfc_dsisr_get(spu);
296 tmp = spu_mfc_sr1_get(spu);
297 crash_spu_info[i].saved_mfc_sr1_RW = tmp;
298
299 tmp &= ~MFC_STATE1_MASTER_RUN_CONTROL_MASK;
300 spu_mfc_sr1_set(spu, tmp);
301
302 __delay(200);
303 }
304}
305
306void crash_register_spus(struct list_head *list)
307{
308 struct spu *spu;
309
310 list_for_each_entry(spu, list, full_list) {
311 if (WARN_ON(spu->number >= CRASH_NUM_SPUS))
312 continue; 241 continue;
313 242
314 crash_spu_info[spu->number].spu = spu; 243 while (paca[i].kexec_state < KEXEC_STATE_REAL_MODE) {
244 barrier();
245 if (!cpu_possible(i)) {
246 break;
247 }
248 if (!cpu_online(i)) {
249 break;
250 }
251 msecs--;
252 mdelay(1);
253 }
315 } 254 }
255 mb();
316} 256}
317
318#else 257#else
319static inline void crash_kexec_stop_spus(void) 258static inline void crash_kexec_wait_realmode(int cpu) {}
320{ 259#endif /* CONFIG_SMP && CONFIG_PPC_STD_MMU_64 */
321}
322#endif /* CONFIG_SPU_BASE */
323 260
324/* 261/*
325 * Register a function to be called on shutdown. Only use this if you 262 * Register a function to be called on shutdown. Only use this if you
@@ -409,23 +346,10 @@ void default_machine_crash_shutdown(struct pt_regs *regs)
409 crashing_cpu = smp_processor_id(); 346 crashing_cpu = smp_processor_id();
410 crash_save_cpu(regs, crashing_cpu); 347 crash_save_cpu(regs, crashing_cpu);
411 crash_kexec_prepare_cpus(crashing_cpu); 348 crash_kexec_prepare_cpus(crashing_cpu);
412 cpu_set(crashing_cpu, cpus_in_crash); 349 cpumask_set_cpu(crashing_cpu, &cpus_in_crash);
413#if defined(CONFIG_PPC_STD_MMU_64) && defined(CONFIG_SMP)
414 crash_kexec_wait_realmode(crashing_cpu); 350 crash_kexec_wait_realmode(crashing_cpu);
415#endif
416
417 for_each_irq(i) {
418 struct irq_desc *desc = irq_to_desc(i);
419
420 if (!desc || !desc->chip || !desc->chip->eoi)
421 continue;
422
423 if (desc->status & IRQ_INPROGRESS)
424 desc->chip->eoi(i);
425 351
426 if (!(desc->status & IRQ_DISABLED)) 352 machine_kexec_mask_interrupts();
427 desc->chip->shutdown(i);
428 }
429 353
430 /* 354 /*
431 * Call registered shutdown routines savely. Swap out 355 * Call registered shutdown routines savely. Swap out
@@ -450,8 +374,6 @@ void default_machine_crash_shutdown(struct pt_regs *regs)
450 crash_shutdown_cpu = -1; 374 crash_shutdown_cpu = -1;
451 __debugger_fault_handler = old_handler; 375 __debugger_fault_handler = old_handler;
452 376
453 crash_kexec_stop_spus();
454
455 if (ppc_md.kexec_cpu_down) 377 if (ppc_md.kexec_cpu_down)
456 ppc_md.kexec_cpu_down(1, 0); 378 ppc_md.kexec_cpu_down(1, 0);
457} 379}
diff --git a/arch/powerpc/kernel/crash_dump.c b/arch/powerpc/kernel/crash_dump.c
index 8e05c16344e4..424afb6b8fba 100644
--- a/arch/powerpc/kernel/crash_dump.c
+++ b/arch/powerpc/kernel/crash_dump.c
@@ -19,6 +19,7 @@
19#include <asm/prom.h> 19#include <asm/prom.h>
20#include <asm/firmware.h> 20#include <asm/firmware.h>
21#include <asm/uaccess.h> 21#include <asm/uaccess.h>
22#include <asm/rtas.h>
22 23
23#ifdef DEBUG 24#ifdef DEBUG
24#include <asm/udbg.h> 25#include <asm/udbg.h>
@@ -27,9 +28,6 @@
27#define DBG(fmt...) 28#define DBG(fmt...)
28#endif 29#endif
29 30
30/* Stores the physical address of elf header of crash image. */
31unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
32
33#ifndef CONFIG_RELOCATABLE 31#ifndef CONFIG_RELOCATABLE
34void __init reserve_kdump_trampoline(void) 32void __init reserve_kdump_trampoline(void)
35{ 33{
@@ -71,20 +69,6 @@ void __init setup_kdump_trampoline(void)
71} 69}
72#endif /* CONFIG_RELOCATABLE */ 70#endif /* CONFIG_RELOCATABLE */
73 71
74/*
75 * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
76 * is_kdump_kernel() to determine if we are booting after a panic. Hence
77 * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
78 */
79static int __init parse_elfcorehdr(char *p)
80{
81 if (p)
82 elfcorehdr_addr = memparse(p, &p);
83
84 return 1;
85}
86__setup("elfcorehdr=", parse_elfcorehdr);
87
88static int __init parse_savemaxmem(char *p) 72static int __init parse_savemaxmem(char *p)
89{ 73{
90 if (p) 74 if (p)
@@ -141,3 +125,35 @@ ssize_t copy_oldmem_page(unsigned long pfn, char *buf,
141 125
142 return csize; 126 return csize;
143} 127}
128
129#ifdef CONFIG_PPC_RTAS
130/*
131 * The crashkernel region will almost always overlap the RTAS region, so
132 * we have to be careful when shrinking the crashkernel region.
133 */
134void crash_free_reserved_phys_range(unsigned long begin, unsigned long end)
135{
136 unsigned long addr;
137 const u32 *basep, *sizep;
138 unsigned int rtas_start = 0, rtas_end = 0;
139
140 basep = of_get_property(rtas.dev, "linux,rtas-base", NULL);
141 sizep = of_get_property(rtas.dev, "rtas-size", NULL);
142
143 if (basep && sizep) {
144 rtas_start = *basep;
145 rtas_end = *basep + *sizep;
146 }
147
148 for (addr = begin; addr < end; addr += PAGE_SIZE) {
149 /* Does this page overlap with the RTAS region? */
150 if (addr <= rtas_end && ((addr + PAGE_SIZE) > rtas_start))
151 continue;
152
153 ClearPageReserved(pfn_to_page(addr >> PAGE_SHIFT));
154 init_page_count(pfn_to_page(addr >> PAGE_SHIFT));
155 free_page((unsigned long)__va(addr));
156 totalram_pages++;
157 }
158}
159#endif
diff --git a/arch/powerpc/kernel/dbell.c b/arch/powerpc/kernel/dbell.c
index 3307a52d797f..2cc451aaaca7 100644
--- a/arch/powerpc/kernel/dbell.c
+++ b/arch/powerpc/kernel/dbell.c
@@ -13,84 +13,35 @@
13#include <linux/kernel.h> 13#include <linux/kernel.h>
14#include <linux/smp.h> 14#include <linux/smp.h>
15#include <linux/threads.h> 15#include <linux/threads.h>
16#include <linux/percpu.h> 16#include <linux/hardirq.h>
17 17
18#include <asm/dbell.h> 18#include <asm/dbell.h>
19#include <asm/irq_regs.h> 19#include <asm/irq_regs.h>
20 20
21#ifdef CONFIG_SMP 21#ifdef CONFIG_SMP
22struct doorbell_cpu_info {
23 unsigned long messages; /* current messages bits */
24 unsigned int tag; /* tag value */
25};
26
27static DEFINE_PER_CPU(struct doorbell_cpu_info, doorbell_cpu_info);
28
29void doorbell_setup_this_cpu(void) 22void doorbell_setup_this_cpu(void)
30{ 23{
31 struct doorbell_cpu_info *info = &__get_cpu_var(doorbell_cpu_info); 24 unsigned long tag = mfspr(SPRN_PIR) & 0x3fff;
32 25
33 info->messages = 0; 26 smp_muxed_ipi_set_data(smp_processor_id(), tag);
34 info->tag = mfspr(SPRN_PIR) & 0x3fff;
35} 27}
36 28
37void doorbell_message_pass(int target, int msg) 29void doorbell_cause_ipi(int cpu, unsigned long data)
38{ 30{
39 struct doorbell_cpu_info *info; 31 ppc_msgsnd(PPC_DBELL, 0, data);
40 int i;
41
42 if (target < NR_CPUS) {
43 info = &per_cpu(doorbell_cpu_info, target);
44 set_bit(msg, &info->messages);
45 ppc_msgsnd(PPC_DBELL, 0, info->tag);
46 }
47 else if (target == MSG_ALL_BUT_SELF) {
48 for_each_online_cpu(i) {
49 if (i == smp_processor_id())
50 continue;
51 info = &per_cpu(doorbell_cpu_info, i);
52 set_bit(msg, &info->messages);
53 ppc_msgsnd(PPC_DBELL, 0, info->tag);
54 }
55 }
56 else { /* target == MSG_ALL */
57 for_each_online_cpu(i) {
58 info = &per_cpu(doorbell_cpu_info, i);
59 set_bit(msg, &info->messages);
60 }
61 ppc_msgsnd(PPC_DBELL, PPC_DBELL_MSG_BRDCAST, 0);
62 }
63} 32}
64 33
65void doorbell_exception(struct pt_regs *regs) 34void doorbell_exception(struct pt_regs *regs)
66{ 35{
67 struct pt_regs *old_regs = set_irq_regs(regs); 36 struct pt_regs *old_regs = set_irq_regs(regs);
68 struct doorbell_cpu_info *info = &__get_cpu_var(doorbell_cpu_info);
69 int msg;
70 37
71 /* Warning: regs can be NULL when called from irq enable */ 38 irq_enter();
72 39
73 if (!info->messages || (num_online_cpus() < 2)) 40 smp_ipi_demux();
74 goto out;
75 41
76 for (msg = 0; msg < 4; msg++) 42 irq_exit();
77 if (test_and_clear_bit(msg, &info->messages))
78 smp_message_recv(msg);
79
80out:
81 set_irq_regs(old_regs); 43 set_irq_regs(old_regs);
82} 44}
83
84void doorbell_check_self(void)
85{
86 struct doorbell_cpu_info *info = &__get_cpu_var(doorbell_cpu_info);
87
88 if (!info->messages)
89 return;
90
91 ppc_msgsnd(PPC_DBELL, 0, info->tag);
92}
93
94#else /* CONFIG_SMP */ 45#else /* CONFIG_SMP */
95void doorbell_exception(struct pt_regs *regs) 46void doorbell_exception(struct pt_regs *regs)
96{ 47{
diff --git a/arch/powerpc/kernel/dma-iommu.c b/arch/powerpc/kernel/dma-iommu.c
index 37771a518119..e7554154a6de 100644
--- a/arch/powerpc/kernel/dma-iommu.c
+++ b/arch/powerpc/kernel/dma-iommu.c
@@ -19,7 +19,7 @@ static void *dma_iommu_alloc_coherent(struct device *dev, size_t size,
19 dma_addr_t *dma_handle, gfp_t flag) 19 dma_addr_t *dma_handle, gfp_t flag)
20{ 20{
21 return iommu_alloc_coherent(dev, get_iommu_table_base(dev), size, 21 return iommu_alloc_coherent(dev, get_iommu_table_base(dev), size,
22 dma_handle, device_to_mask(dev), flag, 22 dma_handle, dev->coherent_dma_mask, flag,
23 dev_to_node(dev)); 23 dev_to_node(dev));
24} 24}
25 25
@@ -74,16 +74,17 @@ static int dma_iommu_dma_supported(struct device *dev, u64 mask)
74{ 74{
75 struct iommu_table *tbl = get_iommu_table_base(dev); 75 struct iommu_table *tbl = get_iommu_table_base(dev);
76 76
77 if (!tbl || tbl->it_offset > mask) { 77 if (!tbl) {
78 printk(KERN_INFO 78 dev_info(dev, "Warning: IOMMU dma not supported: mask 0x%08llx"
79 "Warning: IOMMU offset too big for device mask\n"); 79 ", table unavailable\n", mask);
80 if (tbl) 80 return 0;
81 printk(KERN_INFO 81 }
82 "mask: 0x%08llx, table offset: 0x%08lx\n", 82
83 mask, tbl->it_offset); 83 if ((tbl->it_offset + tbl->it_size) > (mask >> IOMMU_PAGE_SHIFT)) {
84 else 84 dev_info(dev, "Warning: IOMMU window too big for device mask\n");
85 printk(KERN_INFO "mask: 0x%08llx, table unavailable\n", 85 dev_info(dev, "mask: 0x%08llx, table end: 0x%08lx\n",
86 mask); 86 mask, (tbl->it_offset + tbl->it_size) <<
87 IOMMU_PAGE_SHIFT);
87 return 0; 88 return 0;
88 } else 89 } else
89 return 1; 90 return 1;
diff --git a/arch/powerpc/kernel/dma.c b/arch/powerpc/kernel/dma.c
index 84d6367ec003..d238c082c3c5 100644
--- a/arch/powerpc/kernel/dma.c
+++ b/arch/powerpc/kernel/dma.c
@@ -12,6 +12,7 @@
12#include <linux/memblock.h> 12#include <linux/memblock.h>
13#include <asm/bug.h> 13#include <asm/bug.h>
14#include <asm/abs_addr.h> 14#include <asm/abs_addr.h>
15#include <asm/machdep.h>
15 16
16/* 17/*
17 * Generic direct DMA implementation 18 * Generic direct DMA implementation
@@ -89,7 +90,7 @@ static int dma_direct_dma_supported(struct device *dev, u64 mask)
89 /* Could be improved so platforms can set the limit in case 90 /* Could be improved so platforms can set the limit in case
90 * they have limited DMA windows 91 * they have limited DMA windows
91 */ 92 */
92 return mask >= (memblock_end_of_DRAM() - 1); 93 return mask >= get_dma_offset(dev) + (memblock_end_of_DRAM() - 1);
93#else 94#else
94 return 1; 95 return 1;
95#endif 96#endif
@@ -154,6 +155,23 @@ EXPORT_SYMBOL(dma_direct_ops);
154 155
155#define PREALLOC_DMA_DEBUG_ENTRIES (1 << 16) 156#define PREALLOC_DMA_DEBUG_ENTRIES (1 << 16)
156 157
158int dma_set_mask(struct device *dev, u64 dma_mask)
159{
160 struct dma_map_ops *dma_ops = get_dma_ops(dev);
161
162 if (ppc_md.dma_set_mask)
163 return ppc_md.dma_set_mask(dev, dma_mask);
164 if (unlikely(dma_ops == NULL))
165 return -EIO;
166 if (dma_ops->set_dma_mask != NULL)
167 return dma_ops->set_dma_mask(dev, dma_mask);
168 if (!dev->dma_mask || !dma_supported(dev, dma_mask))
169 return -EIO;
170 *dev->dma_mask = dma_mask;
171 return 0;
172}
173EXPORT_SYMBOL(dma_set_mask);
174
157static int __init dma_init(void) 175static int __init dma_init(void)
158{ 176{
159 dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES); 177 dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES);
@@ -161,3 +179,21 @@ static int __init dma_init(void)
161 return 0; 179 return 0;
162} 180}
163fs_initcall(dma_init); 181fs_initcall(dma_init);
182
183int dma_mmap_coherent(struct device *dev, struct vm_area_struct *vma,
184 void *cpu_addr, dma_addr_t handle, size_t size)
185{
186 unsigned long pfn;
187
188#ifdef CONFIG_NOT_COHERENT_CACHE
189 vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
190 pfn = __dma_get_coherent_pfn((unsigned long)cpu_addr);
191#else
192 pfn = page_to_pfn(virt_to_page(cpu_addr));
193#endif
194 return remap_pfn_range(vma, vma->vm_start,
195 pfn + vma->vm_pgoff,
196 vma->vm_end - vma->vm_start,
197 vma->vm_page_prot);
198}
199EXPORT_SYMBOL_GPL(dma_mmap_coherent);
diff --git a/arch/powerpc/kernel/e500-pmu.c b/arch/powerpc/kernel/e500-pmu.c
index 7c07de0d8943..b150b510510f 100644
--- a/arch/powerpc/kernel/e500-pmu.c
+++ b/arch/powerpc/kernel/e500-pmu.c
@@ -126,4 +126,4 @@ static int init_e500_pmu(void)
126 return register_fsl_emb_pmu(&e500_pmu); 126 return register_fsl_emb_pmu(&e500_pmu);
127} 127}
128 128
129arch_initcall(init_e500_pmu); 129early_initcall(init_e500_pmu);
diff --git a/arch/powerpc/kernel/entry_32.S b/arch/powerpc/kernel/entry_32.S
index ed4aeb96398b..56212bc0ab08 100644
--- a/arch/powerpc/kernel/entry_32.S
+++ b/arch/powerpc/kernel/entry_32.S
@@ -31,6 +31,7 @@
31#include <asm/asm-offsets.h> 31#include <asm/asm-offsets.h>
32#include <asm/unistd.h> 32#include <asm/unistd.h>
33#include <asm/ftrace.h> 33#include <asm/ftrace.h>
34#include <asm/ptrace.h>
34 35
35#undef SHOW_SYSCALLS 36#undef SHOW_SYSCALLS
36#undef SHOW_SYSCALLS_TASK 37#undef SHOW_SYSCALLS_TASK
@@ -879,7 +880,18 @@ END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_47x)
879 */ 880 */
880 andi. r10,r9,MSR_EE 881 andi. r10,r9,MSR_EE
881 beq 1f 882 beq 1f
883 /*
884 * Since the ftrace irqsoff latency trace checks CALLER_ADDR1,
885 * which is the stack frame here, we need to force a stack frame
886 * in case we came from user space.
887 */
888 stwu r1,-32(r1)
889 mflr r0
890 stw r0,4(r1)
891 stwu r1,-32(r1)
882 bl trace_hardirqs_on 892 bl trace_hardirqs_on
893 lwz r1,0(r1)
894 lwz r1,0(r1)
883 lwz r9,_MSR(r1) 895 lwz r9,_MSR(r1)
8841: 8961:
885#endif /* CONFIG_TRACE_IRQFLAGS */ 897#endif /* CONFIG_TRACE_IRQFLAGS */
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index 42e9d908914a..d834425186ae 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -97,6 +97,24 @@ system_call_common:
97 addi r9,r1,STACK_FRAME_OVERHEAD 97 addi r9,r1,STACK_FRAME_OVERHEAD
98 ld r11,exception_marker@toc(r2) 98 ld r11,exception_marker@toc(r2)
99 std r11,-16(r9) /* "regshere" marker */ 99 std r11,-16(r9) /* "regshere" marker */
100#if defined(CONFIG_VIRT_CPU_ACCOUNTING) && defined(CONFIG_PPC_SPLPAR)
101BEGIN_FW_FTR_SECTION
102 beq 33f
103 /* if from user, see if there are any DTL entries to process */
104 ld r10,PACALPPACAPTR(r13) /* get ptr to VPA */
105 ld r11,PACA_DTL_RIDX(r13) /* get log read index */
106 ld r10,LPPACA_DTLIDX(r10) /* get log write index */
107 cmpd cr1,r11,r10
108 beq+ cr1,33f
109 bl .accumulate_stolen_time
110 REST_GPR(0,r1)
111 REST_4GPRS(3,r1)
112 REST_2GPRS(7,r1)
113 addi r9,r1,STACK_FRAME_OVERHEAD
11433:
115END_FW_FTR_SECTION_IFSET(FW_FEATURE_SPLPAR)
116#endif /* CONFIG_VIRT_CPU_ACCOUNTING && CONFIG_PPC_SPLPAR */
117
100#ifdef CONFIG_TRACE_IRQFLAGS 118#ifdef CONFIG_TRACE_IRQFLAGS
101 bl .trace_hardirqs_on 119 bl .trace_hardirqs_on
102 REST_GPR(0,r1) 120 REST_GPR(0,r1)
@@ -202,7 +220,9 @@ syscall_exit:
202 bge- syscall_error 220 bge- syscall_error
203syscall_error_cont: 221syscall_error_cont:
204 ld r7,_NIP(r1) 222 ld r7,_NIP(r1)
223BEGIN_FTR_SECTION
205 stdcx. r0,0,r1 /* to clear the reservation */ 224 stdcx. r0,0,r1 /* to clear the reservation */
225END_FTR_SECTION_IFCLR(CPU_FTR_STCX_CHECKS_ADDRESS)
206 andi. r6,r8,MSR_PR 226 andi. r6,r8,MSR_PR
207 ld r4,_LINK(r1) 227 ld r4,_LINK(r1)
208 /* 228 /*
@@ -401,6 +421,12 @@ BEGIN_FTR_SECTION
401 std r24,THREAD_VRSAVE(r3) 421 std r24,THREAD_VRSAVE(r3)
402END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC) 422END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
403#endif /* CONFIG_ALTIVEC */ 423#endif /* CONFIG_ALTIVEC */
424#ifdef CONFIG_PPC64
425BEGIN_FTR_SECTION
426 mfspr r25,SPRN_DSCR
427 std r25,THREAD_DSCR(r3)
428END_FTR_SECTION_IFSET(CPU_FTR_DSCR)
429#endif
404 and. r0,r0,r22 430 and. r0,r0,r22
405 beq+ 1f 431 beq+ 1f
406 andc r22,r22,r0 432 andc r22,r22,r0
@@ -419,6 +445,17 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
419 sync 445 sync
420#endif /* CONFIG_SMP */ 446#endif /* CONFIG_SMP */
421 447
448 /*
449 * If we optimise away the clear of the reservation in system
450 * calls because we know the CPU tracks the address of the
451 * reservation, then we need to clear it here to cover the
452 * case that the kernel context switch path has no larx
453 * instructions.
454 */
455BEGIN_FTR_SECTION
456 ldarx r6,0,r1
457END_FTR_SECTION_IFSET(CPU_FTR_STCX_CHECKS_ADDRESS)
458
422 addi r6,r4,-THREAD /* Convert THREAD to 'current' */ 459 addi r6,r4,-THREAD /* Convert THREAD to 'current' */
423 std r6,PACACURRENT(r13) /* Set new 'current' */ 460 std r6,PACACURRENT(r13) /* Set new 'current' */
424 461
@@ -431,10 +468,10 @@ BEGIN_FTR_SECTION
431 FTR_SECTION_ELSE_NESTED(95) 468 FTR_SECTION_ELSE_NESTED(95)
432 clrrdi r6,r8,40 /* get its 1T ESID */ 469 clrrdi r6,r8,40 /* get its 1T ESID */
433 clrrdi r9,r1,40 /* get current sp 1T ESID */ 470 clrrdi r9,r1,40 /* get current sp 1T ESID */
434 ALT_FTR_SECTION_END_NESTED_IFCLR(CPU_FTR_1T_SEGMENT, 95) 471 ALT_MMU_FTR_SECTION_END_NESTED_IFCLR(MMU_FTR_1T_SEGMENT, 95)
435FTR_SECTION_ELSE 472FTR_SECTION_ELSE
436 b 2f 473 b 2f
437ALT_FTR_SECTION_END_IFSET(CPU_FTR_SLB) 474ALT_MMU_FTR_SECTION_END_IFSET(MMU_FTR_SLB)
438 clrldi. r0,r6,2 /* is new ESID c00000000? */ 475 clrldi. r0,r6,2 /* is new ESID c00000000? */
439 cmpd cr1,r6,r9 /* or is new ESID the same as current ESID? */ 476 cmpd cr1,r6,r9 /* or is new ESID the same as current ESID? */
440 cror eq,4*cr1+eq,eq 477 cror eq,4*cr1+eq,eq
@@ -448,7 +485,7 @@ BEGIN_FTR_SECTION
448 li r9,MMU_SEGSIZE_1T /* insert B field */ 485 li r9,MMU_SEGSIZE_1T /* insert B field */
449 oris r6,r6,(MMU_SEGSIZE_1T << SLBIE_SSIZE_SHIFT)@h 486 oris r6,r6,(MMU_SEGSIZE_1T << SLBIE_SSIZE_SHIFT)@h
450 rldimi r7,r9,SLB_VSID_SSIZE_SHIFT,0 487 rldimi r7,r9,SLB_VSID_SSIZE_SHIFT,0
451END_FTR_SECTION_IFSET(CPU_FTR_1T_SEGMENT) 488END_MMU_FTR_SECTION_IFSET(MMU_FTR_1T_SEGMENT)
452 489
453 /* Update the last bolted SLB. No write barriers are needed 490 /* Update the last bolted SLB. No write barriers are needed
454 * here, provided we only update the current CPU's SLB shadow 491 * here, provided we only update the current CPU's SLB shadow
@@ -460,7 +497,7 @@ END_FTR_SECTION_IFSET(CPU_FTR_1T_SEGMENT)
460 std r7,SLBSHADOW_STACKVSID(r9) /* Save VSID */ 497 std r7,SLBSHADOW_STACKVSID(r9) /* Save VSID */
461 std r0,SLBSHADOW_STACKESID(r9) /* Save ESID */ 498 std r0,SLBSHADOW_STACKESID(r9) /* Save ESID */
462 499
463 /* No need to check for CPU_FTR_NO_SLBIE_B here, since when 500 /* No need to check for MMU_FTR_NO_SLBIE_B here, since when
464 * we have 1TB segments, the only CPUs known to have the errata 501 * we have 1TB segments, the only CPUs known to have the errata
465 * only support less than 1TB of system memory and we'll never 502 * only support less than 1TB of system memory and we'll never
466 * actually hit this code path. 503 * actually hit this code path.
@@ -491,6 +528,15 @@ BEGIN_FTR_SECTION
491 mtspr SPRN_VRSAVE,r0 /* if G4, restore VRSAVE reg */ 528 mtspr SPRN_VRSAVE,r0 /* if G4, restore VRSAVE reg */
492END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC) 529END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
493#endif /* CONFIG_ALTIVEC */ 530#endif /* CONFIG_ALTIVEC */
531#ifdef CONFIG_PPC64
532BEGIN_FTR_SECTION
533 ld r0,THREAD_DSCR(r4)
534 cmpd r0,r25
535 beq 1f
536 mtspr SPRN_DSCR,r0
5371:
538END_FTR_SECTION_IFSET(CPU_FTR_DSCR)
539#endif
494 540
495 /* r3-r13 are destroyed -- Cort */ 541 /* r3-r13 are destroyed -- Cort */
496 REST_8GPRS(14, r1) 542 REST_8GPRS(14, r1)
@@ -576,7 +622,16 @@ ALT_FW_FTR_SECTION_END_IFCLR(FW_FEATURE_ISERIES)
576 andi. r0,r3,MSR_RI 622 andi. r0,r3,MSR_RI
577 beq- unrecov_restore 623 beq- unrecov_restore
578 624
625 /*
626 * Clear the reservation. If we know the CPU tracks the address of
627 * the reservation then we can potentially save some cycles and use
628 * a larx. On POWER6 and POWER7 this is significantly faster.
629 */
630BEGIN_FTR_SECTION
579 stdcx. r0,0,r1 /* to clear the reservation */ 631 stdcx. r0,0,r1 /* to clear the reservation */
632FTR_SECTION_ELSE
633 ldarx r4,0,r1
634ALT_FTR_SECTION_END_IFCLR(CPU_FTR_STCX_CHECKS_ADDRESS)
580 635
581 /* 636 /*
582 * Clear RI before restoring r13. If we are returning to 637 * Clear RI before restoring r13. If we are returning to
@@ -798,7 +853,7 @@ _GLOBAL(enter_rtas)
798 853
799_STATIC(rtas_return_loc) 854_STATIC(rtas_return_loc)
800 /* relocation is off at this point */ 855 /* relocation is off at this point */
801 mfspr r4,SPRN_SPRG_PACA /* Get PACA */ 856 GET_PACA(r4)
802 clrldi r4,r4,2 /* convert to realmode address */ 857 clrldi r4,r4,2 /* convert to realmode address */
803 858
804 bcl 20,31,$+4 859 bcl 20,31,$+4
@@ -829,7 +884,7 @@ _STATIC(rtas_restore_regs)
829 REST_8GPRS(14, r1) /* Restore the non-volatiles */ 884 REST_8GPRS(14, r1) /* Restore the non-volatiles */
830 REST_10GPRS(22, r1) /* ditto */ 885 REST_10GPRS(22, r1) /* ditto */
831 886
832 mfspr r13,SPRN_SPRG_PACA 887 GET_PACA(r13)
833 888
834 ld r4,_CCR(r1) 889 ld r4,_CCR(r1)
835 mtcr r4 890 mtcr r4
diff --git a/arch/powerpc/kernel/exceptions-64e.S b/arch/powerpc/kernel/exceptions-64e.S
index 5c43063d2506..d24d4400cc79 100644
--- a/arch/powerpc/kernel/exceptions-64e.S
+++ b/arch/powerpc/kernel/exceptions-64e.S
@@ -17,6 +17,7 @@
17#include <asm/cputable.h> 17#include <asm/cputable.h>
18#include <asm/setup.h> 18#include <asm/setup.h>
19#include <asm/thread_info.h> 19#include <asm/thread_info.h>
20#include <asm/reg_a2.h>
20#include <asm/exception-64e.h> 21#include <asm/exception-64e.h>
21#include <asm/bug.h> 22#include <asm/bug.h>
22#include <asm/irqflags.h> 23#include <asm/irqflags.h>
@@ -252,9 +253,6 @@ exception_marker:
252 .balign 0x1000 253 .balign 0x1000
253 .globl interrupt_base_book3e 254 .globl interrupt_base_book3e
254interrupt_base_book3e: /* fake trap */ 255interrupt_base_book3e: /* fake trap */
255 /* Note: If real debug exceptions are supported by the HW, the vector
256 * below will have to be patched up to point to an appropriate handler
257 */
258 EXCEPTION_STUB(0x000, machine_check) /* 0x0200 */ 256 EXCEPTION_STUB(0x000, machine_check) /* 0x0200 */
259 EXCEPTION_STUB(0x020, critical_input) /* 0x0580 */ 257 EXCEPTION_STUB(0x020, critical_input) /* 0x0580 */
260 EXCEPTION_STUB(0x040, debug_crit) /* 0x0d00 */ 258 EXCEPTION_STUB(0x040, debug_crit) /* 0x0d00 */
@@ -271,8 +269,13 @@ interrupt_base_book3e: /* fake trap */
271 EXCEPTION_STUB(0x1a0, watchdog) /* 0x09f0 */ 269 EXCEPTION_STUB(0x1a0, watchdog) /* 0x09f0 */
272 EXCEPTION_STUB(0x1c0, data_tlb_miss) 270 EXCEPTION_STUB(0x1c0, data_tlb_miss)
273 EXCEPTION_STUB(0x1e0, instruction_tlb_miss) 271 EXCEPTION_STUB(0x1e0, instruction_tlb_miss)
272 EXCEPTION_STUB(0x260, perfmon)
274 EXCEPTION_STUB(0x280, doorbell) 273 EXCEPTION_STUB(0x280, doorbell)
275 EXCEPTION_STUB(0x2a0, doorbell_crit) 274 EXCEPTION_STUB(0x2a0, doorbell_crit)
275 EXCEPTION_STUB(0x2c0, guest_doorbell)
276 EXCEPTION_STUB(0x2e0, guest_doorbell_crit)
277 EXCEPTION_STUB(0x300, hypercall)
278 EXCEPTION_STUB(0x320, ehpriv)
276 279
277 .globl interrupt_end_book3e 280 .globl interrupt_end_book3e
278interrupt_end_book3e: 281interrupt_end_book3e:
@@ -379,7 +382,7 @@ interrupt_end_book3e:
379 mfspr r13,SPRN_SPRG_PACA /* get our PACA */ 382 mfspr r13,SPRN_SPRG_PACA /* get our PACA */
380 b system_call_common 383 b system_call_common
381 384
382/* Auxillary Processor Unavailable Interrupt */ 385/* Auxiliary Processor Unavailable Interrupt */
383 START_EXCEPTION(ap_unavailable); 386 START_EXCEPTION(ap_unavailable);
384 NORMAL_EXCEPTION_PROLOG(0xf20, PROLOG_ADDITION_NONE) 387 NORMAL_EXCEPTION_PROLOG(0xf20, PROLOG_ADDITION_NONE)
385 EXCEPTION_COMMON(0xf20, PACA_EXGEN, INTS_KEEP) 388 EXCEPTION_COMMON(0xf20, PACA_EXGEN, INTS_KEEP)
@@ -454,6 +457,70 @@ interrupt_end_book3e:
454kernel_dbg_exc: 457kernel_dbg_exc:
455 b . /* NYI */ 458 b . /* NYI */
456 459
460/* Debug exception as a debug interrupt*/
461 START_EXCEPTION(debug_debug);
462 DBG_EXCEPTION_PROLOG(0xd00, PROLOG_ADDITION_2REGS)
463
464 /*
465 * If there is a single step or branch-taken exception in an
466 * exception entry sequence, it was probably meant to apply to
467 * the code where the exception occurred (since exception entry
468 * doesn't turn off DE automatically). We simulate the effect
469 * of turning off DE on entry to an exception handler by turning
470 * off DE in the DSRR1 value and clearing the debug status.
471 */
472
473 mfspr r14,SPRN_DBSR /* check single-step/branch taken */
474 andis. r15,r14,DBSR_IC@h
475 beq+ 1f
476
477 LOAD_REG_IMMEDIATE(r14,interrupt_base_book3e)
478 LOAD_REG_IMMEDIATE(r15,interrupt_end_book3e)
479 cmpld cr0,r10,r14
480 cmpld cr1,r10,r15
481 blt+ cr0,1f
482 bge+ cr1,1f
483
484 /* here it looks like we got an inappropriate debug exception. */
485 lis r14,DBSR_IC@h /* clear the IC event */
486 rlwinm r11,r11,0,~MSR_DE /* clear DE in the DSRR1 value */
487 mtspr SPRN_DBSR,r14
488 mtspr SPRN_DSRR1,r11
489 lwz r10,PACA_EXDBG+EX_CR(r13) /* restore registers */
490 ld r1,PACA_EXDBG+EX_R1(r13)
491 ld r14,PACA_EXDBG+EX_R14(r13)
492 ld r15,PACA_EXDBG+EX_R15(r13)
493 mtcr r10
494 ld r10,PACA_EXDBG+EX_R10(r13) /* restore registers */
495 ld r11,PACA_EXDBG+EX_R11(r13)
496 mfspr r13,SPRN_SPRG_DBG_SCRATCH
497 rfdi
498
499 /* Normal debug exception */
500 /* XXX We only handle coming from userspace for now since we can't
501 * quite save properly an interrupted kernel state yet
502 */
5031: andi. r14,r11,MSR_PR; /* check for userspace again */
504 beq kernel_dbg_exc; /* if from kernel mode */
505
506 /* Now we mash up things to make it look like we are coming on a
507 * normal exception
508 */
509 mfspr r15,SPRN_SPRG_DBG_SCRATCH
510 mtspr SPRN_SPRG_GEN_SCRATCH,r15
511 mfspr r14,SPRN_DBSR
512 EXCEPTION_COMMON(0xd00, PACA_EXDBG, INTS_DISABLE_ALL)
513 std r14,_DSISR(r1)
514 addi r3,r1,STACK_FRAME_OVERHEAD
515 mr r4,r14
516 ld r14,PACA_EXDBG+EX_R14(r13)
517 ld r15,PACA_EXDBG+EX_R15(r13)
518 bl .save_nvgprs
519 bl .DebugException
520 b .ret_from_except
521
522 MASKABLE_EXCEPTION(0x260, perfmon, .performance_monitor_exception, ACK_NONE)
523
457/* Doorbell interrupt */ 524/* Doorbell interrupt */
458 MASKABLE_EXCEPTION(0x2070, doorbell, .doorbell_exception, ACK_NONE) 525 MASKABLE_EXCEPTION(0x2070, doorbell, .doorbell_exception, ACK_NONE)
459 526
@@ -468,6 +535,11 @@ kernel_dbg_exc:
468// b ret_from_crit_except 535// b ret_from_crit_except
469 b . 536 b .
470 537
538 MASKABLE_EXCEPTION(0x2c0, guest_doorbell, .unknown_exception, ACK_NONE)
539 MASKABLE_EXCEPTION(0x2e0, guest_doorbell_crit, .unknown_exception, ACK_NONE)
540 MASKABLE_EXCEPTION(0x310, hypercall, .unknown_exception, ACK_NONE)
541 MASKABLE_EXCEPTION(0x320, ehpriv, .unknown_exception, ACK_NONE)
542
471 543
472/* 544/*
473 * An interrupt came in while soft-disabled; clear EE in SRR1, 545 * An interrupt came in while soft-disabled; clear EE in SRR1,
@@ -587,7 +659,12 @@ fast_exception_return:
587BAD_STACK_TRAMPOLINE(0x000) 659BAD_STACK_TRAMPOLINE(0x000)
588BAD_STACK_TRAMPOLINE(0x100) 660BAD_STACK_TRAMPOLINE(0x100)
589BAD_STACK_TRAMPOLINE(0x200) 661BAD_STACK_TRAMPOLINE(0x200)
662BAD_STACK_TRAMPOLINE(0x260)
663BAD_STACK_TRAMPOLINE(0x2c0)
664BAD_STACK_TRAMPOLINE(0x2e0)
590BAD_STACK_TRAMPOLINE(0x300) 665BAD_STACK_TRAMPOLINE(0x300)
666BAD_STACK_TRAMPOLINE(0x310)
667BAD_STACK_TRAMPOLINE(0x320)
591BAD_STACK_TRAMPOLINE(0x400) 668BAD_STACK_TRAMPOLINE(0x400)
592BAD_STACK_TRAMPOLINE(0x500) 669BAD_STACK_TRAMPOLINE(0x500)
593BAD_STACK_TRAMPOLINE(0x600) 670BAD_STACK_TRAMPOLINE(0x600)
@@ -864,8 +941,23 @@ have_hes:
864 * that will have to be made dependent on whether we are running under 941 * that will have to be made dependent on whether we are running under
865 * a hypervisor I suppose. 942 * a hypervisor I suppose.
866 */ 943 */
867 ori r3,r3,MAS0_HES | MAS0_WQ_ALLWAYS 944
868 mtspr SPRN_MAS0,r3 945 /* BEWARE, MAGIC
946 * This code is called as an ordinary function on the boot CPU. But to
947 * avoid duplication, this code is also used in SCOM bringup of
948 * secondary CPUs. We read the code between the initial_tlb_code_start
949 * and initial_tlb_code_end labels one instruction at a time and RAM it
950 * into the new core via SCOM. That doesn't process branches, so there
951 * must be none between those two labels. It also means if this code
952 * ever takes any parameters, the SCOM code must also be updated to
953 * provide them.
954 */
955 .globl a2_tlbinit_code_start
956a2_tlbinit_code_start:
957
958 ori r11,r3,MAS0_WQ_ALLWAYS
959 oris r11,r11,MAS0_ESEL(3)@h /* Use way 3: workaround A2 erratum 376 */
960 mtspr SPRN_MAS0,r11
869 lis r3,(MAS1_VALID | MAS1_IPROT)@h 961 lis r3,(MAS1_VALID | MAS1_IPROT)@h
870 ori r3,r3,BOOK3E_PAGESZ_1GB << MAS1_TSIZE_SHIFT 962 ori r3,r3,BOOK3E_PAGESZ_1GB << MAS1_TSIZE_SHIFT
871 mtspr SPRN_MAS1,r3 963 mtspr SPRN_MAS1,r3
@@ -879,18 +971,86 @@ have_hes:
879 /* Write the TLB entry */ 971 /* Write the TLB entry */
880 tlbwe 972 tlbwe
881 973
974 .globl a2_tlbinit_after_linear_map
975a2_tlbinit_after_linear_map:
976
882 /* Now we branch the new virtual address mapped by this entry */ 977 /* Now we branch the new virtual address mapped by this entry */
883 LOAD_REG_IMMEDIATE(r3,1f) 978 LOAD_REG_IMMEDIATE(r3,1f)
884 mtctr r3 979 mtctr r3
885 bctr 980 bctr
886 981
8871: /* We are now running at PAGE_OFFSET, clean the TLB of everything 9821: /* We are now running at PAGE_OFFSET, clean the TLB of everything
888 * else (XXX we should scan for bolted crap from the firmware too) 983 * else (including IPROTed things left by firmware)
984 * r4 = TLBnCFG
985 * r3 = current address (more or less)
889 */ 986 */
987
988 li r5,0
989 mtspr SPRN_MAS6,r5
990 tlbsx 0,r3
991
992 rlwinm r9,r4,0,TLBnCFG_N_ENTRY
993 rlwinm r10,r4,8,0xff
994 addi r10,r10,-1 /* Get inner loop mask */
995
996 li r3,1
997
998 mfspr r5,SPRN_MAS1
999 rlwinm r5,r5,0,(~(MAS1_VALID|MAS1_IPROT))
1000
1001 mfspr r6,SPRN_MAS2
1002 rldicr r6,r6,0,51 /* Extract EPN */
1003
1004 mfspr r7,SPRN_MAS0
1005 rlwinm r7,r7,0,0xffff0fff /* Clear HES and WQ */
1006
1007 rlwinm r8,r7,16,0xfff /* Extract ESEL */
1008
10092: add r4,r3,r8
1010 and r4,r4,r10
1011
1012 rlwimi r7,r4,16,MAS0_ESEL_MASK
1013
1014 mtspr SPRN_MAS0,r7
1015 mtspr SPRN_MAS1,r5
1016 mtspr SPRN_MAS2,r6
1017 tlbwe
1018
1019 addi r3,r3,1
1020 and. r4,r3,r10
1021
1022 bne 3f
1023 addis r6,r6,(1<<30)@h
10243:
1025 cmpw r3,r9
1026 blt 2b
1027
1028 .globl a2_tlbinit_after_iprot_flush
1029a2_tlbinit_after_iprot_flush:
1030
1031#ifdef CONFIG_PPC_EARLY_DEBUG_WSP
1032 /* Now establish early debug mappings if applicable */
1033 /* Restore the MAS0 we used for linear mapping load */
1034 mtspr SPRN_MAS0,r11
1035
1036 lis r3,(MAS1_VALID | MAS1_IPROT)@h
1037 ori r3,r3,(BOOK3E_PAGESZ_4K << MAS1_TSIZE_SHIFT)
1038 mtspr SPRN_MAS1,r3
1039 LOAD_REG_IMMEDIATE(r3, WSP_UART_VIRT | MAS2_I | MAS2_G)
1040 mtspr SPRN_MAS2,r3
1041 LOAD_REG_IMMEDIATE(r3, WSP_UART_PHYS | MAS3_SR | MAS3_SW)
1042 mtspr SPRN_MAS7_MAS3,r3
1043 /* re-use the MAS8 value from the linear mapping */
1044 tlbwe
1045#endif /* CONFIG_PPC_EARLY_DEBUG_WSP */
1046
890 PPC_TLBILX(0,0,0) 1047 PPC_TLBILX(0,0,0)
891 sync 1048 sync
892 isync 1049 isync
893 1050
1051 .globl a2_tlbinit_code_end
1052a2_tlbinit_code_end:
1053
894 /* We translate LR and return */ 1054 /* We translate LR and return */
895 mflr r3 1055 mflr r3
896 tovirt(r3,r3) 1056 tovirt(r3,r3)
@@ -1040,3 +1200,33 @@ _GLOBAL(__setup_base_ivors)
1040 sync 1200 sync
1041 1201
1042 blr 1202 blr
1203
1204_GLOBAL(setup_perfmon_ivor)
1205 SET_IVOR(35, 0x260) /* Performance Monitor */
1206 blr
1207
1208_GLOBAL(setup_doorbell_ivors)
1209 SET_IVOR(36, 0x280) /* Processor Doorbell */
1210 SET_IVOR(37, 0x2a0) /* Processor Doorbell Crit */
1211
1212 /* Check MMUCFG[LPIDSIZE] to determine if we have category E.HV */
1213 mfspr r10,SPRN_MMUCFG
1214 rlwinm. r10,r10,0,MMUCFG_LPIDSIZE
1215 beqlr
1216
1217 SET_IVOR(38, 0x2c0) /* Guest Processor Doorbell */
1218 SET_IVOR(39, 0x2e0) /* Guest Processor Doorbell Crit/MC */
1219 blr
1220
1221_GLOBAL(setup_ehv_ivors)
1222 /*
1223 * We may be running as a guest and lack E.HV even on a chip
1224 * that normally has it.
1225 */
1226 mfspr r10,SPRN_MMUCFG
1227 rlwinm. r10,r10,0,MMUCFG_LPIDSIZE
1228 beqlr
1229
1230 SET_IVOR(40, 0x300) /* Embedded Hypervisor System Call */
1231 SET_IVOR(41, 0x320) /* Embedded Hypervisor Privilege */
1232 blr
diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index f53029a01554..a85f4874cba7 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -5,7 +5,7 @@
5 * handling and other fixed offset specific things. 5 * handling and other fixed offset specific things.
6 * 6 *
7 * This file is meant to be #included from head_64.S due to 7 * This file is meant to be #included from head_64.S due to
8 * position dependant assembly. 8 * position dependent assembly.
9 * 9 *
10 * Most of this originates from head_64.S and thus has the same 10 * Most of this originates from head_64.S and thus has the same
11 * copyright history. 11 * copyright history.
@@ -13,6 +13,7 @@
13 */ 13 */
14 14
15#include <asm/exception-64s.h> 15#include <asm/exception-64s.h>
16#include <asm/ptrace.h>
16 17
17/* 18/*
18 * We layout physical memory as follows: 19 * We layout physical memory as follows:
@@ -36,23 +37,51 @@
36 .globl __start_interrupts 37 .globl __start_interrupts
37__start_interrupts: 38__start_interrupts:
38 39
39 STD_EXCEPTION_PSERIES(0x100, system_reset) 40 .globl system_reset_pSeries;
41system_reset_pSeries:
42 HMT_MEDIUM;
43 DO_KVM 0x100;
44 SET_SCRATCH0(r13)
45#ifdef CONFIG_PPC_P7_NAP
46BEGIN_FTR_SECTION
47 /* Running native on arch 2.06 or later, check if we are
48 * waking up from nap. We only handle no state loss and
49 * supervisor state loss. We do -not- handle hypervisor
50 * state loss at this time.
51 */
52 mfspr r13,SPRN_SRR1
53 rlwinm r13,r13,47-31,30,31
54 cmpwi cr0,r13,1
55 bne 1f
56 b .power7_wakeup_noloss
571: cmpwi cr0,r13,2
58 bne 1f
59 b .power7_wakeup_loss
60 /* Total loss of HV state is fatal, we could try to use the
61 * PIR to locate a PACA, then use an emergency stack etc...
62 * but for now, let's just stay stuck here
63 */
641: cmpwi cr0,r13,3
65 beq .
66END_FTR_SECTION_IFSET(CPU_FTR_HVMODE_206)
67#endif /* CONFIG_PPC_P7_NAP */
68 EXCEPTION_PROLOG_PSERIES(PACA_EXGEN, system_reset_common, EXC_STD)
40 69
41 . = 0x200 70 . = 0x200
42_machine_check_pSeries: 71_machine_check_pSeries:
43 HMT_MEDIUM 72 HMT_MEDIUM
44 DO_KVM 0x200 73 DO_KVM 0x200
45 mtspr SPRN_SPRG_SCRATCH0,r13 /* save r13 */ 74 SET_SCRATCH0(r13)
46 EXCEPTION_PROLOG_PSERIES(PACA_EXMC, machine_check_common) 75 EXCEPTION_PROLOG_PSERIES(PACA_EXMC, machine_check_common, EXC_STD)
47 76
48 . = 0x300 77 . = 0x300
49 .globl data_access_pSeries 78 .globl data_access_pSeries
50data_access_pSeries: 79data_access_pSeries:
51 HMT_MEDIUM 80 HMT_MEDIUM
52 DO_KVM 0x300 81 DO_KVM 0x300
53 mtspr SPRN_SPRG_SCRATCH0,r13 82 SET_SCRATCH0(r13)
54BEGIN_FTR_SECTION 83BEGIN_FTR_SECTION
55 mfspr r13,SPRN_SPRG_PACA 84 GET_PACA(r13)
56 std r9,PACA_EXSLB+EX_R9(r13) 85 std r9,PACA_EXSLB+EX_R9(r13)
57 std r10,PACA_EXSLB+EX_R10(r13) 86 std r10,PACA_EXSLB+EX_R10(r13)
58 mfspr r10,SPRN_DAR 87 mfspr r10,SPRN_DAR
@@ -66,22 +95,22 @@ BEGIN_FTR_SECTION
66 std r11,PACA_EXGEN+EX_R11(r13) 95 std r11,PACA_EXGEN+EX_R11(r13)
67 ld r11,PACA_EXSLB+EX_R9(r13) 96 ld r11,PACA_EXSLB+EX_R9(r13)
68 std r12,PACA_EXGEN+EX_R12(r13) 97 std r12,PACA_EXGEN+EX_R12(r13)
69 mfspr r12,SPRN_SPRG_SCRATCH0 98 GET_SCRATCH0(r12)
70 std r10,PACA_EXGEN+EX_R10(r13) 99 std r10,PACA_EXGEN+EX_R10(r13)
71 std r11,PACA_EXGEN+EX_R9(r13) 100 std r11,PACA_EXGEN+EX_R9(r13)
72 std r12,PACA_EXGEN+EX_R13(r13) 101 std r12,PACA_EXGEN+EX_R13(r13)
73 EXCEPTION_PROLOG_PSERIES_1(data_access_common) 102 EXCEPTION_PROLOG_PSERIES_1(data_access_common, EXC_STD)
74FTR_SECTION_ELSE 103FTR_SECTION_ELSE
75 EXCEPTION_PROLOG_PSERIES(PACA_EXGEN, data_access_common) 104 EXCEPTION_PROLOG_PSERIES(PACA_EXGEN, data_access_common, EXC_STD)
76ALT_FTR_SECTION_END_IFCLR(CPU_FTR_SLB) 105ALT_MMU_FTR_SECTION_END_IFCLR(MMU_FTR_SLB)
77 106
78 . = 0x380 107 . = 0x380
79 .globl data_access_slb_pSeries 108 .globl data_access_slb_pSeries
80data_access_slb_pSeries: 109data_access_slb_pSeries:
81 HMT_MEDIUM 110 HMT_MEDIUM
82 DO_KVM 0x380 111 DO_KVM 0x380
83 mtspr SPRN_SPRG_SCRATCH0,r13 112 SET_SCRATCH0(r13)
84 mfspr r13,SPRN_SPRG_PACA /* get paca address into r13 */ 113 GET_PACA(r13)
85 std r3,PACA_EXSLB+EX_R3(r13) 114 std r3,PACA_EXSLB+EX_R3(r13)
86 mfspr r3,SPRN_DAR 115 mfspr r3,SPRN_DAR
87 std r9,PACA_EXSLB+EX_R9(r13) /* save r9 - r12 */ 116 std r9,PACA_EXSLB+EX_R9(r13) /* save r9 - r12 */
@@ -94,7 +123,7 @@ data_access_slb_pSeries:
94 std r10,PACA_EXSLB+EX_R10(r13) 123 std r10,PACA_EXSLB+EX_R10(r13)
95 std r11,PACA_EXSLB+EX_R11(r13) 124 std r11,PACA_EXSLB+EX_R11(r13)
96 std r12,PACA_EXSLB+EX_R12(r13) 125 std r12,PACA_EXSLB+EX_R12(r13)
97 mfspr r10,SPRN_SPRG_SCRATCH0 126 GET_SCRATCH0(r10)
98 std r10,PACA_EXSLB+EX_R13(r13) 127 std r10,PACA_EXSLB+EX_R13(r13)
99 mfspr r12,SPRN_SRR1 /* and SRR1 */ 128 mfspr r12,SPRN_SRR1 /* and SRR1 */
100#ifndef CONFIG_RELOCATABLE 129#ifndef CONFIG_RELOCATABLE
@@ -112,15 +141,15 @@ data_access_slb_pSeries:
112 bctr 141 bctr
113#endif 142#endif
114 143
115 STD_EXCEPTION_PSERIES(0x400, instruction_access) 144 STD_EXCEPTION_PSERIES(0x400, 0x400, instruction_access)
116 145
117 . = 0x480 146 . = 0x480
118 .globl instruction_access_slb_pSeries 147 .globl instruction_access_slb_pSeries
119instruction_access_slb_pSeries: 148instruction_access_slb_pSeries:
120 HMT_MEDIUM 149 HMT_MEDIUM
121 DO_KVM 0x480 150 DO_KVM 0x480
122 mtspr SPRN_SPRG_SCRATCH0,r13 151 SET_SCRATCH0(r13)
123 mfspr r13,SPRN_SPRG_PACA /* get paca address into r13 */ 152 GET_PACA(r13)
124 std r3,PACA_EXSLB+EX_R3(r13) 153 std r3,PACA_EXSLB+EX_R3(r13)
125 mfspr r3,SPRN_SRR0 /* SRR0 is faulting address */ 154 mfspr r3,SPRN_SRR0 /* SRR0 is faulting address */
126 std r9,PACA_EXSLB+EX_R9(r13) /* save r9 - r12 */ 155 std r9,PACA_EXSLB+EX_R9(r13) /* save r9 - r12 */
@@ -133,7 +162,7 @@ instruction_access_slb_pSeries:
133 std r10,PACA_EXSLB+EX_R10(r13) 162 std r10,PACA_EXSLB+EX_R10(r13)
134 std r11,PACA_EXSLB+EX_R11(r13) 163 std r11,PACA_EXSLB+EX_R11(r13)
135 std r12,PACA_EXSLB+EX_R12(r13) 164 std r12,PACA_EXSLB+EX_R12(r13)
136 mfspr r10,SPRN_SPRG_SCRATCH0 165 GET_SCRATCH0(r10)
137 std r10,PACA_EXSLB+EX_R13(r13) 166 std r10,PACA_EXSLB+EX_R13(r13)
138 mfspr r12,SPRN_SRR1 /* and SRR1 */ 167 mfspr r12,SPRN_SRR1 /* and SRR1 */
139#ifndef CONFIG_RELOCATABLE 168#ifndef CONFIG_RELOCATABLE
@@ -146,13 +175,29 @@ instruction_access_slb_pSeries:
146 bctr 175 bctr
147#endif 176#endif
148 177
149 MASKABLE_EXCEPTION_PSERIES(0x500, hardware_interrupt) 178 /* We open code these as we can't have a ". = x" (even with
150 STD_EXCEPTION_PSERIES(0x600, alignment) 179 * x = "." within a feature section
151 STD_EXCEPTION_PSERIES(0x700, program_check) 180 */
152 STD_EXCEPTION_PSERIES(0x800, fp_unavailable) 181 . = 0x500;
153 MASKABLE_EXCEPTION_PSERIES(0x900, decrementer) 182 .globl hardware_interrupt_pSeries;
154 STD_EXCEPTION_PSERIES(0xa00, trap_0a) 183 .globl hardware_interrupt_hv;
155 STD_EXCEPTION_PSERIES(0xb00, trap_0b) 184hardware_interrupt_pSeries:
185hardware_interrupt_hv:
186 BEGIN_FTR_SECTION
187 _MASKABLE_EXCEPTION_PSERIES(0x500, hardware_interrupt, EXC_STD)
188 FTR_SECTION_ELSE
189 _MASKABLE_EXCEPTION_PSERIES(0x502, hardware_interrupt, EXC_HV)
190 ALT_FTR_SECTION_END_IFCLR(CPU_FTR_HVMODE_206)
191
192 STD_EXCEPTION_PSERIES(0x600, 0x600, alignment)
193 STD_EXCEPTION_PSERIES(0x700, 0x700, program_check)
194 STD_EXCEPTION_PSERIES(0x800, 0x800, fp_unavailable)
195
196 MASKABLE_EXCEPTION_PSERIES(0x900, 0x900, decrementer)
197 MASKABLE_EXCEPTION_HV(0x980, 0x980, decrementer)
198
199 STD_EXCEPTION_PSERIES(0xa00, 0xa00, trap_0a)
200 STD_EXCEPTION_PSERIES(0xb00, 0xb00, trap_0b)
156 201
157 . = 0xc00 202 . = 0xc00
158 .globl system_call_pSeries 203 .globl system_call_pSeries
@@ -164,13 +209,13 @@ BEGIN_FTR_SECTION
164 beq- 1f 209 beq- 1f
165END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE) 210END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
166 mr r9,r13 211 mr r9,r13
167 mfspr r13,SPRN_SPRG_PACA 212 GET_PACA(r13)
168 mfspr r11,SPRN_SRR0 213 mfspr r11,SPRN_SRR0
169 ld r12,PACAKBASE(r13)
170 ld r10,PACAKMSR(r13)
171 LOAD_HANDLER(r12, system_call_entry)
172 mtspr SPRN_SRR0,r12
173 mfspr r12,SPRN_SRR1 214 mfspr r12,SPRN_SRR1
215 ld r10,PACAKBASE(r13)
216 LOAD_HANDLER(r10, system_call_entry)
217 mtspr SPRN_SRR0,r10
218 ld r10,PACAKMSR(r13)
174 mtspr SPRN_SRR1,r10 219 mtspr SPRN_SRR1,r10
175 rfid 220 rfid
176 b . /* prevent speculative execution */ 221 b . /* prevent speculative execution */
@@ -182,8 +227,21 @@ END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
182 rfid /* return to userspace */ 227 rfid /* return to userspace */
183 b . 228 b .
184 229
185 STD_EXCEPTION_PSERIES(0xd00, single_step) 230 STD_EXCEPTION_PSERIES(0xd00, 0xd00, single_step)
186 STD_EXCEPTION_PSERIES(0xe00, trap_0e) 231
232 /* At 0xe??? we have a bunch of hypervisor exceptions, we branch
233 * out of line to handle them
234 */
235 . = 0xe00
236 b h_data_storage_hv
237 . = 0xe20
238 b h_instr_storage_hv
239 . = 0xe40
240 b emulation_assist_hv
241 . = 0xe50
242 b hmi_exception_hv
243 . = 0xe60
244 b hmi_exception_hv
187 245
188 /* We need to deal with the Altivec unavailable exception 246 /* We need to deal with the Altivec unavailable exception
189 * here which is at 0xf20, thus in the middle of the 247 * here which is at 0xf20, thus in the middle of the
@@ -192,39 +250,42 @@ END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
192 */ 250 */
193performance_monitor_pSeries_1: 251performance_monitor_pSeries_1:
194 . = 0xf00 252 . = 0xf00
195 DO_KVM 0xf00
196 b performance_monitor_pSeries 253 b performance_monitor_pSeries
197 254
198altivec_unavailable_pSeries_1: 255altivec_unavailable_pSeries_1:
199 . = 0xf20 256 . = 0xf20
200 DO_KVM 0xf20
201 b altivec_unavailable_pSeries 257 b altivec_unavailable_pSeries
202 258
203vsx_unavailable_pSeries_1: 259vsx_unavailable_pSeries_1:
204 . = 0xf40 260 . = 0xf40
205 DO_KVM 0xf40
206 b vsx_unavailable_pSeries 261 b vsx_unavailable_pSeries
207 262
208#ifdef CONFIG_CBE_RAS 263#ifdef CONFIG_CBE_RAS
209 HSTD_EXCEPTION_PSERIES(0x1200, cbe_system_error) 264 STD_EXCEPTION_HV(0x1200, 0x1202, cbe_system_error)
210#endif /* CONFIG_CBE_RAS */ 265#endif /* CONFIG_CBE_RAS */
211 STD_EXCEPTION_PSERIES(0x1300, instruction_breakpoint) 266 STD_EXCEPTION_PSERIES(0x1300, 0x1300, instruction_breakpoint)
212#ifdef CONFIG_CBE_RAS 267#ifdef CONFIG_CBE_RAS
213 HSTD_EXCEPTION_PSERIES(0x1600, cbe_maintenance) 268 STD_EXCEPTION_HV(0x1600, 0x1602, cbe_maintenance)
214#endif /* CONFIG_CBE_RAS */ 269#endif /* CONFIG_CBE_RAS */
215 STD_EXCEPTION_PSERIES(0x1700, altivec_assist) 270 STD_EXCEPTION_PSERIES(0x1700, 0x1700, altivec_assist)
216#ifdef CONFIG_CBE_RAS 271#ifdef CONFIG_CBE_RAS
217 HSTD_EXCEPTION_PSERIES(0x1800, cbe_thermal) 272 STD_EXCEPTION_HV(0x1800, 0x1802, cbe_thermal)
218#endif /* CONFIG_CBE_RAS */ 273#endif /* CONFIG_CBE_RAS */
219 274
220 . = 0x3000 275 . = 0x3000
221 276
222/*** pSeries interrupt support ***/ 277/*** Out of line interrupts support ***/
278
279 /* moved from 0xe00 */
280 STD_EXCEPTION_HV(., 0xe00, h_data_storage)
281 STD_EXCEPTION_HV(., 0xe20, h_instr_storage)
282 STD_EXCEPTION_HV(., 0xe40, emulation_assist)
283 STD_EXCEPTION_HV(., 0xe60, hmi_exception) /* need to flush cache ? */
223 284
224 /* moved from 0xf00 */ 285 /* moved from 0xf00 */
225 STD_EXCEPTION_PSERIES(., performance_monitor) 286 STD_EXCEPTION_PSERIES(., 0xf00, performance_monitor)
226 STD_EXCEPTION_PSERIES(., altivec_unavailable) 287 STD_EXCEPTION_PSERIES(., 0xf20, altivec_unavailable)
227 STD_EXCEPTION_PSERIES(., vsx_unavailable) 288 STD_EXCEPTION_PSERIES(., 0xf40, vsx_unavailable)
228 289
229/* 290/*
230 * An interrupt came in while soft-disabled; clear EE in SRR1, 291 * An interrupt came in while soft-disabled; clear EE in SRR1,
@@ -239,17 +300,30 @@ masked_interrupt:
239 rotldi r10,r10,16 300 rotldi r10,r10,16
240 mtspr SPRN_SRR1,r10 301 mtspr SPRN_SRR1,r10
241 ld r10,PACA_EXGEN+EX_R10(r13) 302 ld r10,PACA_EXGEN+EX_R10(r13)
242 mfspr r13,SPRN_SPRG_SCRATCH0 303 GET_SCRATCH0(r13)
243 rfid 304 rfid
244 b . 305 b .
245 306
307masked_Hinterrupt:
308 stb r10,PACAHARDIRQEN(r13)
309 mtcrf 0x80,r9
310 ld r9,PACA_EXGEN+EX_R9(r13)
311 mfspr r10,SPRN_HSRR1
312 rldicl r10,r10,48,1 /* clear MSR_EE */
313 rotldi r10,r10,16
314 mtspr SPRN_HSRR1,r10
315 ld r10,PACA_EXGEN+EX_R10(r13)
316 GET_SCRATCH0(r13)
317 hrfid
318 b .
319
246 .align 7 320 .align 7
247do_stab_bolted_pSeries: 321do_stab_bolted_pSeries:
248 std r11,PACA_EXSLB+EX_R11(r13) 322 std r11,PACA_EXSLB+EX_R11(r13)
249 std r12,PACA_EXSLB+EX_R12(r13) 323 std r12,PACA_EXSLB+EX_R12(r13)
250 mfspr r10,SPRN_SPRG_SCRATCH0 324 GET_SCRATCH0(r10)
251 std r10,PACA_EXSLB+EX_R13(r13) 325 std r10,PACA_EXSLB+EX_R13(r13)
252 EXCEPTION_PROLOG_PSERIES_1(.do_stab_bolted) 326 EXCEPTION_PROLOG_PSERIES_1(.do_stab_bolted, EXC_STD)
253 327
254#ifdef CONFIG_PPC_PSERIES 328#ifdef CONFIG_PPC_PSERIES
255/* 329/*
@@ -259,15 +333,15 @@ do_stab_bolted_pSeries:
259 .align 7 333 .align 7
260system_reset_fwnmi: 334system_reset_fwnmi:
261 HMT_MEDIUM 335 HMT_MEDIUM
262 mtspr SPRN_SPRG_SCRATCH0,r13 /* save r13 */ 336 SET_SCRATCH0(r13) /* save r13 */
263 EXCEPTION_PROLOG_PSERIES(PACA_EXGEN, system_reset_common) 337 EXCEPTION_PROLOG_PSERIES(PACA_EXGEN, system_reset_common, EXC_STD)
264 338
265 .globl machine_check_fwnmi 339 .globl machine_check_fwnmi
266 .align 7 340 .align 7
267machine_check_fwnmi: 341machine_check_fwnmi:
268 HMT_MEDIUM 342 HMT_MEDIUM
269 mtspr SPRN_SPRG_SCRATCH0,r13 /* save r13 */ 343 SET_SCRATCH0(r13) /* save r13 */
270 EXCEPTION_PROLOG_PSERIES(PACA_EXMC, machine_check_common) 344 EXCEPTION_PROLOG_PSERIES(PACA_EXMC, machine_check_common, EXC_STD)
271 345
272#endif /* CONFIG_PPC_PSERIES */ 346#endif /* CONFIG_PPC_PSERIES */
273 347
@@ -281,7 +355,7 @@ slb_miss_user_pseries:
281 std r10,PACA_EXGEN+EX_R10(r13) 355 std r10,PACA_EXGEN+EX_R10(r13)
282 std r11,PACA_EXGEN+EX_R11(r13) 356 std r11,PACA_EXGEN+EX_R11(r13)
283 std r12,PACA_EXGEN+EX_R12(r13) 357 std r12,PACA_EXGEN+EX_R12(r13)
284 mfspr r10,SPRG_SCRATCH0 358 GET_SCRATCH0(r10)
285 ld r11,PACA_EXSLB+EX_R9(r13) 359 ld r11,PACA_EXSLB+EX_R9(r13)
286 ld r12,PACA_EXSLB+EX_R3(r13) 360 ld r12,PACA_EXSLB+EX_R3(r13)
287 std r10,PACA_EXGEN+EX_R13(r13) 361 std r10,PACA_EXGEN+EX_R13(r13)
@@ -299,6 +373,12 @@ slb_miss_user_pseries:
299 b . /* prevent spec. execution */ 373 b . /* prevent spec. execution */
300#endif /* __DISABLED__ */ 374#endif /* __DISABLED__ */
301 375
376/* KVM's trampoline code needs to be close to the interrupt handlers */
377
378#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
379#include "../kvm/book3s_rmhandlers.S"
380#endif
381
302 .align 7 382 .align 7
303 .globl __end_interrupts 383 .globl __end_interrupts
304__end_interrupts: 384__end_interrupts:
@@ -335,6 +415,8 @@ machine_check_common:
335 STD_EXCEPTION_COMMON(0xb00, trap_0b, .unknown_exception) 415 STD_EXCEPTION_COMMON(0xb00, trap_0b, .unknown_exception)
336 STD_EXCEPTION_COMMON(0xd00, single_step, .single_step_exception) 416 STD_EXCEPTION_COMMON(0xd00, single_step, .single_step_exception)
337 STD_EXCEPTION_COMMON(0xe00, trap_0e, .unknown_exception) 417 STD_EXCEPTION_COMMON(0xe00, trap_0e, .unknown_exception)
418 STD_EXCEPTION_COMMON(0xe40, emulation_assist, .program_check_exception)
419 STD_EXCEPTION_COMMON(0xe60, hmi_exception, .unknown_exception)
338 STD_EXCEPTION_COMMON_IDLE(0xf00, performance_monitor, .performance_monitor_exception) 420 STD_EXCEPTION_COMMON_IDLE(0xf00, performance_monitor, .performance_monitor_exception)
339 STD_EXCEPTION_COMMON(0x1300, instruction_breakpoint, .instruction_breakpoint_exception) 421 STD_EXCEPTION_COMMON(0x1300, instruction_breakpoint, .instruction_breakpoint_exception)
340#ifdef CONFIG_ALTIVEC 422#ifdef CONFIG_ALTIVEC
@@ -379,9 +461,24 @@ bad_stack:
379 std r12,_XER(r1) 461 std r12,_XER(r1)
380 SAVE_GPR(0,r1) 462 SAVE_GPR(0,r1)
381 SAVE_GPR(2,r1) 463 SAVE_GPR(2,r1)
382 SAVE_4GPRS(3,r1) 464 ld r10,EX_R3(r3)
383 SAVE_2GPRS(7,r1) 465 std r10,GPR3(r1)
384 SAVE_10GPRS(12,r1) 466 SAVE_GPR(4,r1)
467 SAVE_4GPRS(5,r1)
468 ld r9,EX_R9(r3)
469 ld r10,EX_R10(r3)
470 SAVE_2GPRS(9,r1)
471 ld r9,EX_R11(r3)
472 ld r10,EX_R12(r3)
473 ld r11,EX_R13(r3)
474 std r9,GPR11(r1)
475 std r10,GPR12(r1)
476 std r11,GPR13(r1)
477BEGIN_FTR_SECTION
478 ld r10,EX_CFAR(r3)
479 std r10,ORIG_GPR3(r1)
480END_FTR_SECTION_IFSET(CPU_FTR_CFAR)
481 SAVE_8GPRS(14,r1)
385 SAVE_10GPRS(22,r1) 482 SAVE_10GPRS(22,r1)
386 lhz r12,PACA_TRAP_SAVE(r13) 483 lhz r12,PACA_TRAP_SAVE(r13)
387 std r12,_TRAP(r1) 484 std r12,_TRAP(r1)
@@ -390,6 +487,9 @@ bad_stack:
390 li r12,0 487 li r12,0
391 std r12,0(r11) 488 std r12,0(r11)
392 ld r2,PACATOC(r13) 489 ld r2,PACATOC(r13)
490 ld r11,exception_marker@toc(r2)
491 std r12,RESULT(r1)
492 std r11,STACK_FRAME_OVERHEAD-16(r1)
3931: addi r3,r1,STACK_FRAME_OVERHEAD 4931: addi r3,r1,STACK_FRAME_OVERHEAD
394 bl .kernel_bad_stack 494 bl .kernel_bad_stack
395 b 1b 495 b 1b
@@ -412,6 +512,19 @@ data_access_common:
412 li r5,0x300 512 li r5,0x300
413 b .do_hash_page /* Try to handle as hpte fault */ 513 b .do_hash_page /* Try to handle as hpte fault */
414 514
515 .align 7
516 .globl h_data_storage_common
517h_data_storage_common:
518 mfspr r10,SPRN_HDAR
519 std r10,PACA_EXGEN+EX_DAR(r13)
520 mfspr r10,SPRN_HDSISR
521 stw r10,PACA_EXGEN+EX_DSISR(r13)
522 EXCEPTION_PROLOG_COMMON(0xe00, PACA_EXGEN)
523 bl .save_nvgprs
524 addi r3,r1,STACK_FRAME_OVERHEAD
525 bl .unknown_exception
526 b .ret_from_except
527
415 .align 7 528 .align 7
416 .globl instruction_access_common 529 .globl instruction_access_common
417instruction_access_common: 530instruction_access_common:
@@ -421,6 +534,8 @@ instruction_access_common:
421 li r5,0x400 534 li r5,0x400
422 b .do_hash_page /* Try to handle as hpte fault */ 535 b .do_hash_page /* Try to handle as hpte fault */
423 536
537 STD_EXCEPTION_COMMON(0xe20, h_instr_storage, .unknown_exception)
538
424/* 539/*
425 * Here is the common SLB miss user that is used when going to virtual 540 * Here is the common SLB miss user that is used when going to virtual
426 * mode for SLB misses, that is currently not used 541 * mode for SLB misses, that is currently not used
@@ -743,7 +858,7 @@ _STATIC(do_hash_page)
743BEGIN_FTR_SECTION 858BEGIN_FTR_SECTION
744 andis. r0,r4,0x0020 /* Is it a segment table fault? */ 859 andis. r0,r4,0x0020 /* Is it a segment table fault? */
745 bne- do_ste_alloc /* If so handle it */ 860 bne- do_ste_alloc /* If so handle it */
746END_FTR_SECTION_IFCLR(CPU_FTR_SLB) 861END_MMU_FTR_SECTION_IFCLR(MMU_FTR_SLB)
747 862
748 clrrdi r11,r1,THREAD_SHIFT 863 clrrdi r11,r1,THREAD_SHIFT
749 lwz r0,TI_PREEMPT(r11) /* If we're in an "NMI" */ 864 lwz r0,TI_PREEMPT(r11) /* If we're in an "NMI" */
@@ -818,12 +933,12 @@ END_FW_FTR_SECTION_IFCLR(FW_FEATURE_ISERIES)
818 933
819 /* 934 /*
820 * hash_page couldn't handle it, set soft interrupt enable back 935 * hash_page couldn't handle it, set soft interrupt enable back
821 * to what it was before the trap. Note that .raw_local_irq_restore 936 * to what it was before the trap. Note that .arch_local_irq_restore
822 * handles any interrupts pending at this point. 937 * handles any interrupts pending at this point.
823 */ 938 */
824 ld r3,SOFTE(r1) 939 ld r3,SOFTE(r1)
825 TRACE_AND_RESTORE_IRQ_PARTIAL(r3, 11f) 940 TRACE_AND_RESTORE_IRQ_PARTIAL(r3, 11f)
826 bl .raw_local_irq_restore 941 bl .arch_local_irq_restore
827 b 11f 942 b 11f
828 943
829/* We have a data breakpoint exception - handle it */ 944/* We have a data breakpoint exception - handle it */
@@ -970,20 +1085,6 @@ _GLOBAL(do_stab_bolted)
970 rfid 1085 rfid
971 b . /* prevent speculative execution */ 1086 b . /* prevent speculative execution */
972 1087
973/*
974 * Space for CPU0's segment table.
975 *
976 * On iSeries, the hypervisor must fill in at least one entry before
977 * we get control (with relocate on). The address is given to the hv
978 * as a page number (see xLparMap below), so this must be at a
979 * fixed address (the linker can't compute (u64)&initial_stab >>
980 * PAGE_SHIFT).
981 */
982 . = STAB0_OFFSET /* 0x6000 */
983 .globl initial_stab
984initial_stab:
985 .space 4096
986
987#ifdef CONFIG_PPC_PSERIES 1088#ifdef CONFIG_PPC_PSERIES
988/* 1089/*
989 * Data area reserved for FWNMI option. 1090 * Data area reserved for FWNMI option.
@@ -1020,3 +1121,17 @@ xLparMap:
1020#ifdef CONFIG_PPC_PSERIES 1121#ifdef CONFIG_PPC_PSERIES
1021 . = 0x8000 1122 . = 0x8000
1022#endif /* CONFIG_PPC_PSERIES */ 1123#endif /* CONFIG_PPC_PSERIES */
1124
1125/*
1126 * Space for CPU0's segment table.
1127 *
1128 * On iSeries, the hypervisor must fill in at least one entry before
1129 * we get control (with relocate on). The address is given to the hv
1130 * as a page number (see xLparMap above), so this must be at a
1131 * fixed address (the linker can't compute (u64)&initial_stab >>
1132 * PAGE_SHIFT).
1133 */
1134 . = STAB0_OFFSET /* 0x8000 */
1135 .globl initial_stab
1136initial_stab:
1137 .space 4096
diff --git a/arch/powerpc/kernel/fpu.S b/arch/powerpc/kernel/fpu.S
index fc8f5b14019c..de369558bf0a 100644
--- a/arch/powerpc/kernel/fpu.S
+++ b/arch/powerpc/kernel/fpu.S
@@ -23,6 +23,7 @@
23#include <asm/thread_info.h> 23#include <asm/thread_info.h>
24#include <asm/ppc_asm.h> 24#include <asm/ppc_asm.h>
25#include <asm/asm-offsets.h> 25#include <asm/asm-offsets.h>
26#include <asm/ptrace.h>
26 27
27#ifdef CONFIG_VSX 28#ifdef CONFIG_VSX
28#define REST_32FPVSRS(n,c,base) \ 29#define REST_32FPVSRS(n,c,base) \
@@ -163,24 +164,14 @@ END_FTR_SECTION_IFSET(CPU_FTR_VSX)
163/* 164/*
164 * These are used in the alignment trap handler when emulating 165 * These are used in the alignment trap handler when emulating
165 * single-precision loads and stores. 166 * single-precision loads and stores.
166 * We restore and save the fpscr so the task gets the same result
167 * and exceptions as if the cpu had performed the load or store.
168 */ 167 */
169 168
170_GLOBAL(cvt_fd) 169_GLOBAL(cvt_fd)
171 lfd 0,THREAD_FPSCR(r5) /* load up fpscr value */
172 MTFSF_L(0)
173 lfs 0,0(r3) 170 lfs 0,0(r3)
174 stfd 0,0(r4) 171 stfd 0,0(r4)
175 mffs 0
176 stfd 0,THREAD_FPSCR(r5) /* save new fpscr value */
177 blr 172 blr
178 173
179_GLOBAL(cvt_df) 174_GLOBAL(cvt_df)
180 lfd 0,THREAD_FPSCR(r5) /* load up fpscr value */
181 MTFSF_L(0)
182 lfd 0,0(r3) 175 lfd 0,0(r3)
183 stfs 0,0(r4) 176 stfs 0,0(r4)
184 mffs 0
185 stfd 0,THREAD_FPSCR(r5) /* save new fpscr value */
186 blr 177 blr
diff --git a/arch/powerpc/kernel/ftrace.c b/arch/powerpc/kernel/ftrace.c
index ce1f3e44c24f..bf99cfa6bbfe 100644
--- a/arch/powerpc/kernel/ftrace.c
+++ b/arch/powerpc/kernel/ftrace.c
@@ -22,6 +22,7 @@
22#include <asm/cacheflush.h> 22#include <asm/cacheflush.h>
23#include <asm/code-patching.h> 23#include <asm/code-patching.h>
24#include <asm/ftrace.h> 24#include <asm/ftrace.h>
25#include <asm/syscall.h>
25 26
26 27
27#ifdef CONFIG_DYNAMIC_FTRACE 28#ifdef CONFIG_DYNAMIC_FTRACE
@@ -600,3 +601,10 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
600 } 601 }
601} 602}
602#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ 603#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
604
605#if defined(CONFIG_FTRACE_SYSCALLS) && defined(CONFIG_PPC64)
606unsigned long __init arch_syscall_addr(int nr)
607{
608 return sys_call_table[nr*2];
609}
610#endif /* CONFIG_FTRACE_SYSCALLS && CONFIG_PPC64 */
diff --git a/arch/powerpc/kernel/head_32.S b/arch/powerpc/kernel/head_32.S
index 98c4b29a56f4..ba250d505e07 100644
--- a/arch/powerpc/kernel/head_32.S
+++ b/arch/powerpc/kernel/head_32.S
@@ -805,19 +805,6 @@ _ENTRY(copy_and_flush)
805 blr 805 blr
806 806
807#ifdef CONFIG_SMP 807#ifdef CONFIG_SMP
808#ifdef CONFIG_GEMINI
809 .globl __secondary_start_gemini
810__secondary_start_gemini:
811 mfspr r4,SPRN_HID0
812 ori r4,r4,HID0_ICFI
813 li r3,0
814 ori r3,r3,HID0_ICE
815 andc r4,r4,r3
816 mtspr SPRN_HID0,r4
817 sync
818 b __secondary_start
819#endif /* CONFIG_GEMINI */
820
821 .globl __secondary_start_mpc86xx 808 .globl __secondary_start_mpc86xx
822__secondary_start_mpc86xx: 809__secondary_start_mpc86xx:
823 mfspr r3, SPRN_PIR 810 mfspr r3, SPRN_PIR
diff --git a/arch/powerpc/kernel/head_40x.S b/arch/powerpc/kernel/head_40x.S
index a90625f9b485..a91626d87fc9 100644
--- a/arch/powerpc/kernel/head_40x.S
+++ b/arch/powerpc/kernel/head_40x.S
@@ -40,6 +40,7 @@
40#include <asm/thread_info.h> 40#include <asm/thread_info.h>
41#include <asm/ppc_asm.h> 41#include <asm/ppc_asm.h>
42#include <asm/asm-offsets.h> 42#include <asm/asm-offsets.h>
43#include <asm/ptrace.h>
43 44
44/* As with the other PowerPC ports, it is expected that when code 45/* As with the other PowerPC ports, it is expected that when code
45 * execution begins here, the following registers contain valid, yet 46 * execution begins here, the following registers contain valid, yet
@@ -765,7 +766,7 @@ DataAccess:
765 * miss get to this point to load the TLB. 766 * miss get to this point to load the TLB.
766 * r10 - TLB_TAG value 767 * r10 - TLB_TAG value
767 * r11 - Linux PTE 768 * r11 - Linux PTE
768 * r12, r9 - avilable to use 769 * r12, r9 - available to use
769 * PID - loaded with proper value when we get here 770 * PID - loaded with proper value when we get here
770 * Upon exit, we reload everything and RFI. 771 * Upon exit, we reload everything and RFI.
771 * Actually, it will fit now, but oh well.....a common place 772 * Actually, it will fit now, but oh well.....a common place
@@ -923,11 +924,7 @@ initial_mmu:
923 mtspr SPRN_PID,r0 924 mtspr SPRN_PID,r0
924 sync 925 sync
925 926
926 /* Configure and load two entries into TLB slots 62 and 63. 927 /* Configure and load one entry into TLB slots 63 */
927 * In case we are pinning TLBs, these are reserved in by the
928 * other TLB functions. If not reserving, then it doesn't
929 * matter where they are loaded.
930 */
931 clrrwi r4,r4,10 /* Mask off the real page number */ 928 clrrwi r4,r4,10 /* Mask off the real page number */
932 ori r4,r4,(TLB_WR | TLB_EX) /* Set the write and execute bits */ 929 ori r4,r4,(TLB_WR | TLB_EX) /* Set the write and execute bits */
933 930
diff --git a/arch/powerpc/kernel/head_44x.S b/arch/powerpc/kernel/head_44x.S
index 562305b40a8e..5e12b741ba5f 100644
--- a/arch/powerpc/kernel/head_44x.S
+++ b/arch/powerpc/kernel/head_44x.S
@@ -37,6 +37,7 @@
37#include <asm/thread_info.h> 37#include <asm/thread_info.h>
38#include <asm/ppc_asm.h> 38#include <asm/ppc_asm.h>
39#include <asm/asm-offsets.h> 39#include <asm/asm-offsets.h>
40#include <asm/ptrace.h>
40#include <asm/synch.h> 41#include <asm/synch.h>
41#include "head_booke.h" 42#include "head_booke.h"
42 43
@@ -177,7 +178,7 @@ interrupt_base:
177 NORMAL_EXCEPTION_PROLOG 178 NORMAL_EXCEPTION_PROLOG
178 EXC_XFER_EE_LITE(0x0c00, DoSyscall) 179 EXC_XFER_EE_LITE(0x0c00, DoSyscall)
179 180
180 /* Auxillary Processor Unavailable Interrupt */ 181 /* Auxiliary Processor Unavailable Interrupt */
181 EXCEPTION(0x2020, AuxillaryProcessorUnavailable, unknown_exception, EXC_XFER_EE) 182 EXCEPTION(0x2020, AuxillaryProcessorUnavailable, unknown_exception, EXC_XFER_EE)
182 183
183 /* Decrementer Interrupt */ 184 /* Decrementer Interrupt */
diff --git a/arch/powerpc/kernel/head_64.S b/arch/powerpc/kernel/head_64.S
index c571cd3c1453..ba504099844a 100644
--- a/arch/powerpc/kernel/head_64.S
+++ b/arch/powerpc/kernel/head_64.S
@@ -38,8 +38,9 @@
38#include <asm/page_64.h> 38#include <asm/page_64.h>
39#include <asm/irqflags.h> 39#include <asm/irqflags.h>
40#include <asm/kvm_book3s_asm.h> 40#include <asm/kvm_book3s_asm.h>
41#include <asm/ptrace.h>
41 42
42/* The physical memory is layed out such that the secondary processor 43/* The physical memory is laid out such that the secondary processor
43 * spin code sits at 0x0000...0x00ff. On server, the vectors follow 44 * spin code sits at 0x0000...0x00ff. On server, the vectors follow
44 * using the layout described in exceptions-64s.S 45 * using the layout described in exceptions-64s.S
45 */ 46 */
@@ -96,7 +97,7 @@ __secondary_hold_acknowledge:
96 .llong hvReleaseData-KERNELBASE 97 .llong hvReleaseData-KERNELBASE
97#endif /* CONFIG_PPC_ISERIES */ 98#endif /* CONFIG_PPC_ISERIES */
98 99
99#ifdef CONFIG_CRASH_DUMP 100#ifdef CONFIG_RELOCATABLE
100 /* This flag is set to 1 by a loader if the kernel should run 101 /* This flag is set to 1 by a loader if the kernel should run
101 * at the loaded address instead of the linked address. This 102 * at the loaded address instead of the linked address. This
102 * is used by kexec-tools to keep the the kdump kernel in the 103 * is used by kexec-tools to keep the the kdump kernel in the
@@ -146,6 +147,8 @@ __secondary_hold:
146 mtctr r4 147 mtctr r4
147 mr r3,r24 148 mr r3,r24
148 li r4,0 149 li r4,0
150 /* Make sure that patched code is visible */
151 isync
149 bctr 152 bctr
150#else 153#else
151 BUG_OPCODE 154 BUG_OPCODE
@@ -166,12 +169,6 @@ exception_marker:
166#include "exceptions-64s.S" 169#include "exceptions-64s.S"
167#endif 170#endif
168 171
169/* KVM trampoline code needs to be close to the interrupt handlers */
170
171#ifdef CONFIG_KVM_BOOK3S_64_HANDLER
172#include "../kvm/book3s_rmhandlers.S"
173#endif
174
175_GLOBAL(generic_secondary_thread_init) 172_GLOBAL(generic_secondary_thread_init)
176 mr r24,r3 173 mr r24,r3
177 174
@@ -221,19 +218,25 @@ generic_secondary_common_init:
221 */ 218 */
222 LOAD_REG_ADDR(r13, paca) /* Load paca pointer */ 219 LOAD_REG_ADDR(r13, paca) /* Load paca pointer */
223 ld r13,0(r13) /* Get base vaddr of paca array */ 220 ld r13,0(r13) /* Get base vaddr of paca array */
221#ifndef CONFIG_SMP
222 addi r13,r13,PACA_SIZE /* know r13 if used accidentally */
223 b .kexec_wait /* wait for next kernel if !SMP */
224#else
225 LOAD_REG_ADDR(r7, nr_cpu_ids) /* Load nr_cpu_ids address */
226 lwz r7,0(r7) /* also the max paca allocated */
224 li r5,0 /* logical cpu id */ 227 li r5,0 /* logical cpu id */
2251: lhz r6,PACAHWCPUID(r13) /* Load HW procid from paca */ 2281: lhz r6,PACAHWCPUID(r13) /* Load HW procid from paca */
226 cmpw r6,r24 /* Compare to our id */ 229 cmpw r6,r24 /* Compare to our id */
227 beq 2f 230 beq 2f
228 addi r13,r13,PACA_SIZE /* Loop to next PACA on miss */ 231 addi r13,r13,PACA_SIZE /* Loop to next PACA on miss */
229 addi r5,r5,1 232 addi r5,r5,1
230 cmpwi r5,NR_CPUS 233 cmpw r5,r7 /* Check if more pacas exist */
231 blt 1b 234 blt 1b
232 235
233 mr r3,r24 /* not found, copy phys to r3 */ 236 mr r3,r24 /* not found, copy phys to r3 */
234 b .kexec_wait /* next kernel might do better */ 237 b .kexec_wait /* next kernel might do better */
235 238
2362: mtspr SPRN_SPRG_PACA,r13 /* Save vaddr of paca in an SPRG */ 2392: SET_PACA(r13)
237#ifdef CONFIG_PPC_BOOK3E 240#ifdef CONFIG_PPC_BOOK3E
238 addi r12,r13,PACA_EXTLB /* and TLB exc frame in another */ 241 addi r12,r13,PACA_EXTLB /* and TLB exc frame in another */
239 mtspr SPRN_SPRG_TLB_EXFRAME,r12 242 mtspr SPRN_SPRG_TLB_EXFRAME,r12
@@ -241,34 +244,39 @@ generic_secondary_common_init:
241 244
242 /* From now on, r24 is expected to be logical cpuid */ 245 /* From now on, r24 is expected to be logical cpuid */
243 mr r24,r5 246 mr r24,r5
2443: HMT_LOW
245 lbz r23,PACAPROCSTART(r13) /* Test if this processor should */
246 /* start. */
247
248#ifndef CONFIG_SMP
249 b 3b /* Never go on non-SMP */
250#else
251 cmpwi 0,r23,0
252 beq 3b /* Loop until told to go */
253
254 sync /* order paca.run and cur_cpu_spec */
255 247
256 /* See if we need to call a cpu state restore handler */ 248 /* See if we need to call a cpu state restore handler */
257 LOAD_REG_ADDR(r23, cur_cpu_spec) 249 LOAD_REG_ADDR(r23, cur_cpu_spec)
258 ld r23,0(r23) 250 ld r23,0(r23)
259 ld r23,CPU_SPEC_RESTORE(r23) 251 ld r23,CPU_SPEC_RESTORE(r23)
260 cmpdi 0,r23,0 252 cmpdi 0,r23,0
261 beq 4f 253 beq 3f
262 ld r23,0(r23) 254 ld r23,0(r23)
263 mtctr r23 255 mtctr r23
264 bctrl 256 bctrl
265 257
2664: /* Create a temp kernel stack for use before relocation is on. */ 2583: LOAD_REG_ADDR(r3, boot_cpu_count) /* Decrement boot_cpu_count */
259 lwarx r4,0,r3
260 subi r4,r4,1
261 stwcx. r4,0,r3
262 bne 3b
263 isync
264
2654: HMT_LOW
266 lbz r23,PACAPROCSTART(r13) /* Test if this processor should */
267 /* start. */
268 cmpwi 0,r23,0
269 beq 4b /* Loop until told to go */
270
271 sync /* order paca.run and cur_cpu_spec */
272 isync /* In case code patching happened */
273
274 /* Create a temp kernel stack for use before relocation is on. */
267 ld r1,PACAEMERGSP(r13) 275 ld r1,PACAEMERGSP(r13)
268 subi r1,r1,STACK_FRAME_OVERHEAD 276 subi r1,r1,STACK_FRAME_OVERHEAD
269 277
270 b __secondary_start 278 b __secondary_start
271#endif 279#endif /* SMP */
272 280
273/* 281/*
274 * Turn the MMU off. 282 * Turn the MMU off.
@@ -390,12 +398,10 @@ _STATIC(__after_prom_start)
390 /* process relocations for the final address of the kernel */ 398 /* process relocations for the final address of the kernel */
391 lis r25,PAGE_OFFSET@highest /* compute virtual base of kernel */ 399 lis r25,PAGE_OFFSET@highest /* compute virtual base of kernel */
392 sldi r25,r25,32 400 sldi r25,r25,32
393#ifdef CONFIG_CRASH_DUMP
394 lwz r7,__run_at_load-_stext(r26) 401 lwz r7,__run_at_load-_stext(r26)
395 cmplwi cr0,r7,1 /* kdump kernel ? - stay where we are */ 402 cmplwi cr0,r7,1 /* flagged to stay where we are ? */
396 bne 1f 403 bne 1f
397 add r25,r25,r26 404 add r25,r25,r26
398#endif
3991: mr r3,r25 4051: mr r3,r25
400 bl .relocate 406 bl .relocate
401#endif 407#endif
@@ -541,7 +547,14 @@ _GLOBAL(pmac_secondary_start)
541 ld r4,0(r4) /* Get base vaddr of paca array */ 547 ld r4,0(r4) /* Get base vaddr of paca array */
542 mulli r13,r24,PACA_SIZE /* Calculate vaddr of right paca */ 548 mulli r13,r24,PACA_SIZE /* Calculate vaddr of right paca */
543 add r13,r13,r4 /* for this processor. */ 549 add r13,r13,r4 /* for this processor. */
544 mtspr SPRN_SPRG_PACA,r13 /* Save vaddr of paca in an SPRG*/ 550 SET_PACA(r13) /* Save vaddr of paca in an SPRG*/
551
552 /* Mark interrupts soft and hard disabled (they might be enabled
553 * in the PACA when doing hotplug)
554 */
555 li r0,0
556 stb r0,PACASOFTIRQEN(r13)
557 stb r0,PACAHARDIRQEN(r13)
545 558
546 /* Create a temp kernel stack for use before relocation is on. */ 559 /* Create a temp kernel stack for use before relocation is on. */
547 ld r1,PACAEMERGSP(r13) 560 ld r1,PACAEMERGSP(r13)
@@ -645,7 +658,7 @@ _GLOBAL(enable_64b_mode)
645 oris r11,r11,0x8000 /* CM bit set, we'll set ICM later */ 658 oris r11,r11,0x8000 /* CM bit set, we'll set ICM later */
646 mtmsr r11 659 mtmsr r11
647#else /* CONFIG_PPC_BOOK3E */ 660#else /* CONFIG_PPC_BOOK3E */
648 li r12,(MSR_SF | MSR_ISF)@highest 661 li r12,(MSR_64BIT | MSR_ISF)@highest
649 sldi r12,r12,48 662 sldi r12,r12,48
650 or r11,r11,r12 663 or r11,r11,r12
651 mtmsrd r11 664 mtmsrd r11
diff --git a/arch/powerpc/kernel/head_8xx.S b/arch/powerpc/kernel/head_8xx.S
index 1f1a04b5c2a4..1cbf64e6b416 100644
--- a/arch/powerpc/kernel/head_8xx.S
+++ b/arch/powerpc/kernel/head_8xx.S
@@ -29,6 +29,7 @@
29#include <asm/thread_info.h> 29#include <asm/thread_info.h>
30#include <asm/ppc_asm.h> 30#include <asm/ppc_asm.h>
31#include <asm/asm-offsets.h> 31#include <asm/asm-offsets.h>
32#include <asm/ptrace.h>
32 33
33/* Macro to make the code more readable. */ 34/* Macro to make the code more readable. */
34#ifdef CONFIG_8xx_CPU6 35#ifdef CONFIG_8xx_CPU6
diff --git a/arch/powerpc/kernel/head_fsl_booke.S b/arch/powerpc/kernel/head_fsl_booke.S
index 4faeba247854..5ecf54cfa7d4 100644
--- a/arch/powerpc/kernel/head_fsl_booke.S
+++ b/arch/powerpc/kernel/head_fsl_booke.S
@@ -41,6 +41,7 @@
41#include <asm/ppc_asm.h> 41#include <asm/ppc_asm.h>
42#include <asm/asm-offsets.h> 42#include <asm/asm-offsets.h>
43#include <asm/cache.h> 43#include <asm/cache.h>
44#include <asm/ptrace.h>
44#include "head_booke.h" 45#include "head_booke.h"
45 46
46/* As with the other PowerPC ports, it is expected that when code 47/* As with the other PowerPC ports, it is expected that when code
@@ -152,8 +153,11 @@ _ENTRY(__early_start)
152 /* Check to see if we're the second processor, and jump 153 /* Check to see if we're the second processor, and jump
153 * to the secondary_start code if so 154 * to the secondary_start code if so
154 */ 155 */
155 mfspr r24,SPRN_PIR 156 lis r24, boot_cpuid@h
156 cmpwi r24,0 157 ori r24, r24, boot_cpuid@l
158 lwz r24, 0(r24)
159 cmpwi r24, -1
160 mfspr r24,SPRN_PIR
157 bne __secondary_start 161 bne __secondary_start
158#endif 162#endif
159 163
@@ -175,6 +179,9 @@ _ENTRY(__early_start)
175 li r0,0 179 li r0,0
176 stwu r0,THREAD_SIZE-STACK_FRAME_OVERHEAD(r1) 180 stwu r0,THREAD_SIZE-STACK_FRAME_OVERHEAD(r1)
177 181
182 rlwinm r22,r1,0,0,31-THREAD_SHIFT /* current thread_info */
183 stw r24, TI_CPU(r22)
184
178 bl early_init 185 bl early_init
179 186
180#ifdef CONFIG_RELOCATABLE 187#ifdef CONFIG_RELOCATABLE
@@ -319,7 +326,7 @@ interrupt_base:
319 NORMAL_EXCEPTION_PROLOG 326 NORMAL_EXCEPTION_PROLOG
320 EXC_XFER_EE_LITE(0x0c00, DoSyscall) 327 EXC_XFER_EE_LITE(0x0c00, DoSyscall)
321 328
322 /* Auxillary Processor Unavailable Interrupt */ 329 /* Auxiliary Processor Unavailable Interrupt */
323 EXCEPTION(0x2900, AuxillaryProcessorUnavailable, unknown_exception, EXC_XFER_EE) 330 EXCEPTION(0x2900, AuxillaryProcessorUnavailable, unknown_exception, EXC_XFER_EE)
324 331
325 /* Decrementer Interrupt */ 332 /* Decrementer Interrupt */
diff --git a/arch/powerpc/kernel/ibmebus.c b/arch/powerpc/kernel/ibmebus.c
index 9b626cfffce1..28581f1ad2c0 100644
--- a/arch/powerpc/kernel/ibmebus.c
+++ b/arch/powerpc/kernel/ibmebus.c
@@ -162,13 +162,10 @@ static int ibmebus_create_device(struct device_node *dn)
162 dev->dev.bus = &ibmebus_bus_type; 162 dev->dev.bus = &ibmebus_bus_type;
163 dev->dev.archdata.dma_ops = &ibmebus_dma_ops; 163 dev->dev.archdata.dma_ops = &ibmebus_dma_ops;
164 164
165 ret = of_device_register(dev); 165 ret = of_device_add(dev);
166 if (ret) { 166 if (ret)
167 of_device_free(dev); 167 platform_device_put(dev);
168 return ret; 168 return ret;
169 }
170
171 return 0;
172} 169}
173 170
174static int ibmebus_create_devices(const struct of_device_id *matches) 171static int ibmebus_create_devices(const struct of_device_id *matches)
@@ -204,13 +201,14 @@ int ibmebus_register_driver(struct of_platform_driver *drv)
204 /* If the driver uses devices that ibmebus doesn't know, add them */ 201 /* If the driver uses devices that ibmebus doesn't know, add them */
205 ibmebus_create_devices(drv->driver.of_match_table); 202 ibmebus_create_devices(drv->driver.of_match_table);
206 203
207 return of_register_driver(drv, &ibmebus_bus_type); 204 drv->driver.bus = &ibmebus_bus_type;
205 return driver_register(&drv->driver);
208} 206}
209EXPORT_SYMBOL(ibmebus_register_driver); 207EXPORT_SYMBOL(ibmebus_register_driver);
210 208
211void ibmebus_unregister_driver(struct of_platform_driver *drv) 209void ibmebus_unregister_driver(struct of_platform_driver *drv)
212{ 210{
213 of_unregister_driver(drv); 211 driver_unregister(&drv->driver);
214} 212}
215EXPORT_SYMBOL(ibmebus_unregister_driver); 213EXPORT_SYMBOL(ibmebus_unregister_driver);
216 214
@@ -311,15 +309,410 @@ static ssize_t ibmebus_store_remove(struct bus_type *bus,
311 } 309 }
312} 310}
313 311
312
314static struct bus_attribute ibmebus_bus_attrs[] = { 313static struct bus_attribute ibmebus_bus_attrs[] = {
315 __ATTR(probe, S_IWUSR, NULL, ibmebus_store_probe), 314 __ATTR(probe, S_IWUSR, NULL, ibmebus_store_probe),
316 __ATTR(remove, S_IWUSR, NULL, ibmebus_store_remove), 315 __ATTR(remove, S_IWUSR, NULL, ibmebus_store_remove),
317 __ATTR_NULL 316 __ATTR_NULL
318}; 317};
319 318
319static int ibmebus_bus_bus_match(struct device *dev, struct device_driver *drv)
320{
321 const struct of_device_id *matches = drv->of_match_table;
322
323 if (!matches)
324 return 0;
325
326 return of_match_device(matches, dev) != NULL;
327}
328
329static int ibmebus_bus_device_probe(struct device *dev)
330{
331 int error = -ENODEV;
332 struct of_platform_driver *drv;
333 struct platform_device *of_dev;
334 const struct of_device_id *match;
335
336 drv = to_of_platform_driver(dev->driver);
337 of_dev = to_platform_device(dev);
338
339 if (!drv->probe)
340 return error;
341
342 of_dev_get(of_dev);
343
344 match = of_match_device(drv->driver.of_match_table, dev);
345 if (match)
346 error = drv->probe(of_dev, match);
347 if (error)
348 of_dev_put(of_dev);
349
350 return error;
351}
352
353static int ibmebus_bus_device_remove(struct device *dev)
354{
355 struct platform_device *of_dev = to_platform_device(dev);
356 struct of_platform_driver *drv = to_of_platform_driver(dev->driver);
357
358 if (dev->driver && drv->remove)
359 drv->remove(of_dev);
360 return 0;
361}
362
363static void ibmebus_bus_device_shutdown(struct device *dev)
364{
365 struct platform_device *of_dev = to_platform_device(dev);
366 struct of_platform_driver *drv = to_of_platform_driver(dev->driver);
367
368 if (dev->driver && drv->shutdown)
369 drv->shutdown(of_dev);
370}
371
372/*
373 * ibmebus_bus_device_attrs
374 */
375static ssize_t devspec_show(struct device *dev,
376 struct device_attribute *attr, char *buf)
377{
378 struct platform_device *ofdev;
379
380 ofdev = to_platform_device(dev);
381 return sprintf(buf, "%s\n", ofdev->dev.of_node->full_name);
382}
383
384static ssize_t name_show(struct device *dev,
385 struct device_attribute *attr, char *buf)
386{
387 struct platform_device *ofdev;
388
389 ofdev = to_platform_device(dev);
390 return sprintf(buf, "%s\n", ofdev->dev.of_node->name);
391}
392
393static ssize_t modalias_show(struct device *dev,
394 struct device_attribute *attr, char *buf)
395{
396 ssize_t len = of_device_get_modalias(dev, buf, PAGE_SIZE - 2);
397 buf[len] = '\n';
398 buf[len+1] = 0;
399 return len+1;
400}
401
402struct device_attribute ibmebus_bus_device_attrs[] = {
403 __ATTR_RO(devspec),
404 __ATTR_RO(name),
405 __ATTR_RO(modalias),
406 __ATTR_NULL
407};
408
409#ifdef CONFIG_PM_SLEEP
410static int ibmebus_bus_legacy_suspend(struct device *dev, pm_message_t mesg)
411{
412 struct platform_device *of_dev = to_platform_device(dev);
413 struct of_platform_driver *drv = to_of_platform_driver(dev->driver);
414 int ret = 0;
415
416 if (dev->driver && drv->suspend)
417 ret = drv->suspend(of_dev, mesg);
418 return ret;
419}
420
421static int ibmebus_bus_legacy_resume(struct device *dev)
422{
423 struct platform_device *of_dev = to_platform_device(dev);
424 struct of_platform_driver *drv = to_of_platform_driver(dev->driver);
425 int ret = 0;
426
427 if (dev->driver && drv->resume)
428 ret = drv->resume(of_dev);
429 return ret;
430}
431
432static int ibmebus_bus_pm_prepare(struct device *dev)
433{
434 struct device_driver *drv = dev->driver;
435 int ret = 0;
436
437 if (drv && drv->pm && drv->pm->prepare)
438 ret = drv->pm->prepare(dev);
439
440 return ret;
441}
442
443static void ibmebus_bus_pm_complete(struct device *dev)
444{
445 struct device_driver *drv = dev->driver;
446
447 if (drv && drv->pm && drv->pm->complete)
448 drv->pm->complete(dev);
449}
450
451#ifdef CONFIG_SUSPEND
452
453static int ibmebus_bus_pm_suspend(struct device *dev)
454{
455 struct device_driver *drv = dev->driver;
456 int ret = 0;
457
458 if (!drv)
459 return 0;
460
461 if (drv->pm) {
462 if (drv->pm->suspend)
463 ret = drv->pm->suspend(dev);
464 } else {
465 ret = ibmebus_bus_legacy_suspend(dev, PMSG_SUSPEND);
466 }
467
468 return ret;
469}
470
471static int ibmebus_bus_pm_suspend_noirq(struct device *dev)
472{
473 struct device_driver *drv = dev->driver;
474 int ret = 0;
475
476 if (!drv)
477 return 0;
478
479 if (drv->pm) {
480 if (drv->pm->suspend_noirq)
481 ret = drv->pm->suspend_noirq(dev);
482 }
483
484 return ret;
485}
486
487static int ibmebus_bus_pm_resume(struct device *dev)
488{
489 struct device_driver *drv = dev->driver;
490 int ret = 0;
491
492 if (!drv)
493 return 0;
494
495 if (drv->pm) {
496 if (drv->pm->resume)
497 ret = drv->pm->resume(dev);
498 } else {
499 ret = ibmebus_bus_legacy_resume(dev);
500 }
501
502 return ret;
503}
504
505static int ibmebus_bus_pm_resume_noirq(struct device *dev)
506{
507 struct device_driver *drv = dev->driver;
508 int ret = 0;
509
510 if (!drv)
511 return 0;
512
513 if (drv->pm) {
514 if (drv->pm->resume_noirq)
515 ret = drv->pm->resume_noirq(dev);
516 }
517
518 return ret;
519}
520
521#else /* !CONFIG_SUSPEND */
522
523#define ibmebus_bus_pm_suspend NULL
524#define ibmebus_bus_pm_resume NULL
525#define ibmebus_bus_pm_suspend_noirq NULL
526#define ibmebus_bus_pm_resume_noirq NULL
527
528#endif /* !CONFIG_SUSPEND */
529
530#ifdef CONFIG_HIBERNATE_CALLBACKS
531
532static int ibmebus_bus_pm_freeze(struct device *dev)
533{
534 struct device_driver *drv = dev->driver;
535 int ret = 0;
536
537 if (!drv)
538 return 0;
539
540 if (drv->pm) {
541 if (drv->pm->freeze)
542 ret = drv->pm->freeze(dev);
543 } else {
544 ret = ibmebus_bus_legacy_suspend(dev, PMSG_FREEZE);
545 }
546
547 return ret;
548}
549
550static int ibmebus_bus_pm_freeze_noirq(struct device *dev)
551{
552 struct device_driver *drv = dev->driver;
553 int ret = 0;
554
555 if (!drv)
556 return 0;
557
558 if (drv->pm) {
559 if (drv->pm->freeze_noirq)
560 ret = drv->pm->freeze_noirq(dev);
561 }
562
563 return ret;
564}
565
566static int ibmebus_bus_pm_thaw(struct device *dev)
567{
568 struct device_driver *drv = dev->driver;
569 int ret = 0;
570
571 if (!drv)
572 return 0;
573
574 if (drv->pm) {
575 if (drv->pm->thaw)
576 ret = drv->pm->thaw(dev);
577 } else {
578 ret = ibmebus_bus_legacy_resume(dev);
579 }
580
581 return ret;
582}
583
584static int ibmebus_bus_pm_thaw_noirq(struct device *dev)
585{
586 struct device_driver *drv = dev->driver;
587 int ret = 0;
588
589 if (!drv)
590 return 0;
591
592 if (drv->pm) {
593 if (drv->pm->thaw_noirq)
594 ret = drv->pm->thaw_noirq(dev);
595 }
596
597 return ret;
598}
599
600static int ibmebus_bus_pm_poweroff(struct device *dev)
601{
602 struct device_driver *drv = dev->driver;
603 int ret = 0;
604
605 if (!drv)
606 return 0;
607
608 if (drv->pm) {
609 if (drv->pm->poweroff)
610 ret = drv->pm->poweroff(dev);
611 } else {
612 ret = ibmebus_bus_legacy_suspend(dev, PMSG_HIBERNATE);
613 }
614
615 return ret;
616}
617
618static int ibmebus_bus_pm_poweroff_noirq(struct device *dev)
619{
620 struct device_driver *drv = dev->driver;
621 int ret = 0;
622
623 if (!drv)
624 return 0;
625
626 if (drv->pm) {
627 if (drv->pm->poweroff_noirq)
628 ret = drv->pm->poweroff_noirq(dev);
629 }
630
631 return ret;
632}
633
634static int ibmebus_bus_pm_restore(struct device *dev)
635{
636 struct device_driver *drv = dev->driver;
637 int ret = 0;
638
639 if (!drv)
640 return 0;
641
642 if (drv->pm) {
643 if (drv->pm->restore)
644 ret = drv->pm->restore(dev);
645 } else {
646 ret = ibmebus_bus_legacy_resume(dev);
647 }
648
649 return ret;
650}
651
652static int ibmebus_bus_pm_restore_noirq(struct device *dev)
653{
654 struct device_driver *drv = dev->driver;
655 int ret = 0;
656
657 if (!drv)
658 return 0;
659
660 if (drv->pm) {
661 if (drv->pm->restore_noirq)
662 ret = drv->pm->restore_noirq(dev);
663 }
664
665 return ret;
666}
667
668#else /* !CONFIG_HIBERNATE_CALLBACKS */
669
670#define ibmebus_bus_pm_freeze NULL
671#define ibmebus_bus_pm_thaw NULL
672#define ibmebus_bus_pm_poweroff NULL
673#define ibmebus_bus_pm_restore NULL
674#define ibmebus_bus_pm_freeze_noirq NULL
675#define ibmebus_bus_pm_thaw_noirq NULL
676#define ibmebus_bus_pm_poweroff_noirq NULL
677#define ibmebus_bus_pm_restore_noirq NULL
678
679#endif /* !CONFIG_HIBERNATE_CALLBACKS */
680
681static struct dev_pm_ops ibmebus_bus_dev_pm_ops = {
682 .prepare = ibmebus_bus_pm_prepare,
683 .complete = ibmebus_bus_pm_complete,
684 .suspend = ibmebus_bus_pm_suspend,
685 .resume = ibmebus_bus_pm_resume,
686 .freeze = ibmebus_bus_pm_freeze,
687 .thaw = ibmebus_bus_pm_thaw,
688 .poweroff = ibmebus_bus_pm_poweroff,
689 .restore = ibmebus_bus_pm_restore,
690 .suspend_noirq = ibmebus_bus_pm_suspend_noirq,
691 .resume_noirq = ibmebus_bus_pm_resume_noirq,
692 .freeze_noirq = ibmebus_bus_pm_freeze_noirq,
693 .thaw_noirq = ibmebus_bus_pm_thaw_noirq,
694 .poweroff_noirq = ibmebus_bus_pm_poweroff_noirq,
695 .restore_noirq = ibmebus_bus_pm_restore_noirq,
696};
697
698#define IBMEBUS_BUS_PM_OPS_PTR (&ibmebus_bus_dev_pm_ops)
699
700#else /* !CONFIG_PM_SLEEP */
701
702#define IBMEBUS_BUS_PM_OPS_PTR NULL
703
704#endif /* !CONFIG_PM_SLEEP */
705
320struct bus_type ibmebus_bus_type = { 706struct bus_type ibmebus_bus_type = {
707 .name = "ibmebus",
321 .uevent = of_device_uevent, 708 .uevent = of_device_uevent,
322 .bus_attrs = ibmebus_bus_attrs 709 .bus_attrs = ibmebus_bus_attrs,
710 .match = ibmebus_bus_bus_match,
711 .probe = ibmebus_bus_device_probe,
712 .remove = ibmebus_bus_device_remove,
713 .shutdown = ibmebus_bus_device_shutdown,
714 .dev_attrs = ibmebus_bus_device_attrs,
715 .pm = IBMEBUS_BUS_PM_OPS_PTR,
323}; 716};
324EXPORT_SYMBOL(ibmebus_bus_type); 717EXPORT_SYMBOL(ibmebus_bus_type);
325 718
@@ -329,7 +722,7 @@ static int __init ibmebus_bus_init(void)
329 722
330 printk(KERN_INFO "IBM eBus Device Driver\n"); 723 printk(KERN_INFO "IBM eBus Device Driver\n");
331 724
332 err = of_bus_type_init(&ibmebus_bus_type, "ibmebus"); 725 err = bus_register(&ibmebus_bus_type);
333 if (err) { 726 if (err) {
334 printk(KERN_ERR "%s: failed to register IBM eBus.\n", 727 printk(KERN_ERR "%s: failed to register IBM eBus.\n",
335 __func__); 728 __func__);
diff --git a/arch/powerpc/kernel/idle_power4.S b/arch/powerpc/kernel/idle_power4.S
index 5328709eeedc..ba3195478600 100644
--- a/arch/powerpc/kernel/idle_power4.S
+++ b/arch/powerpc/kernel/idle_power4.S
@@ -53,24 +53,3 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
53 isync 53 isync
54 b 1b 54 b 1b
55 55
56_GLOBAL(power4_cpu_offline_powersave)
57 /* Go to NAP now */
58 mfmsr r7
59 rldicl r0,r7,48,1
60 rotldi r0,r0,16
61 mtmsrd r0,1 /* hard-disable interrupts */
62 li r0,1
63 li r6,0
64 stb r0,PACAHARDIRQEN(r13) /* we'll hard-enable shortly */
65 stb r6,PACASOFTIRQEN(r13) /* soft-disable irqs */
66BEGIN_FTR_SECTION
67 DSSALL
68 sync
69END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
70 ori r7,r7,MSR_EE
71 oris r7,r7,MSR_POW@h
72 sync
73 isync
74 mtmsrd r7
75 isync
76 blr
diff --git a/arch/powerpc/kernel/idle_power7.S b/arch/powerpc/kernel/idle_power7.S
new file mode 100644
index 000000000000..f8f0bc7f1d4f
--- /dev/null
+++ b/arch/powerpc/kernel/idle_power7.S
@@ -0,0 +1,97 @@
1/*
2 * This file contains the power_save function for 970-family CPUs.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 */
9
10#include <linux/threads.h>
11#include <asm/processor.h>
12#include <asm/page.h>
13#include <asm/cputable.h>
14#include <asm/thread_info.h>
15#include <asm/ppc_asm.h>
16#include <asm/asm-offsets.h>
17#include <asm/ppc-opcode.h>
18
19#undef DEBUG
20
21 .text
22
23_GLOBAL(power7_idle)
24 /* Now check if user or arch enabled NAP mode */
25 LOAD_REG_ADDRBASE(r3,powersave_nap)
26 lwz r4,ADDROFF(powersave_nap)(r3)
27 cmpwi 0,r4,0
28 beqlr
29
30 /* NAP is a state loss, we create a regs frame on the
31 * stack, fill it up with the state we care about and
32 * stick a pointer to it in PACAR1. We really only
33 * need to save PC, some CR bits and the NV GPRs,
34 * but for now an interrupt frame will do.
35 */
36 mflr r0
37 std r0,16(r1)
38 stdu r1,-INT_FRAME_SIZE(r1)
39 std r0,_LINK(r1)
40 std r0,_NIP(r1)
41
42#ifndef CONFIG_SMP
43 /* Make sure FPU, VSX etc... are flushed as we may lose
44 * state when going to nap mode
45 */
46 bl .discard_lazy_cpu_state
47#endif /* CONFIG_SMP */
48
49 /* Hard disable interrupts */
50 mfmsr r9
51 rldicl r9,r9,48,1
52 rotldi r9,r9,16
53 mtmsrd r9,1 /* hard-disable interrupts */
54 li r0,0
55 stb r0,PACASOFTIRQEN(r13) /* we'll hard-enable shortly */
56 stb r0,PACAHARDIRQEN(r13)
57
58 /* Continue saving state */
59 SAVE_GPR(2, r1)
60 SAVE_NVGPRS(r1)
61 mfcr r3
62 std r3,_CCR(r1)
63 std r9,_MSR(r1)
64 std r1,PACAR1(r13)
65
66 /* Magic NAP mode enter sequence */
67 std r0,0(r1)
68 ptesync
69 ld r0,0(r1)
701: cmp cr0,r0,r0
71 bne 1b
72 PPC_NAP
73 b .
74
75_GLOBAL(power7_wakeup_loss)
76 GET_PACA(r13)
77 ld r1,PACAR1(r13)
78 REST_NVGPRS(r1)
79 REST_GPR(2, r1)
80 ld r3,_CCR(r1)
81 ld r4,_MSR(r1)
82 ld r5,_NIP(r1)
83 addi r1,r1,INT_FRAME_SIZE
84 mtcr r3
85 mtspr SPRN_SRR1,r4
86 mtspr SPRN_SRR0,r5
87 rfid
88
89_GLOBAL(power7_wakeup_noloss)
90 GET_PACA(r13)
91 ld r1,PACAR1(r13)
92 ld r4,_MSR(r1)
93 ld r5,_NIP(r1)
94 addi r1,r1,INT_FRAME_SIZE
95 mtspr SPRN_SRR1,r4
96 mtspr SPRN_SRR0,r5
97 rfid
diff --git a/arch/powerpc/kernel/io-workarounds.c b/arch/powerpc/kernel/io-workarounds.c
new file mode 100644
index 000000000000..ffafaea3d261
--- /dev/null
+++ b/arch/powerpc/kernel/io-workarounds.c
@@ -0,0 +1,188 @@
1/*
2 * Support PCI IO workaround
3 *
4 * Copyright (C) 2006 Benjamin Herrenschmidt <benh@kernel.crashing.org>
5 * IBM, Corp.
6 * (C) Copyright 2007-2008 TOSHIBA CORPORATION
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
11 */
12#undef DEBUG
13
14#include <linux/kernel.h>
15
16#include <asm/io.h>
17#include <asm/machdep.h>
18#include <asm/pgtable.h>
19#include <asm/ppc-pci.h>
20#include <asm/io-workarounds.h>
21
22#define IOWA_MAX_BUS 8
23
24static struct iowa_bus iowa_busses[IOWA_MAX_BUS];
25static unsigned int iowa_bus_count;
26
27static struct iowa_bus *iowa_pci_find(unsigned long vaddr, unsigned long paddr)
28{
29 int i, j;
30 struct resource *res;
31 unsigned long vstart, vend;
32
33 for (i = 0; i < iowa_bus_count; i++) {
34 struct iowa_bus *bus = &iowa_busses[i];
35 struct pci_controller *phb = bus->phb;
36
37 if (vaddr) {
38 vstart = (unsigned long)phb->io_base_virt;
39 vend = vstart + phb->pci_io_size - 1;
40 if ((vaddr >= vstart) && (vaddr <= vend))
41 return bus;
42 }
43
44 if (paddr)
45 for (j = 0; j < 3; j++) {
46 res = &phb->mem_resources[j];
47 if (paddr >= res->start && paddr <= res->end)
48 return bus;
49 }
50 }
51
52 return NULL;
53}
54
55struct iowa_bus *iowa_mem_find_bus(const PCI_IO_ADDR addr)
56{
57 struct iowa_bus *bus;
58 int token;
59
60 token = PCI_GET_ADDR_TOKEN(addr);
61
62 if (token && token <= iowa_bus_count)
63 bus = &iowa_busses[token - 1];
64 else {
65 unsigned long vaddr, paddr;
66 pte_t *ptep;
67
68 vaddr = (unsigned long)PCI_FIX_ADDR(addr);
69 if (vaddr < PHB_IO_BASE || vaddr >= PHB_IO_END)
70 return NULL;
71
72 ptep = find_linux_pte(init_mm.pgd, vaddr);
73 if (ptep == NULL)
74 paddr = 0;
75 else
76 paddr = pte_pfn(*ptep) << PAGE_SHIFT;
77 bus = iowa_pci_find(vaddr, paddr);
78
79 if (bus == NULL)
80 return NULL;
81 }
82
83 return bus;
84}
85
86struct iowa_bus *iowa_pio_find_bus(unsigned long port)
87{
88 unsigned long vaddr = (unsigned long)pci_io_base + port;
89 return iowa_pci_find(vaddr, 0);
90}
91
92
93#define DEF_PCI_AC_RET(name, ret, at, al, space, aa) \
94static ret iowa_##name at \
95{ \
96 struct iowa_bus *bus; \
97 bus = iowa_##space##_find_bus(aa); \
98 if (bus && bus->ops && bus->ops->name) \
99 return bus->ops->name al; \
100 return __do_##name al; \
101}
102
103#define DEF_PCI_AC_NORET(name, at, al, space, aa) \
104static void iowa_##name at \
105{ \
106 struct iowa_bus *bus; \
107 bus = iowa_##space##_find_bus(aa); \
108 if (bus && bus->ops && bus->ops->name) { \
109 bus->ops->name al; \
110 return; \
111 } \
112 __do_##name al; \
113}
114
115#include <asm/io-defs.h>
116
117#undef DEF_PCI_AC_RET
118#undef DEF_PCI_AC_NORET
119
120static const struct ppc_pci_io __devinitconst iowa_pci_io = {
121
122#define DEF_PCI_AC_RET(name, ret, at, al, space, aa) .name = iowa_##name,
123#define DEF_PCI_AC_NORET(name, at, al, space, aa) .name = iowa_##name,
124
125#include <asm/io-defs.h>
126
127#undef DEF_PCI_AC_RET
128#undef DEF_PCI_AC_NORET
129
130};
131
132static void __iomem *iowa_ioremap(phys_addr_t addr, unsigned long size,
133 unsigned long flags, void *caller)
134{
135 struct iowa_bus *bus;
136 void __iomem *res = __ioremap_caller(addr, size, flags, caller);
137 int busno;
138
139 bus = iowa_pci_find(0, (unsigned long)addr);
140 if (bus != NULL) {
141 busno = bus - iowa_busses;
142 PCI_SET_ADDR_TOKEN(res, busno + 1);
143 }
144 return res;
145}
146
147/* Enable IO workaround */
148static void __devinit io_workaround_init(void)
149{
150 static int io_workaround_inited;
151
152 if (io_workaround_inited)
153 return;
154 ppc_pci_io = iowa_pci_io;
155 ppc_md.ioremap = iowa_ioremap;
156 io_workaround_inited = 1;
157}
158
159/* Register new bus to support workaround */
160void __devinit iowa_register_bus(struct pci_controller *phb,
161 struct ppc_pci_io *ops,
162 int (*initfunc)(struct iowa_bus *, void *), void *data)
163{
164 struct iowa_bus *bus;
165 struct device_node *np = phb->dn;
166
167 io_workaround_init();
168
169 if (iowa_bus_count >= IOWA_MAX_BUS) {
170 pr_err("IOWA:Too many pci bridges, "
171 "workarounds disabled for %s\n", np->full_name);
172 return;
173 }
174
175 bus = &iowa_busses[iowa_bus_count];
176 bus->phb = phb;
177 bus->ops = ops;
178 bus->private = data;
179
180 if (initfunc)
181 if ((*initfunc)(bus, data))
182 return;
183
184 iowa_bus_count++;
185
186 pr_debug("IOWA:[%d]Add bus, %s.\n", iowa_bus_count-1, np->full_name);
187}
188
diff --git a/arch/powerpc/kernel/iommu.c b/arch/powerpc/kernel/iommu.c
index d5839179ec77..961bb03413f3 100644
--- a/arch/powerpc/kernel/iommu.c
+++ b/arch/powerpc/kernel/iommu.c
@@ -311,8 +311,9 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
311 /* Handle failure */ 311 /* Handle failure */
312 if (unlikely(entry == DMA_ERROR_CODE)) { 312 if (unlikely(entry == DMA_ERROR_CODE)) {
313 if (printk_ratelimit()) 313 if (printk_ratelimit())
314 printk(KERN_INFO "iommu_alloc failed, tbl %p vaddr %lx" 314 dev_info(dev, "iommu_alloc failed, tbl %p "
315 " npages %lx\n", tbl, vaddr, npages); 315 "vaddr %lx npages %lu\n", tbl, vaddr,
316 npages);
316 goto failure; 317 goto failure;
317 } 318 }
318 319
@@ -579,9 +580,9 @@ dma_addr_t iommu_map_page(struct device *dev, struct iommu_table *tbl,
579 attrs); 580 attrs);
580 if (dma_handle == DMA_ERROR_CODE) { 581 if (dma_handle == DMA_ERROR_CODE) {
581 if (printk_ratelimit()) { 582 if (printk_ratelimit()) {
582 printk(KERN_INFO "iommu_alloc failed, " 583 dev_info(dev, "iommu_alloc failed, tbl %p "
583 "tbl %p vaddr %p npages %d\n", 584 "vaddr %p npages %d\n", tbl, vaddr,
584 tbl, vaddr, npages); 585 npages);
585 } 586 }
586 } else 587 } else
587 dma_handle |= (uaddr & ~IOMMU_PAGE_MASK); 588 dma_handle |= (uaddr & ~IOMMU_PAGE_MASK);
@@ -627,7 +628,8 @@ void *iommu_alloc_coherent(struct device *dev, struct iommu_table *tbl,
627 * the tce tables. 628 * the tce tables.
628 */ 629 */
629 if (order >= IOMAP_MAX_ORDER) { 630 if (order >= IOMAP_MAX_ORDER) {
630 printk("iommu_alloc_consistent size too large: 0x%lx\n", size); 631 dev_info(dev, "iommu_alloc_consistent size too large: 0x%lx\n",
632 size);
631 return NULL; 633 return NULL;
632 } 634 }
633 635
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index 4a65386995d7..5b428e308666 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -66,7 +66,6 @@
66#include <asm/ptrace.h> 66#include <asm/ptrace.h>
67#include <asm/machdep.h> 67#include <asm/machdep.h>
68#include <asm/udbg.h> 68#include <asm/udbg.h>
69#include <asm/dbell.h>
70#include <asm/smp.h> 69#include <asm/smp.h>
71 70
72#ifdef CONFIG_PPC64 71#ifdef CONFIG_PPC64
@@ -116,7 +115,7 @@ static inline notrace void set_soft_enabled(unsigned long enable)
116 : : "r" (enable), "i" (offsetof(struct paca_struct, soft_enabled))); 115 : : "r" (enable), "i" (offsetof(struct paca_struct, soft_enabled)));
117} 116}
118 117
119notrace void raw_local_irq_restore(unsigned long en) 118notrace void arch_local_irq_restore(unsigned long en)
120{ 119{
121 /* 120 /*
122 * get_paca()->soft_enabled = en; 121 * get_paca()->soft_enabled = en;
@@ -160,7 +159,8 @@ notrace void raw_local_irq_restore(unsigned long en)
160 159
161#if defined(CONFIG_BOOKE) && defined(CONFIG_SMP) 160#if defined(CONFIG_BOOKE) && defined(CONFIG_SMP)
162 /* Check for pending doorbell interrupts and resend to ourself */ 161 /* Check for pending doorbell interrupts and resend to ourself */
163 doorbell_check_self(); 162 if (cpu_has_feature(CPU_FTR_DBELL))
163 smp_muxed_ipi_resend();
164#endif 164#endif
165 165
166 /* 166 /*
@@ -192,10 +192,10 @@ notrace void raw_local_irq_restore(unsigned long en)
192 192
193 __hard_irq_enable(); 193 __hard_irq_enable();
194} 194}
195EXPORT_SYMBOL(raw_local_irq_restore); 195EXPORT_SYMBOL(arch_local_irq_restore);
196#endif /* CONFIG_PPC64 */ 196#endif /* CONFIG_PPC64 */
197 197
198static int show_other_interrupts(struct seq_file *p, int prec) 198int arch_show_interrupts(struct seq_file *p, int prec)
199{ 199{
200 int j; 200 int j;
201 201
@@ -231,63 +231,6 @@ static int show_other_interrupts(struct seq_file *p, int prec)
231 return 0; 231 return 0;
232} 232}
233 233
234int show_interrupts(struct seq_file *p, void *v)
235{
236 unsigned long flags, any_count = 0;
237 int i = *(loff_t *) v, j, prec;
238 struct irqaction *action;
239 struct irq_desc *desc;
240
241 if (i > nr_irqs)
242 return 0;
243
244 for (prec = 3, j = 1000; prec < 10 && j <= nr_irqs; ++prec)
245 j *= 10;
246
247 if (i == nr_irqs)
248 return show_other_interrupts(p, prec);
249
250 /* print header */
251 if (i == 0) {
252 seq_printf(p, "%*s", prec + 8, "");
253 for_each_online_cpu(j)
254 seq_printf(p, "CPU%-8d", j);
255 seq_putc(p, '\n');
256 }
257
258 desc = irq_to_desc(i);
259 if (!desc)
260 return 0;
261
262 raw_spin_lock_irqsave(&desc->lock, flags);
263 for_each_online_cpu(j)
264 any_count |= kstat_irqs_cpu(i, j);
265 action = desc->action;
266 if (!action && !any_count)
267 goto out;
268
269 seq_printf(p, "%*d: ", prec, i);
270 for_each_online_cpu(j)
271 seq_printf(p, "%10u ", kstat_irqs_cpu(i, j));
272
273 if (desc->chip)
274 seq_printf(p, " %-16s", desc->chip->name);
275 else
276 seq_printf(p, " %-16s", "None");
277 seq_printf(p, " %-8s", (desc->status & IRQ_LEVEL) ? "Level" : "Edge");
278
279 if (action) {
280 seq_printf(p, " %s", action->name);
281 while ((action = action->next) != NULL)
282 seq_printf(p, ", %s", action->name);
283 }
284
285 seq_putc(p, '\n');
286out:
287 raw_spin_unlock_irqrestore(&desc->lock, flags);
288 return 0;
289}
290
291/* 234/*
292 * /proc/stat helpers 235 * /proc/stat helpers
293 */ 236 */
@@ -303,30 +246,37 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
303} 246}
304 247
305#ifdef CONFIG_HOTPLUG_CPU 248#ifdef CONFIG_HOTPLUG_CPU
306void fixup_irqs(const struct cpumask *map) 249void migrate_irqs(void)
307{ 250{
308 struct irq_desc *desc; 251 struct irq_desc *desc;
309 unsigned int irq; 252 unsigned int irq;
310 static int warned; 253 static int warned;
311 cpumask_var_t mask; 254 cpumask_var_t mask;
255 const struct cpumask *map = cpu_online_mask;
312 256
313 alloc_cpumask_var(&mask, GFP_KERNEL); 257 alloc_cpumask_var(&mask, GFP_KERNEL);
314 258
315 for_each_irq(irq) { 259 for_each_irq(irq) {
260 struct irq_data *data;
261 struct irq_chip *chip;
262
316 desc = irq_to_desc(irq); 263 desc = irq_to_desc(irq);
317 if (!desc) 264 if (!desc)
318 continue; 265 continue;
319 266
320 if (desc->status & IRQ_PER_CPU) 267 data = irq_desc_get_irq_data(desc);
268 if (irqd_is_per_cpu(data))
321 continue; 269 continue;
322 270
323 cpumask_and(mask, desc->affinity, map); 271 chip = irq_data_get_irq_chip(data);
272
273 cpumask_and(mask, data->affinity, map);
324 if (cpumask_any(mask) >= nr_cpu_ids) { 274 if (cpumask_any(mask) >= nr_cpu_ids) {
325 printk("Breaking affinity for irq %i\n", irq); 275 printk("Breaking affinity for irq %i\n", irq);
326 cpumask_copy(mask, map); 276 cpumask_copy(mask, map);
327 } 277 }
328 if (desc->chip->set_affinity) 278 if (chip->irq_set_affinity)
329 desc->chip->set_affinity(irq, mask); 279 chip->irq_set_affinity(data, mask, true);
330 else if (desc->action && !(warned++)) 280 else if (desc->action && !(warned++))
331 printk("Cannot set affinity for irq %i\n", irq); 281 printk("Cannot set affinity for irq %i\n", irq);
332 } 282 }
@@ -345,17 +295,20 @@ static inline void handle_one_irq(unsigned int irq)
345 unsigned long saved_sp_limit; 295 unsigned long saved_sp_limit;
346 struct irq_desc *desc; 296 struct irq_desc *desc;
347 297
298 desc = irq_to_desc(irq);
299 if (!desc)
300 return;
301
348 /* Switch to the irq stack to handle this */ 302 /* Switch to the irq stack to handle this */
349 curtp = current_thread_info(); 303 curtp = current_thread_info();
350 irqtp = hardirq_ctx[smp_processor_id()]; 304 irqtp = hardirq_ctx[smp_processor_id()];
351 305
352 if (curtp == irqtp) { 306 if (curtp == irqtp) {
353 /* We're already on the irq stack, just handle it */ 307 /* We're already on the irq stack, just handle it */
354 generic_handle_irq(irq); 308 desc->handle_irq(irq, desc);
355 return; 309 return;
356 } 310 }
357 311
358 desc = irq_to_desc(irq);
359 saved_sp_limit = current->thread.ksp_limit; 312 saved_sp_limit = current->thread.ksp_limit;
360 313
361 irqtp->task = curtp->task; 314 irqtp->task = curtp->task;
@@ -447,24 +400,28 @@ struct thread_info *mcheckirq_ctx[NR_CPUS] __read_mostly;
447void exc_lvl_ctx_init(void) 400void exc_lvl_ctx_init(void)
448{ 401{
449 struct thread_info *tp; 402 struct thread_info *tp;
450 int i, hw_cpu; 403 int i, cpu_nr;
451 404
452 for_each_possible_cpu(i) { 405 for_each_possible_cpu(i) {
453 hw_cpu = get_hard_smp_processor_id(i); 406#ifdef CONFIG_PPC64
454 memset((void *)critirq_ctx[hw_cpu], 0, THREAD_SIZE); 407 cpu_nr = i;
455 tp = critirq_ctx[hw_cpu]; 408#else
456 tp->cpu = i; 409 cpu_nr = get_hard_smp_processor_id(i);
410#endif
411 memset((void *)critirq_ctx[cpu_nr], 0, THREAD_SIZE);
412 tp = critirq_ctx[cpu_nr];
413 tp->cpu = cpu_nr;
457 tp->preempt_count = 0; 414 tp->preempt_count = 0;
458 415
459#ifdef CONFIG_BOOKE 416#ifdef CONFIG_BOOKE
460 memset((void *)dbgirq_ctx[hw_cpu], 0, THREAD_SIZE); 417 memset((void *)dbgirq_ctx[cpu_nr], 0, THREAD_SIZE);
461 tp = dbgirq_ctx[hw_cpu]; 418 tp = dbgirq_ctx[cpu_nr];
462 tp->cpu = i; 419 tp->cpu = cpu_nr;
463 tp->preempt_count = 0; 420 tp->preempt_count = 0;
464 421
465 memset((void *)mcheckirq_ctx[hw_cpu], 0, THREAD_SIZE); 422 memset((void *)mcheckirq_ctx[cpu_nr], 0, THREAD_SIZE);
466 tp = mcheckirq_ctx[hw_cpu]; 423 tp = mcheckirq_ctx[cpu_nr];
467 tp->cpu = i; 424 tp->cpu = cpu_nr;
468 tp->preempt_count = HARDIRQ_OFFSET; 425 tp->preempt_count = HARDIRQ_OFFSET;
469#endif 426#endif
470 } 427 }
@@ -527,20 +484,41 @@ void do_softirq(void)
527 * IRQ controller and virtual interrupts 484 * IRQ controller and virtual interrupts
528 */ 485 */
529 486
487/* The main irq map itself is an array of NR_IRQ entries containing the
488 * associate host and irq number. An entry with a host of NULL is free.
489 * An entry can be allocated if it's free, the allocator always then sets
490 * hwirq first to the host's invalid irq number and then fills ops.
491 */
492struct irq_map_entry {
493 irq_hw_number_t hwirq;
494 struct irq_host *host;
495};
496
530static LIST_HEAD(irq_hosts); 497static LIST_HEAD(irq_hosts);
531static DEFINE_RAW_SPINLOCK(irq_big_lock); 498static DEFINE_RAW_SPINLOCK(irq_big_lock);
532static unsigned int revmap_trees_allocated;
533static DEFINE_MUTEX(revmap_trees_mutex); 499static DEFINE_MUTEX(revmap_trees_mutex);
534struct irq_map_entry irq_map[NR_IRQS]; 500static struct irq_map_entry irq_map[NR_IRQS];
535static unsigned int irq_virq_count = NR_IRQS; 501static unsigned int irq_virq_count = NR_IRQS;
536static struct irq_host *irq_default_host; 502static struct irq_host *irq_default_host;
537 503
504irq_hw_number_t irqd_to_hwirq(struct irq_data *d)
505{
506 return irq_map[d->irq].hwirq;
507}
508EXPORT_SYMBOL_GPL(irqd_to_hwirq);
509
538irq_hw_number_t virq_to_hw(unsigned int virq) 510irq_hw_number_t virq_to_hw(unsigned int virq)
539{ 511{
540 return irq_map[virq].hwirq; 512 return irq_map[virq].hwirq;
541} 513}
542EXPORT_SYMBOL_GPL(virq_to_hw); 514EXPORT_SYMBOL_GPL(virq_to_hw);
543 515
516bool virq_is_host(unsigned int virq, struct irq_host *host)
517{
518 return irq_map[virq].host == host;
519}
520EXPORT_SYMBOL_GPL(virq_is_host);
521
544static int default_irq_host_match(struct irq_host *h, struct device_node *np) 522static int default_irq_host_match(struct irq_host *h, struct device_node *np)
545{ 523{
546 return h->of_node != NULL && h->of_node == np; 524 return h->of_node != NULL && h->of_node == np;
@@ -561,7 +539,7 @@ struct irq_host *irq_alloc_host(struct device_node *of_node,
561 /* Allocate structure and revmap table if using linear mapping */ 539 /* Allocate structure and revmap table if using linear mapping */
562 if (revmap_type == IRQ_HOST_MAP_LINEAR) 540 if (revmap_type == IRQ_HOST_MAP_LINEAR)
563 size += revmap_arg * sizeof(unsigned int); 541 size += revmap_arg * sizeof(unsigned int);
564 host = zalloc_maybe_bootmem(size, GFP_KERNEL); 542 host = kzalloc(size, GFP_KERNEL);
565 if (host == NULL) 543 if (host == NULL)
566 return NULL; 544 return NULL;
567 545
@@ -582,13 +560,8 @@ struct irq_host *irq_alloc_host(struct device_node *of_node,
582 if (revmap_type == IRQ_HOST_MAP_LEGACY) { 560 if (revmap_type == IRQ_HOST_MAP_LEGACY) {
583 if (irq_map[0].host != NULL) { 561 if (irq_map[0].host != NULL) {
584 raw_spin_unlock_irqrestore(&irq_big_lock, flags); 562 raw_spin_unlock_irqrestore(&irq_big_lock, flags);
585 /* If we are early boot, we can't free the structure, 563 of_node_put(host->of_node);
586 * too bad... 564 kfree(host);
587 * this will be fixed once slab is made available early
588 * instead of the current cruft
589 */
590 if (mem_init_done)
591 kfree(host);
592 return NULL; 565 return NULL;
593 } 566 }
594 irq_map[0].host = host; 567 irq_map[0].host = host;
@@ -609,14 +582,14 @@ struct irq_host *irq_alloc_host(struct device_node *of_node,
609 irq_map[i].host = host; 582 irq_map[i].host = host;
610 smp_wmb(); 583 smp_wmb();
611 584
612 /* Clear norequest flags */
613 irq_to_desc(i)->status &= ~IRQ_NOREQUEST;
614
615 /* Legacy flags are left to default at this point, 585 /* Legacy flags are left to default at this point,
616 * one can then use irq_create_mapping() to 586 * one can then use irq_create_mapping() to
617 * explicitly change them 587 * explicitly change them
618 */ 588 */
619 ops->map(host, i, i); 589 ops->map(host, i, i);
590
591 /* Clear norequest flags */
592 irq_clear_status_flags(i, IRQ_NOREQUEST);
620 } 593 }
621 break; 594 break;
622 case IRQ_HOST_MAP_LINEAR: 595 case IRQ_HOST_MAP_LINEAR:
@@ -627,6 +600,9 @@ struct irq_host *irq_alloc_host(struct device_node *of_node,
627 smp_wmb(); 600 smp_wmb();
628 host->revmap_data.linear.revmap = rmap; 601 host->revmap_data.linear.revmap = rmap;
629 break; 602 break;
603 case IRQ_HOST_MAP_TREE:
604 INIT_RADIX_TREE(&host->revmap_data.tree, GFP_KERNEL);
605 break;
630 default: 606 default:
631 break; 607 break;
632 } 608 }
@@ -676,17 +652,14 @@ void irq_set_virq_count(unsigned int count)
676static int irq_setup_virq(struct irq_host *host, unsigned int virq, 652static int irq_setup_virq(struct irq_host *host, unsigned int virq,
677 irq_hw_number_t hwirq) 653 irq_hw_number_t hwirq)
678{ 654{
679 struct irq_desc *desc; 655 int res;
680 656
681 desc = irq_to_desc_alloc_node(virq, 0); 657 res = irq_alloc_desc_at(virq, 0);
682 if (!desc) { 658 if (res != virq) {
683 pr_debug("irq: -> allocating desc failed\n"); 659 pr_debug("irq: -> allocating desc failed\n");
684 goto error; 660 goto error;
685 } 661 }
686 662
687 /* Clear IRQ_NOREQUEST flag */
688 desc->status &= ~IRQ_NOREQUEST;
689
690 /* map it */ 663 /* map it */
691 smp_wmb(); 664 smp_wmb();
692 irq_map[virq].hwirq = hwirq; 665 irq_map[virq].hwirq = hwirq;
@@ -694,11 +667,15 @@ static int irq_setup_virq(struct irq_host *host, unsigned int virq,
694 667
695 if (host->ops->map(host, virq, hwirq)) { 668 if (host->ops->map(host, virq, hwirq)) {
696 pr_debug("irq: -> mapping failed, freeing\n"); 669 pr_debug("irq: -> mapping failed, freeing\n");
697 goto error; 670 goto errdesc;
698 } 671 }
699 672
673 irq_clear_status_flags(virq, IRQ_NOREQUEST);
674
700 return 0; 675 return 0;
701 676
677errdesc:
678 irq_free_descs(virq, 1);
702error: 679error:
703 irq_free_virt(virq, 1); 680 irq_free_virt(virq, 1);
704 return -1; 681 return -1;
@@ -746,13 +723,9 @@ unsigned int irq_create_mapping(struct irq_host *host,
746 } 723 }
747 pr_debug("irq: -> using host @%p\n", host); 724 pr_debug("irq: -> using host @%p\n", host);
748 725
749 /* Check if mapping already exist, if it does, call 726 /* Check if mapping already exists */
750 * host->ops->map() to update the flags
751 */
752 virq = irq_find_mapping(host, hwirq); 727 virq = irq_find_mapping(host, hwirq);
753 if (virq != NO_IRQ) { 728 if (virq != NO_IRQ) {
754 if (host->ops->remap)
755 host->ops->remap(host, virq, hwirq);
756 pr_debug("irq: -> existing mapping on virq %d\n", virq); 729 pr_debug("irq: -> existing mapping on virq %d\n", virq);
757 return virq; 730 return virq;
758 } 731 }
@@ -818,8 +791,8 @@ unsigned int irq_create_of_mapping(struct device_node *controller,
818 791
819 /* Set type if specified and different than the current one */ 792 /* Set type if specified and different than the current one */
820 if (type != IRQ_TYPE_NONE && 793 if (type != IRQ_TYPE_NONE &&
821 type != (irq_to_desc(virq)->status & IRQF_TRIGGER_MASK)) 794 type != (irqd_get_trigger_type(irq_get_irq_data(virq))))
822 set_irq_type(virq, type); 795 irq_set_irq_type(virq, type);
823 return virq; 796 return virq;
824} 797}
825EXPORT_SYMBOL_GPL(irq_create_of_mapping); 798EXPORT_SYMBOL_GPL(irq_create_of_mapping);
@@ -833,16 +806,17 @@ void irq_dispose_mapping(unsigned int virq)
833 return; 806 return;
834 807
835 host = irq_map[virq].host; 808 host = irq_map[virq].host;
836 WARN_ON (host == NULL); 809 if (WARN_ON(host == NULL))
837 if (host == NULL)
838 return; 810 return;
839 811
840 /* Never unmap legacy interrupts */ 812 /* Never unmap legacy interrupts */
841 if (host->revmap_type == IRQ_HOST_MAP_LEGACY) 813 if (host->revmap_type == IRQ_HOST_MAP_LEGACY)
842 return; 814 return;
843 815
816 irq_set_status_flags(virq, IRQ_NOREQUEST);
817
844 /* remove chip and handler */ 818 /* remove chip and handler */
845 set_irq_chip_and_handler(virq, NULL, NULL); 819 irq_set_chip_and_handler(virq, NULL, NULL);
846 820
847 /* Make sure it's completed */ 821 /* Make sure it's completed */
848 synchronize_irq(virq); 822 synchronize_irq(virq);
@@ -860,13 +834,6 @@ void irq_dispose_mapping(unsigned int virq)
860 host->revmap_data.linear.revmap[hwirq] = NO_IRQ; 834 host->revmap_data.linear.revmap[hwirq] = NO_IRQ;
861 break; 835 break;
862 case IRQ_HOST_MAP_TREE: 836 case IRQ_HOST_MAP_TREE:
863 /*
864 * Check if radix tree allocated yet, if not then nothing to
865 * remove.
866 */
867 smp_rmb();
868 if (revmap_trees_allocated < 1)
869 break;
870 mutex_lock(&revmap_trees_mutex); 837 mutex_lock(&revmap_trees_mutex);
871 radix_tree_delete(&host->revmap_data.tree, hwirq); 838 radix_tree_delete(&host->revmap_data.tree, hwirq);
872 mutex_unlock(&revmap_trees_mutex); 839 mutex_unlock(&revmap_trees_mutex);
@@ -877,9 +844,7 @@ void irq_dispose_mapping(unsigned int virq)
877 smp_mb(); 844 smp_mb();
878 irq_map[virq].hwirq = host->inval_irq; 845 irq_map[virq].hwirq = host->inval_irq;
879 846
880 /* Set some flags */ 847 irq_free_descs(virq, 1);
881 irq_to_desc(virq)->status |= IRQ_NOREQUEST;
882
883 /* Free it */ 848 /* Free it */
884 irq_free_virt(virq, 1); 849 irq_free_virt(virq, 1);
885} 850}
@@ -924,21 +889,17 @@ unsigned int irq_radix_revmap_lookup(struct irq_host *host,
924 struct irq_map_entry *ptr; 889 struct irq_map_entry *ptr;
925 unsigned int virq; 890 unsigned int virq;
926 891
927 WARN_ON(host->revmap_type != IRQ_HOST_MAP_TREE); 892 if (WARN_ON_ONCE(host->revmap_type != IRQ_HOST_MAP_TREE))
928
929 /*
930 * Check if the radix tree exists and has bee initialized.
931 * If not, we fallback to slow mode
932 */
933 if (revmap_trees_allocated < 2)
934 return irq_find_mapping(host, hwirq); 893 return irq_find_mapping(host, hwirq);
935 894
936 /* Now try to resolve */
937 /* 895 /*
938 * No rcu_read_lock(ing) needed, the ptr returned can't go under us 896 * The ptr returned references the static global irq_map.
939 * as it's referencing an entry in the static irq_map table. 897 * but freeing an irq can delete nodes along the path to
898 * do the lookup via call_rcu.
940 */ 899 */
900 rcu_read_lock();
941 ptr = radix_tree_lookup(&host->revmap_data.tree, hwirq); 901 ptr = radix_tree_lookup(&host->revmap_data.tree, hwirq);
902 rcu_read_unlock();
942 903
943 /* 904 /*
944 * If found in radix tree, then fine. 905 * If found in radix tree, then fine.
@@ -956,16 +917,7 @@ unsigned int irq_radix_revmap_lookup(struct irq_host *host,
956void irq_radix_revmap_insert(struct irq_host *host, unsigned int virq, 917void irq_radix_revmap_insert(struct irq_host *host, unsigned int virq,
957 irq_hw_number_t hwirq) 918 irq_hw_number_t hwirq)
958{ 919{
959 920 if (WARN_ON(host->revmap_type != IRQ_HOST_MAP_TREE))
960 WARN_ON(host->revmap_type != IRQ_HOST_MAP_TREE);
961
962 /*
963 * Check if the radix tree exists yet.
964 * If not, then the irq will be inserted into the tree when it gets
965 * initialized.
966 */
967 smp_rmb();
968 if (revmap_trees_allocated < 1)
969 return; 921 return;
970 922
971 if (virq != NO_IRQ) { 923 if (virq != NO_IRQ) {
@@ -981,7 +933,8 @@ unsigned int irq_linear_revmap(struct irq_host *host,
981{ 933{
982 unsigned int *revmap; 934 unsigned int *revmap;
983 935
984 WARN_ON(host->revmap_type != IRQ_HOST_MAP_LINEAR); 936 if (WARN_ON_ONCE(host->revmap_type != IRQ_HOST_MAP_LINEAR))
937 return irq_find_mapping(host, hwirq);
985 938
986 /* Check revmap bounds */ 939 /* Check revmap bounds */
987 if (unlikely(hwirq >= host->revmap_data.linear.size)) 940 if (unlikely(hwirq >= host->revmap_data.linear.size))
@@ -1054,14 +1007,23 @@ void irq_free_virt(unsigned int virq, unsigned int count)
1054 WARN_ON (virq < NUM_ISA_INTERRUPTS); 1007 WARN_ON (virq < NUM_ISA_INTERRUPTS);
1055 WARN_ON (count == 0 || (virq + count) > irq_virq_count); 1008 WARN_ON (count == 0 || (virq + count) > irq_virq_count);
1056 1009
1010 if (virq < NUM_ISA_INTERRUPTS) {
1011 if (virq + count < NUM_ISA_INTERRUPTS)
1012 return;
1013 count =- NUM_ISA_INTERRUPTS - virq;
1014 virq = NUM_ISA_INTERRUPTS;
1015 }
1016
1017 if (count > irq_virq_count || virq > irq_virq_count - count) {
1018 if (virq > irq_virq_count)
1019 return;
1020 count = irq_virq_count - virq;
1021 }
1022
1057 raw_spin_lock_irqsave(&irq_big_lock, flags); 1023 raw_spin_lock_irqsave(&irq_big_lock, flags);
1058 for (i = virq; i < (virq + count); i++) { 1024 for (i = virq; i < (virq + count); i++) {
1059 struct irq_host *host; 1025 struct irq_host *host;
1060 1026
1061 if (i < NUM_ISA_INTERRUPTS ||
1062 (virq + count) > irq_virq_count)
1063 continue;
1064
1065 host = irq_map[i].host; 1027 host = irq_map[i].host;
1066 irq_map[i].hwirq = host->inval_irq; 1028 irq_map[i].hwirq = host->inval_irq;
1067 smp_wmb(); 1029 smp_wmb();
@@ -1072,82 +1034,21 @@ void irq_free_virt(unsigned int virq, unsigned int count)
1072 1034
1073int arch_early_irq_init(void) 1035int arch_early_irq_init(void)
1074{ 1036{
1075 struct irq_desc *desc;
1076 int i;
1077
1078 for (i = 0; i < NR_IRQS; i++) {
1079 desc = irq_to_desc(i);
1080 if (desc)
1081 desc->status |= IRQ_NOREQUEST;
1082 }
1083
1084 return 0;
1085}
1086
1087int arch_init_chip_data(struct irq_desc *desc, int node)
1088{
1089 desc->status |= IRQ_NOREQUEST;
1090 return 0; 1037 return 0;
1091} 1038}
1092 1039
1093/* We need to create the radix trees late */
1094static int irq_late_init(void)
1095{
1096 struct irq_host *h;
1097 unsigned int i;
1098
1099 /*
1100 * No mutual exclusion with respect to accessors of the tree is needed
1101 * here as the synchronization is done via the state variable
1102 * revmap_trees_allocated.
1103 */
1104 list_for_each_entry(h, &irq_hosts, link) {
1105 if (h->revmap_type == IRQ_HOST_MAP_TREE)
1106 INIT_RADIX_TREE(&h->revmap_data.tree, GFP_KERNEL);
1107 }
1108
1109 /*
1110 * Make sure the radix trees inits are visible before setting
1111 * the flag
1112 */
1113 smp_wmb();
1114 revmap_trees_allocated = 1;
1115
1116 /*
1117 * Insert the reverse mapping for those interrupts already present
1118 * in irq_map[].
1119 */
1120 mutex_lock(&revmap_trees_mutex);
1121 for (i = 0; i < irq_virq_count; i++) {
1122 if (irq_map[i].host &&
1123 (irq_map[i].host->revmap_type == IRQ_HOST_MAP_TREE))
1124 radix_tree_insert(&irq_map[i].host->revmap_data.tree,
1125 irq_map[i].hwirq, &irq_map[i]);
1126 }
1127 mutex_unlock(&revmap_trees_mutex);
1128
1129 /*
1130 * Make sure the radix trees insertions are visible before setting
1131 * the flag
1132 */
1133 smp_wmb();
1134 revmap_trees_allocated = 2;
1135
1136 return 0;
1137}
1138arch_initcall(irq_late_init);
1139
1140#ifdef CONFIG_VIRQ_DEBUG 1040#ifdef CONFIG_VIRQ_DEBUG
1141static int virq_debug_show(struct seq_file *m, void *private) 1041static int virq_debug_show(struct seq_file *m, void *private)
1142{ 1042{
1143 unsigned long flags; 1043 unsigned long flags;
1144 struct irq_desc *desc; 1044 struct irq_desc *desc;
1145 const char *p; 1045 const char *p;
1146 char none[] = "none"; 1046 static const char none[] = "none";
1047 void *data;
1147 int i; 1048 int i;
1148 1049
1149 seq_printf(m, "%-5s %-7s %-15s %s\n", "virq", "hwirq", 1050 seq_printf(m, "%-5s %-7s %-15s %-18s %s\n", "virq", "hwirq",
1150 "chip name", "host name"); 1051 "chip name", "chip data", "host name");
1151 1052
1152 for (i = 1; i < nr_irqs; i++) { 1053 for (i = 1; i < nr_irqs; i++) {
1153 desc = irq_to_desc(i); 1054 desc = irq_to_desc(i);
@@ -1157,15 +1058,21 @@ static int virq_debug_show(struct seq_file *m, void *private)
1157 raw_spin_lock_irqsave(&desc->lock, flags); 1058 raw_spin_lock_irqsave(&desc->lock, flags);
1158 1059
1159 if (desc->action && desc->action->handler) { 1060 if (desc->action && desc->action->handler) {
1061 struct irq_chip *chip;
1062
1160 seq_printf(m, "%5d ", i); 1063 seq_printf(m, "%5d ", i);
1161 seq_printf(m, "0x%05lx ", virq_to_hw(i)); 1064 seq_printf(m, "0x%05lx ", irq_map[i].hwirq);
1162 1065
1163 if (desc->chip && desc->chip->name) 1066 chip = irq_desc_get_chip(desc);
1164 p = desc->chip->name; 1067 if (chip && chip->name)
1068 p = chip->name;
1165 else 1069 else
1166 p = none; 1070 p = none;
1167 seq_printf(m, "%-15s ", p); 1071 seq_printf(m, "%-15s ", p);
1168 1072
1073 data = irq_desc_get_chip_data(desc);
1074 seq_printf(m, "0x%16p ", data);
1075
1169 if (irq_map[i].host && irq_map[i].host->of_node) 1076 if (irq_map[i].host && irq_map[i].host->of_node)
1170 p = irq_map[i].host->of_node->full_name; 1077 p = irq_map[i].host->of_node->full_name;
1171 else 1078 else
diff --git a/arch/powerpc/kernel/kgdb.c b/arch/powerpc/kernel/kgdb.c
index 7f61a3ac787c..76a6e40a6f7c 100644
--- a/arch/powerpc/kernel/kgdb.c
+++ b/arch/powerpc/kernel/kgdb.c
@@ -109,7 +109,7 @@ static int kgdb_call_nmi_hook(struct pt_regs *regs)
109#ifdef CONFIG_SMP 109#ifdef CONFIG_SMP
110void kgdb_roundup_cpus(unsigned long flags) 110void kgdb_roundup_cpus(unsigned long flags)
111{ 111{
112 smp_send_debugger_break(MSG_ALL_BUT_SELF); 112 smp_send_debugger_break();
113} 113}
114#endif 114#endif
115 115
@@ -142,7 +142,7 @@ static int kgdb_singlestep(struct pt_regs *regs)
142 return 0; 142 return 0;
143 143
144 /* 144 /*
145 * On Book E and perhaps other processsors, singlestep is handled on 145 * On Book E and perhaps other processors, singlestep is handled on
146 * the critical exception stack. This causes current_thread_info() 146 * the critical exception stack. This causes current_thread_info()
147 * to fail, since it it locates the thread_info by masking off 147 * to fail, since it it locates the thread_info by masking off
148 * the low bits of the current stack pointer. We work around 148 * the low bits of the current stack pointer. We work around
@@ -194,40 +194,6 @@ static int kgdb_dabr_match(struct pt_regs *regs)
194 ptr = (unsigned long *)ptr32; \ 194 ptr = (unsigned long *)ptr32; \
195 } while (0) 195 } while (0)
196 196
197
198void pt_regs_to_gdb_regs(unsigned long *gdb_regs, struct pt_regs *regs)
199{
200 unsigned long *ptr = gdb_regs;
201 int reg;
202
203 memset(gdb_regs, 0, NUMREGBYTES);
204
205 for (reg = 0; reg < 32; reg++)
206 PACK64(ptr, regs->gpr[reg]);
207
208#ifdef CONFIG_FSL_BOOKE
209#ifdef CONFIG_SPE
210 for (reg = 0; reg < 32; reg++)
211 PACK64(ptr, current->thread.evr[reg]);
212#else
213 ptr += 32;
214#endif
215#else
216 /* fp registers not used by kernel, leave zero */
217 ptr += 32 * 8 / sizeof(long);
218#endif
219
220 PACK64(ptr, regs->nip);
221 PACK64(ptr, regs->msr);
222 PACK32(ptr, regs->ccr);
223 PACK64(ptr, regs->link);
224 PACK64(ptr, regs->ctr);
225 PACK32(ptr, regs->xer);
226
227 BUG_ON((unsigned long)ptr >
228 (unsigned long)(((void *)gdb_regs) + NUMREGBYTES));
229}
230
231void sleeping_thread_to_gdb_regs(unsigned long *gdb_regs, struct task_struct *p) 197void sleeping_thread_to_gdb_regs(unsigned long *gdb_regs, struct task_struct *p)
232{ 198{
233 struct pt_regs *regs = (struct pt_regs *)(p->thread.ksp + 199 struct pt_regs *regs = (struct pt_regs *)(p->thread.ksp +
@@ -271,44 +237,140 @@ void sleeping_thread_to_gdb_regs(unsigned long *gdb_regs, struct task_struct *p)
271 (unsigned long)(((void *)gdb_regs) + NUMREGBYTES)); 237 (unsigned long)(((void *)gdb_regs) + NUMREGBYTES));
272} 238}
273 239
274#define UNPACK64(dest, ptr) do { dest = *(ptr++); } while (0) 240#define GDB_SIZEOF_REG sizeof(unsigned long)
241#define GDB_SIZEOF_REG_U32 sizeof(u32)
275 242
276#define UNPACK32(dest, ptr) do { \ 243#ifdef CONFIG_FSL_BOOKE
277 u32 *ptr32; \ 244#define GDB_SIZEOF_FLOAT_REG sizeof(unsigned long)
278 ptr32 = (u32 *)ptr; \ 245#else
279 dest = *(ptr32++); \ 246#define GDB_SIZEOF_FLOAT_REG sizeof(u64)
280 ptr = (unsigned long *)ptr32; \ 247#endif
281 } while (0)
282 248
283void gdb_regs_to_pt_regs(unsigned long *gdb_regs, struct pt_regs *regs) 249struct dbg_reg_def_t dbg_reg_def[DBG_MAX_REG_NUM] =
284{ 250{
285 unsigned long *ptr = gdb_regs; 251 { "r0", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[0]) },
286 int reg; 252 { "r1", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[1]) },
287 253 { "r2", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[2]) },
288 for (reg = 0; reg < 32; reg++) 254 { "r3", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[3]) },
289 UNPACK64(regs->gpr[reg], ptr); 255 { "r4", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[4]) },
256 { "r5", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[5]) },
257 { "r6", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[6]) },
258 { "r7", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[7]) },
259 { "r8", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[8]) },
260 { "r9", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[9]) },
261 { "r10", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[10]) },
262 { "r11", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[11]) },
263 { "r12", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[12]) },
264 { "r13", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[13]) },
265 { "r14", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[14]) },
266 { "r15", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[15]) },
267 { "r16", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[16]) },
268 { "r17", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[17]) },
269 { "r18", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[18]) },
270 { "r19", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[19]) },
271 { "r20", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[20]) },
272 { "r21", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[21]) },
273 { "r22", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[22]) },
274 { "r23", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[23]) },
275 { "r24", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[24]) },
276 { "r25", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[25]) },
277 { "r26", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[26]) },
278 { "r27", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[27]) },
279 { "r28", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[28]) },
280 { "r29", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[29]) },
281 { "r30", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[30]) },
282 { "r31", GDB_SIZEOF_REG, offsetof(struct pt_regs, gpr[31]) },
283
284 { "f0", GDB_SIZEOF_FLOAT_REG, 0 },
285 { "f1", GDB_SIZEOF_FLOAT_REG, 1 },
286 { "f2", GDB_SIZEOF_FLOAT_REG, 2 },
287 { "f3", GDB_SIZEOF_FLOAT_REG, 3 },
288 { "f4", GDB_SIZEOF_FLOAT_REG, 4 },
289 { "f5", GDB_SIZEOF_FLOAT_REG, 5 },
290 { "f6", GDB_SIZEOF_FLOAT_REG, 6 },
291 { "f7", GDB_SIZEOF_FLOAT_REG, 7 },
292 { "f8", GDB_SIZEOF_FLOAT_REG, 8 },
293 { "f9", GDB_SIZEOF_FLOAT_REG, 9 },
294 { "f10", GDB_SIZEOF_FLOAT_REG, 10 },
295 { "f11", GDB_SIZEOF_FLOAT_REG, 11 },
296 { "f12", GDB_SIZEOF_FLOAT_REG, 12 },
297 { "f13", GDB_SIZEOF_FLOAT_REG, 13 },
298 { "f14", GDB_SIZEOF_FLOAT_REG, 14 },
299 { "f15", GDB_SIZEOF_FLOAT_REG, 15 },
300 { "f16", GDB_SIZEOF_FLOAT_REG, 16 },
301 { "f17", GDB_SIZEOF_FLOAT_REG, 17 },
302 { "f18", GDB_SIZEOF_FLOAT_REG, 18 },
303 { "f19", GDB_SIZEOF_FLOAT_REG, 19 },
304 { "f20", GDB_SIZEOF_FLOAT_REG, 20 },
305 { "f21", GDB_SIZEOF_FLOAT_REG, 21 },
306 { "f22", GDB_SIZEOF_FLOAT_REG, 22 },
307 { "f23", GDB_SIZEOF_FLOAT_REG, 23 },
308 { "f24", GDB_SIZEOF_FLOAT_REG, 24 },
309 { "f25", GDB_SIZEOF_FLOAT_REG, 25 },
310 { "f26", GDB_SIZEOF_FLOAT_REG, 26 },
311 { "f27", GDB_SIZEOF_FLOAT_REG, 27 },
312 { "f28", GDB_SIZEOF_FLOAT_REG, 28 },
313 { "f29", GDB_SIZEOF_FLOAT_REG, 29 },
314 { "f30", GDB_SIZEOF_FLOAT_REG, 30 },
315 { "f31", GDB_SIZEOF_FLOAT_REG, 31 },
316
317 { "pc", GDB_SIZEOF_REG, offsetof(struct pt_regs, nip) },
318 { "msr", GDB_SIZEOF_REG, offsetof(struct pt_regs, msr) },
319 { "cr", GDB_SIZEOF_REG_U32, offsetof(struct pt_regs, ccr) },
320 { "lr", GDB_SIZEOF_REG, offsetof(struct pt_regs, link) },
321 { "ctr", GDB_SIZEOF_REG_U32, offsetof(struct pt_regs, ctr) },
322 { "xer", GDB_SIZEOF_REG, offsetof(struct pt_regs, xer) },
323};
290 324
291#ifdef CONFIG_FSL_BOOKE 325char *dbg_get_reg(int regno, void *mem, struct pt_regs *regs)
292#ifdef CONFIG_SPE 326{
293 for (reg = 0; reg < 32; reg++) 327 if (regno >= DBG_MAX_REG_NUM || regno < 0)
294 UNPACK64(current->thread.evr[reg], ptr); 328 return NULL;
329
330 if (regno < 32 || regno >= 64)
331 /* First 0 -> 31 gpr registers*/
332 /* pc, msr, ls... registers 64 -> 69 */
333 memcpy(mem, (void *)regs + dbg_reg_def[regno].offset,
334 dbg_reg_def[regno].size);
335
336 if (regno >= 32 && regno < 64) {
337 /* FP registers 32 -> 63 */
338#if defined(CONFIG_FSL_BOOKE) && defined(CONFIG_SPE)
339 if (current)
340 memcpy(mem, &current->thread.evr[regno-32],
341 dbg_reg_def[regno].size);
295#else 342#else
296 ptr += 32; 343 /* fp registers not used by kernel, leave zero */
344 memset(mem, 0, dbg_reg_def[regno].size);
297#endif 345#endif
346 }
347
348 return dbg_reg_def[regno].name;
349}
350
351int dbg_set_reg(int regno, void *mem, struct pt_regs *regs)
352{
353 if (regno >= DBG_MAX_REG_NUM || regno < 0)
354 return -EINVAL;
355
356 if (regno < 32 || regno >= 64)
357 /* First 0 -> 31 gpr registers*/
358 /* pc, msr, ls... registers 64 -> 69 */
359 memcpy((void *)regs + dbg_reg_def[regno].offset, mem,
360 dbg_reg_def[regno].size);
361
362 if (regno >= 32 && regno < 64) {
363 /* FP registers 32 -> 63 */
364#if defined(CONFIG_FSL_BOOKE) && defined(CONFIG_SPE)
365 memcpy(&current->thread.evr[regno-32], mem,
366 dbg_reg_def[regno].size);
298#else 367#else
299 /* fp registers not used by kernel, leave zero */ 368 /* fp registers not used by kernel, leave zero */
300 ptr += 32 * 8 / sizeof(int); 369 return 0;
301#endif 370#endif
371 }
302 372
303 UNPACK64(regs->nip, ptr); 373 return 0;
304 UNPACK64(regs->msr, ptr);
305 UNPACK32(regs->ccr, ptr);
306 UNPACK64(regs->link, ptr);
307 UNPACK64(regs->ctr, ptr);
308 UNPACK32(regs->xer, ptr);
309
310 BUG_ON((unsigned long)ptr >
311 (unsigned long)(((void *)gdb_regs) + NUMREGBYTES));
312} 374}
313 375
314void kgdb_arch_set_pc(struct pt_regs *regs, unsigned long pc) 376void kgdb_arch_set_pc(struct pt_regs *regs, unsigned long pc)
diff --git a/arch/powerpc/kernel/kvm.c b/arch/powerpc/kernel/kvm.c
new file mode 100644
index 000000000000..b06bdae04064
--- /dev/null
+++ b/arch/powerpc/kernel/kvm.c
@@ -0,0 +1,596 @@
1/*
2 * Copyright (C) 2010 SUSE Linux Products GmbH. All rights reserved.
3 *
4 * Authors:
5 * Alexander Graf <agraf@suse.de>
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License, version 2, as
9 * published by the Free Software Foundation.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
19 */
20
21#include <linux/kvm_host.h>
22#include <linux/init.h>
23#include <linux/kvm_para.h>
24#include <linux/slab.h>
25#include <linux/of.h>
26
27#include <asm/reg.h>
28#include <asm/sections.h>
29#include <asm/cacheflush.h>
30#include <asm/disassemble.h>
31
32#define KVM_MAGIC_PAGE (-4096L)
33#define magic_var(x) KVM_MAGIC_PAGE + offsetof(struct kvm_vcpu_arch_shared, x)
34
35#define KVM_INST_LWZ 0x80000000
36#define KVM_INST_STW 0x90000000
37#define KVM_INST_LD 0xe8000000
38#define KVM_INST_STD 0xf8000000
39#define KVM_INST_NOP 0x60000000
40#define KVM_INST_B 0x48000000
41#define KVM_INST_B_MASK 0x03ffffff
42#define KVM_INST_B_MAX 0x01ffffff
43
44#define KVM_MASK_RT 0x03e00000
45#define KVM_RT_30 0x03c00000
46#define KVM_MASK_RB 0x0000f800
47#define KVM_INST_MFMSR 0x7c0000a6
48#define KVM_INST_MFSPR_SPRG0 0x7c1042a6
49#define KVM_INST_MFSPR_SPRG1 0x7c1142a6
50#define KVM_INST_MFSPR_SPRG2 0x7c1242a6
51#define KVM_INST_MFSPR_SPRG3 0x7c1342a6
52#define KVM_INST_MFSPR_SRR0 0x7c1a02a6
53#define KVM_INST_MFSPR_SRR1 0x7c1b02a6
54#define KVM_INST_MFSPR_DAR 0x7c1302a6
55#define KVM_INST_MFSPR_DSISR 0x7c1202a6
56
57#define KVM_INST_MTSPR_SPRG0 0x7c1043a6
58#define KVM_INST_MTSPR_SPRG1 0x7c1143a6
59#define KVM_INST_MTSPR_SPRG2 0x7c1243a6
60#define KVM_INST_MTSPR_SPRG3 0x7c1343a6
61#define KVM_INST_MTSPR_SRR0 0x7c1a03a6
62#define KVM_INST_MTSPR_SRR1 0x7c1b03a6
63#define KVM_INST_MTSPR_DAR 0x7c1303a6
64#define KVM_INST_MTSPR_DSISR 0x7c1203a6
65
66#define KVM_INST_TLBSYNC 0x7c00046c
67#define KVM_INST_MTMSRD_L0 0x7c000164
68#define KVM_INST_MTMSRD_L1 0x7c010164
69#define KVM_INST_MTMSR 0x7c000124
70
71#define KVM_INST_WRTEEI_0 0x7c000146
72#define KVM_INST_WRTEEI_1 0x7c008146
73
74#define KVM_INST_MTSRIN 0x7c0001e4
75
76static bool kvm_patching_worked = true;
77static char kvm_tmp[1024 * 1024];
78static int kvm_tmp_index;
79
80static inline void kvm_patch_ins(u32 *inst, u32 new_inst)
81{
82 *inst = new_inst;
83 flush_icache_range((ulong)inst, (ulong)inst + 4);
84}
85
86static void kvm_patch_ins_ll(u32 *inst, long addr, u32 rt)
87{
88#ifdef CONFIG_64BIT
89 kvm_patch_ins(inst, KVM_INST_LD | rt | (addr & 0x0000fffc));
90#else
91 kvm_patch_ins(inst, KVM_INST_LWZ | rt | (addr & 0x0000fffc));
92#endif
93}
94
95static void kvm_patch_ins_ld(u32 *inst, long addr, u32 rt)
96{
97#ifdef CONFIG_64BIT
98 kvm_patch_ins(inst, KVM_INST_LD | rt | (addr & 0x0000fffc));
99#else
100 kvm_patch_ins(inst, KVM_INST_LWZ | rt | ((addr + 4) & 0x0000fffc));
101#endif
102}
103
104static void kvm_patch_ins_lwz(u32 *inst, long addr, u32 rt)
105{
106 kvm_patch_ins(inst, KVM_INST_LWZ | rt | (addr & 0x0000ffff));
107}
108
109static void kvm_patch_ins_std(u32 *inst, long addr, u32 rt)
110{
111#ifdef CONFIG_64BIT
112 kvm_patch_ins(inst, KVM_INST_STD | rt | (addr & 0x0000fffc));
113#else
114 kvm_patch_ins(inst, KVM_INST_STW | rt | ((addr + 4) & 0x0000fffc));
115#endif
116}
117
118static void kvm_patch_ins_stw(u32 *inst, long addr, u32 rt)
119{
120 kvm_patch_ins(inst, KVM_INST_STW | rt | (addr & 0x0000fffc));
121}
122
123static void kvm_patch_ins_nop(u32 *inst)
124{
125 kvm_patch_ins(inst, KVM_INST_NOP);
126}
127
128static void kvm_patch_ins_b(u32 *inst, int addr)
129{
130#if defined(CONFIG_RELOCATABLE) && defined(CONFIG_PPC_BOOK3S)
131 /* On relocatable kernels interrupts handlers and our code
132 can be in different regions, so we don't patch them */
133
134 extern u32 __end_interrupts;
135 if ((ulong)inst < (ulong)&__end_interrupts)
136 return;
137#endif
138
139 kvm_patch_ins(inst, KVM_INST_B | (addr & KVM_INST_B_MASK));
140}
141
142static u32 *kvm_alloc(int len)
143{
144 u32 *p;
145
146 if ((kvm_tmp_index + len) > ARRAY_SIZE(kvm_tmp)) {
147 printk(KERN_ERR "KVM: No more space (%d + %d)\n",
148 kvm_tmp_index, len);
149 kvm_patching_worked = false;
150 return NULL;
151 }
152
153 p = (void*)&kvm_tmp[kvm_tmp_index];
154 kvm_tmp_index += len;
155
156 return p;
157}
158
159extern u32 kvm_emulate_mtmsrd_branch_offs;
160extern u32 kvm_emulate_mtmsrd_reg_offs;
161extern u32 kvm_emulate_mtmsrd_orig_ins_offs;
162extern u32 kvm_emulate_mtmsrd_len;
163extern u32 kvm_emulate_mtmsrd[];
164
165static void kvm_patch_ins_mtmsrd(u32 *inst, u32 rt)
166{
167 u32 *p;
168 int distance_start;
169 int distance_end;
170 ulong next_inst;
171
172 p = kvm_alloc(kvm_emulate_mtmsrd_len * 4);
173 if (!p)
174 return;
175
176 /* Find out where we are and put everything there */
177 distance_start = (ulong)p - (ulong)inst;
178 next_inst = ((ulong)inst + 4);
179 distance_end = next_inst - (ulong)&p[kvm_emulate_mtmsrd_branch_offs];
180
181 /* Make sure we only write valid b instructions */
182 if (distance_start > KVM_INST_B_MAX) {
183 kvm_patching_worked = false;
184 return;
185 }
186
187 /* Modify the chunk to fit the invocation */
188 memcpy(p, kvm_emulate_mtmsrd, kvm_emulate_mtmsrd_len * 4);
189 p[kvm_emulate_mtmsrd_branch_offs] |= distance_end & KVM_INST_B_MASK;
190 switch (get_rt(rt)) {
191 case 30:
192 kvm_patch_ins_ll(&p[kvm_emulate_mtmsrd_reg_offs],
193 magic_var(scratch2), KVM_RT_30);
194 break;
195 case 31:
196 kvm_patch_ins_ll(&p[kvm_emulate_mtmsrd_reg_offs],
197 magic_var(scratch1), KVM_RT_30);
198 break;
199 default:
200 p[kvm_emulate_mtmsrd_reg_offs] |= rt;
201 break;
202 }
203
204 p[kvm_emulate_mtmsrd_orig_ins_offs] = *inst;
205 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_mtmsrd_len * 4);
206
207 /* Patch the invocation */
208 kvm_patch_ins_b(inst, distance_start);
209}
210
211extern u32 kvm_emulate_mtmsr_branch_offs;
212extern u32 kvm_emulate_mtmsr_reg1_offs;
213extern u32 kvm_emulate_mtmsr_reg2_offs;
214extern u32 kvm_emulate_mtmsr_orig_ins_offs;
215extern u32 kvm_emulate_mtmsr_len;
216extern u32 kvm_emulate_mtmsr[];
217
218static void kvm_patch_ins_mtmsr(u32 *inst, u32 rt)
219{
220 u32 *p;
221 int distance_start;
222 int distance_end;
223 ulong next_inst;
224
225 p = kvm_alloc(kvm_emulate_mtmsr_len * 4);
226 if (!p)
227 return;
228
229 /* Find out where we are and put everything there */
230 distance_start = (ulong)p - (ulong)inst;
231 next_inst = ((ulong)inst + 4);
232 distance_end = next_inst - (ulong)&p[kvm_emulate_mtmsr_branch_offs];
233
234 /* Make sure we only write valid b instructions */
235 if (distance_start > KVM_INST_B_MAX) {
236 kvm_patching_worked = false;
237 return;
238 }
239
240 /* Modify the chunk to fit the invocation */
241 memcpy(p, kvm_emulate_mtmsr, kvm_emulate_mtmsr_len * 4);
242 p[kvm_emulate_mtmsr_branch_offs] |= distance_end & KVM_INST_B_MASK;
243
244 /* Make clobbered registers work too */
245 switch (get_rt(rt)) {
246 case 30:
247 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg1_offs],
248 magic_var(scratch2), KVM_RT_30);
249 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg2_offs],
250 magic_var(scratch2), KVM_RT_30);
251 break;
252 case 31:
253 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg1_offs],
254 magic_var(scratch1), KVM_RT_30);
255 kvm_patch_ins_ll(&p[kvm_emulate_mtmsr_reg2_offs],
256 magic_var(scratch1), KVM_RT_30);
257 break;
258 default:
259 p[kvm_emulate_mtmsr_reg1_offs] |= rt;
260 p[kvm_emulate_mtmsr_reg2_offs] |= rt;
261 break;
262 }
263
264 p[kvm_emulate_mtmsr_orig_ins_offs] = *inst;
265 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_mtmsr_len * 4);
266
267 /* Patch the invocation */
268 kvm_patch_ins_b(inst, distance_start);
269}
270
271#ifdef CONFIG_BOOKE
272
273extern u32 kvm_emulate_wrteei_branch_offs;
274extern u32 kvm_emulate_wrteei_ee_offs;
275extern u32 kvm_emulate_wrteei_len;
276extern u32 kvm_emulate_wrteei[];
277
278static void kvm_patch_ins_wrteei(u32 *inst)
279{
280 u32 *p;
281 int distance_start;
282 int distance_end;
283 ulong next_inst;
284
285 p = kvm_alloc(kvm_emulate_wrteei_len * 4);
286 if (!p)
287 return;
288
289 /* Find out where we are and put everything there */
290 distance_start = (ulong)p - (ulong)inst;
291 next_inst = ((ulong)inst + 4);
292 distance_end = next_inst - (ulong)&p[kvm_emulate_wrteei_branch_offs];
293
294 /* Make sure we only write valid b instructions */
295 if (distance_start > KVM_INST_B_MAX) {
296 kvm_patching_worked = false;
297 return;
298 }
299
300 /* Modify the chunk to fit the invocation */
301 memcpy(p, kvm_emulate_wrteei, kvm_emulate_wrteei_len * 4);
302 p[kvm_emulate_wrteei_branch_offs] |= distance_end & KVM_INST_B_MASK;
303 p[kvm_emulate_wrteei_ee_offs] |= (*inst & MSR_EE);
304 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_wrteei_len * 4);
305
306 /* Patch the invocation */
307 kvm_patch_ins_b(inst, distance_start);
308}
309
310#endif
311
312#ifdef CONFIG_PPC_BOOK3S_32
313
314extern u32 kvm_emulate_mtsrin_branch_offs;
315extern u32 kvm_emulate_mtsrin_reg1_offs;
316extern u32 kvm_emulate_mtsrin_reg2_offs;
317extern u32 kvm_emulate_mtsrin_orig_ins_offs;
318extern u32 kvm_emulate_mtsrin_len;
319extern u32 kvm_emulate_mtsrin[];
320
321static void kvm_patch_ins_mtsrin(u32 *inst, u32 rt, u32 rb)
322{
323 u32 *p;
324 int distance_start;
325 int distance_end;
326 ulong next_inst;
327
328 p = kvm_alloc(kvm_emulate_mtsrin_len * 4);
329 if (!p)
330 return;
331
332 /* Find out where we are and put everything there */
333 distance_start = (ulong)p - (ulong)inst;
334 next_inst = ((ulong)inst + 4);
335 distance_end = next_inst - (ulong)&p[kvm_emulate_mtsrin_branch_offs];
336
337 /* Make sure we only write valid b instructions */
338 if (distance_start > KVM_INST_B_MAX) {
339 kvm_patching_worked = false;
340 return;
341 }
342
343 /* Modify the chunk to fit the invocation */
344 memcpy(p, kvm_emulate_mtsrin, kvm_emulate_mtsrin_len * 4);
345 p[kvm_emulate_mtsrin_branch_offs] |= distance_end & KVM_INST_B_MASK;
346 p[kvm_emulate_mtsrin_reg1_offs] |= (rb << 10);
347 p[kvm_emulate_mtsrin_reg2_offs] |= rt;
348 p[kvm_emulate_mtsrin_orig_ins_offs] = *inst;
349 flush_icache_range((ulong)p, (ulong)p + kvm_emulate_mtsrin_len * 4);
350
351 /* Patch the invocation */
352 kvm_patch_ins_b(inst, distance_start);
353}
354
355#endif
356
357static void kvm_map_magic_page(void *data)
358{
359 u32 *features = data;
360
361 ulong in[8];
362 ulong out[8];
363
364 in[0] = KVM_MAGIC_PAGE;
365 in[1] = KVM_MAGIC_PAGE;
366
367 kvm_hypercall(in, out, HC_VENDOR_KVM | KVM_HC_PPC_MAP_MAGIC_PAGE);
368
369 *features = out[0];
370}
371
372static void kvm_check_ins(u32 *inst, u32 features)
373{
374 u32 _inst = *inst;
375 u32 inst_no_rt = _inst & ~KVM_MASK_RT;
376 u32 inst_rt = _inst & KVM_MASK_RT;
377
378 switch (inst_no_rt) {
379 /* Loads */
380 case KVM_INST_MFMSR:
381 kvm_patch_ins_ld(inst, magic_var(msr), inst_rt);
382 break;
383 case KVM_INST_MFSPR_SPRG0:
384 kvm_patch_ins_ld(inst, magic_var(sprg0), inst_rt);
385 break;
386 case KVM_INST_MFSPR_SPRG1:
387 kvm_patch_ins_ld(inst, magic_var(sprg1), inst_rt);
388 break;
389 case KVM_INST_MFSPR_SPRG2:
390 kvm_patch_ins_ld(inst, magic_var(sprg2), inst_rt);
391 break;
392 case KVM_INST_MFSPR_SPRG3:
393 kvm_patch_ins_ld(inst, magic_var(sprg3), inst_rt);
394 break;
395 case KVM_INST_MFSPR_SRR0:
396 kvm_patch_ins_ld(inst, magic_var(srr0), inst_rt);
397 break;
398 case KVM_INST_MFSPR_SRR1:
399 kvm_patch_ins_ld(inst, magic_var(srr1), inst_rt);
400 break;
401 case KVM_INST_MFSPR_DAR:
402 kvm_patch_ins_ld(inst, magic_var(dar), inst_rt);
403 break;
404 case KVM_INST_MFSPR_DSISR:
405 kvm_patch_ins_lwz(inst, magic_var(dsisr), inst_rt);
406 break;
407
408 /* Stores */
409 case KVM_INST_MTSPR_SPRG0:
410 kvm_patch_ins_std(inst, magic_var(sprg0), inst_rt);
411 break;
412 case KVM_INST_MTSPR_SPRG1:
413 kvm_patch_ins_std(inst, magic_var(sprg1), inst_rt);
414 break;
415 case KVM_INST_MTSPR_SPRG2:
416 kvm_patch_ins_std(inst, magic_var(sprg2), inst_rt);
417 break;
418 case KVM_INST_MTSPR_SPRG3:
419 kvm_patch_ins_std(inst, magic_var(sprg3), inst_rt);
420 break;
421 case KVM_INST_MTSPR_SRR0:
422 kvm_patch_ins_std(inst, magic_var(srr0), inst_rt);
423 break;
424 case KVM_INST_MTSPR_SRR1:
425 kvm_patch_ins_std(inst, magic_var(srr1), inst_rt);
426 break;
427 case KVM_INST_MTSPR_DAR:
428 kvm_patch_ins_std(inst, magic_var(dar), inst_rt);
429 break;
430 case KVM_INST_MTSPR_DSISR:
431 kvm_patch_ins_stw(inst, magic_var(dsisr), inst_rt);
432 break;
433
434 /* Nops */
435 case KVM_INST_TLBSYNC:
436 kvm_patch_ins_nop(inst);
437 break;
438
439 /* Rewrites */
440 case KVM_INST_MTMSRD_L1:
441 kvm_patch_ins_mtmsrd(inst, inst_rt);
442 break;
443 case KVM_INST_MTMSR:
444 case KVM_INST_MTMSRD_L0:
445 kvm_patch_ins_mtmsr(inst, inst_rt);
446 break;
447 }
448
449 switch (inst_no_rt & ~KVM_MASK_RB) {
450#ifdef CONFIG_PPC_BOOK3S_32
451 case KVM_INST_MTSRIN:
452 if (features & KVM_MAGIC_FEAT_SR) {
453 u32 inst_rb = _inst & KVM_MASK_RB;
454 kvm_patch_ins_mtsrin(inst, inst_rt, inst_rb);
455 }
456 break;
457 break;
458#endif
459 }
460
461 switch (_inst) {
462#ifdef CONFIG_BOOKE
463 case KVM_INST_WRTEEI_0:
464 case KVM_INST_WRTEEI_1:
465 kvm_patch_ins_wrteei(inst);
466 break;
467#endif
468 }
469}
470
471static void kvm_use_magic_page(void)
472{
473 u32 *p;
474 u32 *start, *end;
475 u32 tmp;
476 u32 features;
477
478 /* Tell the host to map the magic page to -4096 on all CPUs */
479 on_each_cpu(kvm_map_magic_page, &features, 1);
480
481 /* Quick self-test to see if the mapping works */
482 if (__get_user(tmp, (u32*)KVM_MAGIC_PAGE)) {
483 kvm_patching_worked = false;
484 return;
485 }
486
487 /* Now loop through all code and find instructions */
488 start = (void*)_stext;
489 end = (void*)_etext;
490
491 for (p = start; p < end; p++)
492 kvm_check_ins(p, features);
493
494 printk(KERN_INFO "KVM: Live patching for a fast VM %s\n",
495 kvm_patching_worked ? "worked" : "failed");
496}
497
498unsigned long kvm_hypercall(unsigned long *in,
499 unsigned long *out,
500 unsigned long nr)
501{
502 unsigned long register r0 asm("r0");
503 unsigned long register r3 asm("r3") = in[0];
504 unsigned long register r4 asm("r4") = in[1];
505 unsigned long register r5 asm("r5") = in[2];
506 unsigned long register r6 asm("r6") = in[3];
507 unsigned long register r7 asm("r7") = in[4];
508 unsigned long register r8 asm("r8") = in[5];
509 unsigned long register r9 asm("r9") = in[6];
510 unsigned long register r10 asm("r10") = in[7];
511 unsigned long register r11 asm("r11") = nr;
512 unsigned long register r12 asm("r12");
513
514 asm volatile("bl kvm_hypercall_start"
515 : "=r"(r0), "=r"(r3), "=r"(r4), "=r"(r5), "=r"(r6),
516 "=r"(r7), "=r"(r8), "=r"(r9), "=r"(r10), "=r"(r11),
517 "=r"(r12)
518 : "r"(r3), "r"(r4), "r"(r5), "r"(r6), "r"(r7), "r"(r8),
519 "r"(r9), "r"(r10), "r"(r11)
520 : "memory", "cc", "xer", "ctr", "lr");
521
522 out[0] = r4;
523 out[1] = r5;
524 out[2] = r6;
525 out[3] = r7;
526 out[4] = r8;
527 out[5] = r9;
528 out[6] = r10;
529 out[7] = r11;
530
531 return r3;
532}
533EXPORT_SYMBOL_GPL(kvm_hypercall);
534
535static int kvm_para_setup(void)
536{
537 extern u32 kvm_hypercall_start;
538 struct device_node *hyper_node;
539 u32 *insts;
540 int len, i;
541
542 hyper_node = of_find_node_by_path("/hypervisor");
543 if (!hyper_node)
544 return -1;
545
546 insts = (u32*)of_get_property(hyper_node, "hcall-instructions", &len);
547 if (len % 4)
548 return -1;
549 if (len > (4 * 4))
550 return -1;
551
552 for (i = 0; i < (len / 4); i++)
553 kvm_patch_ins(&(&kvm_hypercall_start)[i], insts[i]);
554
555 return 0;
556}
557
558static __init void kvm_free_tmp(void)
559{
560 unsigned long start, end;
561
562 start = (ulong)&kvm_tmp[kvm_tmp_index + (PAGE_SIZE - 1)] & PAGE_MASK;
563 end = (ulong)&kvm_tmp[ARRAY_SIZE(kvm_tmp)] & PAGE_MASK;
564
565 /* Free the tmp space we don't need */
566 for (; start < end; start += PAGE_SIZE) {
567 ClearPageReserved(virt_to_page(start));
568 init_page_count(virt_to_page(start));
569 free_page(start);
570 totalram_pages++;
571 }
572}
573
574static int __init kvm_guest_init(void)
575{
576 if (!kvm_para_available())
577 goto free_tmp;
578
579 if (kvm_para_setup())
580 goto free_tmp;
581
582 if (kvm_para_has_feature(KVM_FEATURE_MAGIC_PAGE))
583 kvm_use_magic_page();
584
585#ifdef CONFIG_PPC_BOOK3S_64
586 /* Enable napping */
587 powersave_nap = 1;
588#endif
589
590free_tmp:
591 kvm_free_tmp();
592
593 return 0;
594}
595
596postcore_initcall(kvm_guest_init);
diff --git a/arch/powerpc/kernel/kvm_emul.S b/arch/powerpc/kernel/kvm_emul.S
new file mode 100644
index 000000000000..f2b1b2523e61
--- /dev/null
+++ b/arch/powerpc/kernel/kvm_emul.S
@@ -0,0 +1,302 @@
1/*
2 * This program is free software; you can redistribute it and/or modify
3 * it under the terms of the GNU General Public License, version 2, as
4 * published by the Free Software Foundation.
5 *
6 * This program is distributed in the hope that it will be useful,
7 * but WITHOUT ANY WARRANTY; without even the implied warranty of
8 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
9 * GNU General Public License for more details.
10 *
11 * You should have received a copy of the GNU General Public License
12 * along with this program; if not, write to the Free Software
13 * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
14 *
15 * Copyright SUSE Linux Products GmbH 2010
16 *
17 * Authors: Alexander Graf <agraf@suse.de>
18 */
19
20#include <asm/ppc_asm.h>
21#include <asm/kvm_asm.h>
22#include <asm/reg.h>
23#include <asm/page.h>
24#include <asm/asm-offsets.h>
25
26/* Hypercall entry point. Will be patched with device tree instructions. */
27
28.global kvm_hypercall_start
29kvm_hypercall_start:
30 li r3, -1
31 nop
32 nop
33 nop
34 blr
35
36#define KVM_MAGIC_PAGE (-4096)
37
38#ifdef CONFIG_64BIT
39#define LL64(reg, offs, reg2) ld reg, (offs)(reg2)
40#define STL64(reg, offs, reg2) std reg, (offs)(reg2)
41#else
42#define LL64(reg, offs, reg2) lwz reg, (offs + 4)(reg2)
43#define STL64(reg, offs, reg2) stw reg, (offs + 4)(reg2)
44#endif
45
46#define SCRATCH_SAVE \
47 /* Enable critical section. We are critical if \
48 shared->critical == r1 */ \
49 STL64(r1, KVM_MAGIC_PAGE + KVM_MAGIC_CRITICAL, 0); \
50 \
51 /* Save state */ \
52 PPC_STL r31, (KVM_MAGIC_PAGE + KVM_MAGIC_SCRATCH1)(0); \
53 PPC_STL r30, (KVM_MAGIC_PAGE + KVM_MAGIC_SCRATCH2)(0); \
54 mfcr r31; \
55 stw r31, (KVM_MAGIC_PAGE + KVM_MAGIC_SCRATCH3)(0);
56
57#define SCRATCH_RESTORE \
58 /* Restore state */ \
59 PPC_LL r31, (KVM_MAGIC_PAGE + KVM_MAGIC_SCRATCH1)(0); \
60 lwz r30, (KVM_MAGIC_PAGE + KVM_MAGIC_SCRATCH3)(0); \
61 mtcr r30; \
62 PPC_LL r30, (KVM_MAGIC_PAGE + KVM_MAGIC_SCRATCH2)(0); \
63 \
64 /* Disable critical section. We are critical if \
65 shared->critical == r1 and r2 is always != r1 */ \
66 STL64(r2, KVM_MAGIC_PAGE + KVM_MAGIC_CRITICAL, 0);
67
68.global kvm_emulate_mtmsrd
69kvm_emulate_mtmsrd:
70
71 SCRATCH_SAVE
72
73 /* Put MSR & ~(MSR_EE|MSR_RI) in r31 */
74 LL64(r31, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
75 lis r30, (~(MSR_EE | MSR_RI))@h
76 ori r30, r30, (~(MSR_EE | MSR_RI))@l
77 and r31, r31, r30
78
79 /* OR the register's (MSR_EE|MSR_RI) on MSR */
80kvm_emulate_mtmsrd_reg:
81 ori r30, r0, 0
82 andi. r30, r30, (MSR_EE|MSR_RI)
83 or r31, r31, r30
84
85 /* Put MSR back into magic page */
86 STL64(r31, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
87
88 /* Check if we have to fetch an interrupt */
89 lwz r31, (KVM_MAGIC_PAGE + KVM_MAGIC_INT)(0)
90 cmpwi r31, 0
91 beq+ no_check
92
93 /* Check if we may trigger an interrupt */
94 andi. r30, r30, MSR_EE
95 beq no_check
96
97 SCRATCH_RESTORE
98
99 /* Nag hypervisor */
100kvm_emulate_mtmsrd_orig_ins:
101 tlbsync
102
103 b kvm_emulate_mtmsrd_branch
104
105no_check:
106
107 SCRATCH_RESTORE
108
109 /* Go back to caller */
110kvm_emulate_mtmsrd_branch:
111 b .
112kvm_emulate_mtmsrd_end:
113
114.global kvm_emulate_mtmsrd_branch_offs
115kvm_emulate_mtmsrd_branch_offs:
116 .long (kvm_emulate_mtmsrd_branch - kvm_emulate_mtmsrd) / 4
117
118.global kvm_emulate_mtmsrd_reg_offs
119kvm_emulate_mtmsrd_reg_offs:
120 .long (kvm_emulate_mtmsrd_reg - kvm_emulate_mtmsrd) / 4
121
122.global kvm_emulate_mtmsrd_orig_ins_offs
123kvm_emulate_mtmsrd_orig_ins_offs:
124 .long (kvm_emulate_mtmsrd_orig_ins - kvm_emulate_mtmsrd) / 4
125
126.global kvm_emulate_mtmsrd_len
127kvm_emulate_mtmsrd_len:
128 .long (kvm_emulate_mtmsrd_end - kvm_emulate_mtmsrd) / 4
129
130
131#define MSR_SAFE_BITS (MSR_EE | MSR_CE | MSR_ME | MSR_RI)
132#define MSR_CRITICAL_BITS ~MSR_SAFE_BITS
133
134.global kvm_emulate_mtmsr
135kvm_emulate_mtmsr:
136
137 SCRATCH_SAVE
138
139 /* Fetch old MSR in r31 */
140 LL64(r31, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
141
142 /* Find the changed bits between old and new MSR */
143kvm_emulate_mtmsr_reg1:
144 ori r30, r0, 0
145 xor r31, r30, r31
146
147 /* Check if we need to really do mtmsr */
148 LOAD_REG_IMMEDIATE(r30, MSR_CRITICAL_BITS)
149 and. r31, r31, r30
150
151 /* No critical bits changed? Maybe we can stay in the guest. */
152 beq maybe_stay_in_guest
153
154do_mtmsr:
155
156 SCRATCH_RESTORE
157
158 /* Just fire off the mtmsr if it's critical */
159kvm_emulate_mtmsr_orig_ins:
160 mtmsr r0
161
162 b kvm_emulate_mtmsr_branch
163
164maybe_stay_in_guest:
165
166 /* Get the target register in r30 */
167kvm_emulate_mtmsr_reg2:
168 ori r30, r0, 0
169
170 /* Check if we have to fetch an interrupt */
171 lwz r31, (KVM_MAGIC_PAGE + KVM_MAGIC_INT)(0)
172 cmpwi r31, 0
173 beq+ no_mtmsr
174
175 /* Check if we may trigger an interrupt */
176 andi. r31, r30, MSR_EE
177 beq no_mtmsr
178
179 b do_mtmsr
180
181no_mtmsr:
182
183 /* Put MSR into magic page because we don't call mtmsr */
184 STL64(r30, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
185
186 SCRATCH_RESTORE
187
188 /* Go back to caller */
189kvm_emulate_mtmsr_branch:
190 b .
191kvm_emulate_mtmsr_end:
192
193.global kvm_emulate_mtmsr_branch_offs
194kvm_emulate_mtmsr_branch_offs:
195 .long (kvm_emulate_mtmsr_branch - kvm_emulate_mtmsr) / 4
196
197.global kvm_emulate_mtmsr_reg1_offs
198kvm_emulate_mtmsr_reg1_offs:
199 .long (kvm_emulate_mtmsr_reg1 - kvm_emulate_mtmsr) / 4
200
201.global kvm_emulate_mtmsr_reg2_offs
202kvm_emulate_mtmsr_reg2_offs:
203 .long (kvm_emulate_mtmsr_reg2 - kvm_emulate_mtmsr) / 4
204
205.global kvm_emulate_mtmsr_orig_ins_offs
206kvm_emulate_mtmsr_orig_ins_offs:
207 .long (kvm_emulate_mtmsr_orig_ins - kvm_emulate_mtmsr) / 4
208
209.global kvm_emulate_mtmsr_len
210kvm_emulate_mtmsr_len:
211 .long (kvm_emulate_mtmsr_end - kvm_emulate_mtmsr) / 4
212
213
214
215.global kvm_emulate_wrteei
216kvm_emulate_wrteei:
217
218 SCRATCH_SAVE
219
220 /* Fetch old MSR in r31 */
221 LL64(r31, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
222
223 /* Remove MSR_EE from old MSR */
224 li r30, 0
225 ori r30, r30, MSR_EE
226 andc r31, r31, r30
227
228 /* OR new MSR_EE onto the old MSR */
229kvm_emulate_wrteei_ee:
230 ori r31, r31, 0
231
232 /* Write new MSR value back */
233 STL64(r31, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
234
235 SCRATCH_RESTORE
236
237 /* Go back to caller */
238kvm_emulate_wrteei_branch:
239 b .
240kvm_emulate_wrteei_end:
241
242.global kvm_emulate_wrteei_branch_offs
243kvm_emulate_wrteei_branch_offs:
244 .long (kvm_emulate_wrteei_branch - kvm_emulate_wrteei) / 4
245
246.global kvm_emulate_wrteei_ee_offs
247kvm_emulate_wrteei_ee_offs:
248 .long (kvm_emulate_wrteei_ee - kvm_emulate_wrteei) / 4
249
250.global kvm_emulate_wrteei_len
251kvm_emulate_wrteei_len:
252 .long (kvm_emulate_wrteei_end - kvm_emulate_wrteei) / 4
253
254
255.global kvm_emulate_mtsrin
256kvm_emulate_mtsrin:
257
258 SCRATCH_SAVE
259
260 LL64(r31, KVM_MAGIC_PAGE + KVM_MAGIC_MSR, 0)
261 andi. r31, r31, MSR_DR | MSR_IR
262 beq kvm_emulate_mtsrin_reg1
263
264 SCRATCH_RESTORE
265
266kvm_emulate_mtsrin_orig_ins:
267 nop
268 b kvm_emulate_mtsrin_branch
269
270kvm_emulate_mtsrin_reg1:
271 /* rX >> 26 */
272 rlwinm r30,r0,6,26,29
273
274kvm_emulate_mtsrin_reg2:
275 stw r0, (KVM_MAGIC_PAGE + KVM_MAGIC_SR)(r30)
276
277 SCRATCH_RESTORE
278
279 /* Go back to caller */
280kvm_emulate_mtsrin_branch:
281 b .
282kvm_emulate_mtsrin_end:
283
284.global kvm_emulate_mtsrin_branch_offs
285kvm_emulate_mtsrin_branch_offs:
286 .long (kvm_emulate_mtsrin_branch - kvm_emulate_mtsrin) / 4
287
288.global kvm_emulate_mtsrin_reg1_offs
289kvm_emulate_mtsrin_reg1_offs:
290 .long (kvm_emulate_mtsrin_reg1 - kvm_emulate_mtsrin) / 4
291
292.global kvm_emulate_mtsrin_reg2_offs
293kvm_emulate_mtsrin_reg2_offs:
294 .long (kvm_emulate_mtsrin_reg2 - kvm_emulate_mtsrin) / 4
295
296.global kvm_emulate_mtsrin_orig_ins_offs
297kvm_emulate_mtsrin_orig_ins_offs:
298 .long (kvm_emulate_mtsrin_orig_ins - kvm_emulate_mtsrin) / 4
299
300.global kvm_emulate_mtsrin_len
301kvm_emulate_mtsrin_len:
302 .long (kvm_emulate_mtsrin_end - kvm_emulate_mtsrin) / 4
diff --git a/arch/powerpc/kernel/l2cr_6xx.S b/arch/powerpc/kernel/l2cr_6xx.S
index 2a2f3c3f6d80..97ec8557f974 100644
--- a/arch/powerpc/kernel/l2cr_6xx.S
+++ b/arch/powerpc/kernel/l2cr_6xx.S
@@ -151,7 +151,7 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
151 /**** Might be a good idea to set L2DO here - to prevent instructions 151 /**** Might be a good idea to set L2DO here - to prevent instructions
152 from getting into the cache. But since we invalidate 152 from getting into the cache. But since we invalidate
153 the next time we enable the cache it doesn't really matter. 153 the next time we enable the cache it doesn't really matter.
154 Don't do this unless you accomodate all processor variations. 154 Don't do this unless you accommodate all processor variations.
155 The bit moved on the 7450..... 155 The bit moved on the 7450.....
156 ****/ 156 ****/
157 157
diff --git a/arch/powerpc/kernel/legacy_serial.c b/arch/powerpc/kernel/legacy_serial.c
index c1fd0f9658fd..2b97b80d6d7d 100644
--- a/arch/powerpc/kernel/legacy_serial.c
+++ b/arch/powerpc/kernel/legacy_serial.c
@@ -52,14 +52,14 @@ static int __init add_legacy_port(struct device_node *np, int want_index,
52 phys_addr_t taddr, unsigned long irq, 52 phys_addr_t taddr, unsigned long irq,
53 upf_t flags, int irq_check_parent) 53 upf_t flags, int irq_check_parent)
54{ 54{
55 const u32 *clk, *spd; 55 const __be32 *clk, *spd;
56 u32 clock = BASE_BAUD * 16; 56 u32 clock = BASE_BAUD * 16;
57 int index; 57 int index;
58 58
59 /* get clock freq. if present */ 59 /* get clock freq. if present */
60 clk = of_get_property(np, "clock-frequency", NULL); 60 clk = of_get_property(np, "clock-frequency", NULL);
61 if (clk && *clk) 61 if (clk && *clk)
62 clock = *clk; 62 clock = be32_to_cpup(clk);
63 63
64 /* get default speed if present */ 64 /* get default speed if present */
65 spd = of_get_property(np, "current-speed", NULL); 65 spd = of_get_property(np, "current-speed", NULL);
@@ -109,7 +109,7 @@ static int __init add_legacy_port(struct device_node *np, int want_index,
109 legacy_serial_infos[index].taddr = taddr; 109 legacy_serial_infos[index].taddr = taddr;
110 legacy_serial_infos[index].np = of_node_get(np); 110 legacy_serial_infos[index].np = of_node_get(np);
111 legacy_serial_infos[index].clock = clock; 111 legacy_serial_infos[index].clock = clock;
112 legacy_serial_infos[index].speed = spd ? *spd : 0; 112 legacy_serial_infos[index].speed = spd ? be32_to_cpup(spd) : 0;
113 legacy_serial_infos[index].irq_check_parent = irq_check_parent; 113 legacy_serial_infos[index].irq_check_parent = irq_check_parent;
114 114
115 printk(KERN_DEBUG "Found legacy serial port %d for %s\n", 115 printk(KERN_DEBUG "Found legacy serial port %d for %s\n",
@@ -168,7 +168,7 @@ static int __init add_legacy_soc_port(struct device_node *np,
168static int __init add_legacy_isa_port(struct device_node *np, 168static int __init add_legacy_isa_port(struct device_node *np,
169 struct device_node *isa_brg) 169 struct device_node *isa_brg)
170{ 170{
171 const u32 *reg; 171 const __be32 *reg;
172 const char *typep; 172 const char *typep;
173 int index = -1; 173 int index = -1;
174 u64 taddr; 174 u64 taddr;
@@ -181,7 +181,7 @@ static int __init add_legacy_isa_port(struct device_node *np,
181 return -1; 181 return -1;
182 182
183 /* Verify it's an IO port, we don't support anything else */ 183 /* Verify it's an IO port, we don't support anything else */
184 if (!(reg[0] & 0x00000001)) 184 if (!(be32_to_cpu(reg[0]) & 0x00000001))
185 return -1; 185 return -1;
186 186
187 /* Now look for an "ibm,aix-loc" property that gives us ordering 187 /* Now look for an "ibm,aix-loc" property that gives us ordering
@@ -202,7 +202,7 @@ static int __init add_legacy_isa_port(struct device_node *np,
202 taddr = 0; 202 taddr = 0;
203 203
204 /* Add port, irq will be dealt with later */ 204 /* Add port, irq will be dealt with later */
205 return add_legacy_port(np, index, UPIO_PORT, reg[1], taddr, 205 return add_legacy_port(np, index, UPIO_PORT, be32_to_cpu(reg[1]), taddr,
206 NO_IRQ, UPF_BOOT_AUTOCONF, 0); 206 NO_IRQ, UPF_BOOT_AUTOCONF, 0);
207 207
208} 208}
@@ -251,9 +251,9 @@ static int __init add_legacy_pci_port(struct device_node *np,
251 * we get to their "reg" property 251 * we get to their "reg" property
252 */ 252 */
253 if (np != pci_dev) { 253 if (np != pci_dev) {
254 const u32 *reg = of_get_property(np, "reg", NULL); 254 const __be32 *reg = of_get_property(np, "reg", NULL);
255 if (reg && (*reg < 4)) 255 if (reg && (be32_to_cpup(reg) < 4))
256 index = lindex = *reg; 256 index = lindex = be32_to_cpup(reg);
257 } 257 }
258 258
259 /* Local index means it's the Nth port in the PCI chip. Unfortunately 259 /* Local index means it's the Nth port in the PCI chip. Unfortunately
@@ -330,9 +330,11 @@ void __init find_legacy_serial_ports(void)
330 if (!parent) 330 if (!parent)
331 continue; 331 continue;
332 if (of_match_node(legacy_serial_parents, parent) != NULL) { 332 if (of_match_node(legacy_serial_parents, parent) != NULL) {
333 index = add_legacy_soc_port(np, np); 333 if (of_device_is_available(np)) {
334 if (index >= 0 && np == stdout) 334 index = add_legacy_soc_port(np, np);
335 legacy_serial_console = index; 335 if (index >= 0 && np == stdout)
336 legacy_serial_console = index;
337 }
336 } 338 }
337 of_node_put(parent); 339 of_node_put(parent);
338 } 340 }
@@ -507,7 +509,7 @@ static int __init check_legacy_serial_console(void)
507 struct device_node *prom_stdout = NULL; 509 struct device_node *prom_stdout = NULL;
508 int i, speed = 0, offset = 0; 510 int i, speed = 0, offset = 0;
509 const char *name; 511 const char *name;
510 const u32 *spd; 512 const __be32 *spd;
511 513
512 DBG(" -> check_legacy_serial_console()\n"); 514 DBG(" -> check_legacy_serial_console()\n");
513 515
@@ -547,7 +549,7 @@ static int __init check_legacy_serial_console(void)
547 } 549 }
548 spd = of_get_property(prom_stdout, "current-speed", NULL); 550 spd = of_get_property(prom_stdout, "current-speed", NULL);
549 if (spd) 551 if (spd)
550 speed = *spd; 552 speed = be32_to_cpup(spd);
551 553
552 if (strcmp(name, "serial") != 0) 554 if (strcmp(name, "serial") != 0)
553 goto not_found; 555 goto not_found;
diff --git a/arch/powerpc/kernel/lparcfg.c b/arch/powerpc/kernel/lparcfg.c
index 50362b6ef6e9..84daabe2fcba 100644
--- a/arch/powerpc/kernel/lparcfg.c
+++ b/arch/powerpc/kernel/lparcfg.c
@@ -56,7 +56,7 @@ static unsigned long get_purr(void)
56 56
57 for_each_possible_cpu(cpu) { 57 for_each_possible_cpu(cpu) {
58 if (firmware_has_feature(FW_FEATURE_ISERIES)) 58 if (firmware_has_feature(FW_FEATURE_ISERIES))
59 sum_purr += lppaca[cpu].emulated_time_base; 59 sum_purr += lppaca_of(cpu).emulated_time_base;
60 else { 60 else {
61 struct cpu_usage *cu; 61 struct cpu_usage *cu;
62 62
@@ -132,34 +132,6 @@ static int iseries_lparcfg_data(struct seq_file *m, void *v)
132/* 132/*
133 * Methods used to fetch LPAR data when running on a pSeries platform. 133 * Methods used to fetch LPAR data when running on a pSeries platform.
134 */ 134 */
135/**
136 * h_get_mpp
137 * H_GET_MPP hcall returns info in 7 parms
138 */
139int h_get_mpp(struct hvcall_mpp_data *mpp_data)
140{
141 int rc;
142 unsigned long retbuf[PLPAR_HCALL9_BUFSIZE];
143
144 rc = plpar_hcall9(H_GET_MPP, retbuf);
145
146 mpp_data->entitled_mem = retbuf[0];
147 mpp_data->mapped_mem = retbuf[1];
148
149 mpp_data->group_num = (retbuf[2] >> 2 * 8) & 0xffff;
150 mpp_data->pool_num = retbuf[2] & 0xffff;
151
152 mpp_data->mem_weight = (retbuf[3] >> 7 * 8) & 0xff;
153 mpp_data->unallocated_mem_weight = (retbuf[3] >> 6 * 8) & 0xff;
154 mpp_data->unallocated_entitlement = retbuf[3] & 0xffffffffffff;
155
156 mpp_data->pool_size = retbuf[4];
157 mpp_data->loan_request = retbuf[5];
158 mpp_data->backing_mem = retbuf[6];
159
160 return rc;
161}
162EXPORT_SYMBOL(h_get_mpp);
163 135
164struct hvcall_ppp_data { 136struct hvcall_ppp_data {
165 u64 entitlement; 137 u64 entitlement;
@@ -262,8 +234,8 @@ static void parse_ppp_data(struct seq_file *m)
262 seq_printf(m, "system_active_processors=%d\n", 234 seq_printf(m, "system_active_processors=%d\n",
263 ppp_data.active_system_procs); 235 ppp_data.active_system_procs);
264 236
265 /* pool related entries are apropriate for shared configs */ 237 /* pool related entries are appropriate for shared configs */
266 if (lppaca[0].shared_proc) { 238 if (lppaca_of(0).shared_proc) {
267 unsigned long pool_idle_time, pool_procs; 239 unsigned long pool_idle_time, pool_procs;
268 240
269 seq_printf(m, "pool=%d\n", ppp_data.pool_num); 241 seq_printf(m, "pool=%d\n", ppp_data.pool_num);
@@ -345,6 +317,30 @@ static void parse_mpp_data(struct seq_file *m)
345 seq_printf(m, "backing_memory=%ld bytes\n", mpp_data.backing_mem); 317 seq_printf(m, "backing_memory=%ld bytes\n", mpp_data.backing_mem);
346} 318}
347 319
320/**
321 * parse_mpp_x_data
322 * Parse out data returned from h_get_mpp_x
323 */
324static void parse_mpp_x_data(struct seq_file *m)
325{
326 struct hvcall_mpp_x_data mpp_x_data;
327
328 if (!firmware_has_feature(FW_FEATURE_XCMO))
329 return;
330 if (h_get_mpp_x(&mpp_x_data))
331 return;
332
333 seq_printf(m, "coalesced_bytes=%ld\n", mpp_x_data.coalesced_bytes);
334
335 if (mpp_x_data.pool_coalesced_bytes)
336 seq_printf(m, "pool_coalesced_bytes=%ld\n",
337 mpp_x_data.pool_coalesced_bytes);
338 if (mpp_x_data.pool_purr_cycles)
339 seq_printf(m, "coalesce_pool_purr=%ld\n", mpp_x_data.pool_purr_cycles);
340 if (mpp_x_data.pool_spurr_cycles)
341 seq_printf(m, "coalesce_pool_spurr=%ld\n", mpp_x_data.pool_spurr_cycles);
342}
343
348#define SPLPAR_CHARACTERISTICS_TOKEN 20 344#define SPLPAR_CHARACTERISTICS_TOKEN 20
349#define SPLPAR_MAXLENGTH 1026*(sizeof(char)) 345#define SPLPAR_MAXLENGTH 1026*(sizeof(char))
350 346
@@ -460,8 +456,8 @@ static void pseries_cmo_data(struct seq_file *m)
460 return; 456 return;
461 457
462 for_each_possible_cpu(cpu) { 458 for_each_possible_cpu(cpu) {
463 cmo_faults += lppaca[cpu].cmo_faults; 459 cmo_faults += lppaca_of(cpu).cmo_faults;
464 cmo_fault_time += lppaca[cpu].cmo_fault_time; 460 cmo_fault_time += lppaca_of(cpu).cmo_fault_time;
465 } 461 }
466 462
467 seq_printf(m, "cmo_faults=%lu\n", cmo_faults); 463 seq_printf(m, "cmo_faults=%lu\n", cmo_faults);
@@ -479,8 +475,8 @@ static void splpar_dispatch_data(struct seq_file *m)
479 unsigned long dispatch_dispersions = 0; 475 unsigned long dispatch_dispersions = 0;
480 476
481 for_each_possible_cpu(cpu) { 477 for_each_possible_cpu(cpu) {
482 dispatches += lppaca[cpu].yield_count; 478 dispatches += lppaca_of(cpu).yield_count;
483 dispatch_dispersions += lppaca[cpu].dispersion_count; 479 dispatch_dispersions += lppaca_of(cpu).dispersion_count;
484 } 480 }
485 481
486 seq_printf(m, "dispatches=%lu\n", dispatches); 482 seq_printf(m, "dispatches=%lu\n", dispatches);
@@ -520,6 +516,7 @@ static int pseries_lparcfg_data(struct seq_file *m, void *v)
520 parse_system_parameter_string(m); 516 parse_system_parameter_string(m);
521 parse_ppp_data(m); 517 parse_ppp_data(m);
522 parse_mpp_data(m); 518 parse_mpp_data(m);
519 parse_mpp_x_data(m);
523 pseries_cmo_data(m); 520 pseries_cmo_data(m);
524 splpar_dispatch_data(m); 521 splpar_dispatch_data(m);
525 522
@@ -545,7 +542,7 @@ static int pseries_lparcfg_data(struct seq_file *m, void *v)
545 seq_printf(m, "partition_potential_processors=%d\n", 542 seq_printf(m, "partition_potential_processors=%d\n",
546 partition_potential_processors); 543 partition_potential_processors);
547 544
548 seq_printf(m, "shared_processor_mode=%d\n", lppaca[0].shared_proc); 545 seq_printf(m, "shared_processor_mode=%d\n", lppaca_of(0).shared_proc);
549 546
550 seq_printf(m, "slb_size=%d\n", mmu_slb_size); 547 seq_printf(m, "slb_size=%d\n", mmu_slb_size);
551 548
@@ -780,6 +777,7 @@ static const struct file_operations lparcfg_fops = {
780 .write = lparcfg_write, 777 .write = lparcfg_write,
781 .open = lparcfg_open, 778 .open = lparcfg_open,
782 .release = single_release, 779 .release = single_release,
780 .llseek = seq_lseek,
783}; 781};
784 782
785static int __init lparcfg_init(void) 783static int __init lparcfg_init(void)
diff --git a/arch/powerpc/kernel/machine_kexec.c b/arch/powerpc/kernel/machine_kexec.c
index dd6c141f1662..7ee50f0547cb 100644
--- a/arch/powerpc/kernel/machine_kexec.c
+++ b/arch/powerpc/kernel/machine_kexec.c
@@ -14,16 +14,41 @@
14#include <linux/threads.h> 14#include <linux/threads.h>
15#include <linux/memblock.h> 15#include <linux/memblock.h>
16#include <linux/of.h> 16#include <linux/of.h>
17#include <linux/irq.h>
18#include <linux/ftrace.h>
19
17#include <asm/machdep.h> 20#include <asm/machdep.h>
18#include <asm/prom.h> 21#include <asm/prom.h>
19#include <asm/sections.h> 22#include <asm/sections.h>
20 23
24void machine_kexec_mask_interrupts(void) {
25 unsigned int i;
26
27 for_each_irq(i) {
28 struct irq_desc *desc = irq_to_desc(i);
29 struct irq_chip *chip;
30
31 if (!desc)
32 continue;
33
34 chip = irq_desc_get_chip(desc);
35 if (!chip)
36 continue;
37
38 if (chip->irq_eoi && irqd_irq_inprogress(&desc->irq_data))
39 chip->irq_eoi(&desc->irq_data);
40
41 if (chip->irq_mask)
42 chip->irq_mask(&desc->irq_data);
43
44 if (chip->irq_disable && !irqd_irq_disabled(&desc->irq_data))
45 chip->irq_disable(&desc->irq_data);
46 }
47}
48
21void machine_crash_shutdown(struct pt_regs *regs) 49void machine_crash_shutdown(struct pt_regs *regs)
22{ 50{
23 if (ppc_md.machine_crash_shutdown) 51 default_machine_crash_shutdown(regs);
24 ppc_md.machine_crash_shutdown(regs);
25 else
26 default_machine_crash_shutdown(regs);
27} 52}
28 53
29/* 54/*
@@ -41,8 +66,6 @@ int machine_kexec_prepare(struct kimage *image)
41 66
42void machine_kexec_cleanup(struct kimage *image) 67void machine_kexec_cleanup(struct kimage *image)
43{ 68{
44 if (ppc_md.machine_kexec_cleanup)
45 ppc_md.machine_kexec_cleanup(image);
46} 69}
47 70
48void arch_crash_save_vmcoreinfo(void) 71void arch_crash_save_vmcoreinfo(void)
@@ -63,11 +86,17 @@ void arch_crash_save_vmcoreinfo(void)
63 */ 86 */
64void machine_kexec(struct kimage *image) 87void machine_kexec(struct kimage *image)
65{ 88{
89 int save_ftrace_enabled;
90
91 save_ftrace_enabled = __ftrace_enabled_save();
92
66 if (ppc_md.machine_kexec) 93 if (ppc_md.machine_kexec)
67 ppc_md.machine_kexec(image); 94 ppc_md.machine_kexec(image);
68 else 95 else
69 default_machine_kexec(image); 96 default_machine_kexec(image);
70 97
98 __ftrace_enabled_restore(save_ftrace_enabled);
99
71 /* Fall back to normal restart if we're still alive. */ 100 /* Fall back to normal restart if we're still alive. */
72 machine_restart(NULL); 101 machine_restart(NULL);
73 for(;;); 102 for(;;);
diff --git a/arch/powerpc/kernel/machine_kexec_32.c b/arch/powerpc/kernel/machine_kexec_32.c
index ae63a964b858..e63f2e7d2efb 100644
--- a/arch/powerpc/kernel/machine_kexec_32.c
+++ b/arch/powerpc/kernel/machine_kexec_32.c
@@ -39,6 +39,10 @@ void default_machine_kexec(struct kimage *image)
39 /* Interrupts aren't acceptable while we reboot */ 39 /* Interrupts aren't acceptable while we reboot */
40 local_irq_disable(); 40 local_irq_disable();
41 41
42 /* mask each interrupt so we are in a more sane state for the
43 * kexec kernel */
44 machine_kexec_mask_interrupts();
45
42 page_list = image->head; 46 page_list = image->head;
43 47
44 /* we need both effective and real address here */ 48 /* we need both effective and real address here */
diff --git a/arch/powerpc/kernel/misc.S b/arch/powerpc/kernel/misc.S
index 2d29752cbe16..b69463ec2010 100644
--- a/arch/powerpc/kernel/misc.S
+++ b/arch/powerpc/kernel/misc.S
@@ -122,8 +122,3 @@ _GLOBAL(longjmp)
122 mtlr r0 122 mtlr r0
123 mr r3,r4 123 mr r3,r4
124 blr 124 blr
125
126_GLOBAL(__setup_cpu_power7)
127_GLOBAL(__restore_cpu_power7)
128 /* place holder */
129 blr
diff --git a/arch/powerpc/kernel/misc_32.S b/arch/powerpc/kernel/misc_32.S
index a7a570dcdd57..998a10028608 100644
--- a/arch/powerpc/kernel/misc_32.S
+++ b/arch/powerpc/kernel/misc_32.S
@@ -30,6 +30,7 @@
30#include <asm/processor.h> 30#include <asm/processor.h>
31#include <asm/kexec.h> 31#include <asm/kexec.h>
32#include <asm/bug.h> 32#include <asm/bug.h>
33#include <asm/ptrace.h>
33 34
34 .text 35 .text
35 36
@@ -693,6 +694,17 @@ _GLOBAL(kernel_thread)
693 addi r1,r1,16 694 addi r1,r1,16
694 blr 695 blr
695 696
697#ifdef CONFIG_SMP
698_GLOBAL(start_secondary_resume)
699 /* Reset stack */
700 rlwinm r1,r1,0,0,(31-THREAD_SHIFT) /* current_thread_info() */
701 addi r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
702 li r3,0
703 stw r3,0(r1) /* Zero the stack frame pointer */
704 bl start_secondary
705 b .
706#endif /* CONFIG_SMP */
707
696/* 708/*
697 * This routine is just here to keep GCC happy - sigh... 709 * This routine is just here to keep GCC happy - sigh...
698 */ 710 */
diff --git a/arch/powerpc/kernel/misc_64.S b/arch/powerpc/kernel/misc_64.S
index e5144906a56d..e89df59cdc5a 100644
--- a/arch/powerpc/kernel/misc_64.S
+++ b/arch/powerpc/kernel/misc_64.S
@@ -25,6 +25,7 @@
25#include <asm/cputable.h> 25#include <asm/cputable.h>
26#include <asm/thread_info.h> 26#include <asm/thread_info.h>
27#include <asm/kexec.h> 27#include <asm/kexec.h>
28#include <asm/ptrace.h>
28 29
29 .text 30 .text
30 31
@@ -461,7 +462,8 @@ _GLOBAL(disable_kernel_fp)
461 * wait for the flag to change, indicating this kernel is going away but 462 * wait for the flag to change, indicating this kernel is going away but
462 * the slave code for the next one is at addresses 0 to 100. 463 * the slave code for the next one is at addresses 0 to 100.
463 * 464 *
464 * This is used by all slaves. 465 * This is used by all slaves, even those that did not find a matching
466 * paca in the secondary startup code.
465 * 467 *
466 * Physical (hardware) cpu id should be in r3. 468 * Physical (hardware) cpu id should be in r3.
467 */ 469 */
@@ -470,10 +472,6 @@ _GLOBAL(kexec_wait)
4701: mflr r5 4721: mflr r5
471 addi r5,r5,kexec_flag-1b 473 addi r5,r5,kexec_flag-1b
472 474
473 li r4,KEXEC_STATE_REAL_MODE
474 stb r4,PACAKEXECSTATE(r13)
475 SYNC
476
47799: HMT_LOW 47599: HMT_LOW
478#ifdef CONFIG_KEXEC /* use no memory without kexec */ 476#ifdef CONFIG_KEXEC /* use no memory without kexec */
479 lwz r4,0(r5) 477 lwz r4,0(r5)
@@ -498,11 +496,17 @@ kexec_flag:
498 * 496 *
499 * get phys id from paca 497 * get phys id from paca
500 * switch to real mode 498 * switch to real mode
499 * mark the paca as no longer used
501 * join other cpus in kexec_wait(phys_id) 500 * join other cpus in kexec_wait(phys_id)
502 */ 501 */
503_GLOBAL(kexec_smp_wait) 502_GLOBAL(kexec_smp_wait)
504 lhz r3,PACAHWCPUID(r13) 503 lhz r3,PACAHWCPUID(r13)
505 bl real_mode 504 bl real_mode
505
506 li r4,KEXEC_STATE_REAL_MODE
507 stb r4,PACAKEXECSTATE(r13)
508 SYNC
509
506 b .kexec_wait 510 b .kexec_wait
507 511
508/* 512/*
diff --git a/arch/powerpc/kernel/mpc7450-pmu.c b/arch/powerpc/kernel/mpc7450-pmu.c
index 09d72028f317..2cc5e0301d0b 100644
--- a/arch/powerpc/kernel/mpc7450-pmu.c
+++ b/arch/powerpc/kernel/mpc7450-pmu.c
@@ -414,4 +414,4 @@ static int init_mpc7450_pmu(void)
414 return register_power_pmu(&mpc7450_pmu); 414 return register_power_pmu(&mpc7450_pmu);
415} 415}
416 416
417arch_initcall(init_mpc7450_pmu); 417early_initcall(init_mpc7450_pmu);
diff --git a/arch/powerpc/kernel/nvram_64.c b/arch/powerpc/kernel/nvram_64.c
index 9cf197f01e94..bec1e930ed73 100644
--- a/arch/powerpc/kernel/nvram_64.c
+++ b/arch/powerpc/kernel/nvram_64.c
@@ -34,15 +34,26 @@
34 34
35#undef DEBUG_NVRAM 35#undef DEBUG_NVRAM
36 36
37static struct nvram_partition * nvram_part; 37#define NVRAM_HEADER_LEN sizeof(struct nvram_header)
38static long nvram_error_log_index = -1; 38#define NVRAM_BLOCK_LEN NVRAM_HEADER_LEN
39static long nvram_error_log_size = 0; 39
40/* If change this size, then change the size of NVNAME_LEN */
41struct nvram_header {
42 unsigned char signature;
43 unsigned char checksum;
44 unsigned short length;
45 /* Terminating null required only for names < 12 chars. */
46 char name[12];
47};
40 48
41struct err_log_info { 49struct nvram_partition {
42 int error_type; 50 struct list_head partition;
43 unsigned int seq_num; 51 struct nvram_header header;
52 unsigned int index;
44}; 53};
45 54
55static LIST_HEAD(nvram_partitions);
56
46static loff_t dev_nvram_llseek(struct file *file, loff_t offset, int origin) 57static loff_t dev_nvram_llseek(struct file *file, loff_t offset, int origin)
47{ 58{
48 int size; 59 int size;
@@ -186,14 +197,12 @@ static struct miscdevice nvram_dev = {
186#ifdef DEBUG_NVRAM 197#ifdef DEBUG_NVRAM
187static void __init nvram_print_partitions(char * label) 198static void __init nvram_print_partitions(char * label)
188{ 199{
189 struct list_head * p;
190 struct nvram_partition * tmp_part; 200 struct nvram_partition * tmp_part;
191 201
192 printk(KERN_WARNING "--------%s---------\n", label); 202 printk(KERN_WARNING "--------%s---------\n", label);
193 printk(KERN_WARNING "indx\t\tsig\tchks\tlen\tname\n"); 203 printk(KERN_WARNING "indx\t\tsig\tchks\tlen\tname\n");
194 list_for_each(p, &nvram_part->partition) { 204 list_for_each_entry(tmp_part, &nvram_partitions, partition) {
195 tmp_part = list_entry(p, struct nvram_partition, partition); 205 printk(KERN_WARNING "%4d \t%02x\t%02x\t%d\t%12s\n",
196 printk(KERN_WARNING "%4d \t%02x\t%02x\t%d\t%s\n",
197 tmp_part->index, tmp_part->header.signature, 206 tmp_part->index, tmp_part->header.signature,
198 tmp_part->header.checksum, tmp_part->header.length, 207 tmp_part->header.checksum, tmp_part->header.length,
199 tmp_part->header.name); 208 tmp_part->header.name);
@@ -228,95 +237,136 @@ static unsigned char __init nvram_checksum(struct nvram_header *p)
228 return c_sum; 237 return c_sum;
229} 238}
230 239
231static int __init nvram_remove_os_partition(void) 240/*
241 * Per the criteria passed via nvram_remove_partition(), should this
242 * partition be removed? 1=remove, 0=keep
243 */
244static int nvram_can_remove_partition(struct nvram_partition *part,
245 const char *name, int sig, const char *exceptions[])
246{
247 if (part->header.signature != sig)
248 return 0;
249 if (name) {
250 if (strncmp(name, part->header.name, 12))
251 return 0;
252 } else if (exceptions) {
253 const char **except;
254 for (except = exceptions; *except; except++) {
255 if (!strncmp(*except, part->header.name, 12))
256 return 0;
257 }
258 }
259 return 1;
260}
261
262/**
263 * nvram_remove_partition - Remove one or more partitions in nvram
264 * @name: name of the partition to remove, or NULL for a
265 * signature only match
266 * @sig: signature of the partition(s) to remove
267 * @exceptions: When removing all partitions with a matching signature,
268 * leave these alone.
269 */
270
271int __init nvram_remove_partition(const char *name, int sig,
272 const char *exceptions[])
232{ 273{
233 struct list_head *i; 274 struct nvram_partition *part, *prev, *tmp;
234 struct list_head *j;
235 struct nvram_partition * part;
236 struct nvram_partition * cur_part;
237 int rc; 275 int rc;
238 276
239 list_for_each(i, &nvram_part->partition) { 277 list_for_each_entry(part, &nvram_partitions, partition) {
240 part = list_entry(i, struct nvram_partition, partition); 278 if (!nvram_can_remove_partition(part, name, sig, exceptions))
241 if (part->header.signature != NVRAM_SIG_OS)
242 continue; 279 continue;
243 280
244 /* Make os partition a free partition */ 281 /* Make partition a free partition */
245 part->header.signature = NVRAM_SIG_FREE; 282 part->header.signature = NVRAM_SIG_FREE;
246 sprintf(part->header.name, "wwwwwwwwwwww"); 283 strncpy(part->header.name, "wwwwwwwwwwww", 12);
247 part->header.checksum = nvram_checksum(&part->header); 284 part->header.checksum = nvram_checksum(&part->header);
248
249 /* Merge contiguous free partitions backwards */
250 list_for_each_prev(j, &part->partition) {
251 cur_part = list_entry(j, struct nvram_partition, partition);
252 if (cur_part == nvram_part || cur_part->header.signature != NVRAM_SIG_FREE) {
253 break;
254 }
255
256 part->header.length += cur_part->header.length;
257 part->header.checksum = nvram_checksum(&part->header);
258 part->index = cur_part->index;
259
260 list_del(&cur_part->partition);
261 kfree(cur_part);
262 j = &part->partition; /* fixup our loop */
263 }
264
265 /* Merge contiguous free partitions forwards */
266 list_for_each(j, &part->partition) {
267 cur_part = list_entry(j, struct nvram_partition, partition);
268 if (cur_part == nvram_part || cur_part->header.signature != NVRAM_SIG_FREE) {
269 break;
270 }
271
272 part->header.length += cur_part->header.length;
273 part->header.checksum = nvram_checksum(&part->header);
274
275 list_del(&cur_part->partition);
276 kfree(cur_part);
277 j = &part->partition; /* fixup our loop */
278 }
279
280 rc = nvram_write_header(part); 285 rc = nvram_write_header(part);
281 if (rc <= 0) { 286 if (rc <= 0) {
282 printk(KERN_ERR "nvram_remove_os_partition: nvram_write failed (%d)\n", rc); 287 printk(KERN_ERR "nvram_remove_partition: nvram_write failed (%d)\n", rc);
283 return rc; 288 return rc;
284 } 289 }
290 }
285 291
292 /* Merge contiguous ones */
293 prev = NULL;
294 list_for_each_entry_safe(part, tmp, &nvram_partitions, partition) {
295 if (part->header.signature != NVRAM_SIG_FREE) {
296 prev = NULL;
297 continue;
298 }
299 if (prev) {
300 prev->header.length += part->header.length;
301 prev->header.checksum = nvram_checksum(&part->header);
302 rc = nvram_write_header(part);
303 if (rc <= 0) {
304 printk(KERN_ERR "nvram_remove_partition: nvram_write failed (%d)\n", rc);
305 return rc;
306 }
307 list_del(&part->partition);
308 kfree(part);
309 } else
310 prev = part;
286 } 311 }
287 312
288 return 0; 313 return 0;
289} 314}
290 315
291/* nvram_create_os_partition 316/**
317 * nvram_create_partition - Create a partition in nvram
318 * @name: name of the partition to create
319 * @sig: signature of the partition to create
320 * @req_size: size of data to allocate in bytes
321 * @min_size: minimum acceptable size (0 means req_size)
292 * 322 *
293 * Create a OS linux partition to buffer error logs. 323 * Returns a negative error code or a positive nvram index
294 * Will create a partition starting at the first free 324 * of the beginning of the data area of the newly created
295 * space found if space has enough room. 325 * partition. If you provided a min_size smaller than req_size
326 * you need to query for the actual size yourself after the
327 * call using nvram_partition_get_size().
296 */ 328 */
297static int __init nvram_create_os_partition(void) 329loff_t __init nvram_create_partition(const char *name, int sig,
330 int req_size, int min_size)
298{ 331{
299 struct nvram_partition *part; 332 struct nvram_partition *part;
300 struct nvram_partition *new_part; 333 struct nvram_partition *new_part;
301 struct nvram_partition *free_part = NULL; 334 struct nvram_partition *free_part = NULL;
302 int seq_init[2] = { 0, 0 }; 335 static char nv_init_vals[16];
303 loff_t tmp_index; 336 loff_t tmp_index;
304 long size = 0; 337 long size = 0;
305 int rc; 338 int rc;
306 339
340 /* Convert sizes from bytes to blocks */
341 req_size = _ALIGN_UP(req_size, NVRAM_BLOCK_LEN) / NVRAM_BLOCK_LEN;
342 min_size = _ALIGN_UP(min_size, NVRAM_BLOCK_LEN) / NVRAM_BLOCK_LEN;
343
344 /* If no minimum size specified, make it the same as the
345 * requested size
346 */
347 if (min_size == 0)
348 min_size = req_size;
349 if (min_size > req_size)
350 return -EINVAL;
351
352 /* Now add one block to each for the header */
353 req_size += 1;
354 min_size += 1;
355
307 /* Find a free partition that will give us the maximum needed size 356 /* Find a free partition that will give us the maximum needed size
308 If can't find one that will give us the minimum size needed */ 357 If can't find one that will give us the minimum size needed */
309 list_for_each_entry(part, &nvram_part->partition, partition) { 358 list_for_each_entry(part, &nvram_partitions, partition) {
310 if (part->header.signature != NVRAM_SIG_FREE) 359 if (part->header.signature != NVRAM_SIG_FREE)
311 continue; 360 continue;
312 361
313 if (part->header.length >= NVRAM_MAX_REQ) { 362 if (part->header.length >= req_size) {
314 size = NVRAM_MAX_REQ; 363 size = req_size;
315 free_part = part; 364 free_part = part;
316 break; 365 break;
317 } 366 }
318 if (!size && part->header.length >= NVRAM_MIN_REQ) { 367 if (part->header.length > size &&
319 size = NVRAM_MIN_REQ; 368 part->header.length >= min_size) {
369 size = part->header.length;
320 free_part = part; 370 free_part = part;
321 } 371 }
322 } 372 }
@@ -326,136 +376,95 @@ static int __init nvram_create_os_partition(void)
326 /* Create our OS partition */ 376 /* Create our OS partition */
327 new_part = kmalloc(sizeof(*new_part), GFP_KERNEL); 377 new_part = kmalloc(sizeof(*new_part), GFP_KERNEL);
328 if (!new_part) { 378 if (!new_part) {
329 printk(KERN_ERR "nvram_create_os_partition: kmalloc failed\n"); 379 pr_err("nvram_create_os_partition: kmalloc failed\n");
330 return -ENOMEM; 380 return -ENOMEM;
331 } 381 }
332 382
333 new_part->index = free_part->index; 383 new_part->index = free_part->index;
334 new_part->header.signature = NVRAM_SIG_OS; 384 new_part->header.signature = sig;
335 new_part->header.length = size; 385 new_part->header.length = size;
336 strcpy(new_part->header.name, "ppc64,linux"); 386 strncpy(new_part->header.name, name, 12);
337 new_part->header.checksum = nvram_checksum(&new_part->header); 387 new_part->header.checksum = nvram_checksum(&new_part->header);
338 388
339 rc = nvram_write_header(new_part); 389 rc = nvram_write_header(new_part);
340 if (rc <= 0) { 390 if (rc <= 0) {
341 printk(KERN_ERR "nvram_create_os_partition: nvram_write_header " 391 pr_err("nvram_create_os_partition: nvram_write_header "
342 "failed (%d)\n", rc);
343 return rc;
344 }
345
346 /* make sure and initialize to zero the sequence number and the error
347 type logged */
348 tmp_index = new_part->index + NVRAM_HEADER_LEN;
349 rc = ppc_md.nvram_write((char *)&seq_init, sizeof(seq_init), &tmp_index);
350 if (rc <= 0) {
351 printk(KERN_ERR "nvram_create_os_partition: nvram_write "
352 "failed (%d)\n", rc); 392 "failed (%d)\n", rc);
353 return rc; 393 return rc;
354 } 394 }
355
356 nvram_error_log_index = new_part->index + NVRAM_HEADER_LEN;
357 nvram_error_log_size = ((part->header.length - 1) *
358 NVRAM_BLOCK_LEN) - sizeof(struct err_log_info);
359
360 list_add_tail(&new_part->partition, &free_part->partition); 395 list_add_tail(&new_part->partition, &free_part->partition);
361 396
362 if (free_part->header.length <= size) { 397 /* Adjust or remove the partition we stole the space from */
398 if (free_part->header.length > size) {
399 free_part->index += size * NVRAM_BLOCK_LEN;
400 free_part->header.length -= size;
401 free_part->header.checksum = nvram_checksum(&free_part->header);
402 rc = nvram_write_header(free_part);
403 if (rc <= 0) {
404 pr_err("nvram_create_os_partition: nvram_write_header "
405 "failed (%d)\n", rc);
406 return rc;
407 }
408 } else {
363 list_del(&free_part->partition); 409 list_del(&free_part->partition);
364 kfree(free_part); 410 kfree(free_part);
365 return 0;
366 } 411 }
367 412
368 /* Adjust the partition we stole the space from */ 413 /* Clear the new partition */
369 free_part->index += size * NVRAM_BLOCK_LEN; 414 for (tmp_index = new_part->index + NVRAM_HEADER_LEN;
370 free_part->header.length -= size; 415 tmp_index < ((size - 1) * NVRAM_BLOCK_LEN);
371 free_part->header.checksum = nvram_checksum(&free_part->header); 416 tmp_index += NVRAM_BLOCK_LEN) {
372 417 rc = ppc_md.nvram_write(nv_init_vals, NVRAM_BLOCK_LEN, &tmp_index);
373 rc = nvram_write_header(free_part); 418 if (rc <= 0) {
374 if (rc <= 0) { 419 pr_err("nvram_create_partition: nvram_write failed (%d)\n", rc);
375 printk(KERN_ERR "nvram_create_os_partition: nvram_write_header " 420 return rc;
376 "failed (%d)\n", rc); 421 }
377 return rc;
378 } 422 }
379 423
380 return 0; 424 return new_part->index + NVRAM_HEADER_LEN;
381} 425}
382 426
383 427/**
384/* nvram_setup_partition 428 * nvram_get_partition_size - Get the data size of an nvram partition
385 * 429 * @data_index: This is the offset of the start of the data of
386 * This will setup the partition we need for buffering the 430 * the partition. The same value that is returned by
387 * error logs and cleanup partitions if needed. 431 * nvram_create_partition().
388 *
389 * The general strategy is the following:
390 * 1.) If there is ppc64,linux partition large enough then use it.
391 * 2.) If there is not a ppc64,linux partition large enough, search
392 * for a free partition that is large enough.
393 * 3.) If there is not a free partition large enough remove
394 * _all_ OS partitions and consolidate the space.
395 * 4.) Will first try getting a chunk that will satisfy the maximum
396 * error log size (NVRAM_MAX_REQ).
397 * 5.) If the max chunk cannot be allocated then try finding a chunk
398 * that will satisfy the minum needed (NVRAM_MIN_REQ).
399 */ 432 */
400static int __init nvram_setup_partition(void) 433int nvram_get_partition_size(loff_t data_index)
401{ 434{
402 struct list_head * p; 435 struct nvram_partition *part;
403 struct nvram_partition * part; 436
404 int rc; 437 list_for_each_entry(part, &nvram_partitions, partition) {
405 438 if (part->index + NVRAM_HEADER_LEN == data_index)
406 /* For now, we don't do any of this on pmac, until I 439 return (part->header.length - 1) * NVRAM_BLOCK_LEN;
407 * have figured out if it's worth killing some unused stuffs 440 }
408 * in our nvram, as Apple defined partitions use pretty much 441 return -1;
409 * all of the space 442}
410 */
411 if (machine_is(powermac))
412 return -ENOSPC;
413
414 /* see if we have an OS partition that meets our needs.
415 will try getting the max we need. If not we'll delete
416 partitions and try again. */
417 list_for_each(p, &nvram_part->partition) {
418 part = list_entry(p, struct nvram_partition, partition);
419 if (part->header.signature != NVRAM_SIG_OS)
420 continue;
421 443
422 if (strcmp(part->header.name, "ppc64,linux"))
423 continue;
424 444
425 if (part->header.length >= NVRAM_MIN_REQ) { 445/**
426 /* found our partition */ 446 * nvram_find_partition - Find an nvram partition by signature and name
427 nvram_error_log_index = part->index + NVRAM_HEADER_LEN; 447 * @name: Name of the partition or NULL for any name
428 nvram_error_log_size = ((part->header.length - 1) * 448 * @sig: Signature to test against
429 NVRAM_BLOCK_LEN) - sizeof(struct err_log_info); 449 * @out_size: if non-NULL, returns the size of the data part of the partition
430 return 0; 450 */
451loff_t nvram_find_partition(const char *name, int sig, int *out_size)
452{
453 struct nvram_partition *p;
454
455 list_for_each_entry(p, &nvram_partitions, partition) {
456 if (p->header.signature == sig &&
457 (!name || !strncmp(p->header.name, name, 12))) {
458 if (out_size)
459 *out_size = (p->header.length - 1) *
460 NVRAM_BLOCK_LEN;
461 return p->index + NVRAM_HEADER_LEN;
431 } 462 }
432 } 463 }
433
434 /* try creating a partition with the free space we have */
435 rc = nvram_create_os_partition();
436 if (!rc) {
437 return 0;
438 }
439
440 /* need to free up some space */
441 rc = nvram_remove_os_partition();
442 if (rc) {
443 return rc;
444 }
445
446 /* create a partition in this new space */
447 rc = nvram_create_os_partition();
448 if (rc) {
449 printk(KERN_ERR "nvram_create_os_partition: Could not find a "
450 "NVRAM partition large enough\n");
451 return rc;
452 }
453
454 return 0; 464 return 0;
455} 465}
456 466
457 467int __init nvram_scan_partitions(void)
458static int __init nvram_scan_partitions(void)
459{ 468{
460 loff_t cur_index = 0; 469 loff_t cur_index = 0;
461 struct nvram_header phead; 470 struct nvram_header phead;
@@ -465,7 +474,7 @@ static int __init nvram_scan_partitions(void)
465 int total_size; 474 int total_size;
466 int err; 475 int err;
467 476
468 if (ppc_md.nvram_size == NULL) 477 if (ppc_md.nvram_size == NULL || ppc_md.nvram_size() <= 0)
469 return -ENODEV; 478 return -ENODEV;
470 total_size = ppc_md.nvram_size(); 479 total_size = ppc_md.nvram_size();
471 480
@@ -512,12 +521,16 @@ static int __init nvram_scan_partitions(void)
512 521
513 memcpy(&tmp_part->header, &phead, NVRAM_HEADER_LEN); 522 memcpy(&tmp_part->header, &phead, NVRAM_HEADER_LEN);
514 tmp_part->index = cur_index; 523 tmp_part->index = cur_index;
515 list_add_tail(&tmp_part->partition, &nvram_part->partition); 524 list_add_tail(&tmp_part->partition, &nvram_partitions);
516 525
517 cur_index += phead.length * NVRAM_BLOCK_LEN; 526 cur_index += phead.length * NVRAM_BLOCK_LEN;
518 } 527 }
519 err = 0; 528 err = 0;
520 529
530#ifdef DEBUG_NVRAM
531 nvram_print_partitions("NVRAM Partitions");
532#endif
533
521 out: 534 out:
522 kfree(header); 535 kfree(header);
523 return err; 536 return err;
@@ -525,9 +538,10 @@ static int __init nvram_scan_partitions(void)
525 538
526static int __init nvram_init(void) 539static int __init nvram_init(void)
527{ 540{
528 int error;
529 int rc; 541 int rc;
530 542
543 BUILD_BUG_ON(NVRAM_BLOCK_LEN != 16);
544
531 if (ppc_md.nvram_size == NULL || ppc_md.nvram_size() <= 0) 545 if (ppc_md.nvram_size == NULL || ppc_md.nvram_size() <= 0)
532 return -ENODEV; 546 return -ENODEV;
533 547
@@ -537,29 +551,6 @@ static int __init nvram_init(void)
537 return rc; 551 return rc;
538 } 552 }
539 553
540 /* initialize our anchor for the nvram partition list */
541 nvram_part = kmalloc(sizeof(struct nvram_partition), GFP_KERNEL);
542 if (!nvram_part) {
543 printk(KERN_ERR "nvram_init: Failed kmalloc\n");
544 return -ENOMEM;
545 }
546 INIT_LIST_HEAD(&nvram_part->partition);
547
548 /* Get all the NVRAM partitions */
549 error = nvram_scan_partitions();
550 if (error) {
551 printk(KERN_ERR "nvram_init: Failed nvram_scan_partitions\n");
552 return error;
553 }
554
555 if(nvram_setup_partition())
556 printk(KERN_WARNING "nvram_init: Could not find nvram partition"
557 " for nvram buffered error logging.\n");
558
559#ifdef DEBUG_NVRAM
560 nvram_print_partitions("NVRAM Partitions");
561#endif
562
563 return rc; 554 return rc;
564} 555}
565 556
@@ -568,135 +559,6 @@ void __exit nvram_cleanup(void)
568 misc_deregister( &nvram_dev ); 559 misc_deregister( &nvram_dev );
569} 560}
570 561
571
572#ifdef CONFIG_PPC_PSERIES
573
574/* nvram_write_error_log
575 *
576 * We need to buffer the error logs into nvram to ensure that we have
577 * the failure information to decode. If we have a severe error there
578 * is no way to guarantee that the OS or the machine is in a state to
579 * get back to user land and write the error to disk. For example if
580 * the SCSI device driver causes a Machine Check by writing to a bad
581 * IO address, there is no way of guaranteeing that the device driver
582 * is in any state that is would also be able to write the error data
583 * captured to disk, thus we buffer it in NVRAM for analysis on the
584 * next boot.
585 *
586 * In NVRAM the partition containing the error log buffer will looks like:
587 * Header (in bytes):
588 * +-----------+----------+--------+------------+------------------+
589 * | signature | checksum | length | name | data |
590 * |0 |1 |2 3|4 15|16 length-1|
591 * +-----------+----------+--------+------------+------------------+
592 *
593 * The 'data' section would look like (in bytes):
594 * +--------------+------------+-----------------------------------+
595 * | event_logged | sequence # | error log |
596 * |0 3|4 7|8 nvram_error_log_size-1|
597 * +--------------+------------+-----------------------------------+
598 *
599 * event_logged: 0 if event has not been logged to syslog, 1 if it has
600 * sequence #: The unique sequence # for each event. (until it wraps)
601 * error log: The error log from event_scan
602 */
603int nvram_write_error_log(char * buff, int length,
604 unsigned int err_type, unsigned int error_log_cnt)
605{
606 int rc;
607 loff_t tmp_index;
608 struct err_log_info info;
609
610 if (nvram_error_log_index == -1) {
611 return -ESPIPE;
612 }
613
614 if (length > nvram_error_log_size) {
615 length = nvram_error_log_size;
616 }
617
618 info.error_type = err_type;
619 info.seq_num = error_log_cnt;
620
621 tmp_index = nvram_error_log_index;
622
623 rc = ppc_md.nvram_write((char *)&info, sizeof(struct err_log_info), &tmp_index);
624 if (rc <= 0) {
625 printk(KERN_ERR "nvram_write_error_log: Failed nvram_write (%d)\n", rc);
626 return rc;
627 }
628
629 rc = ppc_md.nvram_write(buff, length, &tmp_index);
630 if (rc <= 0) {
631 printk(KERN_ERR "nvram_write_error_log: Failed nvram_write (%d)\n", rc);
632 return rc;
633 }
634
635 return 0;
636}
637
638/* nvram_read_error_log
639 *
640 * Reads nvram for error log for at most 'length'
641 */
642int nvram_read_error_log(char * buff, int length,
643 unsigned int * err_type, unsigned int * error_log_cnt)
644{
645 int rc;
646 loff_t tmp_index;
647 struct err_log_info info;
648
649 if (nvram_error_log_index == -1)
650 return -1;
651
652 if (length > nvram_error_log_size)
653 length = nvram_error_log_size;
654
655 tmp_index = nvram_error_log_index;
656
657 rc = ppc_md.nvram_read((char *)&info, sizeof(struct err_log_info), &tmp_index);
658 if (rc <= 0) {
659 printk(KERN_ERR "nvram_read_error_log: Failed nvram_read (%d)\n", rc);
660 return rc;
661 }
662
663 rc = ppc_md.nvram_read(buff, length, &tmp_index);
664 if (rc <= 0) {
665 printk(KERN_ERR "nvram_read_error_log: Failed nvram_read (%d)\n", rc);
666 return rc;
667 }
668
669 *error_log_cnt = info.seq_num;
670 *err_type = info.error_type;
671
672 return 0;
673}
674
675/* This doesn't actually zero anything, but it sets the event_logged
676 * word to tell that this event is safely in syslog.
677 */
678int nvram_clear_error_log(void)
679{
680 loff_t tmp_index;
681 int clear_word = ERR_FLAG_ALREADY_LOGGED;
682 int rc;
683
684 if (nvram_error_log_index == -1)
685 return -1;
686
687 tmp_index = nvram_error_log_index;
688
689 rc = ppc_md.nvram_write((char *)&clear_word, sizeof(int), &tmp_index);
690 if (rc <= 0) {
691 printk(KERN_ERR "nvram_clear_error_log: Failed nvram_write (%d)\n", rc);
692 return rc;
693 }
694
695 return 0;
696}
697
698#endif /* CONFIG_PPC_PSERIES */
699
700module_init(nvram_init); 562module_init(nvram_init);
701module_exit(nvram_cleanup); 563module_exit(nvram_cleanup);
702MODULE_LICENSE("GPL"); 564MODULE_LICENSE("GPL");
diff --git a/arch/powerpc/kernel/of_platform.c b/arch/powerpc/kernel/of_platform.c
index b2c363ef38ad..24582181b6ec 100644
--- a/arch/powerpc/kernel/of_platform.c
+++ b/arch/powerpc/kernel/of_platform.c
@@ -36,8 +36,7 @@
36 * lacking some bits needed here. 36 * lacking some bits needed here.
37 */ 37 */
38 38
39static int __devinit of_pci_phb_probe(struct platform_device *dev, 39static int __devinit of_pci_phb_probe(struct platform_device *dev)
40 const struct of_device_id *match)
41{ 40{
42 struct pci_controller *phb; 41 struct pci_controller *phb;
43 42
@@ -74,7 +73,7 @@ static int __devinit of_pci_phb_probe(struct platform_device *dev,
74#endif /* CONFIG_EEH */ 73#endif /* CONFIG_EEH */
75 74
76 /* Scan the bus */ 75 /* Scan the bus */
77 pcibios_scan_phb(phb, dev->dev.of_node); 76 pcibios_scan_phb(phb);
78 if (phb->bus == NULL) 77 if (phb->bus == NULL)
79 return -ENXIO; 78 return -ENXIO;
80 79
@@ -104,7 +103,7 @@ static struct of_device_id of_pci_phb_ids[] = {
104 {} 103 {}
105}; 104};
106 105
107static struct of_platform_driver of_pci_phb_driver = { 106static struct platform_driver of_pci_phb_driver = {
108 .probe = of_pci_phb_probe, 107 .probe = of_pci_phb_probe,
109 .driver = { 108 .driver = {
110 .name = "of-pci", 109 .name = "of-pci",
@@ -115,7 +114,7 @@ static struct of_platform_driver of_pci_phb_driver = {
115 114
116static __init int of_pci_phb_init(void) 115static __init int of_pci_phb_init(void)
117{ 116{
118 return of_register_platform_driver(&of_pci_phb_driver); 117 return platform_driver_register(&of_pci_phb_driver);
119} 118}
120 119
121device_initcall(of_pci_phb_init); 120device_initcall(of_pci_phb_init);
diff --git a/arch/powerpc/kernel/paca.c b/arch/powerpc/kernel/paca.c
index d0a26f1770fe..efeb88184182 100644
--- a/arch/powerpc/kernel/paca.c
+++ b/arch/powerpc/kernel/paca.c
@@ -7,7 +7,7 @@
7 * 2 of the License, or (at your option) any later version. 7 * 2 of the License, or (at your option) any later version.
8 */ 8 */
9 9
10#include <linux/threads.h> 10#include <linux/smp.h>
11#include <linux/module.h> 11#include <linux/module.h>
12#include <linux/memblock.h> 12#include <linux/memblock.h>
13 13
@@ -36,7 +36,7 @@ extern unsigned long __toc_start;
36 * will suffice to ensure that it doesn't cross a page boundary. 36 * will suffice to ensure that it doesn't cross a page boundary.
37 */ 37 */
38struct lppaca lppaca[] = { 38struct lppaca lppaca[] = {
39 [0 ... (NR_CPUS-1)] = { 39 [0 ... (NR_LPPACAS-1)] = {
40 .desc = 0xd397d781, /* "LpPa" */ 40 .desc = 0xd397d781, /* "LpPa" */
41 .size = sizeof(struct lppaca), 41 .size = sizeof(struct lppaca),
42 .dyn_proc_status = 2, 42 .dyn_proc_status = 2,
@@ -49,6 +49,54 @@ struct lppaca lppaca[] = {
49 }, 49 },
50}; 50};
51 51
52static struct lppaca *extra_lppacas;
53static long __initdata lppaca_size;
54
55static void allocate_lppacas(int nr_cpus, unsigned long limit)
56{
57 if (nr_cpus <= NR_LPPACAS)
58 return;
59
60 lppaca_size = PAGE_ALIGN(sizeof(struct lppaca) *
61 (nr_cpus - NR_LPPACAS));
62 extra_lppacas = __va(memblock_alloc_base(lppaca_size,
63 PAGE_SIZE, limit));
64}
65
66static struct lppaca *new_lppaca(int cpu)
67{
68 struct lppaca *lp;
69
70 if (cpu < NR_LPPACAS)
71 return &lppaca[cpu];
72
73 lp = extra_lppacas + (cpu - NR_LPPACAS);
74 *lp = lppaca[0];
75
76 return lp;
77}
78
79static void free_lppacas(void)
80{
81 long new_size = 0, nr;
82
83 if (!lppaca_size)
84 return;
85 nr = num_possible_cpus() - NR_LPPACAS;
86 if (nr > 0)
87 new_size = PAGE_ALIGN(nr * sizeof(struct lppaca));
88 if (new_size >= lppaca_size)
89 return;
90
91 memblock_free(__pa(extra_lppacas) + new_size, lppaca_size - new_size);
92 lppaca_size = new_size;
93}
94
95#else
96
97static inline void allocate_lppacas(int nr_cpus, unsigned long limit) { }
98static inline void free_lppacas(void) { }
99
52#endif /* CONFIG_PPC_BOOK3S */ 100#endif /* CONFIG_PPC_BOOK3S */
53 101
54#ifdef CONFIG_PPC_STD_MMU_64 102#ifdef CONFIG_PPC_STD_MMU_64
@@ -88,7 +136,7 @@ void __init initialise_paca(struct paca_struct *new_paca, int cpu)
88 unsigned long kernel_toc = (unsigned long)(&__toc_start) + 0x8000UL; 136 unsigned long kernel_toc = (unsigned long)(&__toc_start) + 0x8000UL;
89 137
90#ifdef CONFIG_PPC_BOOK3S 138#ifdef CONFIG_PPC_BOOK3S
91 new_paca->lppaca_ptr = &lppaca[cpu]; 139 new_paca->lppaca_ptr = new_lppaca(cpu);
92#else 140#else
93 new_paca->kernel_pgd = swapper_pg_dir; 141 new_paca->kernel_pgd = swapper_pg_dir;
94#endif 142#endif
@@ -108,18 +156,29 @@ void __init initialise_paca(struct paca_struct *new_paca, int cpu)
108/* Put the paca pointer into r13 and SPRG_PACA */ 156/* Put the paca pointer into r13 and SPRG_PACA */
109void setup_paca(struct paca_struct *new_paca) 157void setup_paca(struct paca_struct *new_paca)
110{ 158{
159 /* Setup r13 */
111 local_paca = new_paca; 160 local_paca = new_paca;
112 mtspr(SPRN_SPRG_PACA, local_paca); 161
113#ifdef CONFIG_PPC_BOOK3E 162#ifdef CONFIG_PPC_BOOK3E
163 /* On Book3E, initialize the TLB miss exception frames */
114 mtspr(SPRN_SPRG_TLB_EXFRAME, local_paca->extlb); 164 mtspr(SPRN_SPRG_TLB_EXFRAME, local_paca->extlb);
165#else
166 /* In HV mode, we setup both HPACA and PACA to avoid problems
167 * if we do a GET_PACA() before the feature fixups have been
168 * applied
169 */
170 if (cpu_has_feature(CPU_FTR_HVMODE_206))
171 mtspr(SPRN_SPRG_HPACA, local_paca);
115#endif 172#endif
173 mtspr(SPRN_SPRG_PACA, local_paca);
174
116} 175}
117 176
118static int __initdata paca_size; 177static int __initdata paca_size;
119 178
120void __init allocate_pacas(void) 179void __init allocate_pacas(void)
121{ 180{
122 int nr_cpus, cpu, limit; 181 int cpu, limit;
123 182
124 /* 183 /*
125 * We can't take SLB misses on the paca, and we want to access them 184 * We can't take SLB misses on the paca, and we want to access them
@@ -127,25 +186,22 @@ void __init allocate_pacas(void)
127 * the first segment. On iSeries they must be within the area mapped 186 * the first segment. On iSeries they must be within the area mapped
128 * by the HV, which is HvPagesToMap * HVPAGESIZE bytes. 187 * by the HV, which is HvPagesToMap * HVPAGESIZE bytes.
129 */ 188 */
130 limit = min(0x10000000ULL, memblock.rmo_size); 189 limit = min(0x10000000ULL, ppc64_rma_size);
131 if (firmware_has_feature(FW_FEATURE_ISERIES)) 190 if (firmware_has_feature(FW_FEATURE_ISERIES))
132 limit = min(limit, HvPagesToMap * HVPAGESIZE); 191 limit = min(limit, HvPagesToMap * HVPAGESIZE);
133 192
134 nr_cpus = NR_CPUS; 193 paca_size = PAGE_ALIGN(sizeof(struct paca_struct) * nr_cpu_ids);
135 /* On iSeries we know we can never have more than 64 cpus */
136 if (firmware_has_feature(FW_FEATURE_ISERIES))
137 nr_cpus = min(64, nr_cpus);
138
139 paca_size = PAGE_ALIGN(sizeof(struct paca_struct) * nr_cpus);
140 194
141 paca = __va(memblock_alloc_base(paca_size, PAGE_SIZE, limit)); 195 paca = __va(memblock_alloc_base(paca_size, PAGE_SIZE, limit));
142 memset(paca, 0, paca_size); 196 memset(paca, 0, paca_size);
143 197
144 printk(KERN_DEBUG "Allocated %u bytes for %d pacas at %p\n", 198 printk(KERN_DEBUG "Allocated %u bytes for %d pacas at %p\n",
145 paca_size, nr_cpus, paca); 199 paca_size, nr_cpu_ids, paca);
200
201 allocate_lppacas(nr_cpu_ids, limit);
146 202
147 /* Can't use for_each_*_cpu, as they aren't functional yet */ 203 /* Can't use for_each_*_cpu, as they aren't functional yet */
148 for (cpu = 0; cpu < nr_cpus; cpu++) 204 for (cpu = 0; cpu < nr_cpu_ids; cpu++)
149 initialise_paca(&paca[cpu], cpu); 205 initialise_paca(&paca[cpu], cpu);
150} 206}
151 207
@@ -153,7 +209,7 @@ void __init free_unused_pacas(void)
153{ 209{
154 int new_size; 210 int new_size;
155 211
156 new_size = PAGE_ALIGN(sizeof(struct paca_struct) * num_possible_cpus()); 212 new_size = PAGE_ALIGN(sizeof(struct paca_struct) * nr_cpu_ids);
157 213
158 if (new_size >= paca_size) 214 if (new_size >= paca_size)
159 return; 215 return;
@@ -164,4 +220,6 @@ void __init free_unused_pacas(void)
164 paca_size - new_size); 220 paca_size - new_size);
165 221
166 paca_size = new_size; 222 paca_size = new_size;
223
224 free_lppacas();
167} 225}
diff --git a/arch/powerpc/kernel/pci-common.c b/arch/powerpc/kernel/pci-common.c
index 9021c4ad4bbd..893af2a9cd03 100644
--- a/arch/powerpc/kernel/pci-common.c
+++ b/arch/powerpc/kernel/pci-common.c
@@ -22,6 +22,7 @@
22#include <linux/init.h> 22#include <linux/init.h>
23#include <linux/bootmem.h> 23#include <linux/bootmem.h>
24#include <linux/of_address.h> 24#include <linux/of_address.h>
25#include <linux/of_pci.h>
25#include <linux/mm.h> 26#include <linux/mm.h>
26#include <linux/list.h> 27#include <linux/list.h>
27#include <linux/syscalls.h> 28#include <linux/syscalls.h>
@@ -260,7 +261,7 @@ int pci_read_irq_line(struct pci_dev *pci_dev)
260 261
261 virq = irq_create_mapping(NULL, line); 262 virq = irq_create_mapping(NULL, line);
262 if (virq != NO_IRQ) 263 if (virq != NO_IRQ)
263 set_irq_type(virq, IRQ_TYPE_LEVEL_LOW); 264 irq_set_irq_type(virq, IRQ_TYPE_LEVEL_LOW);
264 } else { 265 } else {
265 pr_debug(" Got one, spec %d cells (0x%08x 0x%08x...) on %s\n", 266 pr_debug(" Got one, spec %d cells (0x%08x 0x%08x...) on %s\n",
266 oirq.size, oirq.specifier[0], oirq.specifier[1], 267 oirq.size, oirq.specifier[0], oirq.specifier[1],
@@ -1090,8 +1091,6 @@ void __devinit pcibios_setup_bus_devices(struct pci_bus *bus)
1090 bus->number, bus->self ? pci_name(bus->self) : "PHB"); 1091 bus->number, bus->self ? pci_name(bus->self) : "PHB");
1091 1092
1092 list_for_each_entry(dev, &bus->devices, bus_list) { 1093 list_for_each_entry(dev, &bus->devices, bus_list) {
1093 struct dev_archdata *sd = &dev->dev.archdata;
1094
1095 /* Cardbus can call us to add new devices to a bus, so ignore 1094 /* Cardbus can call us to add new devices to a bus, so ignore
1096 * those who are already fully discovered 1095 * those who are already fully discovered
1097 */ 1096 */
@@ -1107,7 +1106,7 @@ void __devinit pcibios_setup_bus_devices(struct pci_bus *bus)
1107 set_dev_node(&dev->dev, pcibus_to_node(dev->bus)); 1106 set_dev_node(&dev->dev, pcibus_to_node(dev->bus));
1108 1107
1109 /* Hook up default DMA ops */ 1108 /* Hook up default DMA ops */
1110 sd->dma_ops = pci_dma_ops; 1109 set_dma_ops(&dev->dev, pci_dma_ops);
1111 set_dma_offset(&dev->dev, PCI_DRAM_OFFSET); 1110 set_dma_offset(&dev->dev, PCI_DRAM_OFFSET);
1112 1111
1113 /* Additional platform DMA/iommu setup */ 1112 /* Additional platform DMA/iommu setup */
@@ -1689,13 +1688,8 @@ int early_find_capability(struct pci_controller *hose, int bus, int devfn,
1689/** 1688/**
1690 * pci_scan_phb - Given a pci_controller, setup and scan the PCI bus 1689 * pci_scan_phb - Given a pci_controller, setup and scan the PCI bus
1691 * @hose: Pointer to the PCI host controller instance structure 1690 * @hose: Pointer to the PCI host controller instance structure
1692 * @sysdata: value to use for sysdata pointer. ppc32 and ppc64 differ here
1693 *
1694 * Note: the 'data' pointer is a temporary measure. As 32 and 64 bit
1695 * pci code gets merged, this parameter should become unnecessary because
1696 * both will use the same value.
1697 */ 1691 */
1698void __devinit pcibios_scan_phb(struct pci_controller *hose, void *sysdata) 1692void __devinit pcibios_scan_phb(struct pci_controller *hose)
1699{ 1693{
1700 struct pci_bus *bus; 1694 struct pci_bus *bus;
1701 struct device_node *node = hose->dn; 1695 struct device_node *node = hose->dn;
@@ -1705,13 +1699,13 @@ void __devinit pcibios_scan_phb(struct pci_controller *hose, void *sysdata)
1705 node ? node->full_name : "<NO NAME>"); 1699 node ? node->full_name : "<NO NAME>");
1706 1700
1707 /* Create an empty bus for the toplevel */ 1701 /* Create an empty bus for the toplevel */
1708 bus = pci_create_bus(hose->parent, hose->first_busno, hose->ops, 1702 bus = pci_create_bus(hose->parent, hose->first_busno, hose->ops, hose);
1709 sysdata);
1710 if (bus == NULL) { 1703 if (bus == NULL) {
1711 pr_err("Failed to create bus for PCI domain %04x\n", 1704 pr_err("Failed to create bus for PCI domain %04x\n",
1712 hose->global_number); 1705 hose->global_number);
1713 return; 1706 return;
1714 } 1707 }
1708 bus->dev.of_node = of_node_get(node);
1715 bus->secondary = hose->first_busno; 1709 bus->secondary = hose->first_busno;
1716 hose->bus = bus; 1710 hose->bus = bus;
1717 1711
diff --git a/arch/powerpc/kernel/pci_32.c b/arch/powerpc/kernel/pci_32.c
index e7db5b48004a..bedb370459f2 100644
--- a/arch/powerpc/kernel/pci_32.c
+++ b/arch/powerpc/kernel/pci_32.c
@@ -381,7 +381,7 @@ static int __init pcibios_init(void)
381 if (pci_assign_all_buses) 381 if (pci_assign_all_buses)
382 hose->first_busno = next_busno; 382 hose->first_busno = next_busno;
383 hose->last_busno = 0xff; 383 hose->last_busno = 0xff;
384 pcibios_scan_phb(hose, hose); 384 pcibios_scan_phb(hose);
385 pci_bus_add_devices(hose->bus); 385 pci_bus_add_devices(hose->bus);
386 if (pci_assign_all_buses || next_busno <= hose->last_busno) 386 if (pci_assign_all_buses || next_busno <= hose->last_busno)
387 next_busno = hose->last_busno + pcibios_assign_bus_offset; 387 next_busno = hose->last_busno + pcibios_assign_bus_offset;
diff --git a/arch/powerpc/kernel/pci_64.c b/arch/powerpc/kernel/pci_64.c
index d43fc65749c1..fc6452b6be9f 100644
--- a/arch/powerpc/kernel/pci_64.c
+++ b/arch/powerpc/kernel/pci_64.c
@@ -64,7 +64,7 @@ static int __init pcibios_init(void)
64 64
65 /* Scan all of the recorded PCI controllers. */ 65 /* Scan all of the recorded PCI controllers. */
66 list_for_each_entry_safe(hose, tmp, &hose_list, list_node) { 66 list_for_each_entry_safe(hose, tmp, &hose_list, list_node) {
67 pcibios_scan_phb(hose, hose->dn); 67 pcibios_scan_phb(hose);
68 pci_bus_add_devices(hose->bus); 68 pci_bus_add_devices(hose->bus);
69 } 69 }
70 70
@@ -193,8 +193,7 @@ int __devinit pcibios_map_io_space(struct pci_bus *bus)
193 hose->io_resource.start += io_virt_offset; 193 hose->io_resource.start += io_virt_offset;
194 hose->io_resource.end += io_virt_offset; 194 hose->io_resource.end += io_virt_offset;
195 195
196 pr_debug(" hose->io_resource=0x%016llx...0x%016llx\n", 196 pr_debug(" hose->io_resource=%pR\n", &hose->io_resource);
197 hose->io_resource.start, hose->io_resource.end);
198 197
199 return 0; 198 return 0;
200} 199}
@@ -243,10 +242,10 @@ long sys_pciconfig_iobase(long which, unsigned long in_bus,
243 break; 242 break;
244 bus = NULL; 243 bus = NULL;
245 } 244 }
246 if (bus == NULL || bus->sysdata == NULL) 245 if (bus == NULL || bus->dev.of_node == NULL)
247 return -ENODEV; 246 return -ENODEV;
248 247
249 hose_node = (struct device_node *)bus->sysdata; 248 hose_node = bus->dev.of_node;
250 hose = PCI_DN(hose_node)->phb; 249 hose = PCI_DN(hose_node)->phb;
251 250
252 switch (which) { 251 switch (which) {
diff --git a/arch/powerpc/kernel/pci_dn.c b/arch/powerpc/kernel/pci_dn.c
index d56b35ee7f74..6baabc13306a 100644
--- a/arch/powerpc/kernel/pci_dn.c
+++ b/arch/powerpc/kernel/pci_dn.c
@@ -43,10 +43,9 @@ void * __devinit update_dn_pci_info(struct device_node *dn, void *data)
43 const u32 *regs; 43 const u32 *regs;
44 struct pci_dn *pdn; 44 struct pci_dn *pdn;
45 45
46 pdn = alloc_maybe_bootmem(sizeof(*pdn), GFP_KERNEL); 46 pdn = zalloc_maybe_bootmem(sizeof(*pdn), GFP_KERNEL);
47 if (pdn == NULL) 47 if (pdn == NULL)
48 return NULL; 48 return NULL;
49 memset(pdn, 0, sizeof(*pdn));
50 dn->data = pdn; 49 dn->data = pdn;
51 pdn->node = dn; 50 pdn->node = dn;
52 pdn->phb = phb; 51 pdn->phb = phb;
@@ -161,7 +160,7 @@ static void *is_devfn_node(struct device_node *dn, void *data)
161/* 160/*
162 * This is the "slow" path for looking up a device_node from a 161 * This is the "slow" path for looking up a device_node from a
163 * pci_dev. It will hunt for the device under its parent's 162 * pci_dev. It will hunt for the device under its parent's
164 * phb and then update sysdata for a future fastpath. 163 * phb and then update of_node pointer.
165 * 164 *
166 * It may also do fixups on the actual device since this happens 165 * It may also do fixups on the actual device since this happens
167 * on the first read/write. 166 * on the first read/write.
@@ -170,16 +169,22 @@ static void *is_devfn_node(struct device_node *dn, void *data)
170 * In this case it may probe for real hardware ("just in case") 169 * In this case it may probe for real hardware ("just in case")
171 * and add a device_node to the device tree if necessary. 170 * and add a device_node to the device tree if necessary.
172 * 171 *
172 * Is this function necessary anymore now that dev->dev.of_node is
173 * used to store the node pointer?
174 *
173 */ 175 */
174struct device_node *fetch_dev_dn(struct pci_dev *dev) 176struct device_node *fetch_dev_dn(struct pci_dev *dev)
175{ 177{
176 struct device_node *orig_dn = dev->sysdata; 178 struct pci_controller *phb = dev->sysdata;
177 struct device_node *dn; 179 struct device_node *dn;
178 unsigned long searchval = (dev->bus->number << 8) | dev->devfn; 180 unsigned long searchval = (dev->bus->number << 8) | dev->devfn;
179 181
180 dn = traverse_pci_devices(orig_dn, is_devfn_node, (void *)searchval); 182 if (WARN_ON(!phb))
183 return NULL;
184
185 dn = traverse_pci_devices(phb->dn, is_devfn_node, (void *)searchval);
181 if (dn) 186 if (dn)
182 dev->sysdata = dn; 187 dev->dev.of_node = dn;
183 return dn; 188 return dn;
184} 189}
185EXPORT_SYMBOL(fetch_dev_dn); 190EXPORT_SYMBOL(fetch_dev_dn);
diff --git a/arch/powerpc/kernel/pci_of_scan.c b/arch/powerpc/kernel/pci_of_scan.c
index e751506323b4..1e89a72fd030 100644
--- a/arch/powerpc/kernel/pci_of_scan.c
+++ b/arch/powerpc/kernel/pci_of_scan.c
@@ -135,7 +135,7 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
135 pr_debug(" create device, devfn: %x, type: %s\n", devfn, type); 135 pr_debug(" create device, devfn: %x, type: %s\n", devfn, type);
136 136
137 dev->bus = bus; 137 dev->bus = bus;
138 dev->sysdata = node; 138 dev->dev.of_node = of_node_get(node);
139 dev->dev.parent = bus->bridge; 139 dev->dev.parent = bus->bridge;
140 dev->dev.bus = &pci_bus_type; 140 dev->dev.bus = &pci_bus_type;
141 dev->devfn = devfn; 141 dev->devfn = devfn;
@@ -238,7 +238,7 @@ void __devinit of_scan_pci_bridge(struct device_node *node,
238 bus->primary = dev->bus->number; 238 bus->primary = dev->bus->number;
239 bus->subordinate = busrange[1]; 239 bus->subordinate = busrange[1];
240 bus->bridge_ctl = 0; 240 bus->bridge_ctl = 0;
241 bus->sysdata = node; 241 bus->dev.of_node = of_node_get(node);
242 242
243 /* parse ranges property */ 243 /* parse ranges property */
244 /* PCI #address-cells == 3 and #size-cells == 2 always */ 244 /* PCI #address-cells == 3 and #size-cells == 2 always */
diff --git a/arch/powerpc/kernel/perf_callchain.c b/arch/powerpc/kernel/perf_callchain.c
index 95ad9dad298e..d05ae4204bbf 100644
--- a/arch/powerpc/kernel/perf_callchain.c
+++ b/arch/powerpc/kernel/perf_callchain.c
@@ -23,18 +23,6 @@
23#include "ppc32.h" 23#include "ppc32.h"
24#endif 24#endif
25 25
26/*
27 * Store another value in a callchain_entry.
28 */
29static inline void callchain_store(struct perf_callchain_entry *entry, u64 ip)
30{
31 unsigned int nr = entry->nr;
32
33 if (nr < PERF_MAX_STACK_DEPTH) {
34 entry->ip[nr] = ip;
35 entry->nr = nr + 1;
36 }
37}
38 26
39/* 27/*
40 * Is sp valid as the address of the next kernel stack frame after prev_sp? 28 * Is sp valid as the address of the next kernel stack frame after prev_sp?
@@ -58,8 +46,8 @@ static int valid_next_sp(unsigned long sp, unsigned long prev_sp)
58 return 0; 46 return 0;
59} 47}
60 48
61static void perf_callchain_kernel(struct pt_regs *regs, 49void
62 struct perf_callchain_entry *entry) 50perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs)
63{ 51{
64 unsigned long sp, next_sp; 52 unsigned long sp, next_sp;
65 unsigned long next_ip; 53 unsigned long next_ip;
@@ -69,8 +57,7 @@ static void perf_callchain_kernel(struct pt_regs *regs,
69 57
70 lr = regs->link; 58 lr = regs->link;
71 sp = regs->gpr[1]; 59 sp = regs->gpr[1];
72 callchain_store(entry, PERF_CONTEXT_KERNEL); 60 perf_callchain_store(entry, regs->nip);
73 callchain_store(entry, regs->nip);
74 61
75 if (!validate_sp(sp, current, STACK_FRAME_OVERHEAD)) 62 if (!validate_sp(sp, current, STACK_FRAME_OVERHEAD))
76 return; 63 return;
@@ -89,7 +76,7 @@ static void perf_callchain_kernel(struct pt_regs *regs,
89 next_ip = regs->nip; 76 next_ip = regs->nip;
90 lr = regs->link; 77 lr = regs->link;
91 level = 0; 78 level = 0;
92 callchain_store(entry, PERF_CONTEXT_KERNEL); 79 perf_callchain_store(entry, PERF_CONTEXT_KERNEL);
93 80
94 } else { 81 } else {
95 if (level == 0) 82 if (level == 0)
@@ -111,7 +98,7 @@ static void perf_callchain_kernel(struct pt_regs *regs,
111 ++level; 98 ++level;
112 } 99 }
113 100
114 callchain_store(entry, next_ip); 101 perf_callchain_store(entry, next_ip);
115 if (!valid_next_sp(next_sp, sp)) 102 if (!valid_next_sp(next_sp, sp))
116 return; 103 return;
117 sp = next_sp; 104 sp = next_sp;
@@ -233,8 +220,8 @@ static int sane_signal_64_frame(unsigned long sp)
233 puc == (unsigned long) &sf->uc; 220 puc == (unsigned long) &sf->uc;
234} 221}
235 222
236static void perf_callchain_user_64(struct pt_regs *regs, 223static void perf_callchain_user_64(struct perf_callchain_entry *entry,
237 struct perf_callchain_entry *entry) 224 struct pt_regs *regs)
238{ 225{
239 unsigned long sp, next_sp; 226 unsigned long sp, next_sp;
240 unsigned long next_ip; 227 unsigned long next_ip;
@@ -246,8 +233,7 @@ static void perf_callchain_user_64(struct pt_regs *regs,
246 next_ip = regs->nip; 233 next_ip = regs->nip;
247 lr = regs->link; 234 lr = regs->link;
248 sp = regs->gpr[1]; 235 sp = regs->gpr[1];
249 callchain_store(entry, PERF_CONTEXT_USER); 236 perf_callchain_store(entry, next_ip);
250 callchain_store(entry, next_ip);
251 237
252 for (;;) { 238 for (;;) {
253 fp = (unsigned long __user *) sp; 239 fp = (unsigned long __user *) sp;
@@ -276,14 +262,14 @@ static void perf_callchain_user_64(struct pt_regs *regs,
276 read_user_stack_64(&uregs[PT_R1], &sp)) 262 read_user_stack_64(&uregs[PT_R1], &sp))
277 return; 263 return;
278 level = 0; 264 level = 0;
279 callchain_store(entry, PERF_CONTEXT_USER); 265 perf_callchain_store(entry, PERF_CONTEXT_USER);
280 callchain_store(entry, next_ip); 266 perf_callchain_store(entry, next_ip);
281 continue; 267 continue;
282 } 268 }
283 269
284 if (level == 0) 270 if (level == 0)
285 next_ip = lr; 271 next_ip = lr;
286 callchain_store(entry, next_ip); 272 perf_callchain_store(entry, next_ip);
287 ++level; 273 ++level;
288 sp = next_sp; 274 sp = next_sp;
289 } 275 }
@@ -315,8 +301,8 @@ static int read_user_stack_32(unsigned int __user *ptr, unsigned int *ret)
315 return __get_user_inatomic(*ret, ptr); 301 return __get_user_inatomic(*ret, ptr);
316} 302}
317 303
318static inline void perf_callchain_user_64(struct pt_regs *regs, 304static inline void perf_callchain_user_64(struct perf_callchain_entry *entry,
319 struct perf_callchain_entry *entry) 305 struct pt_regs *regs)
320{ 306{
321} 307}
322 308
@@ -435,8 +421,8 @@ static unsigned int __user *signal_frame_32_regs(unsigned int sp,
435 return mctx->mc_gregs; 421 return mctx->mc_gregs;
436} 422}
437 423
438static void perf_callchain_user_32(struct pt_regs *regs, 424static void perf_callchain_user_32(struct perf_callchain_entry *entry,
439 struct perf_callchain_entry *entry) 425 struct pt_regs *regs)
440{ 426{
441 unsigned int sp, next_sp; 427 unsigned int sp, next_sp;
442 unsigned int next_ip; 428 unsigned int next_ip;
@@ -447,8 +433,7 @@ static void perf_callchain_user_32(struct pt_regs *regs,
447 next_ip = regs->nip; 433 next_ip = regs->nip;
448 lr = regs->link; 434 lr = regs->link;
449 sp = regs->gpr[1]; 435 sp = regs->gpr[1];
450 callchain_store(entry, PERF_CONTEXT_USER); 436 perf_callchain_store(entry, next_ip);
451 callchain_store(entry, next_ip);
452 437
453 while (entry->nr < PERF_MAX_STACK_DEPTH) { 438 while (entry->nr < PERF_MAX_STACK_DEPTH) {
454 fp = (unsigned int __user *) (unsigned long) sp; 439 fp = (unsigned int __user *) (unsigned long) sp;
@@ -470,45 +455,24 @@ static void perf_callchain_user_32(struct pt_regs *regs,
470 read_user_stack_32(&uregs[PT_R1], &sp)) 455 read_user_stack_32(&uregs[PT_R1], &sp))
471 return; 456 return;
472 level = 0; 457 level = 0;
473 callchain_store(entry, PERF_CONTEXT_USER); 458 perf_callchain_store(entry, PERF_CONTEXT_USER);
474 callchain_store(entry, next_ip); 459 perf_callchain_store(entry, next_ip);
475 continue; 460 continue;
476 } 461 }
477 462
478 if (level == 0) 463 if (level == 0)
479 next_ip = lr; 464 next_ip = lr;
480 callchain_store(entry, next_ip); 465 perf_callchain_store(entry, next_ip);
481 ++level; 466 ++level;
482 sp = next_sp; 467 sp = next_sp;
483 } 468 }
484} 469}
485 470
486/* 471void
487 * Since we can't get PMU interrupts inside a PMU interrupt handler, 472perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
488 * we don't need separate irq and nmi entries here.
489 */
490static DEFINE_PER_CPU(struct perf_callchain_entry, cpu_perf_callchain);
491
492struct perf_callchain_entry *perf_callchain(struct pt_regs *regs)
493{ 473{
494 struct perf_callchain_entry *entry = &__get_cpu_var(cpu_perf_callchain); 474 if (current_is_64bit())
495 475 perf_callchain_user_64(entry, regs);
496 entry->nr = 0; 476 else
497 477 perf_callchain_user_32(entry, regs);
498 if (!user_mode(regs)) {
499 perf_callchain_kernel(regs, entry);
500 if (current->mm)
501 regs = task_pt_regs(current);
502 else
503 regs = NULL;
504 }
505
506 if (regs) {
507 if (current_is_64bit())
508 perf_callchain_user_64(regs, entry);
509 else
510 perf_callchain_user_32(regs, entry);
511 }
512
513 return entry;
514} 478}
diff --git a/arch/powerpc/kernel/perf_event.c b/arch/powerpc/kernel/perf_event.c
index d301a30445e0..822f63008ae1 100644
--- a/arch/powerpc/kernel/perf_event.c
+++ b/arch/powerpc/kernel/perf_event.c
@@ -398,10 +398,32 @@ static int check_excludes(struct perf_event **ctrs, unsigned int cflags[],
398 return 0; 398 return 0;
399} 399}
400 400
401static u64 check_and_compute_delta(u64 prev, u64 val)
402{
403 u64 delta = (val - prev) & 0xfffffffful;
404
405 /*
406 * POWER7 can roll back counter values, if the new value is smaller
407 * than the previous value it will cause the delta and the counter to
408 * have bogus values unless we rolled a counter over. If a coutner is
409 * rolled back, it will be smaller, but within 256, which is the maximum
410 * number of events to rollback at once. If we dectect a rollback
411 * return 0. This can lead to a small lack of precision in the
412 * counters.
413 */
414 if (prev > val && (prev - val) < 256)
415 delta = 0;
416
417 return delta;
418}
419
401static void power_pmu_read(struct perf_event *event) 420static void power_pmu_read(struct perf_event *event)
402{ 421{
403 s64 val, delta, prev; 422 s64 val, delta, prev;
404 423
424 if (event->hw.state & PERF_HES_STOPPED)
425 return;
426
405 if (!event->hw.idx) 427 if (!event->hw.idx)
406 return; 428 return;
407 /* 429 /*
@@ -413,10 +435,11 @@ static void power_pmu_read(struct perf_event *event)
413 prev = local64_read(&event->hw.prev_count); 435 prev = local64_read(&event->hw.prev_count);
414 barrier(); 436 barrier();
415 val = read_pmc(event->hw.idx); 437 val = read_pmc(event->hw.idx);
438 delta = check_and_compute_delta(prev, val);
439 if (!delta)
440 return;
416 } while (local64_cmpxchg(&event->hw.prev_count, prev, val) != prev); 441 } while (local64_cmpxchg(&event->hw.prev_count, prev, val) != prev);
417 442
418 /* The counters are only 32 bits wide */
419 delta = (val - prev) & 0xfffffffful;
420 local64_add(delta, &event->count); 443 local64_add(delta, &event->count);
421 local64_sub(delta, &event->hw.period_left); 444 local64_sub(delta, &event->hw.period_left);
422} 445}
@@ -446,8 +469,9 @@ static void freeze_limited_counters(struct cpu_hw_events *cpuhw,
446 val = (event->hw.idx == 5) ? pmc5 : pmc6; 469 val = (event->hw.idx == 5) ? pmc5 : pmc6;
447 prev = local64_read(&event->hw.prev_count); 470 prev = local64_read(&event->hw.prev_count);
448 event->hw.idx = 0; 471 event->hw.idx = 0;
449 delta = (val - prev) & 0xfffffffful; 472 delta = check_and_compute_delta(prev, val);
450 local64_add(delta, &event->count); 473 if (delta)
474 local64_add(delta, &event->count);
451 } 475 }
452} 476}
453 477
@@ -455,14 +479,16 @@ static void thaw_limited_counters(struct cpu_hw_events *cpuhw,
455 unsigned long pmc5, unsigned long pmc6) 479 unsigned long pmc5, unsigned long pmc6)
456{ 480{
457 struct perf_event *event; 481 struct perf_event *event;
458 u64 val; 482 u64 val, prev;
459 int i; 483 int i;
460 484
461 for (i = 0; i < cpuhw->n_limited; ++i) { 485 for (i = 0; i < cpuhw->n_limited; ++i) {
462 event = cpuhw->limited_counter[i]; 486 event = cpuhw->limited_counter[i];
463 event->hw.idx = cpuhw->limited_hwidx[i]; 487 event->hw.idx = cpuhw->limited_hwidx[i];
464 val = (event->hw.idx == 5) ? pmc5 : pmc6; 488 val = (event->hw.idx == 5) ? pmc5 : pmc6;
465 local64_set(&event->hw.prev_count, val); 489 prev = local64_read(&event->hw.prev_count);
490 if (check_and_compute_delta(prev, val))
491 local64_set(&event->hw.prev_count, val);
466 perf_event_update_userpage(event); 492 perf_event_update_userpage(event);
467 } 493 }
468} 494}
@@ -517,7 +543,7 @@ static void write_mmcr0(struct cpu_hw_events *cpuhw, unsigned long mmcr0)
517 * Disable all events to prevent PMU interrupts and to allow 543 * Disable all events to prevent PMU interrupts and to allow
518 * events to be added or removed. 544 * events to be added or removed.
519 */ 545 */
520void hw_perf_disable(void) 546static void power_pmu_disable(struct pmu *pmu)
521{ 547{
522 struct cpu_hw_events *cpuhw; 548 struct cpu_hw_events *cpuhw;
523 unsigned long flags; 549 unsigned long flags;
@@ -565,7 +591,7 @@ void hw_perf_disable(void)
565 * If we were previously disabled and events were added, then 591 * If we were previously disabled and events were added, then
566 * put the new config on the PMU. 592 * put the new config on the PMU.
567 */ 593 */
568void hw_perf_enable(void) 594static void power_pmu_enable(struct pmu *pmu)
569{ 595{
570 struct perf_event *event; 596 struct perf_event *event;
571 struct cpu_hw_events *cpuhw; 597 struct cpu_hw_events *cpuhw;
@@ -672,6 +698,8 @@ void hw_perf_enable(void)
672 } 698 }
673 local64_set(&event->hw.prev_count, val); 699 local64_set(&event->hw.prev_count, val);
674 event->hw.idx = idx; 700 event->hw.idx = idx;
701 if (event->hw.state & PERF_HES_STOPPED)
702 val = 0;
675 write_pmc(idx, val); 703 write_pmc(idx, val);
676 perf_event_update_userpage(event); 704 perf_event_update_userpage(event);
677 } 705 }
@@ -727,7 +755,7 @@ static int collect_events(struct perf_event *group, int max_count,
727 * re-enable the PMU in order to get hw_perf_enable to do the 755 * re-enable the PMU in order to get hw_perf_enable to do the
728 * actual work of reconfiguring the PMU. 756 * actual work of reconfiguring the PMU.
729 */ 757 */
730static int power_pmu_enable(struct perf_event *event) 758static int power_pmu_add(struct perf_event *event, int ef_flags)
731{ 759{
732 struct cpu_hw_events *cpuhw; 760 struct cpu_hw_events *cpuhw;
733 unsigned long flags; 761 unsigned long flags;
@@ -735,7 +763,7 @@ static int power_pmu_enable(struct perf_event *event)
735 int ret = -EAGAIN; 763 int ret = -EAGAIN;
736 764
737 local_irq_save(flags); 765 local_irq_save(flags);
738 perf_disable(); 766 perf_pmu_disable(event->pmu);
739 767
740 /* 768 /*
741 * Add the event to the list (if there is room) 769 * Add the event to the list (if there is room)
@@ -749,9 +777,12 @@ static int power_pmu_enable(struct perf_event *event)
749 cpuhw->events[n0] = event->hw.config; 777 cpuhw->events[n0] = event->hw.config;
750 cpuhw->flags[n0] = event->hw.event_base; 778 cpuhw->flags[n0] = event->hw.event_base;
751 779
780 if (!(ef_flags & PERF_EF_START))
781 event->hw.state = PERF_HES_STOPPED | PERF_HES_UPTODATE;
782
752 /* 783 /*
753 * If group events scheduling transaction was started, 784 * If group events scheduling transaction was started,
754 * skip the schedulability test here, it will be peformed 785 * skip the schedulability test here, it will be performed
755 * at commit time(->commit_txn) as a whole 786 * at commit time(->commit_txn) as a whole
756 */ 787 */
757 if (cpuhw->group_flag & PERF_EVENT_TXN) 788 if (cpuhw->group_flag & PERF_EVENT_TXN)
@@ -769,7 +800,7 @@ nocheck:
769 800
770 ret = 0; 801 ret = 0;
771 out: 802 out:
772 perf_enable(); 803 perf_pmu_enable(event->pmu);
773 local_irq_restore(flags); 804 local_irq_restore(flags);
774 return ret; 805 return ret;
775} 806}
@@ -777,14 +808,14 @@ nocheck:
777/* 808/*
778 * Remove a event from the PMU. 809 * Remove a event from the PMU.
779 */ 810 */
780static void power_pmu_disable(struct perf_event *event) 811static void power_pmu_del(struct perf_event *event, int ef_flags)
781{ 812{
782 struct cpu_hw_events *cpuhw; 813 struct cpu_hw_events *cpuhw;
783 long i; 814 long i;
784 unsigned long flags; 815 unsigned long flags;
785 816
786 local_irq_save(flags); 817 local_irq_save(flags);
787 perf_disable(); 818 perf_pmu_disable(event->pmu);
788 819
789 power_pmu_read(event); 820 power_pmu_read(event);
790 821
@@ -821,34 +852,60 @@ static void power_pmu_disable(struct perf_event *event)
821 cpuhw->mmcr[0] &= ~(MMCR0_PMXE | MMCR0_FCECE); 852 cpuhw->mmcr[0] &= ~(MMCR0_PMXE | MMCR0_FCECE);
822 } 853 }
823 854
824 perf_enable(); 855 perf_pmu_enable(event->pmu);
825 local_irq_restore(flags); 856 local_irq_restore(flags);
826} 857}
827 858
828/* 859/*
829 * Re-enable interrupts on a event after they were throttled 860 * POWER-PMU does not support disabling individual counters, hence
830 * because they were coming too fast. 861 * program their cycle counter to their max value and ignore the interrupts.
831 */ 862 */
832static void power_pmu_unthrottle(struct perf_event *event) 863
864static void power_pmu_start(struct perf_event *event, int ef_flags)
865{
866 unsigned long flags;
867 s64 left;
868
869 if (!event->hw.idx || !event->hw.sample_period)
870 return;
871
872 if (!(event->hw.state & PERF_HES_STOPPED))
873 return;
874
875 if (ef_flags & PERF_EF_RELOAD)
876 WARN_ON_ONCE(!(event->hw.state & PERF_HES_UPTODATE));
877
878 local_irq_save(flags);
879 perf_pmu_disable(event->pmu);
880
881 event->hw.state = 0;
882 left = local64_read(&event->hw.period_left);
883 write_pmc(event->hw.idx, left);
884
885 perf_event_update_userpage(event);
886 perf_pmu_enable(event->pmu);
887 local_irq_restore(flags);
888}
889
890static void power_pmu_stop(struct perf_event *event, int ef_flags)
833{ 891{
834 s64 val, left;
835 unsigned long flags; 892 unsigned long flags;
836 893
837 if (!event->hw.idx || !event->hw.sample_period) 894 if (!event->hw.idx || !event->hw.sample_period)
838 return; 895 return;
896
897 if (event->hw.state & PERF_HES_STOPPED)
898 return;
899
839 local_irq_save(flags); 900 local_irq_save(flags);
840 perf_disable(); 901 perf_pmu_disable(event->pmu);
902
841 power_pmu_read(event); 903 power_pmu_read(event);
842 left = event->hw.sample_period; 904 event->hw.state |= PERF_HES_STOPPED | PERF_HES_UPTODATE;
843 event->hw.last_period = left; 905 write_pmc(event->hw.idx, 0);
844 val = 0; 906
845 if (left < 0x80000000L)
846 val = 0x80000000L - left;
847 write_pmc(event->hw.idx, val);
848 local64_set(&event->hw.prev_count, val);
849 local64_set(&event->hw.period_left, left);
850 perf_event_update_userpage(event); 907 perf_event_update_userpage(event);
851 perf_enable(); 908 perf_pmu_enable(event->pmu);
852 local_irq_restore(flags); 909 local_irq_restore(flags);
853} 910}
854 911
@@ -857,10 +914,11 @@ static void power_pmu_unthrottle(struct perf_event *event)
857 * Set the flag to make pmu::enable() not perform the 914 * Set the flag to make pmu::enable() not perform the
858 * schedulability test, it will be performed at commit time 915 * schedulability test, it will be performed at commit time
859 */ 916 */
860void power_pmu_start_txn(const struct pmu *pmu) 917void power_pmu_start_txn(struct pmu *pmu)
861{ 918{
862 struct cpu_hw_events *cpuhw = &__get_cpu_var(cpu_hw_events); 919 struct cpu_hw_events *cpuhw = &__get_cpu_var(cpu_hw_events);
863 920
921 perf_pmu_disable(pmu);
864 cpuhw->group_flag |= PERF_EVENT_TXN; 922 cpuhw->group_flag |= PERF_EVENT_TXN;
865 cpuhw->n_txn_start = cpuhw->n_events; 923 cpuhw->n_txn_start = cpuhw->n_events;
866} 924}
@@ -870,11 +928,12 @@ void power_pmu_start_txn(const struct pmu *pmu)
870 * Clear the flag and pmu::enable() will perform the 928 * Clear the flag and pmu::enable() will perform the
871 * schedulability test. 929 * schedulability test.
872 */ 930 */
873void power_pmu_cancel_txn(const struct pmu *pmu) 931void power_pmu_cancel_txn(struct pmu *pmu)
874{ 932{
875 struct cpu_hw_events *cpuhw = &__get_cpu_var(cpu_hw_events); 933 struct cpu_hw_events *cpuhw = &__get_cpu_var(cpu_hw_events);
876 934
877 cpuhw->group_flag &= ~PERF_EVENT_TXN; 935 cpuhw->group_flag &= ~PERF_EVENT_TXN;
936 perf_pmu_enable(pmu);
878} 937}
879 938
880/* 939/*
@@ -882,7 +941,7 @@ void power_pmu_cancel_txn(const struct pmu *pmu)
882 * Perform the group schedulability test as a whole 941 * Perform the group schedulability test as a whole
883 * Return 0 if success 942 * Return 0 if success
884 */ 943 */
885int power_pmu_commit_txn(const struct pmu *pmu) 944int power_pmu_commit_txn(struct pmu *pmu)
886{ 945{
887 struct cpu_hw_events *cpuhw; 946 struct cpu_hw_events *cpuhw;
888 long i, n; 947 long i, n;
@@ -901,19 +960,10 @@ int power_pmu_commit_txn(const struct pmu *pmu)
901 cpuhw->event[i]->hw.config = cpuhw->events[i]; 960 cpuhw->event[i]->hw.config = cpuhw->events[i];
902 961
903 cpuhw->group_flag &= ~PERF_EVENT_TXN; 962 cpuhw->group_flag &= ~PERF_EVENT_TXN;
963 perf_pmu_enable(pmu);
904 return 0; 964 return 0;
905} 965}
906 966
907struct pmu power_pmu = {
908 .enable = power_pmu_enable,
909 .disable = power_pmu_disable,
910 .read = power_pmu_read,
911 .unthrottle = power_pmu_unthrottle,
912 .start_txn = power_pmu_start_txn,
913 .cancel_txn = power_pmu_cancel_txn,
914 .commit_txn = power_pmu_commit_txn,
915};
916
917/* 967/*
918 * Return 1 if we might be able to put event on a limited PMC, 968 * Return 1 if we might be able to put event on a limited PMC,
919 * or 0 if not. 969 * or 0 if not.
@@ -1014,7 +1064,7 @@ static int hw_perf_cache_event(u64 config, u64 *eventp)
1014 return 0; 1064 return 0;
1015} 1065}
1016 1066
1017const struct pmu *hw_perf_event_init(struct perf_event *event) 1067static int power_pmu_event_init(struct perf_event *event)
1018{ 1068{
1019 u64 ev; 1069 u64 ev;
1020 unsigned long flags; 1070 unsigned long flags;
@@ -1026,25 +1076,27 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
1026 struct cpu_hw_events *cpuhw; 1076 struct cpu_hw_events *cpuhw;
1027 1077
1028 if (!ppmu) 1078 if (!ppmu)
1029 return ERR_PTR(-ENXIO); 1079 return -ENOENT;
1080
1030 switch (event->attr.type) { 1081 switch (event->attr.type) {
1031 case PERF_TYPE_HARDWARE: 1082 case PERF_TYPE_HARDWARE:
1032 ev = event->attr.config; 1083 ev = event->attr.config;
1033 if (ev >= ppmu->n_generic || ppmu->generic_events[ev] == 0) 1084 if (ev >= ppmu->n_generic || ppmu->generic_events[ev] == 0)
1034 return ERR_PTR(-EOPNOTSUPP); 1085 return -EOPNOTSUPP;
1035 ev = ppmu->generic_events[ev]; 1086 ev = ppmu->generic_events[ev];
1036 break; 1087 break;
1037 case PERF_TYPE_HW_CACHE: 1088 case PERF_TYPE_HW_CACHE:
1038 err = hw_perf_cache_event(event->attr.config, &ev); 1089 err = hw_perf_cache_event(event->attr.config, &ev);
1039 if (err) 1090 if (err)
1040 return ERR_PTR(err); 1091 return err;
1041 break; 1092 break;
1042 case PERF_TYPE_RAW: 1093 case PERF_TYPE_RAW:
1043 ev = event->attr.config; 1094 ev = event->attr.config;
1044 break; 1095 break;
1045 default: 1096 default:
1046 return ERR_PTR(-EINVAL); 1097 return -ENOENT;
1047 } 1098 }
1099
1048 event->hw.config_base = ev; 1100 event->hw.config_base = ev;
1049 event->hw.idx = 0; 1101 event->hw.idx = 0;
1050 1102
@@ -1063,7 +1115,7 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
1063 * XXX we should check if the task is an idle task. 1115 * XXX we should check if the task is an idle task.
1064 */ 1116 */
1065 flags = 0; 1117 flags = 0;
1066 if (event->ctx->task) 1118 if (event->attach_state & PERF_ATTACH_TASK)
1067 flags |= PPMU_ONLY_COUNT_RUN; 1119 flags |= PPMU_ONLY_COUNT_RUN;
1068 1120
1069 /* 1121 /*
@@ -1081,7 +1133,7 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
1081 */ 1133 */
1082 ev = normal_pmc_alternative(ev, flags); 1134 ev = normal_pmc_alternative(ev, flags);
1083 if (!ev) 1135 if (!ev)
1084 return ERR_PTR(-EINVAL); 1136 return -EINVAL;
1085 } 1137 }
1086 } 1138 }
1087 1139
@@ -1095,19 +1147,19 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
1095 n = collect_events(event->group_leader, ppmu->n_counter - 1, 1147 n = collect_events(event->group_leader, ppmu->n_counter - 1,
1096 ctrs, events, cflags); 1148 ctrs, events, cflags);
1097 if (n < 0) 1149 if (n < 0)
1098 return ERR_PTR(-EINVAL); 1150 return -EINVAL;
1099 } 1151 }
1100 events[n] = ev; 1152 events[n] = ev;
1101 ctrs[n] = event; 1153 ctrs[n] = event;
1102 cflags[n] = flags; 1154 cflags[n] = flags;
1103 if (check_excludes(ctrs, cflags, n, 1)) 1155 if (check_excludes(ctrs, cflags, n, 1))
1104 return ERR_PTR(-EINVAL); 1156 return -EINVAL;
1105 1157
1106 cpuhw = &get_cpu_var(cpu_hw_events); 1158 cpuhw = &get_cpu_var(cpu_hw_events);
1107 err = power_check_constraints(cpuhw, events, cflags, n + 1); 1159 err = power_check_constraints(cpuhw, events, cflags, n + 1);
1108 put_cpu_var(cpu_hw_events); 1160 put_cpu_var(cpu_hw_events);
1109 if (err) 1161 if (err)
1110 return ERR_PTR(-EINVAL); 1162 return -EINVAL;
1111 1163
1112 event->hw.config = events[n]; 1164 event->hw.config = events[n];
1113 event->hw.event_base = cflags[n]; 1165 event->hw.event_base = cflags[n];
@@ -1132,11 +1184,23 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
1132 } 1184 }
1133 event->destroy = hw_perf_event_destroy; 1185 event->destroy = hw_perf_event_destroy;
1134 1186
1135 if (err) 1187 return err;
1136 return ERR_PTR(err);
1137 return &power_pmu;
1138} 1188}
1139 1189
1190struct pmu power_pmu = {
1191 .pmu_enable = power_pmu_enable,
1192 .pmu_disable = power_pmu_disable,
1193 .event_init = power_pmu_event_init,
1194 .add = power_pmu_add,
1195 .del = power_pmu_del,
1196 .start = power_pmu_start,
1197 .stop = power_pmu_stop,
1198 .read = power_pmu_read,
1199 .start_txn = power_pmu_start_txn,
1200 .cancel_txn = power_pmu_cancel_txn,
1201 .commit_txn = power_pmu_commit_txn,
1202};
1203
1140/* 1204/*
1141 * A counter has overflowed; update its count and record 1205 * A counter has overflowed; update its count and record
1142 * things if requested. Note that interrupts are hard-disabled 1206 * things if requested. Note that interrupts are hard-disabled
@@ -1149,9 +1213,14 @@ static void record_and_restart(struct perf_event *event, unsigned long val,
1149 s64 prev, delta, left; 1213 s64 prev, delta, left;
1150 int record = 0; 1214 int record = 0;
1151 1215
1216 if (event->hw.state & PERF_HES_STOPPED) {
1217 write_pmc(event->hw.idx, 0);
1218 return;
1219 }
1220
1152 /* we don't have to worry about interrupts here */ 1221 /* we don't have to worry about interrupts here */
1153 prev = local64_read(&event->hw.prev_count); 1222 prev = local64_read(&event->hw.prev_count);
1154 delta = (val - prev) & 0xfffffffful; 1223 delta = check_and_compute_delta(prev, val);
1155 local64_add(delta, &event->count); 1224 local64_add(delta, &event->count);
1156 1225
1157 /* 1226 /*
@@ -1166,11 +1235,17 @@ static void record_and_restart(struct perf_event *event, unsigned long val,
1166 if (left <= 0) 1235 if (left <= 0)
1167 left = period; 1236 left = period;
1168 record = 1; 1237 record = 1;
1238 event->hw.last_period = event->hw.sample_period;
1169 } 1239 }
1170 if (left < 0x80000000LL) 1240 if (left < 0x80000000LL)
1171 val = 0x80000000LL - left; 1241 val = 0x80000000LL - left;
1172 } 1242 }
1173 1243
1244 write_pmc(event->hw.idx, val);
1245 local64_set(&event->hw.prev_count, val);
1246 local64_set(&event->hw.period_left, left);
1247 perf_event_update_userpage(event);
1248
1174 /* 1249 /*
1175 * Finally record data if requested. 1250 * Finally record data if requested.
1176 */ 1251 */
@@ -1183,23 +1258,9 @@ static void record_and_restart(struct perf_event *event, unsigned long val,
1183 if (event->attr.sample_type & PERF_SAMPLE_ADDR) 1258 if (event->attr.sample_type & PERF_SAMPLE_ADDR)
1184 perf_get_data_addr(regs, &data.addr); 1259 perf_get_data_addr(regs, &data.addr);
1185 1260
1186 if (perf_event_overflow(event, nmi, &data, regs)) { 1261 if (perf_event_overflow(event, nmi, &data, regs))
1187 /* 1262 power_pmu_stop(event, 0);
1188 * Interrupts are coming too fast - throttle them
1189 * by setting the event to 0, so it will be
1190 * at least 2^30 cycles until the next interrupt
1191 * (assuming each event counts at most 2 counts
1192 * per cycle).
1193 */
1194 val = 0;
1195 left = ~0ULL >> 1;
1196 }
1197 } 1263 }
1198
1199 write_pmc(event->hw.idx, val);
1200 local64_set(&event->hw.prev_count, val);
1201 local64_set(&event->hw.period_left, left);
1202 perf_event_update_userpage(event);
1203} 1264}
1204 1265
1205/* 1266/*
@@ -1231,6 +1292,28 @@ unsigned long perf_instruction_pointer(struct pt_regs *regs)
1231 return ip; 1292 return ip;
1232} 1293}
1233 1294
1295static bool pmc_overflow(unsigned long val)
1296{
1297 if ((int)val < 0)
1298 return true;
1299
1300 /*
1301 * Events on POWER7 can roll back if a speculative event doesn't
1302 * eventually complete. Unfortunately in some rare cases they will
1303 * raise a performance monitor exception. We need to catch this to
1304 * ensure we reset the PMC. In all cases the PMC will be 256 or less
1305 * cycles from overflow.
1306 *
1307 * We only do this if the first pass fails to find any overflowing
1308 * PMCs because a user might set a period of less than 256 and we
1309 * don't want to mistakenly reset them.
1310 */
1311 if (__is_processor(PV_POWER7) && ((0x80000000 - val) <= 256))
1312 return true;
1313
1314 return false;
1315}
1316
1234/* 1317/*
1235 * Performance monitor interrupt stuff 1318 * Performance monitor interrupt stuff
1236 */ 1319 */
@@ -1278,7 +1361,7 @@ static void perf_event_interrupt(struct pt_regs *regs)
1278 if (is_limited_pmc(i + 1)) 1361 if (is_limited_pmc(i + 1))
1279 continue; 1362 continue;
1280 val = read_pmc(i + 1); 1363 val = read_pmc(i + 1);
1281 if ((int)val < 0) 1364 if (pmc_overflow(val))
1282 write_pmc(i + 1, 0); 1365 write_pmc(i + 1, 0);
1283 } 1366 }
1284 } 1367 }
@@ -1342,6 +1425,7 @@ int register_power_pmu(struct power_pmu *pmu)
1342 freeze_events_kernel = MMCR0_FCHV; 1425 freeze_events_kernel = MMCR0_FCHV;
1343#endif /* CONFIG_PPC64 */ 1426#endif /* CONFIG_PPC64 */
1344 1427
1428 perf_pmu_register(&power_pmu, "cpu", PERF_TYPE_RAW);
1345 perf_cpu_notifier(power_pmu_notifier); 1429 perf_cpu_notifier(power_pmu_notifier);
1346 1430
1347 return 0; 1431 return 0;
diff --git a/arch/powerpc/kernel/perf_event_fsl_emb.c b/arch/powerpc/kernel/perf_event_fsl_emb.c
index 1ba45471ae43..b0dc8f7069cd 100644
--- a/arch/powerpc/kernel/perf_event_fsl_emb.c
+++ b/arch/powerpc/kernel/perf_event_fsl_emb.c
@@ -156,6 +156,9 @@ static void fsl_emb_pmu_read(struct perf_event *event)
156{ 156{
157 s64 val, delta, prev; 157 s64 val, delta, prev;
158 158
159 if (event->hw.state & PERF_HES_STOPPED)
160 return;
161
159 /* 162 /*
160 * Performance monitor interrupts come even when interrupts 163 * Performance monitor interrupts come even when interrupts
161 * are soft-disabled, as long as interrupts are hard-enabled. 164 * are soft-disabled, as long as interrupts are hard-enabled.
@@ -177,7 +180,7 @@ static void fsl_emb_pmu_read(struct perf_event *event)
177 * Disable all events to prevent PMU interrupts and to allow 180 * Disable all events to prevent PMU interrupts and to allow
178 * events to be added or removed. 181 * events to be added or removed.
179 */ 182 */
180void hw_perf_disable(void) 183static void fsl_emb_pmu_disable(struct pmu *pmu)
181{ 184{
182 struct cpu_hw_events *cpuhw; 185 struct cpu_hw_events *cpuhw;
183 unsigned long flags; 186 unsigned long flags;
@@ -216,7 +219,7 @@ void hw_perf_disable(void)
216 * If we were previously disabled and events were added, then 219 * If we were previously disabled and events were added, then
217 * put the new config on the PMU. 220 * put the new config on the PMU.
218 */ 221 */
219void hw_perf_enable(void) 222static void fsl_emb_pmu_enable(struct pmu *pmu)
220{ 223{
221 struct cpu_hw_events *cpuhw; 224 struct cpu_hw_events *cpuhw;
222 unsigned long flags; 225 unsigned long flags;
@@ -262,8 +265,8 @@ static int collect_events(struct perf_event *group, int max_count,
262 return n; 265 return n;
263} 266}
264 267
265/* perf must be disabled, context locked on entry */ 268/* context locked on entry */
266static int fsl_emb_pmu_enable(struct perf_event *event) 269static int fsl_emb_pmu_add(struct perf_event *event, int flags)
267{ 270{
268 struct cpu_hw_events *cpuhw; 271 struct cpu_hw_events *cpuhw;
269 int ret = -EAGAIN; 272 int ret = -EAGAIN;
@@ -271,6 +274,7 @@ static int fsl_emb_pmu_enable(struct perf_event *event)
271 u64 val; 274 u64 val;
272 int i; 275 int i;
273 276
277 perf_pmu_disable(event->pmu);
274 cpuhw = &get_cpu_var(cpu_hw_events); 278 cpuhw = &get_cpu_var(cpu_hw_events);
275 279
276 if (event->hw.config & FSL_EMB_EVENT_RESTRICTED) 280 if (event->hw.config & FSL_EMB_EVENT_RESTRICTED)
@@ -301,6 +305,12 @@ static int fsl_emb_pmu_enable(struct perf_event *event)
301 val = 0x80000000L - left; 305 val = 0x80000000L - left;
302 } 306 }
303 local64_set(&event->hw.prev_count, val); 307 local64_set(&event->hw.prev_count, val);
308
309 if (!(flags & PERF_EF_START)) {
310 event->hw.state = PERF_HES_STOPPED | PERF_HES_UPTODATE;
311 val = 0;
312 }
313
304 write_pmc(i, val); 314 write_pmc(i, val);
305 perf_event_update_userpage(event); 315 perf_event_update_userpage(event);
306 316
@@ -310,15 +320,17 @@ static int fsl_emb_pmu_enable(struct perf_event *event)
310 ret = 0; 320 ret = 0;
311 out: 321 out:
312 put_cpu_var(cpu_hw_events); 322 put_cpu_var(cpu_hw_events);
323 perf_pmu_enable(event->pmu);
313 return ret; 324 return ret;
314} 325}
315 326
316/* perf must be disabled, context locked on entry */ 327/* context locked on entry */
317static void fsl_emb_pmu_disable(struct perf_event *event) 328static void fsl_emb_pmu_del(struct perf_event *event, int flags)
318{ 329{
319 struct cpu_hw_events *cpuhw; 330 struct cpu_hw_events *cpuhw;
320 int i = event->hw.idx; 331 int i = event->hw.idx;
321 332
333 perf_pmu_disable(event->pmu);
322 if (i < 0) 334 if (i < 0)
323 goto out; 335 goto out;
324 336
@@ -346,44 +358,57 @@ static void fsl_emb_pmu_disable(struct perf_event *event)
346 cpuhw->n_events--; 358 cpuhw->n_events--;
347 359
348 out: 360 out:
361 perf_pmu_enable(event->pmu);
349 put_cpu_var(cpu_hw_events); 362 put_cpu_var(cpu_hw_events);
350} 363}
351 364
352/* 365static void fsl_emb_pmu_start(struct perf_event *event, int ef_flags)
353 * Re-enable interrupts on a event after they were throttled
354 * because they were coming too fast.
355 *
356 * Context is locked on entry, but perf is not disabled.
357 */
358static void fsl_emb_pmu_unthrottle(struct perf_event *event)
359{ 366{
360 s64 val, left;
361 unsigned long flags; 367 unsigned long flags;
368 s64 left;
362 369
363 if (event->hw.idx < 0 || !event->hw.sample_period) 370 if (event->hw.idx < 0 || !event->hw.sample_period)
364 return; 371 return;
372
373 if (!(event->hw.state & PERF_HES_STOPPED))
374 return;
375
376 if (ef_flags & PERF_EF_RELOAD)
377 WARN_ON_ONCE(!(event->hw.state & PERF_HES_UPTODATE));
378
365 local_irq_save(flags); 379 local_irq_save(flags);
366 perf_disable(); 380 perf_pmu_disable(event->pmu);
367 fsl_emb_pmu_read(event); 381
368 left = event->hw.sample_period; 382 event->hw.state = 0;
369 event->hw.last_period = left; 383 left = local64_read(&event->hw.period_left);
370 val = 0; 384 write_pmc(event->hw.idx, left);
371 if (left < 0x80000000L) 385
372 val = 0x80000000L - left;
373 write_pmc(event->hw.idx, val);
374 local64_set(&event->hw.prev_count, val);
375 local64_set(&event->hw.period_left, left);
376 perf_event_update_userpage(event); 386 perf_event_update_userpage(event);
377 perf_enable(); 387 perf_pmu_enable(event->pmu);
378 local_irq_restore(flags); 388 local_irq_restore(flags);
379} 389}
380 390
381static struct pmu fsl_emb_pmu = { 391static void fsl_emb_pmu_stop(struct perf_event *event, int ef_flags)
382 .enable = fsl_emb_pmu_enable, 392{
383 .disable = fsl_emb_pmu_disable, 393 unsigned long flags;
384 .read = fsl_emb_pmu_read, 394
385 .unthrottle = fsl_emb_pmu_unthrottle, 395 if (event->hw.idx < 0 || !event->hw.sample_period)
386}; 396 return;
397
398 if (event->hw.state & PERF_HES_STOPPED)
399 return;
400
401 local_irq_save(flags);
402 perf_pmu_disable(event->pmu);
403
404 fsl_emb_pmu_read(event);
405 event->hw.state |= PERF_HES_STOPPED | PERF_HES_UPTODATE;
406 write_pmc(event->hw.idx, 0);
407
408 perf_event_update_userpage(event);
409 perf_pmu_enable(event->pmu);
410 local_irq_restore(flags);
411}
387 412
388/* 413/*
389 * Release the PMU if this is the last perf_event. 414 * Release the PMU if this is the last perf_event.
@@ -428,7 +453,7 @@ static int hw_perf_cache_event(u64 config, u64 *eventp)
428 return 0; 453 return 0;
429} 454}
430 455
431const struct pmu *hw_perf_event_init(struct perf_event *event) 456static int fsl_emb_pmu_event_init(struct perf_event *event)
432{ 457{
433 u64 ev; 458 u64 ev;
434 struct perf_event *events[MAX_HWEVENTS]; 459 struct perf_event *events[MAX_HWEVENTS];
@@ -441,14 +466,14 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
441 case PERF_TYPE_HARDWARE: 466 case PERF_TYPE_HARDWARE:
442 ev = event->attr.config; 467 ev = event->attr.config;
443 if (ev >= ppmu->n_generic || ppmu->generic_events[ev] == 0) 468 if (ev >= ppmu->n_generic || ppmu->generic_events[ev] == 0)
444 return ERR_PTR(-EOPNOTSUPP); 469 return -EOPNOTSUPP;
445 ev = ppmu->generic_events[ev]; 470 ev = ppmu->generic_events[ev];
446 break; 471 break;
447 472
448 case PERF_TYPE_HW_CACHE: 473 case PERF_TYPE_HW_CACHE:
449 err = hw_perf_cache_event(event->attr.config, &ev); 474 err = hw_perf_cache_event(event->attr.config, &ev);
450 if (err) 475 if (err)
451 return ERR_PTR(err); 476 return err;
452 break; 477 break;
453 478
454 case PERF_TYPE_RAW: 479 case PERF_TYPE_RAW:
@@ -456,12 +481,12 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
456 break; 481 break;
457 482
458 default: 483 default:
459 return ERR_PTR(-EINVAL); 484 return -ENOENT;
460 } 485 }
461 486
462 event->hw.config = ppmu->xlate_event(ev); 487 event->hw.config = ppmu->xlate_event(ev);
463 if (!(event->hw.config & FSL_EMB_EVENT_VALID)) 488 if (!(event->hw.config & FSL_EMB_EVENT_VALID))
464 return ERR_PTR(-EINVAL); 489 return -EINVAL;
465 490
466 /* 491 /*
467 * If this is in a group, check if it can go on with all the 492 * If this is in a group, check if it can go on with all the
@@ -473,7 +498,7 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
473 n = collect_events(event->group_leader, 498 n = collect_events(event->group_leader,
474 ppmu->n_counter - 1, events); 499 ppmu->n_counter - 1, events);
475 if (n < 0) 500 if (n < 0)
476 return ERR_PTR(-EINVAL); 501 return -EINVAL;
477 } 502 }
478 503
479 if (event->hw.config & FSL_EMB_EVENT_RESTRICTED) { 504 if (event->hw.config & FSL_EMB_EVENT_RESTRICTED) {
@@ -484,7 +509,7 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
484 } 509 }
485 510
486 if (num_restricted >= ppmu->n_restricted) 511 if (num_restricted >= ppmu->n_restricted)
487 return ERR_PTR(-EINVAL); 512 return -EINVAL;
488 } 513 }
489 514
490 event->hw.idx = -1; 515 event->hw.idx = -1;
@@ -497,7 +522,7 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
497 if (event->attr.exclude_kernel) 522 if (event->attr.exclude_kernel)
498 event->hw.config_base |= PMLCA_FCS; 523 event->hw.config_base |= PMLCA_FCS;
499 if (event->attr.exclude_idle) 524 if (event->attr.exclude_idle)
500 return ERR_PTR(-ENOTSUPP); 525 return -ENOTSUPP;
501 526
502 event->hw.last_period = event->hw.sample_period; 527 event->hw.last_period = event->hw.sample_period;
503 local64_set(&event->hw.period_left, event->hw.last_period); 528 local64_set(&event->hw.period_left, event->hw.last_period);
@@ -523,11 +548,20 @@ const struct pmu *hw_perf_event_init(struct perf_event *event)
523 } 548 }
524 event->destroy = hw_perf_event_destroy; 549 event->destroy = hw_perf_event_destroy;
525 550
526 if (err) 551 return err;
527 return ERR_PTR(err);
528 return &fsl_emb_pmu;
529} 552}
530 553
554static struct pmu fsl_emb_pmu = {
555 .pmu_enable = fsl_emb_pmu_enable,
556 .pmu_disable = fsl_emb_pmu_disable,
557 .event_init = fsl_emb_pmu_event_init,
558 .add = fsl_emb_pmu_add,
559 .del = fsl_emb_pmu_del,
560 .start = fsl_emb_pmu_start,
561 .stop = fsl_emb_pmu_stop,
562 .read = fsl_emb_pmu_read,
563};
564
531/* 565/*
532 * A counter has overflowed; update its count and record 566 * A counter has overflowed; update its count and record
533 * things if requested. Note that interrupts are hard-disabled 567 * things if requested. Note that interrupts are hard-disabled
@@ -540,6 +574,11 @@ static void record_and_restart(struct perf_event *event, unsigned long val,
540 s64 prev, delta, left; 574 s64 prev, delta, left;
541 int record = 0; 575 int record = 0;
542 576
577 if (event->hw.state & PERF_HES_STOPPED) {
578 write_pmc(event->hw.idx, 0);
579 return;
580 }
581
543 /* we don't have to worry about interrupts here */ 582 /* we don't have to worry about interrupts here */
544 prev = local64_read(&event->hw.prev_count); 583 prev = local64_read(&event->hw.prev_count);
545 delta = (val - prev) & 0xfffffffful; 584 delta = (val - prev) & 0xfffffffful;
@@ -557,11 +596,17 @@ static void record_and_restart(struct perf_event *event, unsigned long val,
557 if (left <= 0) 596 if (left <= 0)
558 left = period; 597 left = period;
559 record = 1; 598 record = 1;
599 event->hw.last_period = event->hw.sample_period;
560 } 600 }
561 if (left < 0x80000000LL) 601 if (left < 0x80000000LL)
562 val = 0x80000000LL - left; 602 val = 0x80000000LL - left;
563 } 603 }
564 604
605 write_pmc(event->hw.idx, val);
606 local64_set(&event->hw.prev_count, val);
607 local64_set(&event->hw.period_left, left);
608 perf_event_update_userpage(event);
609
565 /* 610 /*
566 * Finally record data if requested. 611 * Finally record data if requested.
567 */ 612 */
@@ -571,23 +616,9 @@ static void record_and_restart(struct perf_event *event, unsigned long val,
571 perf_sample_data_init(&data, 0); 616 perf_sample_data_init(&data, 0);
572 data.period = event->hw.last_period; 617 data.period = event->hw.last_period;
573 618
574 if (perf_event_overflow(event, nmi, &data, regs)) { 619 if (perf_event_overflow(event, nmi, &data, regs))
575 /* 620 fsl_emb_pmu_stop(event, 0);
576 * Interrupts are coming too fast - throttle them
577 * by setting the event to 0, so it will be
578 * at least 2^30 cycles until the next interrupt
579 * (assuming each event counts at most 2 counts
580 * per cycle).
581 */
582 val = 0;
583 left = ~0ULL >> 1;
584 }
585 } 621 }
586
587 write_pmc(event->hw.idx, val);
588 local64_set(&event->hw.prev_count, val);
589 local64_set(&event->hw.period_left, left);
590 perf_event_update_userpage(event);
591} 622}
592 623
593static void perf_event_interrupt(struct pt_regs *regs) 624static void perf_event_interrupt(struct pt_regs *regs)
@@ -651,5 +682,7 @@ int register_fsl_emb_pmu(struct fsl_emb_pmu *pmu)
651 pr_info("%s performance monitor hardware support registered\n", 682 pr_info("%s performance monitor hardware support registered\n",
652 pmu->name); 683 pmu->name);
653 684
685 perf_pmu_register(&fsl_emb_pmu, "cpu", PERF_TYPE_RAW);
686
654 return 0; 687 return 0;
655} 688}
diff --git a/arch/powerpc/kernel/power4-pmu.c b/arch/powerpc/kernel/power4-pmu.c
index 2a361cdda635..ead8b3c2649e 100644
--- a/arch/powerpc/kernel/power4-pmu.c
+++ b/arch/powerpc/kernel/power4-pmu.c
@@ -613,4 +613,4 @@ static int init_power4_pmu(void)
613 return register_power_pmu(&power4_pmu); 613 return register_power_pmu(&power4_pmu);
614} 614}
615 615
616arch_initcall(init_power4_pmu); 616early_initcall(init_power4_pmu);
diff --git a/arch/powerpc/kernel/power5+-pmu.c b/arch/powerpc/kernel/power5+-pmu.c
index 199de527d411..eca0ac595cb6 100644
--- a/arch/powerpc/kernel/power5+-pmu.c
+++ b/arch/powerpc/kernel/power5+-pmu.c
@@ -682,4 +682,4 @@ static int init_power5p_pmu(void)
682 return register_power_pmu(&power5p_pmu); 682 return register_power_pmu(&power5p_pmu);
683} 683}
684 684
685arch_initcall(init_power5p_pmu); 685early_initcall(init_power5p_pmu);
diff --git a/arch/powerpc/kernel/power5-pmu.c b/arch/powerpc/kernel/power5-pmu.c
index 98b6a729a9dd..d5ff0f64a5e6 100644
--- a/arch/powerpc/kernel/power5-pmu.c
+++ b/arch/powerpc/kernel/power5-pmu.c
@@ -621,4 +621,4 @@ static int init_power5_pmu(void)
621 return register_power_pmu(&power5_pmu); 621 return register_power_pmu(&power5_pmu);
622} 622}
623 623
624arch_initcall(init_power5_pmu); 624early_initcall(init_power5_pmu);
diff --git a/arch/powerpc/kernel/power6-pmu.c b/arch/powerpc/kernel/power6-pmu.c
index 84a607bda8fb..31603927e376 100644
--- a/arch/powerpc/kernel/power6-pmu.c
+++ b/arch/powerpc/kernel/power6-pmu.c
@@ -544,4 +544,4 @@ static int init_power6_pmu(void)
544 return register_power_pmu(&power6_pmu); 544 return register_power_pmu(&power6_pmu);
545} 545}
546 546
547arch_initcall(init_power6_pmu); 547early_initcall(init_power6_pmu);
diff --git a/arch/powerpc/kernel/power7-pmu.c b/arch/powerpc/kernel/power7-pmu.c
index 852f7b7f6b40..593740fcb799 100644
--- a/arch/powerpc/kernel/power7-pmu.c
+++ b/arch/powerpc/kernel/power7-pmu.c
@@ -369,4 +369,4 @@ static int init_power7_pmu(void)
369 return register_power_pmu(&power7_pmu); 369 return register_power_pmu(&power7_pmu);
370} 370}
371 371
372arch_initcall(init_power7_pmu); 372early_initcall(init_power7_pmu);
diff --git a/arch/powerpc/kernel/ppc970-pmu.c b/arch/powerpc/kernel/ppc970-pmu.c
index 8eff48e20dba..9a6e093858fe 100644
--- a/arch/powerpc/kernel/ppc970-pmu.c
+++ b/arch/powerpc/kernel/ppc970-pmu.c
@@ -169,9 +169,11 @@ static int p970_marked_instr_event(u64 event)
169 switch (unit) { 169 switch (unit) {
170 case PM_VPU: 170 case PM_VPU:
171 mask = 0x4c; /* byte 0 bits 2,3,6 */ 171 mask = 0x4c; /* byte 0 bits 2,3,6 */
172 break;
172 case PM_LSU0: 173 case PM_LSU0:
173 /* byte 2 bits 0,2,3,4,6; all of byte 1 */ 174 /* byte 2 bits 0,2,3,4,6; all of byte 1 */
174 mask = 0x085dff00; 175 mask = 0x085dff00;
176 break;
175 case PM_LSU1L: 177 case PM_LSU1L:
176 mask = 0x50 << 24; /* byte 3 bits 4,6 */ 178 mask = 0x50 << 24; /* byte 3 bits 4,6 */
177 break; 179 break;
@@ -492,4 +494,4 @@ static int init_ppc970_pmu(void)
492 return register_power_pmu(&ppc970_pmu); 494 return register_power_pmu(&ppc970_pmu);
493} 495}
494 496
495arch_initcall(init_ppc970_pmu); 497early_initcall(init_ppc970_pmu);
diff --git a/arch/powerpc/kernel/ppc_ksyms.c b/arch/powerpc/kernel/ppc_ksyms.c
index ab3e392ac63c..7d28f540200c 100644
--- a/arch/powerpc/kernel/ppc_ksyms.c
+++ b/arch/powerpc/kernel/ppc_ksyms.c
@@ -54,7 +54,6 @@ extern void single_step_exception(struct pt_regs *regs);
54extern int sys_sigreturn(struct pt_regs *regs); 54extern int sys_sigreturn(struct pt_regs *regs);
55 55
56EXPORT_SYMBOL(clear_pages); 56EXPORT_SYMBOL(clear_pages);
57EXPORT_SYMBOL(copy_page);
58EXPORT_SYMBOL(ISA_DMA_THRESHOLD); 57EXPORT_SYMBOL(ISA_DMA_THRESHOLD);
59EXPORT_SYMBOL(DMA_MODE_READ); 58EXPORT_SYMBOL(DMA_MODE_READ);
60EXPORT_SYMBOL(DMA_MODE_WRITE); 59EXPORT_SYMBOL(DMA_MODE_WRITE);
@@ -88,9 +87,7 @@ EXPORT_SYMBOL(__copy_tofrom_user);
88EXPORT_SYMBOL(__clear_user); 87EXPORT_SYMBOL(__clear_user);
89EXPORT_SYMBOL(__strncpy_from_user); 88EXPORT_SYMBOL(__strncpy_from_user);
90EXPORT_SYMBOL(__strnlen_user); 89EXPORT_SYMBOL(__strnlen_user);
91#ifdef CONFIG_PPC64 90EXPORT_SYMBOL(copy_page);
92EXPORT_SYMBOL(copy_4K_page);
93#endif
94 91
95#if defined(CONFIG_PCI) && defined(CONFIG_PPC32) 92#if defined(CONFIG_PCI) && defined(CONFIG_PPC32)
96EXPORT_SYMBOL(isa_io_base); 93EXPORT_SYMBOL(isa_io_base);
@@ -186,3 +183,10 @@ EXPORT_SYMBOL(__mtdcr);
186EXPORT_SYMBOL(__mfdcr); 183EXPORT_SYMBOL(__mfdcr);
187#endif 184#endif
188EXPORT_SYMBOL(empty_zero_page); 185EXPORT_SYMBOL(empty_zero_page);
186
187#ifdef CONFIG_PPC64
188EXPORT_SYMBOL(__arch_hweight8);
189EXPORT_SYMBOL(__arch_hweight16);
190EXPORT_SYMBOL(__arch_hweight32);
191EXPORT_SYMBOL(__arch_hweight64);
192#endif
diff --git a/arch/powerpc/kernel/ppc_save_regs.S b/arch/powerpc/kernel/ppc_save_regs.S
index 5113bd2285e1..1b1787d52896 100644
--- a/arch/powerpc/kernel/ppc_save_regs.S
+++ b/arch/powerpc/kernel/ppc_save_regs.S
@@ -11,10 +11,11 @@
11#include <asm/processor.h> 11#include <asm/processor.h>
12#include <asm/ppc_asm.h> 12#include <asm/ppc_asm.h>
13#include <asm/asm-offsets.h> 13#include <asm/asm-offsets.h>
14#include <asm/ptrace.h>
14 15
15/* 16/*
16 * Grab the register values as they are now. 17 * Grab the register values as they are now.
17 * This won't do a particularily good job because we really 18 * This won't do a particularly good job because we really
18 * want our caller's caller's registers, and our caller has 19 * want our caller's caller's registers, and our caller has
19 * already executed its prologue. 20 * already executed its prologue.
20 * ToDo: We could reach back into the caller's save area to do 21 * ToDo: We could reach back into the caller's save area to do
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index b1c648a36b03..91e52df3d81d 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -353,6 +353,7 @@ static void switch_booke_debug_regs(struct thread_struct *new_thread)
353 prime_debug_regs(new_thread); 353 prime_debug_regs(new_thread);
354} 354}
355#else /* !CONFIG_PPC_ADV_DEBUG_REGS */ 355#else /* !CONFIG_PPC_ADV_DEBUG_REGS */
356#ifndef CONFIG_HAVE_HW_BREAKPOINT
356static void set_debug_reg_defaults(struct thread_struct *thread) 357static void set_debug_reg_defaults(struct thread_struct *thread)
357{ 358{
358 if (thread->dabr) { 359 if (thread->dabr) {
@@ -360,6 +361,7 @@ static void set_debug_reg_defaults(struct thread_struct *thread)
360 set_dabr(0); 361 set_dabr(0);
361 } 362 }
362} 363}
364#endif /* !CONFIG_HAVE_HW_BREAKPOINT */
363#endif /* CONFIG_PPC_ADV_DEBUG_REGS */ 365#endif /* CONFIG_PPC_ADV_DEBUG_REGS */
364 366
365int set_dabr(unsigned long dabr) 367int set_dabr(unsigned long dabr)
@@ -393,6 +395,9 @@ struct task_struct *__switch_to(struct task_struct *prev,
393 struct thread_struct *new_thread, *old_thread; 395 struct thread_struct *new_thread, *old_thread;
394 unsigned long flags; 396 unsigned long flags;
395 struct task_struct *last; 397 struct task_struct *last;
398#ifdef CONFIG_PPC_BOOK3S_64
399 struct ppc64_tlb_batch *batch;
400#endif
396 401
397#ifdef CONFIG_SMP 402#ifdef CONFIG_SMP
398 /* avoid complexity of lazy save/restore of fpu 403 /* avoid complexity of lazy save/restore of fpu
@@ -511,13 +516,22 @@ struct task_struct *__switch_to(struct task_struct *prev,
511 old_thread->accum_tb += (current_tb - start_tb); 516 old_thread->accum_tb += (current_tb - start_tb);
512 new_thread->start_tb = current_tb; 517 new_thread->start_tb = current_tb;
513 } 518 }
514#endif 519#endif /* CONFIG_PPC64 */
520
521#ifdef CONFIG_PPC_BOOK3S_64
522 batch = &__get_cpu_var(ppc64_tlb_batch);
523 if (batch->active) {
524 current_thread_info()->local_flags |= _TLF_LAZY_MMU;
525 if (batch->index)
526 __flush_tlb_pending(batch);
527 batch->active = 0;
528 }
529#endif /* CONFIG_PPC_BOOK3S_64 */
515 530
516 local_irq_save(flags); 531 local_irq_save(flags);
517 532
518 account_system_vtime(current); 533 account_system_vtime(current);
519 account_process_vtime(current); 534 account_process_vtime(current);
520 calculate_steal_time();
521 535
522 /* 536 /*
523 * We can't take a PMU exception inside _switch() since there is a 537 * We can't take a PMU exception inside _switch() since there is a
@@ -527,6 +541,14 @@ struct task_struct *__switch_to(struct task_struct *prev,
527 hard_irq_disable(); 541 hard_irq_disable();
528 last = _switch(old_thread, new_thread); 542 last = _switch(old_thread, new_thread);
529 543
544#ifdef CONFIG_PPC_BOOK3S_64
545 if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
546 current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
547 batch = &__get_cpu_var(ppc64_tlb_batch);
548 batch->active = 1;
549 }
550#endif /* CONFIG_PPC_BOOK3S_64 */
551
530 local_irq_restore(flags); 552 local_irq_restore(flags);
531 553
532 return last; 554 return last;
@@ -632,7 +654,7 @@ void show_regs(struct pt_regs * regs)
632#ifdef CONFIG_PPC_ADV_DEBUG_REGS 654#ifdef CONFIG_PPC_ADV_DEBUG_REGS
633 printk("DEAR: "REG", ESR: "REG"\n", regs->dar, regs->dsisr); 655 printk("DEAR: "REG", ESR: "REG"\n", regs->dar, regs->dsisr);
634#else 656#else
635 printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr); 657 printk("DAR: "REG", DSISR: %08lx\n", regs->dar, regs->dsisr);
636#endif 658#endif
637 printk("TASK = %p[%d] '%s' THREAD: %p", 659 printk("TASK = %p[%d] '%s' THREAD: %p",
638 current, task_pid_nr(current), current->comm, task_thread_info(current)); 660 current, task_pid_nr(current), current->comm, task_thread_info(current));
@@ -671,11 +693,11 @@ void flush_thread(void)
671{ 693{
672 discard_lazy_cpu_state(); 694 discard_lazy_cpu_state();
673 695
674#ifdef CONFIG_HAVE_HW_BREAKPOINTS 696#ifdef CONFIG_HAVE_HW_BREAKPOINT
675 flush_ptrace_hw_breakpoint(current); 697 flush_ptrace_hw_breakpoint(current);
676#else /* CONFIG_HAVE_HW_BREAKPOINTS */ 698#else /* CONFIG_HAVE_HW_BREAKPOINT */
677 set_debug_reg_defaults(&current->thread); 699 set_debug_reg_defaults(&current->thread);
678#endif /* CONFIG_HAVE_HW_BREAKPOINTS */ 700#endif /* CONFIG_HAVE_HW_BREAKPOINT */
679} 701}
680 702
681void 703void
@@ -701,6 +723,8 @@ void prepare_to_copy(struct task_struct *tsk)
701/* 723/*
702 * Copy a thread.. 724 * Copy a thread..
703 */ 725 */
726extern unsigned long dscr_default; /* defined in arch/powerpc/kernel/sysfs.c */
727
704int copy_thread(unsigned long clone_flags, unsigned long usp, 728int copy_thread(unsigned long clone_flags, unsigned long usp,
705 unsigned long unused, struct task_struct *p, 729 unsigned long unused, struct task_struct *p,
706 struct pt_regs *regs) 730 struct pt_regs *regs)
@@ -754,11 +778,11 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
754 _ALIGN_UP(sizeof(struct thread_info), 16); 778 _ALIGN_UP(sizeof(struct thread_info), 16);
755 779
756#ifdef CONFIG_PPC_STD_MMU_64 780#ifdef CONFIG_PPC_STD_MMU_64
757 if (cpu_has_feature(CPU_FTR_SLB)) { 781 if (mmu_has_feature(MMU_FTR_SLB)) {
758 unsigned long sp_vsid; 782 unsigned long sp_vsid;
759 unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp; 783 unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
760 784
761 if (cpu_has_feature(CPU_FTR_1T_SEGMENT)) 785 if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
762 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_1T) 786 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_1T)
763 << SLB_VSID_SHIFT_1T; 787 << SLB_VSID_SHIFT_1T;
764 else 788 else
@@ -768,6 +792,20 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
768 p->thread.ksp_vsid = sp_vsid; 792 p->thread.ksp_vsid = sp_vsid;
769 } 793 }
770#endif /* CONFIG_PPC_STD_MMU_64 */ 794#endif /* CONFIG_PPC_STD_MMU_64 */
795#ifdef CONFIG_PPC64
796 if (cpu_has_feature(CPU_FTR_DSCR)) {
797 if (current->thread.dscr_inherit) {
798 p->thread.dscr_inherit = 1;
799 p->thread.dscr = current->thread.dscr;
800 } else if (0 != dscr_default) {
801 p->thread.dscr_inherit = 1;
802 p->thread.dscr = dscr_default;
803 } else {
804 p->thread.dscr_inherit = 0;
805 p->thread.dscr = 0;
806 }
807 }
808#endif
771 809
772 /* 810 /*
773 * The PPC64 ABI makes use of a TOC to contain function 811 * The PPC64 ABI makes use of a TOC to contain function
@@ -1217,11 +1255,11 @@ void __ppc64_runlatch_off(void)
1217 1255
1218static struct kmem_cache *thread_info_cache; 1256static struct kmem_cache *thread_info_cache;
1219 1257
1220struct thread_info *alloc_thread_info(struct task_struct *tsk) 1258struct thread_info *alloc_thread_info_node(struct task_struct *tsk, int node)
1221{ 1259{
1222 struct thread_info *ti; 1260 struct thread_info *ti;
1223 1261
1224 ti = kmem_cache_alloc(thread_info_cache, GFP_KERNEL); 1262 ti = kmem_cache_alloc_node(thread_info_cache, GFP_KERNEL, node);
1225 if (unlikely(ti == NULL)) 1263 if (unlikely(ti == NULL))
1226 return NULL; 1264 return NULL;
1227#ifdef CONFIG_DEBUG_STACK_USAGE 1265#ifdef CONFIG_DEBUG_STACK_USAGE
@@ -1298,14 +1336,3 @@ unsigned long randomize_et_dyn(unsigned long base)
1298 1336
1299 return ret; 1337 return ret;
1300} 1338}
1301
1302#ifdef CONFIG_SMP
1303int arch_sd_sibling_asym_packing(void)
1304{
1305 if (cpu_has_feature(CPU_FTR_ASYM_SMT)) {
1306 printk_once(KERN_INFO "Enabling Asymmetric SMT scheduling\n");
1307 return SD_ASYM_PACKING;
1308 }
1309 return 0;
1310}
1311#endif
diff --git a/arch/powerpc/kernel/prom.c b/arch/powerpc/kernel/prom.c
index fed9bf6187d1..8c3112a57cf2 100644
--- a/arch/powerpc/kernel/prom.c
+++ b/arch/powerpc/kernel/prom.c
@@ -66,7 +66,9 @@
66int __initdata iommu_is_off; 66int __initdata iommu_is_off;
67int __initdata iommu_force_on; 67int __initdata iommu_force_on;
68unsigned long tce_alloc_start, tce_alloc_end; 68unsigned long tce_alloc_start, tce_alloc_end;
69u64 ppc64_rma_size;
69#endif 70#endif
71static phys_addr_t first_memblock_size;
70 72
71static int __init early_parse_mem(char *p) 73static int __init early_parse_mem(char *p)
72{ 74{
@@ -80,11 +82,29 @@ static int __init early_parse_mem(char *p)
80} 82}
81early_param("mem", early_parse_mem); 83early_param("mem", early_parse_mem);
82 84
85/*
86 * overlaps_initrd - check for overlap with page aligned extension of
87 * initrd.
88 */
89static inline int overlaps_initrd(unsigned long start, unsigned long size)
90{
91#ifdef CONFIG_BLK_DEV_INITRD
92 if (!initrd_start)
93 return 0;
94
95 return (start + size) > _ALIGN_DOWN(initrd_start, PAGE_SIZE) &&
96 start <= _ALIGN_UP(initrd_end, PAGE_SIZE);
97#else
98 return 0;
99#endif
100}
101
83/** 102/**
84 * move_device_tree - move tree to an unused area, if needed. 103 * move_device_tree - move tree to an unused area, if needed.
85 * 104 *
86 * The device tree may be allocated beyond our memory limit, or inside the 105 * The device tree may be allocated beyond our memory limit, or inside the
87 * crash kernel region for kdump. If so, move it out of the way. 106 * crash kernel region for kdump, or within the page aligned range of initrd.
107 * If so, move it out of the way.
88 */ 108 */
89static void __init move_device_tree(void) 109static void __init move_device_tree(void)
90{ 110{
@@ -96,9 +116,10 @@ static void __init move_device_tree(void)
96 start = __pa(initial_boot_params); 116 start = __pa(initial_boot_params);
97 size = be32_to_cpu(initial_boot_params->totalsize); 117 size = be32_to_cpu(initial_boot_params->totalsize);
98 118
99 if ((memory_limit && (start + size) > memory_limit) || 119 if ((memory_limit && (start + size) > PHYSICAL_START + memory_limit) ||
100 overlaps_crashkernel(start, size)) { 120 overlaps_crashkernel(start, size) ||
101 p = __va(memblock_alloc_base(size, PAGE_SIZE, memblock.rmo_size)); 121 overlaps_initrd(start, size)) {
122 p = __va(memblock_alloc(size, PAGE_SIZE));
102 memcpy(p, initial_boot_params, size); 123 memcpy(p, initial_boot_params, size);
103 initial_boot_params = (struct boot_param_header *)p; 124 initial_boot_params = (struct boot_param_header *)p;
104 DBG("Moved device tree to 0x%p\n", p); 125 DBG("Moved device tree to 0x%p\n", p);
@@ -122,18 +143,19 @@ static void __init move_device_tree(void)
122 */ 143 */
123static struct ibm_pa_feature { 144static struct ibm_pa_feature {
124 unsigned long cpu_features; /* CPU_FTR_xxx bit */ 145 unsigned long cpu_features; /* CPU_FTR_xxx bit */
146 unsigned long mmu_features; /* MMU_FTR_xxx bit */
125 unsigned int cpu_user_ftrs; /* PPC_FEATURE_xxx bit */ 147 unsigned int cpu_user_ftrs; /* PPC_FEATURE_xxx bit */
126 unsigned char pabyte; /* byte number in ibm,pa-features */ 148 unsigned char pabyte; /* byte number in ibm,pa-features */
127 unsigned char pabit; /* bit number (big-endian) */ 149 unsigned char pabit; /* bit number (big-endian) */
128 unsigned char invert; /* if 1, pa bit set => clear feature */ 150 unsigned char invert; /* if 1, pa bit set => clear feature */
129} ibm_pa_features[] __initdata = { 151} ibm_pa_features[] __initdata = {
130 {0, PPC_FEATURE_HAS_MMU, 0, 0, 0}, 152 {0, 0, PPC_FEATURE_HAS_MMU, 0, 0, 0},
131 {0, PPC_FEATURE_HAS_FPU, 0, 1, 0}, 153 {0, 0, PPC_FEATURE_HAS_FPU, 0, 1, 0},
132 {CPU_FTR_SLB, 0, 0, 2, 0}, 154 {0, MMU_FTR_SLB, 0, 0, 2, 0},
133 {CPU_FTR_CTRL, 0, 0, 3, 0}, 155 {CPU_FTR_CTRL, 0, 0, 0, 3, 0},
134 {CPU_FTR_NOEXECUTE, 0, 0, 6, 0}, 156 {CPU_FTR_NOEXECUTE, 0, 0, 0, 6, 0},
135 {CPU_FTR_NODSISRALIGN, 0, 1, 1, 1}, 157 {CPU_FTR_NODSISRALIGN, 0, 0, 1, 1, 1},
136 {CPU_FTR_CI_LARGE_PAGE, 0, 1, 2, 0}, 158 {0, MMU_FTR_CI_LARGE_PAGE, 0, 1, 2, 0},
137 {CPU_FTR_REAL_LE, PPC_FEATURE_TRUE_LE, 5, 0, 0}, 159 {CPU_FTR_REAL_LE, PPC_FEATURE_TRUE_LE, 5, 0, 0},
138}; 160};
139 161
@@ -165,9 +187,11 @@ static void __init scan_features(unsigned long node, unsigned char *ftrs,
165 if (bit ^ fp->invert) { 187 if (bit ^ fp->invert) {
166 cur_cpu_spec->cpu_features |= fp->cpu_features; 188 cur_cpu_spec->cpu_features |= fp->cpu_features;
167 cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftrs; 189 cur_cpu_spec->cpu_user_features |= fp->cpu_user_ftrs;
190 cur_cpu_spec->mmu_features |= fp->mmu_features;
168 } else { 191 } else {
169 cur_cpu_spec->cpu_features &= ~fp->cpu_features; 192 cur_cpu_spec->cpu_features &= ~fp->cpu_features;
170 cur_cpu_spec->cpu_user_features &= ~fp->cpu_user_ftrs; 193 cur_cpu_spec->cpu_user_features &= ~fp->cpu_user_ftrs;
194 cur_cpu_spec->mmu_features &= ~fp->mmu_features;
171 } 195 }
172 } 196 }
173} 197}
@@ -267,13 +291,13 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
267 const char *uname, int depth, 291 const char *uname, int depth,
268 void *data) 292 void *data)
269{ 293{
270 static int logical_cpuid = 0;
271 char *type = of_get_flat_dt_prop(node, "device_type", NULL); 294 char *type = of_get_flat_dt_prop(node, "device_type", NULL);
272 const u32 *prop; 295 const u32 *prop;
273 const u32 *intserv; 296 const u32 *intserv;
274 int i, nthreads; 297 int i, nthreads;
275 unsigned long len; 298 unsigned long len;
276 int found = 0; 299 int found = -1;
300 int found_thread = 0;
277 301
278 /* We are scanning "cpu" nodes only */ 302 /* We are scanning "cpu" nodes only */
279 if (type == NULL || strcmp(type, "cpu") != 0) 303 if (type == NULL || strcmp(type, "cpu") != 0)
@@ -297,11 +321,10 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
297 * version 2 of the kexec param format adds the phys cpuid of 321 * version 2 of the kexec param format adds the phys cpuid of
298 * booted proc. 322 * booted proc.
299 */ 323 */
300 if (initial_boot_params && initial_boot_params->version >= 2) { 324 if (initial_boot_params->version >= 2) {
301 if (intserv[i] == 325 if (intserv[i] == initial_boot_params->boot_cpuid_phys) {
302 initial_boot_params->boot_cpuid_phys) { 326 found = boot_cpu_count;
303 found = 1; 327 found_thread = i;
304 break;
305 } 328 }
306 } else { 329 } else {
307 /* 330 /*
@@ -310,23 +333,20 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
310 * off secondary threads. 333 * off secondary threads.
311 */ 334 */
312 if (of_get_flat_dt_prop(node, 335 if (of_get_flat_dt_prop(node,
313 "linux,boot-cpu", NULL) != NULL) { 336 "linux,boot-cpu", NULL) != NULL)
314 found = 1; 337 found = boot_cpu_count;
315 break;
316 }
317 } 338 }
318
319#ifdef CONFIG_SMP 339#ifdef CONFIG_SMP
320 /* logical cpu id is always 0 on UP kernels */ 340 /* logical cpu id is always 0 on UP kernels */
321 logical_cpuid++; 341 boot_cpu_count++;
322#endif 342#endif
323 } 343 }
324 344
325 if (found) { 345 if (found >= 0) {
326 DBG("boot cpu: logical %d physical %d\n", logical_cpuid, 346 DBG("boot cpu: logical %d physical %d\n", found,
327 intserv[i]); 347 intserv[found_thread]);
328 boot_cpuid = logical_cpuid; 348 boot_cpuid = found;
329 set_hard_smp_processor_id(boot_cpuid, intserv[i]); 349 set_hard_smp_processor_id(found, intserv[found_thread]);
330 350
331 /* 351 /*
332 * PAPR defines "logical" PVR values for cpus that 352 * PAPR defines "logical" PVR values for cpus that
@@ -363,10 +383,15 @@ static int __init early_init_dt_scan_cpus(unsigned long node,
363 return 0; 383 return 0;
364} 384}
365 385
366void __init early_init_dt_scan_chosen_arch(unsigned long node) 386int __init early_init_dt_scan_chosen_ppc(unsigned long node, const char *uname,
387 int depth, void *data)
367{ 388{
368 unsigned long *lprop; 389 unsigned long *lprop;
369 390
391 /* Use common scan routine to determine if this is the chosen node */
392 if (early_init_dt_scan_chosen(node, uname, depth, data) == 0)
393 return 0;
394
370#ifdef CONFIG_PPC64 395#ifdef CONFIG_PPC64
371 /* check if iommu is forced on or off */ 396 /* check if iommu is forced on or off */
372 if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL) 397 if (of_get_flat_dt_prop(node, "linux,iommu-off", NULL) != NULL)
@@ -398,6 +423,9 @@ void __init early_init_dt_scan_chosen_arch(unsigned long node)
398 if (lprop) 423 if (lprop)
399 crashk_res.end = crashk_res.start + *lprop - 1; 424 crashk_res.end = crashk_res.start + *lprop - 1;
400#endif 425#endif
426
427 /* break now */
428 return 1;
401} 429}
402 430
403#ifdef CONFIG_PPC_PSERIES 431#ifdef CONFIG_PPC_PSERIES
@@ -492,7 +520,7 @@ static int __init early_init_dt_scan_memory_ppc(unsigned long node,
492 520
493void __init early_init_dt_add_memory_arch(u64 base, u64 size) 521void __init early_init_dt_add_memory_arch(u64 base, u64 size)
494{ 522{
495#if defined(CONFIG_PPC64) 523#ifdef CONFIG_PPC64
496 if (iommu_is_off) { 524 if (iommu_is_off) {
497 if (base >= 0x80000000ul) 525 if (base >= 0x80000000ul)
498 return; 526 return;
@@ -500,15 +528,22 @@ void __init early_init_dt_add_memory_arch(u64 base, u64 size)
500 size = 0x80000000ul - base; 528 size = 0x80000000ul - base;
501 } 529 }
502#endif 530#endif
531 /* Keep track of the beginning of memory -and- the size of
532 * the very first block in the device-tree as it represents
533 * the RMA on ppc64 server
534 */
535 if (base < memstart_addr) {
536 memstart_addr = base;
537 first_memblock_size = size;
538 }
503 539
540 /* Add the chunk to the MEMBLOCK list */
504 memblock_add(base, size); 541 memblock_add(base, size);
505
506 memstart_addr = min((u64)memstart_addr, base);
507} 542}
508 543
509u64 __init early_init_dt_alloc_memory_arch(u64 size, u64 align) 544void * __init early_init_dt_alloc_memory_arch(u64 size, u64 align)
510{ 545{
511 return memblock_alloc(size, align); 546 return __va(memblock_alloc(size, align));
512} 547}
513 548
514#ifdef CONFIG_BLK_DEV_INITRD 549#ifdef CONFIG_BLK_DEV_INITRD
@@ -539,7 +574,9 @@ static void __init early_reserve_mem(void)
539#ifdef CONFIG_BLK_DEV_INITRD 574#ifdef CONFIG_BLK_DEV_INITRD
540 /* then reserve the initrd, if any */ 575 /* then reserve the initrd, if any */
541 if (initrd_start && (initrd_end > initrd_start)) 576 if (initrd_start && (initrd_end > initrd_start))
542 memblock_reserve(__pa(initrd_start), initrd_end - initrd_start); 577 memblock_reserve(_ALIGN_DOWN(__pa(initrd_start), PAGE_SIZE),
578 _ALIGN_UP(initrd_end, PAGE_SIZE) -
579 _ALIGN_DOWN(initrd_start, PAGE_SIZE));
543#endif /* CONFIG_BLK_DEV_INITRD */ 580#endif /* CONFIG_BLK_DEV_INITRD */
544 581
545#ifdef CONFIG_PPC32 582#ifdef CONFIG_PPC32
@@ -655,7 +692,6 @@ static void __init phyp_dump_reserve_mem(void)
655static inline void __init phyp_dump_reserve_mem(void) {} 692static inline void __init phyp_dump_reserve_mem(void) {}
656#endif /* CONFIG_PHYP_DUMP && CONFIG_PPC_RTAS */ 693#endif /* CONFIG_PHYP_DUMP && CONFIG_PPC_RTAS */
657 694
658
659void __init early_init_devtree(void *params) 695void __init early_init_devtree(void *params)
660{ 696{
661 phys_addr_t limit; 697 phys_addr_t limit;
@@ -671,7 +707,7 @@ void __init early_init_devtree(void *params)
671#endif 707#endif
672 708
673#ifdef CONFIG_PHYP_DUMP 709#ifdef CONFIG_PHYP_DUMP
674 /* scan tree to see if dump occured during last boot */ 710 /* scan tree to see if dump occurred during last boot */
675 of_scan_flat_dt(early_init_dt_scan_phyp_dump, NULL); 711 of_scan_flat_dt(early_init_dt_scan_phyp_dump, NULL);
676#endif 712#endif
677 713
@@ -679,12 +715,14 @@ void __init early_init_devtree(void *params)
679 * device-tree, including the platform type, initrd location and 715 * device-tree, including the platform type, initrd location and
680 * size, TCE reserve, and more ... 716 * size, TCE reserve, and more ...
681 */ 717 */
682 of_scan_flat_dt(early_init_dt_scan_chosen, NULL); 718 of_scan_flat_dt(early_init_dt_scan_chosen_ppc, cmd_line);
683 719
684 /* Scan memory nodes and rebuild MEMBLOCKs */ 720 /* Scan memory nodes and rebuild MEMBLOCKs */
685 memblock_init(); 721 memblock_init();
722
686 of_scan_flat_dt(early_init_dt_scan_root, NULL); 723 of_scan_flat_dt(early_init_dt_scan_root, NULL);
687 of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL); 724 of_scan_flat_dt(early_init_dt_scan_memory_ppc, NULL);
725 setup_initial_memory_limit(memstart_addr, first_memblock_size);
688 726
689 /* Save command line for /proc/cmdline and then parse parameters */ 727 /* Save command line for /proc/cmdline and then parse parameters */
690 strlcpy(boot_command_line, cmd_line, COMMAND_LINE_SIZE); 728 strlcpy(boot_command_line, cmd_line, COMMAND_LINE_SIZE);
@@ -726,7 +764,7 @@ void __init early_init_devtree(void *params)
726 764
727 DBG("Scanning CPUs ...\n"); 765 DBG("Scanning CPUs ...\n");
728 766
729 /* Retreive CPU related informations from the flat tree 767 /* Retrieve CPU related informations from the flat tree
730 * (altivec support, boot CPU ID, ...) 768 * (altivec support, boot CPU ID, ...)
731 */ 769 */
732 of_scan_flat_dt(early_init_dt_scan_cpus, NULL); 770 of_scan_flat_dt(early_init_dt_scan_cpus, NULL);
diff --git a/arch/powerpc/kernel/prom_init.c b/arch/powerpc/kernel/prom_init.c
index 941ff4dbc567..c016033ba78d 100644
--- a/arch/powerpc/kernel/prom_init.c
+++ b/arch/powerpc/kernel/prom_init.c
@@ -335,6 +335,7 @@ static void __init prom_printf(const char *format, ...)
335 const char *p, *q, *s; 335 const char *p, *q, *s;
336 va_list args; 336 va_list args;
337 unsigned long v; 337 unsigned long v;
338 long vs;
338 struct prom_t *_prom = &RELOC(prom); 339 struct prom_t *_prom = &RELOC(prom);
339 340
340 va_start(args, format); 341 va_start(args, format);
@@ -368,12 +369,35 @@ static void __init prom_printf(const char *format, ...)
368 v = va_arg(args, unsigned long); 369 v = va_arg(args, unsigned long);
369 prom_print_hex(v); 370 prom_print_hex(v);
370 break; 371 break;
372 case 'd':
373 ++q;
374 vs = va_arg(args, int);
375 if (vs < 0) {
376 prom_print(RELOC("-"));
377 vs = -vs;
378 }
379 prom_print_dec(vs);
380 break;
371 case 'l': 381 case 'l':
372 ++q; 382 ++q;
373 if (*q == 'u') { /* '%lu' */ 383 if (*q == 0)
384 break;
385 else if (*q == 'x') {
386 ++q;
387 v = va_arg(args, unsigned long);
388 prom_print_hex(v);
389 } else if (*q == 'u') { /* '%lu' */
374 ++q; 390 ++q;
375 v = va_arg(args, unsigned long); 391 v = va_arg(args, unsigned long);
376 prom_print_dec(v); 392 prom_print_dec(v);
393 } else if (*q == 'd') { /* %ld */
394 ++q;
395 vs = va_arg(args, long);
396 if (vs < 0) {
397 prom_print(RELOC("-"));
398 vs = -vs;
399 }
400 prom_print_dec(vs);
377 } 401 }
378 break; 402 break;
379 } 403 }
@@ -676,8 +700,10 @@ static void __init early_cmdline_parse(void)
676#endif /* CONFIG_PCI_MSI */ 700#endif /* CONFIG_PCI_MSI */
677#ifdef CONFIG_PPC_SMLPAR 701#ifdef CONFIG_PPC_SMLPAR
678#define OV5_CMO 0x80 /* Cooperative Memory Overcommitment */ 702#define OV5_CMO 0x80 /* Cooperative Memory Overcommitment */
703#define OV5_XCMO 0x40 /* Page Coalescing */
679#else 704#else
680#define OV5_CMO 0x00 705#define OV5_CMO 0x00
706#define OV5_XCMO 0x00
681#endif 707#endif
682#define OV5_TYPE1_AFFINITY 0x80 /* Type 1 NUMA affinity */ 708#define OV5_TYPE1_AFFINITY 0x80 /* Type 1 NUMA affinity */
683 709
@@ -732,7 +758,7 @@ static unsigned char ibm_architecture_vec[] = {
732 OV5_LPAR | OV5_SPLPAR | OV5_LARGE_PAGES | OV5_DRCONF_MEMORY | 758 OV5_LPAR | OV5_SPLPAR | OV5_LARGE_PAGES | OV5_DRCONF_MEMORY |
733 OV5_DONATE_DEDICATE_CPU | OV5_MSI, 759 OV5_DONATE_DEDICATE_CPU | OV5_MSI,
734 0, 760 0,
735 OV5_CMO, 761 OV5_CMO | OV5_XCMO,
736 OV5_TYPE1_AFFINITY, 762 OV5_TYPE1_AFFINITY,
737 0, 763 0,
738 0, 764 0,
diff --git a/arch/powerpc/kernel/prom_parse.c b/arch/powerpc/kernel/prom_parse.c
index 88334af038e5..47187cc2cf00 100644
--- a/arch/powerpc/kernel/prom_parse.c
+++ b/arch/powerpc/kernel/prom_parse.c
@@ -2,95 +2,11 @@
2 2
3#include <linux/kernel.h> 3#include <linux/kernel.h>
4#include <linux/string.h> 4#include <linux/string.h>
5#include <linux/pci_regs.h>
6#include <linux/module.h> 5#include <linux/module.h>
7#include <linux/ioport.h> 6#include <linux/ioport.h>
8#include <linux/etherdevice.h> 7#include <linux/etherdevice.h>
9#include <linux/of_address.h> 8#include <linux/of_address.h>
10#include <asm/prom.h> 9#include <asm/prom.h>
11#include <asm/pci-bridge.h>
12
13#ifdef CONFIG_PCI
14int of_irq_map_pci(struct pci_dev *pdev, struct of_irq *out_irq)
15{
16 struct device_node *dn, *ppnode;
17 struct pci_dev *ppdev;
18 u32 lspec;
19 u32 laddr[3];
20 u8 pin;
21 int rc;
22
23 /* Check if we have a device node, if yes, fallback to standard OF
24 * parsing
25 */
26 dn = pci_device_to_OF_node(pdev);
27 if (dn) {
28 rc = of_irq_map_one(dn, 0, out_irq);
29 if (!rc)
30 return rc;
31 }
32
33 /* Ok, we don't, time to have fun. Let's start by building up an
34 * interrupt spec. we assume #interrupt-cells is 1, which is standard
35 * for PCI. If you do different, then don't use that routine.
36 */
37 rc = pci_read_config_byte(pdev, PCI_INTERRUPT_PIN, &pin);
38 if (rc != 0)
39 return rc;
40 /* No pin, exit */
41 if (pin == 0)
42 return -ENODEV;
43
44 /* Now we walk up the PCI tree */
45 lspec = pin;
46 for (;;) {
47 /* Get the pci_dev of our parent */
48 ppdev = pdev->bus->self;
49
50 /* Ouch, it's a host bridge... */
51 if (ppdev == NULL) {
52#ifdef CONFIG_PPC64
53 ppnode = pci_bus_to_OF_node(pdev->bus);
54#else
55 struct pci_controller *host;
56 host = pci_bus_to_host(pdev->bus);
57 ppnode = host ? host->dn : NULL;
58#endif
59 /* No node for host bridge ? give up */
60 if (ppnode == NULL)
61 return -EINVAL;
62 } else
63 /* We found a P2P bridge, check if it has a node */
64 ppnode = pci_device_to_OF_node(ppdev);
65
66 /* Ok, we have found a parent with a device-node, hand over to
67 * the OF parsing code.
68 * We build a unit address from the linux device to be used for
69 * resolution. Note that we use the linux bus number which may
70 * not match your firmware bus numbering.
71 * Fortunately, in most cases, interrupt-map-mask doesn't include
72 * the bus number as part of the matching.
73 * You should still be careful about that though if you intend
74 * to rely on this function (you ship a firmware that doesn't
75 * create device nodes for all PCI devices).
76 */
77 if (ppnode)
78 break;
79
80 /* We can only get here if we hit a P2P bridge with no node,
81 * let's do standard swizzling and try again
82 */
83 lspec = pci_swizzle_interrupt_pin(pdev, lspec);
84 pdev = ppdev;
85 }
86
87 laddr[0] = (pdev->bus->number << 16)
88 | (pdev->devfn << 8);
89 laddr[1] = laddr[2] = 0;
90 return of_irq_map_raw(ppnode, &lspec, 1, laddr, out_irq);
91}
92EXPORT_SYMBOL_GPL(of_irq_map_pci);
93#endif /* CONFIG_PCI */
94 10
95void of_parse_dma_window(struct device_node *dn, const void *dma_window_prop, 11void of_parse_dma_window(struct device_node *dn, const void *dma_window_prop,
96 unsigned long *busno, unsigned long *phys, unsigned long *size) 12 unsigned long *busno, unsigned long *phys, unsigned long *size)
@@ -117,41 +33,3 @@ void of_parse_dma_window(struct device_node *dn, const void *dma_window_prop,
117 cells = prop ? *(u32 *)prop : of_n_size_cells(dn); 33 cells = prop ? *(u32 *)prop : of_n_size_cells(dn);
118 *size = of_read_number(dma_window, cells); 34 *size = of_read_number(dma_window, cells);
119} 35}
120
121/**
122 * Search the device tree for the best MAC address to use. 'mac-address' is
123 * checked first, because that is supposed to contain to "most recent" MAC
124 * address. If that isn't set, then 'local-mac-address' is checked next,
125 * because that is the default address. If that isn't set, then the obsolete
126 * 'address' is checked, just in case we're using an old device tree.
127 *
128 * Note that the 'address' property is supposed to contain a virtual address of
129 * the register set, but some DTS files have redefined that property to be the
130 * MAC address.
131 *
132 * All-zero MAC addresses are rejected, because those could be properties that
133 * exist in the device tree, but were not set by U-Boot. For example, the
134 * DTS could define 'mac-address' and 'local-mac-address', with zero MAC
135 * addresses. Some older U-Boots only initialized 'local-mac-address'. In
136 * this case, the real MAC is in 'local-mac-address', and 'mac-address' exists
137 * but is all zeros.
138*/
139const void *of_get_mac_address(struct device_node *np)
140{
141 struct property *pp;
142
143 pp = of_find_property(np, "mac-address", NULL);
144 if (pp && (pp->length == 6) && is_valid_ether_addr(pp->value))
145 return pp->value;
146
147 pp = of_find_property(np, "local-mac-address", NULL);
148 if (pp && (pp->length == 6) && is_valid_ether_addr(pp->value))
149 return pp->value;
150
151 pp = of_find_property(np, "address", NULL);
152 if (pp && (pp->length == 6) && is_valid_ether_addr(pp->value))
153 return pp->value;
154
155 return NULL;
156}
157EXPORT_SYMBOL(of_get_mac_address);
diff --git a/arch/powerpc/kernel/ptrace.c b/arch/powerpc/kernel/ptrace.c
index 11f3cd9c832f..cb22024f2b42 100644
--- a/arch/powerpc/kernel/ptrace.c
+++ b/arch/powerpc/kernel/ptrace.c
@@ -29,6 +29,7 @@
29#include <linux/signal.h> 29#include <linux/signal.h>
30#include <linux/seccomp.h> 30#include <linux/seccomp.h>
31#include <linux/audit.h> 31#include <linux/audit.h>
32#include <trace/syscall.h>
32#ifdef CONFIG_PPC32 33#ifdef CONFIG_PPC32
33#include <linux/module.h> 34#include <linux/module.h>
34#endif 35#endif
@@ -40,6 +41,9 @@
40#include <asm/pgtable.h> 41#include <asm/pgtable.h>
41#include <asm/system.h> 42#include <asm/system.h>
42 43
44#define CREATE_TRACE_POINTS
45#include <trace/events/syscalls.h>
46
43/* 47/*
44 * The parameter save area on the stack is used to store arguments being passed 48 * The parameter save area on the stack is used to store arguments being passed
45 * to callee function and is located at fixed offset from stack pointer. 49 * to callee function and is located at fixed offset from stack pointer.
@@ -229,12 +233,16 @@ static int gpr_get(struct task_struct *target, const struct user_regset *regset,
229 unsigned int pos, unsigned int count, 233 unsigned int pos, unsigned int count,
230 void *kbuf, void __user *ubuf) 234 void *kbuf, void __user *ubuf)
231{ 235{
232 int ret; 236 int i, ret;
233 237
234 if (target->thread.regs == NULL) 238 if (target->thread.regs == NULL)
235 return -EIO; 239 return -EIO;
236 240
237 CHECK_FULL_REGS(target->thread.regs); 241 if (!FULL_REGS(target->thread.regs)) {
242 /* We have a partial register set. Fill 14-31 with bogus values */
243 for (i = 14; i < 32; i++)
244 target->thread.regs->gpr[i] = NV_REG_POISON;
245 }
238 246
239 ret = user_regset_copyout(&pos, &count, &kbuf, &ubuf, 247 ret = user_regset_copyout(&pos, &count, &kbuf, &ubuf,
240 target->thread.regs, 248 target->thread.regs,
@@ -459,7 +467,7 @@ static int vr_set(struct task_struct *target, const struct user_regset *regset,
459#ifdef CONFIG_VSX 467#ifdef CONFIG_VSX
460/* 468/*
461 * Currently to set and and get all the vsx state, you need to call 469 * Currently to set and and get all the vsx state, you need to call
462 * the fp and VMX calls aswell. This only get/sets the lower 32 470 * the fp and VMX calls as well. This only get/sets the lower 32
463 * 128bit VSX registers. 471 * 128bit VSX registers.
464 */ 472 */
465 473
@@ -641,11 +649,16 @@ static int gpr32_get(struct task_struct *target,
641 compat_ulong_t *k = kbuf; 649 compat_ulong_t *k = kbuf;
642 compat_ulong_t __user *u = ubuf; 650 compat_ulong_t __user *u = ubuf;
643 compat_ulong_t reg; 651 compat_ulong_t reg;
652 int i;
644 653
645 if (target->thread.regs == NULL) 654 if (target->thread.regs == NULL)
646 return -EIO; 655 return -EIO;
647 656
648 CHECK_FULL_REGS(target->thread.regs); 657 if (!FULL_REGS(target->thread.regs)) {
658 /* We have a partial register set. Fill 14-31 with bogus values */
659 for (i = 14; i < 32; i++)
660 target->thread.regs->gpr[i] = NV_REG_POISON;
661 }
649 662
650 pos /= sizeof(reg); 663 pos /= sizeof(reg);
651 count /= sizeof(reg); 664 count /= sizeof(reg);
@@ -924,12 +937,16 @@ int ptrace_set_debugreg(struct task_struct *task, unsigned long addr,
924 if (data && !(data & DABR_TRANSLATION)) 937 if (data && !(data & DABR_TRANSLATION))
925 return -EIO; 938 return -EIO;
926#ifdef CONFIG_HAVE_HW_BREAKPOINT 939#ifdef CONFIG_HAVE_HW_BREAKPOINT
940 if (ptrace_get_breakpoints(task) < 0)
941 return -ESRCH;
942
927 bp = thread->ptrace_bps[0]; 943 bp = thread->ptrace_bps[0];
928 if ((!data) || !(data & (DABR_DATA_WRITE | DABR_DATA_READ))) { 944 if ((!data) || !(data & (DABR_DATA_WRITE | DABR_DATA_READ))) {
929 if (bp) { 945 if (bp) {
930 unregister_hw_breakpoint(bp); 946 unregister_hw_breakpoint(bp);
931 thread->ptrace_bps[0] = NULL; 947 thread->ptrace_bps[0] = NULL;
932 } 948 }
949 ptrace_put_breakpoints(task);
933 return 0; 950 return 0;
934 } 951 }
935 if (bp) { 952 if (bp) {
@@ -939,9 +956,12 @@ int ptrace_set_debugreg(struct task_struct *task, unsigned long addr,
939 (DABR_DATA_WRITE | DABR_DATA_READ), 956 (DABR_DATA_WRITE | DABR_DATA_READ),
940 &attr.bp_type); 957 &attr.bp_type);
941 ret = modify_user_hw_breakpoint(bp, &attr); 958 ret = modify_user_hw_breakpoint(bp, &attr);
942 if (ret) 959 if (ret) {
960 ptrace_put_breakpoints(task);
943 return ret; 961 return ret;
962 }
944 thread->ptrace_bps[0] = bp; 963 thread->ptrace_bps[0] = bp;
964 ptrace_put_breakpoints(task);
945 thread->dabr = data; 965 thread->dabr = data;
946 return 0; 966 return 0;
947 } 967 }
@@ -956,9 +976,12 @@ int ptrace_set_debugreg(struct task_struct *task, unsigned long addr,
956 ptrace_triggered, task); 976 ptrace_triggered, task);
957 if (IS_ERR(bp)) { 977 if (IS_ERR(bp)) {
958 thread->ptrace_bps[0] = NULL; 978 thread->ptrace_bps[0] = NULL;
979 ptrace_put_breakpoints(task);
959 return PTR_ERR(bp); 980 return PTR_ERR(bp);
960 } 981 }
961 982
983 ptrace_put_breakpoints(task);
984
962#endif /* CONFIG_HAVE_HW_BREAKPOINT */ 985#endif /* CONFIG_HAVE_HW_BREAKPOINT */
963 986
964 /* Move contents to the DABR register */ 987 /* Move contents to the DABR register */
@@ -1316,6 +1339,10 @@ static int set_dac_range(struct task_struct *child,
1316static long ppc_set_hwdebug(struct task_struct *child, 1339static long ppc_set_hwdebug(struct task_struct *child,
1317 struct ppc_hw_breakpoint *bp_info) 1340 struct ppc_hw_breakpoint *bp_info)
1318{ 1341{
1342#ifndef CONFIG_PPC_ADV_DEBUG_REGS
1343 unsigned long dabr;
1344#endif
1345
1319 if (bp_info->version != 1) 1346 if (bp_info->version != 1)
1320 return -ENOTSUPP; 1347 return -ENOTSUPP;
1321#ifdef CONFIG_PPC_ADV_DEBUG_REGS 1348#ifdef CONFIG_PPC_ADV_DEBUG_REGS
@@ -1353,11 +1380,10 @@ static long ppc_set_hwdebug(struct task_struct *child,
1353 /* 1380 /*
1354 * We only support one data breakpoint 1381 * We only support one data breakpoint
1355 */ 1382 */
1356 if (((bp_info->trigger_type & PPC_BREAKPOINT_TRIGGER_RW) == 0) || 1383 if ((bp_info->trigger_type & PPC_BREAKPOINT_TRIGGER_RW) == 0 ||
1357 ((bp_info->trigger_type & ~PPC_BREAKPOINT_TRIGGER_RW) != 0) || 1384 (bp_info->trigger_type & ~PPC_BREAKPOINT_TRIGGER_RW) != 0 ||
1358 (bp_info->trigger_type != PPC_BREAKPOINT_TRIGGER_WRITE) || 1385 bp_info->addr_mode != PPC_BREAKPOINT_MODE_EXACT ||
1359 (bp_info->addr_mode != PPC_BREAKPOINT_MODE_EXACT) || 1386 bp_info->condition_mode != PPC_BREAKPOINT_CONDITION_NONE)
1360 (bp_info->condition_mode != PPC_BREAKPOINT_CONDITION_NONE))
1361 return -EINVAL; 1387 return -EINVAL;
1362 1388
1363 if (child->thread.dabr) 1389 if (child->thread.dabr)
@@ -1366,7 +1392,14 @@ static long ppc_set_hwdebug(struct task_struct *child,
1366 if ((unsigned long)bp_info->addr >= TASK_SIZE) 1392 if ((unsigned long)bp_info->addr >= TASK_SIZE)
1367 return -EIO; 1393 return -EIO;
1368 1394
1369 child->thread.dabr = (unsigned long)bp_info->addr; 1395 dabr = (unsigned long)bp_info->addr & ~7UL;
1396 dabr |= DABR_TRANSLATION;
1397 if (bp_info->trigger_type & PPC_BREAKPOINT_TRIGGER_READ)
1398 dabr |= DABR_DATA_READ;
1399 if (bp_info->trigger_type & PPC_BREAKPOINT_TRIGGER_WRITE)
1400 dabr |= DABR_DATA_WRITE;
1401
1402 child->thread.dabr = dabr;
1370 1403
1371 return 1; 1404 return 1;
1372#endif /* !CONFIG_PPC_ADV_DEBUG_DVCS */ 1405#endif /* !CONFIG_PPC_ADV_DEBUG_DVCS */
@@ -1406,37 +1439,42 @@ static long ppc_del_hwdebug(struct task_struct *child, long addr, long data)
1406 * Here are the old "legacy" powerpc specific getregs/setregs ptrace calls, 1439 * Here are the old "legacy" powerpc specific getregs/setregs ptrace calls,
1407 * we mark them as obsolete now, they will be removed in a future version 1440 * we mark them as obsolete now, they will be removed in a future version
1408 */ 1441 */
1409static long arch_ptrace_old(struct task_struct *child, long request, long addr, 1442static long arch_ptrace_old(struct task_struct *child, long request,
1410 long data) 1443 unsigned long addr, unsigned long data)
1411{ 1444{
1445 void __user *datavp = (void __user *) data;
1446
1412 switch (request) { 1447 switch (request) {
1413 case PPC_PTRACE_GETREGS: /* Get GPRs 0 - 31. */ 1448 case PPC_PTRACE_GETREGS: /* Get GPRs 0 - 31. */
1414 return copy_regset_to_user(child, &user_ppc_native_view, 1449 return copy_regset_to_user(child, &user_ppc_native_view,
1415 REGSET_GPR, 0, 32 * sizeof(long), 1450 REGSET_GPR, 0, 32 * sizeof(long),
1416 (void __user *) data); 1451 datavp);
1417 1452
1418 case PPC_PTRACE_SETREGS: /* Set GPRs 0 - 31. */ 1453 case PPC_PTRACE_SETREGS: /* Set GPRs 0 - 31. */
1419 return copy_regset_from_user(child, &user_ppc_native_view, 1454 return copy_regset_from_user(child, &user_ppc_native_view,
1420 REGSET_GPR, 0, 32 * sizeof(long), 1455 REGSET_GPR, 0, 32 * sizeof(long),
1421 (const void __user *) data); 1456 datavp);
1422 1457
1423 case PPC_PTRACE_GETFPREGS: /* Get FPRs 0 - 31. */ 1458 case PPC_PTRACE_GETFPREGS: /* Get FPRs 0 - 31. */
1424 return copy_regset_to_user(child, &user_ppc_native_view, 1459 return copy_regset_to_user(child, &user_ppc_native_view,
1425 REGSET_FPR, 0, 32 * sizeof(double), 1460 REGSET_FPR, 0, 32 * sizeof(double),
1426 (void __user *) data); 1461 datavp);
1427 1462
1428 case PPC_PTRACE_SETFPREGS: /* Set FPRs 0 - 31. */ 1463 case PPC_PTRACE_SETFPREGS: /* Set FPRs 0 - 31. */
1429 return copy_regset_from_user(child, &user_ppc_native_view, 1464 return copy_regset_from_user(child, &user_ppc_native_view,
1430 REGSET_FPR, 0, 32 * sizeof(double), 1465 REGSET_FPR, 0, 32 * sizeof(double),
1431 (const void __user *) data); 1466 datavp);
1432 } 1467 }
1433 1468
1434 return -EPERM; 1469 return -EPERM;
1435} 1470}
1436 1471
1437long arch_ptrace(struct task_struct *child, long request, long addr, long data) 1472long arch_ptrace(struct task_struct *child, long request,
1473 unsigned long addr, unsigned long data)
1438{ 1474{
1439 int ret = -EPERM; 1475 int ret = -EPERM;
1476 void __user *datavp = (void __user *) data;
1477 unsigned long __user *datalp = datavp;
1440 1478
1441 switch (request) { 1479 switch (request) {
1442 /* read the word at location addr in the USER area. */ 1480 /* read the word at location addr in the USER area. */
@@ -1446,11 +1484,11 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1446 ret = -EIO; 1484 ret = -EIO;
1447 /* convert to index and check */ 1485 /* convert to index and check */
1448#ifdef CONFIG_PPC32 1486#ifdef CONFIG_PPC32
1449 index = (unsigned long) addr >> 2; 1487 index = addr >> 2;
1450 if ((addr & 3) || (index > PT_FPSCR) 1488 if ((addr & 3) || (index > PT_FPSCR)
1451 || (child->thread.regs == NULL)) 1489 || (child->thread.regs == NULL))
1452#else 1490#else
1453 index = (unsigned long) addr >> 3; 1491 index = addr >> 3;
1454 if ((addr & 7) || (index > PT_FPSCR)) 1492 if ((addr & 7) || (index > PT_FPSCR))
1455#endif 1493#endif
1456 break; 1494 break;
@@ -1463,7 +1501,7 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1463 tmp = ((unsigned long *)child->thread.fpr) 1501 tmp = ((unsigned long *)child->thread.fpr)
1464 [TS_FPRWIDTH * (index - PT_FPR0)]; 1502 [TS_FPRWIDTH * (index - PT_FPR0)];
1465 } 1503 }
1466 ret = put_user(tmp,(unsigned long __user *) data); 1504 ret = put_user(tmp, datalp);
1467 break; 1505 break;
1468 } 1506 }
1469 1507
@@ -1474,11 +1512,11 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1474 ret = -EIO; 1512 ret = -EIO;
1475 /* convert to index and check */ 1513 /* convert to index and check */
1476#ifdef CONFIG_PPC32 1514#ifdef CONFIG_PPC32
1477 index = (unsigned long) addr >> 2; 1515 index = addr >> 2;
1478 if ((addr & 3) || (index > PT_FPSCR) 1516 if ((addr & 3) || (index > PT_FPSCR)
1479 || (child->thread.regs == NULL)) 1517 || (child->thread.regs == NULL))
1480#else 1518#else
1481 index = (unsigned long) addr >> 3; 1519 index = addr >> 3;
1482 if ((addr & 7) || (index > PT_FPSCR)) 1520 if ((addr & 7) || (index > PT_FPSCR))
1483#endif 1521#endif
1484 break; 1522 break;
@@ -1525,11 +1563,11 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1525 dbginfo.features = 0; 1563 dbginfo.features = 0;
1526#endif /* CONFIG_PPC_ADV_DEBUG_REGS */ 1564#endif /* CONFIG_PPC_ADV_DEBUG_REGS */
1527 1565
1528 if (!access_ok(VERIFY_WRITE, data, 1566 if (!access_ok(VERIFY_WRITE, datavp,
1529 sizeof(struct ppc_debug_info))) 1567 sizeof(struct ppc_debug_info)))
1530 return -EFAULT; 1568 return -EFAULT;
1531 ret = __copy_to_user((struct ppc_debug_info __user *)data, 1569 ret = __copy_to_user(datavp, &dbginfo,
1532 &dbginfo, sizeof(struct ppc_debug_info)) ? 1570 sizeof(struct ppc_debug_info)) ?
1533 -EFAULT : 0; 1571 -EFAULT : 0;
1534 break; 1572 break;
1535 } 1573 }
@@ -1537,11 +1575,10 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1537 case PPC_PTRACE_SETHWDEBUG: { 1575 case PPC_PTRACE_SETHWDEBUG: {
1538 struct ppc_hw_breakpoint bp_info; 1576 struct ppc_hw_breakpoint bp_info;
1539 1577
1540 if (!access_ok(VERIFY_READ, data, 1578 if (!access_ok(VERIFY_READ, datavp,
1541 sizeof(struct ppc_hw_breakpoint))) 1579 sizeof(struct ppc_hw_breakpoint)))
1542 return -EFAULT; 1580 return -EFAULT;
1543 ret = __copy_from_user(&bp_info, 1581 ret = __copy_from_user(&bp_info, datavp,
1544 (struct ppc_hw_breakpoint __user *)data,
1545 sizeof(struct ppc_hw_breakpoint)) ? 1582 sizeof(struct ppc_hw_breakpoint)) ?
1546 -EFAULT : 0; 1583 -EFAULT : 0;
1547 if (!ret) 1584 if (!ret)
@@ -1560,11 +1597,9 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1560 if (addr > 0) 1597 if (addr > 0)
1561 break; 1598 break;
1562#ifdef CONFIG_PPC_ADV_DEBUG_REGS 1599#ifdef CONFIG_PPC_ADV_DEBUG_REGS
1563 ret = put_user(child->thread.dac1, 1600 ret = put_user(child->thread.dac1, datalp);
1564 (unsigned long __user *)data);
1565#else 1601#else
1566 ret = put_user(child->thread.dabr, 1602 ret = put_user(child->thread.dabr, datalp);
1567 (unsigned long __user *)data);
1568#endif 1603#endif
1569 break; 1604 break;
1570 } 1605 }
@@ -1580,7 +1615,7 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1580 return copy_regset_to_user(child, &user_ppc_native_view, 1615 return copy_regset_to_user(child, &user_ppc_native_view,
1581 REGSET_GPR, 1616 REGSET_GPR,
1582 0, sizeof(struct pt_regs), 1617 0, sizeof(struct pt_regs),
1583 (void __user *) data); 1618 datavp);
1584 1619
1585#ifdef CONFIG_PPC64 1620#ifdef CONFIG_PPC64
1586 case PTRACE_SETREGS64: 1621 case PTRACE_SETREGS64:
@@ -1589,19 +1624,19 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1589 return copy_regset_from_user(child, &user_ppc_native_view, 1624 return copy_regset_from_user(child, &user_ppc_native_view,
1590 REGSET_GPR, 1625 REGSET_GPR,
1591 0, sizeof(struct pt_regs), 1626 0, sizeof(struct pt_regs),
1592 (const void __user *) data); 1627 datavp);
1593 1628
1594 case PTRACE_GETFPREGS: /* Get the child FPU state (FPR0...31 + FPSCR) */ 1629 case PTRACE_GETFPREGS: /* Get the child FPU state (FPR0...31 + FPSCR) */
1595 return copy_regset_to_user(child, &user_ppc_native_view, 1630 return copy_regset_to_user(child, &user_ppc_native_view,
1596 REGSET_FPR, 1631 REGSET_FPR,
1597 0, sizeof(elf_fpregset_t), 1632 0, sizeof(elf_fpregset_t),
1598 (void __user *) data); 1633 datavp);
1599 1634
1600 case PTRACE_SETFPREGS: /* Set the child FPU state (FPR0...31 + FPSCR) */ 1635 case PTRACE_SETFPREGS: /* Set the child FPU state (FPR0...31 + FPSCR) */
1601 return copy_regset_from_user(child, &user_ppc_native_view, 1636 return copy_regset_from_user(child, &user_ppc_native_view,
1602 REGSET_FPR, 1637 REGSET_FPR,
1603 0, sizeof(elf_fpregset_t), 1638 0, sizeof(elf_fpregset_t),
1604 (const void __user *) data); 1639 datavp);
1605 1640
1606#ifdef CONFIG_ALTIVEC 1641#ifdef CONFIG_ALTIVEC
1607 case PTRACE_GETVRREGS: 1642 case PTRACE_GETVRREGS:
@@ -1609,40 +1644,40 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
1609 REGSET_VMX, 1644 REGSET_VMX,
1610 0, (33 * sizeof(vector128) + 1645 0, (33 * sizeof(vector128) +
1611 sizeof(u32)), 1646 sizeof(u32)),
1612 (void __user *) data); 1647 datavp);
1613 1648
1614 case PTRACE_SETVRREGS: 1649 case PTRACE_SETVRREGS:
1615 return copy_regset_from_user(child, &user_ppc_native_view, 1650 return copy_regset_from_user(child, &user_ppc_native_view,
1616 REGSET_VMX, 1651 REGSET_VMX,
1617 0, (33 * sizeof(vector128) + 1652 0, (33 * sizeof(vector128) +
1618 sizeof(u32)), 1653 sizeof(u32)),
1619 (const void __user *) data); 1654 datavp);
1620#endif 1655#endif
1621#ifdef CONFIG_VSX 1656#ifdef CONFIG_VSX
1622 case PTRACE_GETVSRREGS: 1657 case PTRACE_GETVSRREGS:
1623 return copy_regset_to_user(child, &user_ppc_native_view, 1658 return copy_regset_to_user(child, &user_ppc_native_view,
1624 REGSET_VSX, 1659 REGSET_VSX,
1625 0, 32 * sizeof(double), 1660 0, 32 * sizeof(double),
1626 (void __user *) data); 1661 datavp);
1627 1662
1628 case PTRACE_SETVSRREGS: 1663 case PTRACE_SETVSRREGS:
1629 return copy_regset_from_user(child, &user_ppc_native_view, 1664 return copy_regset_from_user(child, &user_ppc_native_view,
1630 REGSET_VSX, 1665 REGSET_VSX,
1631 0, 32 * sizeof(double), 1666 0, 32 * sizeof(double),
1632 (const void __user *) data); 1667 datavp);
1633#endif 1668#endif
1634#ifdef CONFIG_SPE 1669#ifdef CONFIG_SPE
1635 case PTRACE_GETEVRREGS: 1670 case PTRACE_GETEVRREGS:
1636 /* Get the child spe register state. */ 1671 /* Get the child spe register state. */
1637 return copy_regset_to_user(child, &user_ppc_native_view, 1672 return copy_regset_to_user(child, &user_ppc_native_view,
1638 REGSET_SPE, 0, 35 * sizeof(u32), 1673 REGSET_SPE, 0, 35 * sizeof(u32),
1639 (void __user *) data); 1674 datavp);
1640 1675
1641 case PTRACE_SETEVRREGS: 1676 case PTRACE_SETEVRREGS:
1642 /* Set the child spe register state. */ 1677 /* Set the child spe register state. */
1643 return copy_regset_from_user(child, &user_ppc_native_view, 1678 return copy_regset_from_user(child, &user_ppc_native_view,
1644 REGSET_SPE, 0, 35 * sizeof(u32), 1679 REGSET_SPE, 0, 35 * sizeof(u32),
1645 (const void __user *) data); 1680 datavp);
1646#endif 1681#endif
1647 1682
1648 /* Old reverse args ptrace callss */ 1683 /* Old reverse args ptrace callss */
@@ -1679,9 +1714,12 @@ long do_syscall_trace_enter(struct pt_regs *regs)
1679 */ 1714 */
1680 ret = -1L; 1715 ret = -1L;
1681 1716
1717 if (unlikely(test_thread_flag(TIF_SYSCALL_TRACEPOINT)))
1718 trace_sys_enter(regs, regs->gpr[0]);
1719
1682 if (unlikely(current->audit_context)) { 1720 if (unlikely(current->audit_context)) {
1683#ifdef CONFIG_PPC64 1721#ifdef CONFIG_PPC64
1684 if (!test_thread_flag(TIF_32BIT)) 1722 if (!is_32bit_task())
1685 audit_syscall_entry(AUDIT_ARCH_PPC64, 1723 audit_syscall_entry(AUDIT_ARCH_PPC64,
1686 regs->gpr[0], 1724 regs->gpr[0],
1687 regs->gpr[3], regs->gpr[4], 1725 regs->gpr[3], regs->gpr[4],
@@ -1707,6 +1745,9 @@ void do_syscall_trace_leave(struct pt_regs *regs)
1707 audit_syscall_exit((regs->ccr&0x10000000)?AUDITSC_FAILURE:AUDITSC_SUCCESS, 1745 audit_syscall_exit((regs->ccr&0x10000000)?AUDITSC_FAILURE:AUDITSC_SUCCESS,
1708 regs->result); 1746 regs->result);
1709 1747
1748 if (unlikely(test_thread_flag(TIF_SYSCALL_TRACEPOINT)))
1749 trace_sys_exit(regs, regs->result);
1750
1710 step = test_thread_flag(TIF_SINGLESTEP); 1751 step = test_thread_flag(TIF_SINGLESTEP);
1711 if (step || test_thread_flag(TIF_SYSCALL_TRACE)) 1752 if (step || test_thread_flag(TIF_SYSCALL_TRACE))
1712 tracehook_report_syscall_exit(regs, step); 1753 tracehook_report_syscall_exit(regs, step);
diff --git a/arch/powerpc/kernel/ptrace32.c b/arch/powerpc/kernel/ptrace32.c
index 8a6daf4129f6..69c4be917d07 100644
--- a/arch/powerpc/kernel/ptrace32.c
+++ b/arch/powerpc/kernel/ptrace32.c
@@ -280,7 +280,11 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
280 /* We only support one DABR and no IABRS at the moment */ 280 /* We only support one DABR and no IABRS at the moment */
281 if (addr > 0) 281 if (addr > 0)
282 break; 282 break;
283#ifdef CONFIG_PPC_ADV_DEBUG_REGS
284 ret = put_user(child->thread.dac1, (u32 __user *)data);
285#else
283 ret = put_user(child->thread.dabr, (u32 __user *)data); 286 ret = put_user(child->thread.dabr, (u32 __user *)data);
287#endif
284 break; 288 break;
285 } 289 }
286 290
@@ -312,6 +316,9 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
312 case PTRACE_SET_DEBUGREG: 316 case PTRACE_SET_DEBUGREG:
313 case PTRACE_SYSCALL: 317 case PTRACE_SYSCALL:
314 case PTRACE_CONT: 318 case PTRACE_CONT:
319 case PPC_PTRACE_GETHWDBGINFO:
320 case PPC_PTRACE_SETHWDEBUG:
321 case PPC_PTRACE_DELHWDEBUG:
315 ret = arch_ptrace(child, request, addr, data); 322 ret = arch_ptrace(child, request, addr, data);
316 break; 323 break;
317 324
diff --git a/arch/powerpc/kernel/rtas-rtc.c b/arch/powerpc/kernel/rtas-rtc.c
index 77578c093dda..c57c19358a26 100644
--- a/arch/powerpc/kernel/rtas-rtc.c
+++ b/arch/powerpc/kernel/rtas-rtc.c
@@ -4,6 +4,7 @@
4#include <linux/init.h> 4#include <linux/init.h>
5#include <linux/rtc.h> 5#include <linux/rtc.h>
6#include <linux/delay.h> 6#include <linux/delay.h>
7#include <linux/ratelimit.h>
7#include <asm/prom.h> 8#include <asm/prom.h>
8#include <asm/rtas.h> 9#include <asm/rtas.h>
9#include <asm/time.h> 10#include <asm/time.h>
@@ -29,9 +30,10 @@ unsigned long __init rtas_get_boot_time(void)
29 } 30 }
30 } while (wait_time && (get_tb() < max_wait_tb)); 31 } while (wait_time && (get_tb() < max_wait_tb));
31 32
32 if (error != 0 && printk_ratelimit()) { 33 if (error != 0) {
33 printk(KERN_WARNING "error: reading the clock failed (%d)\n", 34 printk_ratelimited(KERN_WARNING
34 error); 35 "error: reading the clock failed (%d)\n",
36 error);
35 return 0; 37 return 0;
36 } 38 }
37 39
@@ -55,19 +57,21 @@ void rtas_get_rtc_time(struct rtc_time *rtc_tm)
55 57
56 wait_time = rtas_busy_delay_time(error); 58 wait_time = rtas_busy_delay_time(error);
57 if (wait_time) { 59 if (wait_time) {
58 if (in_interrupt() && printk_ratelimit()) { 60 if (in_interrupt()) {
59 memset(rtc_tm, 0, sizeof(struct rtc_time)); 61 memset(rtc_tm, 0, sizeof(struct rtc_time));
60 printk(KERN_WARNING "error: reading clock" 62 printk_ratelimited(KERN_WARNING
61 " would delay interrupt\n"); 63 "error: reading clock "
64 "would delay interrupt\n");
62 return; /* delay not allowed */ 65 return; /* delay not allowed */
63 } 66 }
64 msleep(wait_time); 67 msleep(wait_time);
65 } 68 }
66 } while (wait_time && (get_tb() < max_wait_tb)); 69 } while (wait_time && (get_tb() < max_wait_tb));
67 70
68 if (error != 0 && printk_ratelimit()) { 71 if (error != 0) {
69 printk(KERN_WARNING "error: reading the clock failed (%d)\n", 72 printk_ratelimited(KERN_WARNING
70 error); 73 "error: reading the clock failed (%d)\n",
74 error);
71 return; 75 return;
72 } 76 }
73 77
@@ -99,9 +103,10 @@ int rtas_set_rtc_time(struct rtc_time *tm)
99 } 103 }
100 } while (wait_time && (get_tb() < max_wait_tb)); 104 } while (wait_time && (get_tb() < max_wait_tb));
101 105
102 if (error != 0 && printk_ratelimit()) 106 if (error != 0)
103 printk(KERN_WARNING "error: setting the clock failed (%d)\n", 107 printk_ratelimited(KERN_WARNING
104 error); 108 "error: setting the clock failed (%d)\n",
109 error);
105 110
106 return 0; 111 return 0;
107} 112}
diff --git a/arch/powerpc/kernel/rtas.c b/arch/powerpc/kernel/rtas.c
index 41048de3c6c3..271ff6318eda 100644
--- a/arch/powerpc/kernel/rtas.c
+++ b/arch/powerpc/kernel/rtas.c
@@ -41,6 +41,8 @@
41#include <asm/atomic.h> 41#include <asm/atomic.h>
42#include <asm/time.h> 42#include <asm/time.h>
43#include <asm/mmu.h> 43#include <asm/mmu.h>
44#include <asm/topology.h>
45#include <asm/pSeries_reconfig.h>
44 46
45struct rtas_t rtas = { 47struct rtas_t rtas = {
46 .lock = __ARCH_SPIN_LOCK_UNLOCKED 48 .lock = __ARCH_SPIN_LOCK_UNLOCKED
@@ -493,7 +495,7 @@ unsigned int rtas_busy_delay(int status)
493 495
494 might_sleep(); 496 might_sleep();
495 ms = rtas_busy_delay_time(status); 497 ms = rtas_busy_delay_time(status);
496 if (ms) 498 if (ms && need_resched())
497 msleep(ms); 499 msleep(ms);
498 500
499 return ms; 501 return ms;
@@ -713,6 +715,7 @@ static int __rtas_suspend_last_cpu(struct rtas_suspend_me_data *data, int wake_w
713 int cpu; 715 int cpu;
714 716
715 slb_set_size(SLB_MIN_SIZE); 717 slb_set_size(SLB_MIN_SIZE);
718 stop_topology_update();
716 printk(KERN_DEBUG "calling ibm,suspend-me on cpu %i\n", smp_processor_id()); 719 printk(KERN_DEBUG "calling ibm,suspend-me on cpu %i\n", smp_processor_id());
717 720
718 while (rc == H_MULTI_THREADS_ACTIVE && !atomic_read(&data->done) && 721 while (rc == H_MULTI_THREADS_ACTIVE && !atomic_read(&data->done) &&
@@ -728,6 +731,8 @@ static int __rtas_suspend_last_cpu(struct rtas_suspend_me_data *data, int wake_w
728 rc = atomic_read(&data->error); 731 rc = atomic_read(&data->error);
729 732
730 atomic_set(&data->error, rc); 733 atomic_set(&data->error, rc);
734 start_topology_update();
735 pSeries_coalesce_init();
731 736
732 if (wake_when_done) { 737 if (wake_when_done) {
733 atomic_set(&data->done, 1); 738 atomic_set(&data->done, 1);
@@ -805,7 +810,7 @@ static void rtas_percpu_suspend_me(void *info)
805 __rtas_suspend_cpu((struct rtas_suspend_me_data *)info, 1); 810 __rtas_suspend_cpu((struct rtas_suspend_me_data *)info, 1);
806} 811}
807 812
808static int rtas_ibm_suspend_me(struct rtas_args *args) 813int rtas_ibm_suspend_me(struct rtas_args *args)
809{ 814{
810 long state; 815 long state;
811 long rc; 816 long rc;
@@ -855,7 +860,7 @@ static int rtas_ibm_suspend_me(struct rtas_args *args)
855 return atomic_read(&data.error); 860 return atomic_read(&data.error);
856} 861}
857#else /* CONFIG_PPC_PSERIES */ 862#else /* CONFIG_PPC_PSERIES */
858static int rtas_ibm_suspend_me(struct rtas_args *args) 863int rtas_ibm_suspend_me(struct rtas_args *args)
859{ 864{
860 return -ENOSYS; 865 return -ENOSYS;
861} 866}
@@ -969,7 +974,7 @@ void __init rtas_initialize(void)
969 */ 974 */
970#ifdef CONFIG_PPC64 975#ifdef CONFIG_PPC64
971 if (machine_is(pseries) && firmware_has_feature(FW_FEATURE_LPAR)) { 976 if (machine_is(pseries) && firmware_has_feature(FW_FEATURE_LPAR)) {
972 rtas_region = min(memblock.rmo_size, RTAS_INSTANTIATE_MAX); 977 rtas_region = min(ppc64_rma_size, RTAS_INSTANTIATE_MAX);
973 ibm_suspend_me_token = rtas_token("ibm,suspend-me"); 978 ibm_suspend_me_token = rtas_token("ibm,suspend-me");
974 } 979 }
975#endif 980#endif
diff --git a/arch/powerpc/kernel/rtas_flash.c b/arch/powerpc/kernel/rtas_flash.c
index 67a84d8f118d..bf5f5ce3a7bd 100644
--- a/arch/powerpc/kernel/rtas_flash.c
+++ b/arch/powerpc/kernel/rtas_flash.c
@@ -256,31 +256,16 @@ static ssize_t rtas_flash_read(struct file *file, char __user *buf,
256 struct proc_dir_entry *dp = PDE(file->f_path.dentry->d_inode); 256 struct proc_dir_entry *dp = PDE(file->f_path.dentry->d_inode);
257 struct rtas_update_flash_t *uf; 257 struct rtas_update_flash_t *uf;
258 char msg[RTAS_MSG_MAXLEN]; 258 char msg[RTAS_MSG_MAXLEN];
259 int msglen;
260 259
261 uf = (struct rtas_update_flash_t *) dp->data; 260 uf = dp->data;
262 261
263 if (!strcmp(dp->name, FIRMWARE_FLASH_NAME)) { 262 if (!strcmp(dp->name, FIRMWARE_FLASH_NAME)) {
264 get_flash_status_msg(uf->status, msg); 263 get_flash_status_msg(uf->status, msg);
265 } else { /* FIRMWARE_UPDATE_NAME */ 264 } else { /* FIRMWARE_UPDATE_NAME */
266 sprintf(msg, "%d\n", uf->status); 265 sprintf(msg, "%d\n", uf->status);
267 } 266 }
268 msglen = strlen(msg);
269 if (msglen > count)
270 msglen = count;
271
272 if (ppos && *ppos != 0)
273 return 0; /* be cheap */
274
275 if (!access_ok(VERIFY_WRITE, buf, msglen))
276 return -EINVAL;
277 267
278 if (copy_to_user(buf, msg, msglen)) 268 return simple_read_from_buffer(buf, count, ppos, msg, strlen(msg));
279 return -EFAULT;
280
281 if (ppos)
282 *ppos = msglen;
283 return msglen;
284} 269}
285 270
286/* constructor for flash_block_cache */ 271/* constructor for flash_block_cache */
@@ -394,26 +379,13 @@ static ssize_t manage_flash_read(struct file *file, char __user *buf,
394 char msg[RTAS_MSG_MAXLEN]; 379 char msg[RTAS_MSG_MAXLEN];
395 int msglen; 380 int msglen;
396 381
397 args_buf = (struct rtas_manage_flash_t *) dp->data; 382 args_buf = dp->data;
398 if (args_buf == NULL) 383 if (args_buf == NULL)
399 return 0; 384 return 0;
400 385
401 msglen = sprintf(msg, "%d\n", args_buf->status); 386 msglen = sprintf(msg, "%d\n", args_buf->status);
402 if (msglen > count)
403 msglen = count;
404 387
405 if (ppos && *ppos != 0) 388 return simple_read_from_buffer(buf, count, ppos, msg, msglen);
406 return 0; /* be cheap */
407
408 if (!access_ok(VERIFY_WRITE, buf, msglen))
409 return -EINVAL;
410
411 if (copy_to_user(buf, msg, msglen))
412 return -EFAULT;
413
414 if (ppos)
415 *ppos = msglen;
416 return msglen;
417} 389}
418 390
419static ssize_t manage_flash_write(struct file *file, const char __user *buf, 391static ssize_t manage_flash_write(struct file *file, const char __user *buf,
@@ -495,24 +467,11 @@ static ssize_t validate_flash_read(struct file *file, char __user *buf,
495 char msg[RTAS_MSG_MAXLEN]; 467 char msg[RTAS_MSG_MAXLEN];
496 int msglen; 468 int msglen;
497 469
498 args_buf = (struct rtas_validate_flash_t *) dp->data; 470 args_buf = dp->data;
499 471
500 if (ppos && *ppos != 0)
501 return 0; /* be cheap */
502
503 msglen = get_validate_flash_msg(args_buf, msg); 472 msglen = get_validate_flash_msg(args_buf, msg);
504 if (msglen > count)
505 msglen = count;
506
507 if (!access_ok(VERIFY_WRITE, buf, msglen))
508 return -EINVAL;
509
510 if (copy_to_user(buf, msg, msglen))
511 return -EFAULT;
512 473
513 if (ppos) 474 return simple_read_from_buffer(buf, count, ppos, msg, msglen);
514 *ppos = msglen;
515 return msglen;
516} 475}
517 476
518static ssize_t validate_flash_write(struct file *file, const char __user *buf, 477static ssize_t validate_flash_write(struct file *file, const char __user *buf,
@@ -716,6 +675,7 @@ static const struct file_operations rtas_flash_operations = {
716 .write = rtas_flash_write, 675 .write = rtas_flash_write,
717 .open = rtas_excl_open, 676 .open = rtas_excl_open,
718 .release = rtas_flash_release, 677 .release = rtas_flash_release,
678 .llseek = default_llseek,
719}; 679};
720 680
721static const struct file_operations manage_flash_operations = { 681static const struct file_operations manage_flash_operations = {
@@ -724,6 +684,7 @@ static const struct file_operations manage_flash_operations = {
724 .write = manage_flash_write, 684 .write = manage_flash_write,
725 .open = rtas_excl_open, 685 .open = rtas_excl_open,
726 .release = rtas_excl_release, 686 .release = rtas_excl_release,
687 .llseek = default_llseek,
727}; 688};
728 689
729static const struct file_operations validate_flash_operations = { 690static const struct file_operations validate_flash_operations = {
@@ -732,6 +693,7 @@ static const struct file_operations validate_flash_operations = {
732 .write = validate_flash_write, 693 .write = validate_flash_write,
733 .open = rtas_excl_open, 694 .open = rtas_excl_open,
734 .release = validate_flash_release, 695 .release = validate_flash_release,
696 .llseek = default_llseek,
735}; 697};
736 698
737static int __init rtas_flash_init(void) 699static int __init rtas_flash_init(void)
diff --git a/arch/powerpc/kernel/rtasd.c b/arch/powerpc/kernel/rtasd.c
index 638883e23e3a..67f6c3b51357 100644
--- a/arch/powerpc/kernel/rtasd.c
+++ b/arch/powerpc/kernel/rtasd.c
@@ -160,7 +160,7 @@ static int log_rtas_len(char * buf)
160 /* rtas fixed header */ 160 /* rtas fixed header */
161 len = 8; 161 len = 8;
162 err = (struct rtas_error_log *)buf; 162 err = (struct rtas_error_log *)buf;
163 if (err->extended_log_length) { 163 if (err->extended && err->extended_log_length) {
164 164
165 /* extended header */ 165 /* extended header */
166 len += err->extended_log_length; 166 len += err->extended_log_length;
@@ -354,6 +354,7 @@ static const struct file_operations proc_rtas_log_operations = {
354 .poll = rtas_log_poll, 354 .poll = rtas_log_poll,
355 .open = rtas_log_open, 355 .open = rtas_log_open,
356 .release = rtas_log_release, 356 .release = rtas_log_release,
357 .llseek = noop_llseek,
357}; 358};
358 359
359static int enable_surveillance(int timeout) 360static int enable_surveillance(int timeout)
@@ -411,7 +412,8 @@ static void rtas_event_scan(struct work_struct *w)
411 412
412 get_online_cpus(); 413 get_online_cpus();
413 414
414 cpu = cpumask_next(smp_processor_id(), cpu_online_mask); 415 /* raw_ OK because just using CPU as starting point. */
416 cpu = cpumask_next(raw_smp_processor_id(), cpu_online_mask);
415 if (cpu >= nr_cpu_ids) { 417 if (cpu >= nr_cpu_ids) {
416 cpu = cpumask_first(cpu_online_mask); 418 cpu = cpumask_first(cpu_online_mask);
417 419
@@ -463,7 +465,7 @@ static void start_event_scan(void)
463 pr_debug("rtasd: will sleep for %d milliseconds\n", 465 pr_debug("rtasd: will sleep for %d milliseconds\n",
464 (30000 / rtas_event_scan_rate)); 466 (30000 / rtas_event_scan_rate));
465 467
466 /* Retreive errors from nvram if any */ 468 /* Retrieve errors from nvram if any */
467 retreive_nvram_error_log(); 469 retreive_nvram_error_log();
468 470
469 schedule_delayed_work_on(cpumask_first(cpu_online_mask), 471 schedule_delayed_work_on(cpumask_first(cpu_online_mask),
diff --git a/arch/powerpc/kernel/setup-common.c b/arch/powerpc/kernel/setup-common.c
index 9d4882a46647..79fca2651b65 100644
--- a/arch/powerpc/kernel/setup-common.c
+++ b/arch/powerpc/kernel/setup-common.c
@@ -381,7 +381,7 @@ static void __init cpu_init_thread_core_maps(int tpc)
381 int i; 381 int i;
382 382
383 threads_per_core = tpc; 383 threads_per_core = tpc;
384 threads_core_mask = CPU_MASK_NONE; 384 cpumask_clear(&threads_core_mask);
385 385
386 /* This implementation only supports power of 2 number of threads 386 /* This implementation only supports power of 2 number of threads
387 * for simplicity and performance 387 * for simplicity and performance
@@ -390,7 +390,7 @@ static void __init cpu_init_thread_core_maps(int tpc)
390 BUG_ON(tpc != (1 << threads_shift)); 390 BUG_ON(tpc != (1 << threads_shift));
391 391
392 for (i = 0; i < tpc; i++) 392 for (i = 0; i < tpc; i++)
393 cpu_set(i, threads_core_mask); 393 cpumask_set_cpu(i, &threads_core_mask);
394 394
395 printk(KERN_INFO "CPU maps initialized for %d thread%s per core\n", 395 printk(KERN_INFO "CPU maps initialized for %d thread%s per core\n",
396 tpc, tpc > 1 ? "s" : ""); 396 tpc, tpc > 1 ? "s" : "");
@@ -404,7 +404,7 @@ static void __init cpu_init_thread_core_maps(int tpc)
404 * cpu_present_mask 404 * cpu_present_mask
405 * 405 *
406 * Having the possible map set up early allows us to restrict allocations 406 * Having the possible map set up early allows us to restrict allocations
407 * of things like irqstacks to num_possible_cpus() rather than NR_CPUS. 407 * of things like irqstacks to nr_cpu_ids rather than NR_CPUS.
408 * 408 *
409 * We do not initialize the online map here; cpus set their own bits in 409 * We do not initialize the online map here; cpus set their own bits in
410 * cpu_online_mask as they come up. 410 * cpu_online_mask as they come up.
@@ -424,7 +424,7 @@ void __init smp_setup_cpu_maps(void)
424 424
425 DBG("smp_setup_cpu_maps()\n"); 425 DBG("smp_setup_cpu_maps()\n");
426 426
427 while ((dn = of_find_node_by_type(dn, "cpu")) && cpu < NR_CPUS) { 427 while ((dn = of_find_node_by_type(dn, "cpu")) && cpu < nr_cpu_ids) {
428 const int *intserv; 428 const int *intserv;
429 int j, len; 429 int j, len;
430 430
@@ -443,7 +443,7 @@ void __init smp_setup_cpu_maps(void)
443 intserv = &cpu; /* assume logical == phys */ 443 intserv = &cpu; /* assume logical == phys */
444 } 444 }
445 445
446 for (j = 0; j < nthreads && cpu < NR_CPUS; j++) { 446 for (j = 0; j < nthreads && cpu < nr_cpu_ids; j++) {
447 DBG(" thread %d -> cpu %d (hard id %d)\n", 447 DBG(" thread %d -> cpu %d (hard id %d)\n",
448 j, cpu, intserv[j]); 448 j, cpu, intserv[j]);
449 set_cpu_present(cpu, true); 449 set_cpu_present(cpu, true);
@@ -483,12 +483,12 @@ void __init smp_setup_cpu_maps(void)
483 if (cpu_has_feature(CPU_FTR_SMT)) 483 if (cpu_has_feature(CPU_FTR_SMT))
484 maxcpus *= nthreads; 484 maxcpus *= nthreads;
485 485
486 if (maxcpus > NR_CPUS) { 486 if (maxcpus > nr_cpu_ids) {
487 printk(KERN_WARNING 487 printk(KERN_WARNING
488 "Partition configured for %d cpus, " 488 "Partition configured for %d cpus, "
489 "operating system maximum is %d.\n", 489 "operating system maximum is %d.\n",
490 maxcpus, NR_CPUS); 490 maxcpus, nr_cpu_ids);
491 maxcpus = NR_CPUS; 491 maxcpus = nr_cpu_ids;
492 } else 492 } else
493 printk(KERN_INFO "Partition configured for %d cpus.\n", 493 printk(KERN_INFO "Partition configured for %d cpus.\n",
494 maxcpus); 494 maxcpus);
@@ -509,6 +509,9 @@ void __init smp_setup_cpu_maps(void)
509 */ 509 */
510 cpu_init_thread_core_maps(nthreads); 510 cpu_init_thread_core_maps(nthreads);
511 511
512 /* Now that possible cpus are set, set nr_cpu_ids for later use */
513 setup_nr_cpu_ids();
514
512 free_unused_pacas(); 515 free_unused_pacas();
513} 516}
514#endif /* CONFIG_SMP */ 517#endif /* CONFIG_SMP */
@@ -599,6 +602,10 @@ int check_legacy_ioport(unsigned long base_port)
599 * name instead */ 602 * name instead */
600 if (!np) 603 if (!np)
601 np = of_find_node_by_name(NULL, "8042"); 604 np = of_find_node_by_name(NULL, "8042");
605 if (np) {
606 of_i8042_kbd_irq = 1;
607 of_i8042_aux_irq = 12;
608 }
602 break; 609 break;
603 case FDC_BASE: /* FDC1 */ 610 case FDC_BASE: /* FDC1 */
604 np = of_find_node_by_type(NULL, "fdc"); 611 np = of_find_node_by_type(NULL, "fdc");
diff --git a/arch/powerpc/kernel/setup_32.c b/arch/powerpc/kernel/setup_32.c
index 93666f9cabf1..620d792b52e4 100644
--- a/arch/powerpc/kernel/setup_32.c
+++ b/arch/powerpc/kernel/setup_32.c
@@ -46,8 +46,9 @@
46 46
47extern void bootx_init(unsigned long r4, unsigned long phys); 47extern void bootx_init(unsigned long r4, unsigned long phys);
48 48
49int boot_cpuid; 49int boot_cpuid = -1;
50EXPORT_SYMBOL_GPL(boot_cpuid); 50EXPORT_SYMBOL_GPL(boot_cpuid);
51int __initdata boot_cpu_count;
51int boot_cpuid_phys; 52int boot_cpuid_phys;
52 53
53int smp_hw_index[NR_CPUS]; 54int smp_hw_index[NR_CPUS];
@@ -246,7 +247,7 @@ static void __init irqstack_early_init(void)
246 unsigned int i; 247 unsigned int i;
247 248
248 /* interrupt stacks must be in lowmem, we get that for free on ppc32 249 /* interrupt stacks must be in lowmem, we get that for free on ppc32
249 * as the memblock is limited to lowmem by MEMBLOCK_REAL_LIMIT */ 250 * as the memblock is limited to lowmem by default */
250 for_each_possible_cpu(i) { 251 for_each_possible_cpu(i) {
251 softirq_ctx[i] = (struct thread_info *) 252 softirq_ctx[i] = (struct thread_info *)
252 __va(memblock_alloc(THREAD_SIZE, THREAD_SIZE)); 253 __va(memblock_alloc(THREAD_SIZE, THREAD_SIZE));
diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
index e72690ec9b87..a88bf2713d41 100644
--- a/arch/powerpc/kernel/setup_64.c
+++ b/arch/powerpc/kernel/setup_64.c
@@ -62,6 +62,7 @@
62#include <asm/udbg.h> 62#include <asm/udbg.h>
63#include <asm/kexec.h> 63#include <asm/kexec.h>
64#include <asm/mmu_context.h> 64#include <asm/mmu_context.h>
65#include <asm/code-patching.h>
65 66
66#include "setup.h" 67#include "setup.h"
67 68
@@ -72,6 +73,7 @@
72#endif 73#endif
73 74
74int boot_cpuid = 0; 75int boot_cpuid = 0;
76int __initdata boot_cpu_count;
75u64 ppc64_pft_size; 77u64 ppc64_pft_size;
76 78
77/* Pick defaults since we might want to patch instructions 79/* Pick defaults since we might want to patch instructions
@@ -233,6 +235,7 @@ void early_setup_secondary(void)
233void smp_release_cpus(void) 235void smp_release_cpus(void)
234{ 236{
235 unsigned long *ptr; 237 unsigned long *ptr;
238 int i;
236 239
237 DBG(" -> smp_release_cpus()\n"); 240 DBG(" -> smp_release_cpus()\n");
238 241
@@ -245,7 +248,16 @@ void smp_release_cpus(void)
245 ptr = (unsigned long *)((unsigned long)&__secondary_hold_spinloop 248 ptr = (unsigned long *)((unsigned long)&__secondary_hold_spinloop
246 - PHYSICAL_START); 249 - PHYSICAL_START);
247 *ptr = __pa(generic_secondary_smp_init); 250 *ptr = __pa(generic_secondary_smp_init);
248 mb(); 251
252 /* And wait a bit for them to catch up */
253 for (i = 0; i < 100000; i++) {
254 mb();
255 HMT_low();
256 if (boot_cpu_count == 0)
257 break;
258 udelay(1);
259 }
260 DBG("boot_cpu_count = %d\n", boot_cpu_count);
249 261
250 DBG(" <- smp_release_cpus()\n"); 262 DBG(" <- smp_release_cpus()\n");
251} 263}
@@ -423,22 +435,35 @@ void __init setup_system(void)
423 DBG(" <- setup_system()\n"); 435 DBG(" <- setup_system()\n");
424} 436}
425 437
426static u64 slb0_limit(void) 438/* This returns the limit below which memory accesses to the linear
439 * mapping are guarnateed not to cause a TLB or SLB miss. This is
440 * used to allocate interrupt or emergency stacks for which our
441 * exception entry path doesn't deal with being interrupted.
442 */
443static u64 safe_stack_limit(void)
427{ 444{
428 if (cpu_has_feature(CPU_FTR_1T_SEGMENT)) { 445#ifdef CONFIG_PPC_BOOK3E
446 /* Freescale BookE bolts the entire linear mapping */
447 if (mmu_has_feature(MMU_FTR_TYPE_FSL_E))
448 return linear_map_top;
449 /* Other BookE, we assume the first GB is bolted */
450 return 1ul << 30;
451#else
452 /* BookS, the first segment is bolted */
453 if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
429 return 1UL << SID_SHIFT_1T; 454 return 1UL << SID_SHIFT_1T;
430 }
431 return 1UL << SID_SHIFT; 455 return 1UL << SID_SHIFT;
456#endif
432} 457}
433 458
434static void __init irqstack_early_init(void) 459static void __init irqstack_early_init(void)
435{ 460{
436 u64 limit = slb0_limit(); 461 u64 limit = safe_stack_limit();
437 unsigned int i; 462 unsigned int i;
438 463
439 /* 464 /*
440 * interrupt stacks must be under 256MB, we cannot afford to take 465 * Interrupt stacks must be in the first segment since we
441 * SLB misses on them. 466 * cannot afford to take SLB misses on them.
442 */ 467 */
443 for_each_possible_cpu(i) { 468 for_each_possible_cpu(i) {
444 softirq_ctx[i] = (struct thread_info *) 469 softirq_ctx[i] = (struct thread_info *)
@@ -453,6 +478,9 @@ static void __init irqstack_early_init(void)
453#ifdef CONFIG_PPC_BOOK3E 478#ifdef CONFIG_PPC_BOOK3E
454static void __init exc_lvl_early_init(void) 479static void __init exc_lvl_early_init(void)
455{ 480{
481 extern unsigned int interrupt_base_book3e;
482 extern unsigned int exc_debug_debug_book3e;
483
456 unsigned int i; 484 unsigned int i;
457 485
458 for_each_possible_cpu(i) { 486 for_each_possible_cpu(i) {
@@ -463,6 +491,10 @@ static void __init exc_lvl_early_init(void)
463 mcheckirq_ctx[i] = (struct thread_info *) 491 mcheckirq_ctx[i] = (struct thread_info *)
464 __va(memblock_alloc(THREAD_SIZE, THREAD_SIZE)); 492 __va(memblock_alloc(THREAD_SIZE, THREAD_SIZE));
465 } 493 }
494
495 if (cpu_has_feature(CPU_FTR_DEBUG_LVL_EXC))
496 patch_branch(&interrupt_base_book3e + (0x040 / 4) + 1,
497 (unsigned long)&exc_debug_debug_book3e, 0);
466} 498}
467#else 499#else
468#define exc_lvl_early_init() 500#define exc_lvl_early_init()
@@ -486,7 +518,7 @@ static void __init emergency_stack_init(void)
486 * bringup, we need to get at them in real mode. This means they 518 * bringup, we need to get at them in real mode. This means they
487 * must also be within the RMO region. 519 * must also be within the RMO region.
488 */ 520 */
489 limit = min(slb0_limit(), memblock.rmo_size); 521 limit = min(safe_stack_limit(), ppc64_rma_size);
490 522
491 for_each_possible_cpu(i) { 523 for_each_possible_cpu(i) {
492 unsigned long sp; 524 unsigned long sp;
@@ -497,9 +529,8 @@ static void __init emergency_stack_init(void)
497} 529}
498 530
499/* 531/*
500 * Called into from start_kernel, after lock_kernel has been called. 532 * Called into from start_kernel this initializes bootmem, which is used
501 * Initializes bootmem, which is unsed to manage page allocation until 533 * to manage page allocation until mem_init is called.
502 * mem_init is called.
503 */ 534 */
504void __init setup_arch(char **cmdline_p) 535void __init setup_arch(char **cmdline_p)
505{ 536{
diff --git a/arch/powerpc/kernel/signal_32.c b/arch/powerpc/kernel/signal_32.c
index b96a3a010c26..78b76dc54dfb 100644
--- a/arch/powerpc/kernel/signal_32.c
+++ b/arch/powerpc/kernel/signal_32.c
@@ -25,6 +25,7 @@
25#include <linux/errno.h> 25#include <linux/errno.h>
26#include <linux/elf.h> 26#include <linux/elf.h>
27#include <linux/ptrace.h> 27#include <linux/ptrace.h>
28#include <linux/ratelimit.h>
28#ifdef CONFIG_PPC64 29#ifdef CONFIG_PPC64
29#include <linux/syscalls.h> 30#include <linux/syscalls.h>
30#include <linux/compat.h> 31#include <linux/compat.h>
@@ -892,11 +893,12 @@ badframe:
892 printk("badframe in handle_rt_signal, regs=%p frame=%p newsp=%lx\n", 893 printk("badframe in handle_rt_signal, regs=%p frame=%p newsp=%lx\n",
893 regs, frame, newsp); 894 regs, frame, newsp);
894#endif 895#endif
895 if (show_unhandled_signals && printk_ratelimit()) 896 if (show_unhandled_signals)
896 printk(KERN_INFO "%s[%d]: bad frame in handle_rt_signal32: " 897 printk_ratelimited(KERN_INFO
897 "%p nip %08lx lr %08lx\n", 898 "%s[%d]: bad frame in handle_rt_signal32: "
898 current->comm, current->pid, 899 "%p nip %08lx lr %08lx\n",
899 addr, regs->nip, regs->link); 900 current->comm, current->pid,
901 addr, regs->nip, regs->link);
900 902
901 force_sigsegv(sig, current); 903 force_sigsegv(sig, current);
902 return 0; 904 return 0;
@@ -1058,11 +1060,12 @@ long sys_rt_sigreturn(int r3, int r4, int r5, int r6, int r7, int r8,
1058 return 0; 1060 return 0;
1059 1061
1060 bad: 1062 bad:
1061 if (show_unhandled_signals && printk_ratelimit()) 1063 if (show_unhandled_signals)
1062 printk(KERN_INFO "%s[%d]: bad frame in sys_rt_sigreturn: " 1064 printk_ratelimited(KERN_INFO
1063 "%p nip %08lx lr %08lx\n", 1065 "%s[%d]: bad frame in sys_rt_sigreturn: "
1064 current->comm, current->pid, 1066 "%p nip %08lx lr %08lx\n",
1065 rt_sf, regs->nip, regs->link); 1067 current->comm, current->pid,
1068 rt_sf, regs->nip, regs->link);
1066 1069
1067 force_sig(SIGSEGV, current); 1070 force_sig(SIGSEGV, current);
1068 return 0; 1071 return 0;
@@ -1149,12 +1152,12 @@ int sys_debug_setcontext(struct ucontext __user *ctx,
1149 * We kill the task with a SIGSEGV in this situation. 1152 * We kill the task with a SIGSEGV in this situation.
1150 */ 1153 */
1151 if (do_setcontext(ctx, regs, 1)) { 1154 if (do_setcontext(ctx, regs, 1)) {
1152 if (show_unhandled_signals && printk_ratelimit()) 1155 if (show_unhandled_signals)
1153 printk(KERN_INFO "%s[%d]: bad frame in " 1156 printk_ratelimited(KERN_INFO "%s[%d]: bad frame in "
1154 "sys_debug_setcontext: %p nip %08lx " 1157 "sys_debug_setcontext: %p nip %08lx "
1155 "lr %08lx\n", 1158 "lr %08lx\n",
1156 current->comm, current->pid, 1159 current->comm, current->pid,
1157 ctx, regs->nip, regs->link); 1160 ctx, regs->nip, regs->link);
1158 1161
1159 force_sig(SIGSEGV, current); 1162 force_sig(SIGSEGV, current);
1160 goto out; 1163 goto out;
@@ -1236,11 +1239,12 @@ badframe:
1236 printk("badframe in handle_signal, regs=%p frame=%p newsp=%lx\n", 1239 printk("badframe in handle_signal, regs=%p frame=%p newsp=%lx\n",
1237 regs, frame, newsp); 1240 regs, frame, newsp);
1238#endif 1241#endif
1239 if (show_unhandled_signals && printk_ratelimit()) 1242 if (show_unhandled_signals)
1240 printk(KERN_INFO "%s[%d]: bad frame in handle_signal32: " 1243 printk_ratelimited(KERN_INFO
1241 "%p nip %08lx lr %08lx\n", 1244 "%s[%d]: bad frame in handle_signal32: "
1242 current->comm, current->pid, 1245 "%p nip %08lx lr %08lx\n",
1243 frame, regs->nip, regs->link); 1246 current->comm, current->pid,
1247 frame, regs->nip, regs->link);
1244 1248
1245 force_sigsegv(sig, current); 1249 force_sigsegv(sig, current);
1246 return 0; 1250 return 0;
@@ -1288,11 +1292,12 @@ long sys_sigreturn(int r3, int r4, int r5, int r6, int r7, int r8,
1288 return 0; 1292 return 0;
1289 1293
1290badframe: 1294badframe:
1291 if (show_unhandled_signals && printk_ratelimit()) 1295 if (show_unhandled_signals)
1292 printk(KERN_INFO "%s[%d]: bad frame in sys_sigreturn: " 1296 printk_ratelimited(KERN_INFO
1293 "%p nip %08lx lr %08lx\n", 1297 "%s[%d]: bad frame in sys_sigreturn: "
1294 current->comm, current->pid, 1298 "%p nip %08lx lr %08lx\n",
1295 addr, regs->nip, regs->link); 1299 current->comm, current->pid,
1300 addr, regs->nip, regs->link);
1296 1301
1297 force_sig(SIGSEGV, current); 1302 force_sig(SIGSEGV, current);
1298 return 0; 1303 return 0;
diff --git a/arch/powerpc/kernel/signal_64.c b/arch/powerpc/kernel/signal_64.c
index 27c4a4584f80..e91c736cc842 100644
--- a/arch/powerpc/kernel/signal_64.c
+++ b/arch/powerpc/kernel/signal_64.c
@@ -24,6 +24,7 @@
24#include <linux/elf.h> 24#include <linux/elf.h>
25#include <linux/ptrace.h> 25#include <linux/ptrace.h>
26#include <linux/module.h> 26#include <linux/module.h>
27#include <linux/ratelimit.h>
27 28
28#include <asm/sigcontext.h> 29#include <asm/sigcontext.h>
29#include <asm/ucontext.h> 30#include <asm/ucontext.h>
@@ -380,10 +381,10 @@ badframe:
380 printk("badframe in sys_rt_sigreturn, regs=%p uc=%p &uc->uc_mcontext=%p\n", 381 printk("badframe in sys_rt_sigreturn, regs=%p uc=%p &uc->uc_mcontext=%p\n",
381 regs, uc, &uc->uc_mcontext); 382 regs, uc, &uc->uc_mcontext);
382#endif 383#endif
383 if (show_unhandled_signals && printk_ratelimit()) 384 if (show_unhandled_signals)
384 printk(regs->msr & MSR_SF ? fmt64 : fmt32, 385 printk_ratelimited(regs->msr & MSR_64BIT ? fmt64 : fmt32,
385 current->comm, current->pid, "rt_sigreturn", 386 current->comm, current->pid, "rt_sigreturn",
386 (long)uc, regs->nip, regs->link); 387 (long)uc, regs->nip, regs->link);
387 388
388 force_sig(SIGSEGV, current); 389 force_sig(SIGSEGV, current);
389 return 0; 390 return 0;
@@ -468,10 +469,10 @@ badframe:
468 printk("badframe in setup_rt_frame, regs=%p frame=%p newsp=%lx\n", 469 printk("badframe in setup_rt_frame, regs=%p frame=%p newsp=%lx\n",
469 regs, frame, newsp); 470 regs, frame, newsp);
470#endif 471#endif
471 if (show_unhandled_signals && printk_ratelimit()) 472 if (show_unhandled_signals)
472 printk(regs->msr & MSR_SF ? fmt64 : fmt32, 473 printk_ratelimited(regs->msr & MSR_64BIT ? fmt64 : fmt32,
473 current->comm, current->pid, "setup_rt_frame", 474 current->comm, current->pid, "setup_rt_frame",
474 (long)frame, regs->nip, regs->link); 475 (long)frame, regs->nip, regs->link);
475 476
476 force_sigsegv(signr, current); 477 force_sigsegv(signr, current);
477 return 0; 478 return 0;
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 0008bc58e826..8ebc6700b98d 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -57,6 +57,25 @@
57#define DBG(fmt...) 57#define DBG(fmt...)
58#endif 58#endif
59 59
60
61/* Store all idle threads, this can be reused instead of creating
62* a new thread. Also avoids complicated thread destroy functionality
63* for idle threads.
64*/
65#ifdef CONFIG_HOTPLUG_CPU
66/*
67 * Needed only for CONFIG_HOTPLUG_CPU because __cpuinitdata is
68 * removed after init for !CONFIG_HOTPLUG_CPU.
69 */
70static DEFINE_PER_CPU(struct task_struct *, idle_thread_array);
71#define get_idle_for_cpu(x) (per_cpu(idle_thread_array, x))
72#define set_idle_for_cpu(x, p) (per_cpu(idle_thread_array, x) = (p))
73#else
74static struct task_struct *idle_thread_array[NR_CPUS] __cpuinitdata ;
75#define get_idle_for_cpu(x) (idle_thread_array[(x)])
76#define set_idle_for_cpu(x, p) (idle_thread_array[(x)] = (p))
77#endif
78
60struct thread_info *secondary_ti; 79struct thread_info *secondary_ti;
61 80
62DEFINE_PER_CPU(cpumask_var_t, cpu_sibling_map); 81DEFINE_PER_CPU(cpumask_var_t, cpu_sibling_map);
@@ -76,7 +95,7 @@ int smt_enabled_at_boot = 1;
76static void (*crash_ipi_function_ptr)(struct pt_regs *) = NULL; 95static void (*crash_ipi_function_ptr)(struct pt_regs *) = NULL;
77 96
78#ifdef CONFIG_PPC64 97#ifdef CONFIG_PPC64
79void __devinit smp_generic_kick_cpu(int nr) 98int __devinit smp_generic_kick_cpu(int nr)
80{ 99{
81 BUG_ON(nr < 0 || nr >= NR_CPUS); 100 BUG_ON(nr < 0 || nr >= NR_CPUS);
82 101
@@ -87,37 +106,10 @@ void __devinit smp_generic_kick_cpu(int nr)
87 */ 106 */
88 paca[nr].cpu_start = 1; 107 paca[nr].cpu_start = 1;
89 smp_mb(); 108 smp_mb();
90}
91#endif
92 109
93void smp_message_recv(int msg) 110 return 0;
94{
95 switch(msg) {
96 case PPC_MSG_CALL_FUNCTION:
97 generic_smp_call_function_interrupt();
98 break;
99 case PPC_MSG_RESCHEDULE:
100 /* we notice need_resched on exit */
101 break;
102 case PPC_MSG_CALL_FUNC_SINGLE:
103 generic_smp_call_function_single_interrupt();
104 break;
105 case PPC_MSG_DEBUGGER_BREAK:
106 if (crash_ipi_function_ptr) {
107 crash_ipi_function_ptr(get_irq_regs());
108 break;
109 }
110#ifdef CONFIG_DEBUGGER
111 debugger_ipi(get_irq_regs());
112 break;
113#endif /* CONFIG_DEBUGGER */
114 /* FALLTHROUGH */
115 default:
116 printk("SMP %d: smp_message_recv(): unknown msg %d\n",
117 smp_processor_id(), msg);
118 break;
119 }
120} 111}
112#endif
121 113
122static irqreturn_t call_function_action(int irq, void *data) 114static irqreturn_t call_function_action(int irq, void *data)
123{ 115{
@@ -127,7 +119,7 @@ static irqreturn_t call_function_action(int irq, void *data)
127 119
128static irqreturn_t reschedule_action(int irq, void *data) 120static irqreturn_t reschedule_action(int irq, void *data)
129{ 121{
130 /* we just need the return path side effect of checking need_resched */ 122 scheduler_ipi();
131 return IRQ_HANDLED; 123 return IRQ_HANDLED;
132} 124}
133 125
@@ -139,7 +131,15 @@ static irqreturn_t call_function_single_action(int irq, void *data)
139 131
140static irqreturn_t debug_ipi_action(int irq, void *data) 132static irqreturn_t debug_ipi_action(int irq, void *data)
141{ 133{
142 smp_message_recv(PPC_MSG_DEBUGGER_BREAK); 134 if (crash_ipi_function_ptr) {
135 crash_ipi_function_ptr(get_irq_regs());
136 return IRQ_HANDLED;
137 }
138
139#ifdef CONFIG_DEBUGGER
140 debugger_ipi(get_irq_regs());
141#endif /* CONFIG_DEBUGGER */
142
143 return IRQ_HANDLED; 143 return IRQ_HANDLED;
144} 144}
145 145
@@ -178,6 +178,66 @@ int smp_request_message_ipi(int virq, int msg)
178 return err; 178 return err;
179} 179}
180 180
181#ifdef CONFIG_PPC_SMP_MUXED_IPI
182struct cpu_messages {
183 int messages; /* current messages */
184 unsigned long data; /* data for cause ipi */
185};
186static DEFINE_PER_CPU_SHARED_ALIGNED(struct cpu_messages, ipi_message);
187
188void smp_muxed_ipi_set_data(int cpu, unsigned long data)
189{
190 struct cpu_messages *info = &per_cpu(ipi_message, cpu);
191
192 info->data = data;
193}
194
195void smp_muxed_ipi_message_pass(int cpu, int msg)
196{
197 struct cpu_messages *info = &per_cpu(ipi_message, cpu);
198 char *message = (char *)&info->messages;
199
200 message[msg] = 1;
201 mb();
202 smp_ops->cause_ipi(cpu, info->data);
203}
204
205void smp_muxed_ipi_resend(void)
206{
207 struct cpu_messages *info = &__get_cpu_var(ipi_message);
208
209 if (info->messages)
210 smp_ops->cause_ipi(smp_processor_id(), info->data);
211}
212
213irqreturn_t smp_ipi_demux(void)
214{
215 struct cpu_messages *info = &__get_cpu_var(ipi_message);
216 unsigned int all;
217
218 mb(); /* order any irq clear */
219
220 do {
221 all = xchg_local(&info->messages, 0);
222
223#ifdef __BIG_ENDIAN
224 if (all & (1 << (24 - 8 * PPC_MSG_CALL_FUNCTION)))
225 generic_smp_call_function_interrupt();
226 if (all & (1 << (24 - 8 * PPC_MSG_RESCHEDULE)))
227 scheduler_ipi();
228 if (all & (1 << (24 - 8 * PPC_MSG_CALL_FUNC_SINGLE)))
229 generic_smp_call_function_single_interrupt();
230 if (all & (1 << (24 - 8 * PPC_MSG_DEBUGGER_BREAK)))
231 debug_ipi_action(0, NULL);
232#else
233#error Unsupported ENDIAN
234#endif
235 } while (info->messages);
236
237 return IRQ_HANDLED;
238}
239#endif /* CONFIG_PPC_SMP_MUXED_IPI */
240
181void smp_send_reschedule(int cpu) 241void smp_send_reschedule(int cpu)
182{ 242{
183 if (likely(smp_ops)) 243 if (likely(smp_ops))
@@ -197,11 +257,18 @@ void arch_send_call_function_ipi_mask(const struct cpumask *mask)
197 smp_ops->message_pass(cpu, PPC_MSG_CALL_FUNCTION); 257 smp_ops->message_pass(cpu, PPC_MSG_CALL_FUNCTION);
198} 258}
199 259
200#ifdef CONFIG_DEBUGGER 260#if defined(CONFIG_DEBUGGER) || defined(CONFIG_KEXEC)
201void smp_send_debugger_break(int cpu) 261void smp_send_debugger_break(void)
202{ 262{
203 if (likely(smp_ops)) 263 int cpu;
204 smp_ops->message_pass(cpu, PPC_MSG_DEBUGGER_BREAK); 264 int me = raw_smp_processor_id();
265
266 if (unlikely(!smp_ops))
267 return;
268
269 for_each_online_cpu(cpu)
270 if (cpu != me)
271 smp_ops->message_pass(cpu, PPC_MSG_DEBUGGER_BREAK);
205} 272}
206#endif 273#endif
207 274
@@ -209,9 +276,9 @@ void smp_send_debugger_break(int cpu)
209void crash_send_ipi(void (*crash_ipi_callback)(struct pt_regs *)) 276void crash_send_ipi(void (*crash_ipi_callback)(struct pt_regs *))
210{ 277{
211 crash_ipi_function_ptr = crash_ipi_callback; 278 crash_ipi_function_ptr = crash_ipi_callback;
212 if (crash_ipi_callback && smp_ops) { 279 if (crash_ipi_callback) {
213 mb(); 280 mb();
214 smp_ops->message_pass(MSG_ALL_BUT_SELF, PPC_MSG_DEBUGGER_BREAK); 281 smp_send_debugger_break();
215 } 282 }
216} 283}
217#endif 284#endif
@@ -238,23 +305,6 @@ static void __devinit smp_store_cpu_info(int id)
238 per_cpu(cpu_pvr, id) = mfspr(SPRN_PVR); 305 per_cpu(cpu_pvr, id) = mfspr(SPRN_PVR);
239} 306}
240 307
241static void __init smp_create_idle(unsigned int cpu)
242{
243 struct task_struct *p;
244
245 /* create a process for the processor */
246 p = fork_idle(cpu);
247 if (IS_ERR(p))
248 panic("failed fork for CPU %u: %li", cpu, PTR_ERR(p));
249#ifdef CONFIG_PPC64
250 paca[cpu].__current = p;
251 paca[cpu].kstack = (unsigned long) task_thread_info(p)
252 + THREAD_SIZE - STACK_FRAME_OVERHEAD;
253#endif
254 current_set[cpu] = task_thread_info(p);
255 task_thread_info(p)->cpu = cpu;
256}
257
258void __init smp_prepare_cpus(unsigned int max_cpus) 308void __init smp_prepare_cpus(unsigned int max_cpus)
259{ 309{
260 unsigned int cpu; 310 unsigned int cpu;
@@ -288,10 +338,6 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
288 max_cpus = NR_CPUS; 338 max_cpus = NR_CPUS;
289 else 339 else
290 max_cpus = 1; 340 max_cpus = 1;
291
292 for_each_possible_cpu(cpu)
293 if (cpu != boot_cpuid)
294 smp_create_idle(cpu);
295} 341}
296 342
297void __devinit smp_prepare_boot_cpu(void) 343void __devinit smp_prepare_boot_cpu(void)
@@ -305,7 +351,7 @@ void __devinit smp_prepare_boot_cpu(void)
305 351
306#ifdef CONFIG_HOTPLUG_CPU 352#ifdef CONFIG_HOTPLUG_CPU
307/* State of each CPU during hotplug phases */ 353/* State of each CPU during hotplug phases */
308DEFINE_PER_CPU(int, cpu_state) = { 0 }; 354static DEFINE_PER_CPU(int, cpu_state) = { 0 };
309 355
310int generic_cpu_disable(void) 356int generic_cpu_disable(void)
311{ 357{
@@ -317,30 +363,8 @@ int generic_cpu_disable(void)
317 set_cpu_online(cpu, false); 363 set_cpu_online(cpu, false);
318#ifdef CONFIG_PPC64 364#ifdef CONFIG_PPC64
319 vdso_data->processorCount--; 365 vdso_data->processorCount--;
320 fixup_irqs(cpu_online_mask);
321#endif
322 return 0;
323}
324
325int generic_cpu_enable(unsigned int cpu)
326{
327 /* Do the normal bootup if we haven't
328 * already bootstrapped. */
329 if (system_state != SYSTEM_RUNNING)
330 return -ENOSYS;
331
332 /* get the target out of it's holding state */
333 per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
334 smp_wmb();
335
336 while (!cpu_online(cpu))
337 cpu_relax();
338
339#ifdef CONFIG_PPC64
340 fixup_irqs(cpu_online_mask);
341 /* counter the irq disable in fixup_irqs */
342 local_irq_enable();
343#endif 366#endif
367 migrate_irqs();
344 return 0; 368 return 0;
345} 369}
346 370
@@ -362,37 +386,89 @@ void generic_mach_cpu_die(void)
362 unsigned int cpu; 386 unsigned int cpu;
363 387
364 local_irq_disable(); 388 local_irq_disable();
389 idle_task_exit();
365 cpu = smp_processor_id(); 390 cpu = smp_processor_id();
366 printk(KERN_DEBUG "CPU%d offline\n", cpu); 391 printk(KERN_DEBUG "CPU%d offline\n", cpu);
367 __get_cpu_var(cpu_state) = CPU_DEAD; 392 __get_cpu_var(cpu_state) = CPU_DEAD;
368 smp_wmb(); 393 smp_wmb();
369 while (__get_cpu_var(cpu_state) != CPU_UP_PREPARE) 394 while (__get_cpu_var(cpu_state) != CPU_UP_PREPARE)
370 cpu_relax(); 395 cpu_relax();
371 set_cpu_online(cpu, true); 396}
372 local_irq_enable(); 397
398void generic_set_cpu_dead(unsigned int cpu)
399{
400 per_cpu(cpu_state, cpu) = CPU_DEAD;
373} 401}
374#endif 402#endif
375 403
376static int __devinit cpu_enable(unsigned int cpu) 404struct create_idle {
405 struct work_struct work;
406 struct task_struct *idle;
407 struct completion done;
408 int cpu;
409};
410
411static void __cpuinit do_fork_idle(struct work_struct *work)
377{ 412{
378 if (smp_ops && smp_ops->cpu_enable) 413 struct create_idle *c_idle =
379 return smp_ops->cpu_enable(cpu); 414 container_of(work, struct create_idle, work);
380 415
381 return -ENOSYS; 416 c_idle->idle = fork_idle(c_idle->cpu);
417 complete(&c_idle->done);
382} 418}
383 419
384int __cpuinit __cpu_up(unsigned int cpu) 420static int __cpuinit create_idle(unsigned int cpu)
385{ 421{
386 int c; 422 struct thread_info *ti;
423 struct create_idle c_idle = {
424 .cpu = cpu,
425 .done = COMPLETION_INITIALIZER_ONSTACK(c_idle.done),
426 };
427 INIT_WORK_ONSTACK(&c_idle.work, do_fork_idle);
387 428
388 secondary_ti = current_set[cpu]; 429 c_idle.idle = get_idle_for_cpu(cpu);
389 if (!cpu_enable(cpu)) 430
390 return 0; 431 /* We can't use kernel_thread since we must avoid to
432 * reschedule the child. We use a workqueue because
433 * we want to fork from a kernel thread, not whatever
434 * userspace process happens to be trying to online us.
435 */
436 if (!c_idle.idle) {
437 schedule_work(&c_idle.work);
438 wait_for_completion(&c_idle.done);
439 } else
440 init_idle(c_idle.idle, cpu);
441 if (IS_ERR(c_idle.idle)) {
442 pr_err("Failed fork for CPU %u: %li", cpu, PTR_ERR(c_idle.idle));
443 return PTR_ERR(c_idle.idle);
444 }
445 ti = task_thread_info(c_idle.idle);
446
447#ifdef CONFIG_PPC64
448 paca[cpu].__current = c_idle.idle;
449 paca[cpu].kstack = (unsigned long)ti + THREAD_SIZE - STACK_FRAME_OVERHEAD;
450#endif
451 ti->cpu = cpu;
452 current_set[cpu] = ti;
453
454 return 0;
455}
456
457int __cpuinit __cpu_up(unsigned int cpu)
458{
459 int rc, c;
391 460
392 if (smp_ops == NULL || 461 if (smp_ops == NULL ||
393 (smp_ops->cpu_bootable && !smp_ops->cpu_bootable(cpu))) 462 (smp_ops->cpu_bootable && !smp_ops->cpu_bootable(cpu)))
394 return -EINVAL; 463 return -EINVAL;
395 464
465 /* Make sure we have an idle thread */
466 rc = create_idle(cpu);
467 if (rc)
468 return rc;
469
470 secondary_ti = current_set[cpu];
471
396 /* Make sure callin-map entry is 0 (can be leftover a CPU 472 /* Make sure callin-map entry is 0 (can be leftover a CPU
397 * hotplug 473 * hotplug
398 */ 474 */
@@ -406,7 +482,11 @@ int __cpuinit __cpu_up(unsigned int cpu)
406 482
407 /* wake up cpus */ 483 /* wake up cpus */
408 DBG("smp: kicking cpu %d\n", cpu); 484 DBG("smp: kicking cpu %d\n", cpu);
409 smp_ops->kick_cpu(cpu); 485 rc = smp_ops->kick_cpu(cpu);
486 if (rc) {
487 pr_err("smp: failed starting cpu %d (rc %d)\n", cpu, rc);
488 return rc;
489 }
410 490
411 /* 491 /*
412 * wait to see if the cpu made a callin (is actually up). 492 * wait to see if the cpu made a callin (is actually up).
@@ -466,6 +546,19 @@ out:
466 return id; 546 return id;
467} 547}
468 548
549/* Helper routines for cpu to core mapping */
550int cpu_core_index_of_thread(int cpu)
551{
552 return cpu >> threads_shift;
553}
554EXPORT_SYMBOL_GPL(cpu_core_index_of_thread);
555
556int cpu_first_thread_of_core(int core)
557{
558 return core << threads_shift;
559}
560EXPORT_SYMBOL_GPL(cpu_first_thread_of_core);
561
469/* Must be called when no change can occur to cpu_present_mask, 562/* Must be called when no change can occur to cpu_present_mask,
470 * i.e. during cpu online or offline. 563 * i.e. during cpu online or offline.
471 */ 564 */
@@ -489,7 +582,7 @@ static struct device_node *cpu_to_l2cache(int cpu)
489} 582}
490 583
491/* Activate a secondary processor. */ 584/* Activate a secondary processor. */
492int __devinit start_secondary(void *unused) 585void __devinit start_secondary(void *unused)
493{ 586{
494 unsigned int cpu = smp_processor_id(); 587 unsigned int cpu = smp_processor_id();
495 struct device_node *l2_cache; 588 struct device_node *l2_cache;
@@ -508,16 +601,17 @@ int __devinit start_secondary(void *unused)
508 if (smp_ops->take_timebase) 601 if (smp_ops->take_timebase)
509 smp_ops->take_timebase(); 602 smp_ops->take_timebase();
510 603
511 if (system_state > SYSTEM_BOOTING)
512 snapshot_timebase();
513
514 secondary_cpu_time_init(); 604 secondary_cpu_time_init();
515 605
606#ifdef CONFIG_PPC64
607 if (system_state == SYSTEM_RUNNING)
608 vdso_data->processorCount++;
609#endif
516 ipi_call_lock(); 610 ipi_call_lock();
517 notify_cpu_starting(cpu); 611 notify_cpu_starting(cpu);
518 set_cpu_online(cpu, true); 612 set_cpu_online(cpu, true);
519 /* Update sibling maps */ 613 /* Update sibling maps */
520 base = cpu_first_thread_in_core(cpu); 614 base = cpu_first_thread_sibling(cpu);
521 for (i = 0; i < threads_per_core; i++) { 615 for (i = 0; i < threads_per_core; i++) {
522 if (cpu_is_offline(base + i)) 616 if (cpu_is_offline(base + i))
523 continue; 617 continue;
@@ -548,7 +642,8 @@ int __devinit start_secondary(void *unused)
548 local_irq_enable(); 642 local_irq_enable();
549 643
550 cpu_idle(); 644 cpu_idle();
551 return 0; 645
646 BUG();
552} 647}
553 648
554int setup_profiling_timer(unsigned int multiplier) 649int setup_profiling_timer(unsigned int multiplier)
@@ -565,7 +660,7 @@ void __init smp_cpus_done(unsigned int max_cpus)
565 * se we pin us down to CPU 0 for a short while 660 * se we pin us down to CPU 0 for a short while
566 */ 661 */
567 alloc_cpumask_var(&old_mask, GFP_NOWAIT); 662 alloc_cpumask_var(&old_mask, GFP_NOWAIT);
568 cpumask_copy(old_mask, &current->cpus_allowed); 663 cpumask_copy(old_mask, tsk_cpus_allowed(current));
569 set_cpus_allowed_ptr(current, cpumask_of(boot_cpuid)); 664 set_cpus_allowed_ptr(current, cpumask_of(boot_cpuid));
570 665
571 if (smp_ops && smp_ops->setup_cpu) 666 if (smp_ops && smp_ops->setup_cpu)
@@ -575,9 +670,20 @@ void __init smp_cpus_done(unsigned int max_cpus)
575 670
576 free_cpumask_var(old_mask); 671 free_cpumask_var(old_mask);
577 672
578 snapshot_timebases(); 673 if (smp_ops && smp_ops->bringup_done)
674 smp_ops->bringup_done();
579 675
580 dump_numa_cpu_topology(); 676 dump_numa_cpu_topology();
677
678}
679
680int arch_sd_sibling_asym_packing(void)
681{
682 if (cpu_has_feature(CPU_FTR_ASYM_SMT)) {
683 printk_once(KERN_INFO "Enabling Asymmetric SMT scheduling\n");
684 return SD_ASYM_PACKING;
685 }
686 return 0;
581} 687}
582 688
583#ifdef CONFIG_HOTPLUG_CPU 689#ifdef CONFIG_HOTPLUG_CPU
@@ -596,7 +702,7 @@ int __cpu_disable(void)
596 return err; 702 return err;
597 703
598 /* Update sibling maps */ 704 /* Update sibling maps */
599 base = cpu_first_thread_in_core(cpu); 705 base = cpu_first_thread_sibling(cpu);
600 for (i = 0; i < threads_per_core; i++) { 706 for (i = 0; i < threads_per_core; i++) {
601 cpumask_clear_cpu(cpu, cpu_sibling_mask(base + i)); 707 cpumask_clear_cpu(cpu, cpu_sibling_mask(base + i));
602 cpumask_clear_cpu(base + i, cpu_sibling_mask(cpu)); 708 cpumask_clear_cpu(base + i, cpu_sibling_mask(cpu));
@@ -643,5 +749,9 @@ void cpu_die(void)
643{ 749{
644 if (ppc_md.cpu_die) 750 if (ppc_md.cpu_die)
645 ppc_md.cpu_die(); 751 ppc_md.cpu_die();
752
753 /* If we return, we re-enter start_secondary */
754 start_secondary_resume();
646} 755}
756
647#endif 757#endif
diff --git a/arch/powerpc/kernel/swsusp.c b/arch/powerpc/kernel/swsusp.c
index 560c96119501..aa17b76dd427 100644
--- a/arch/powerpc/kernel/swsusp.c
+++ b/arch/powerpc/kernel/swsusp.c
@@ -10,7 +10,6 @@
10 */ 10 */
11 11
12#include <linux/sched.h> 12#include <linux/sched.h>
13#include <asm/suspend.h>
14#include <asm/system.h> 13#include <asm/system.h>
15#include <asm/current.h> 14#include <asm/current.h>
16#include <asm/mmu_context.h> 15#include <asm/mmu_context.h>
diff --git a/arch/powerpc/kernel/swsusp_32.S b/arch/powerpc/kernel/swsusp_32.S
index b0754e237438..ba4dee3d233f 100644
--- a/arch/powerpc/kernel/swsusp_32.S
+++ b/arch/powerpc/kernel/swsusp_32.S
@@ -143,7 +143,7 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
143 143
144 /* Disable MSR:DR to make sure we don't take a TLB or 144 /* Disable MSR:DR to make sure we don't take a TLB or
145 * hash miss during the copy, as our hash table will 145 * hash miss during the copy, as our hash table will
146 * for a while be unuseable. For .text, we assume we are 146 * for a while be unusable. For .text, we assume we are
147 * covered by a BAT. This works only for non-G5 at this 147 * covered by a BAT. This works only for non-G5 at this
148 * point. G5 will need a better approach, possibly using 148 * point. G5 will need a better approach, possibly using
149 * a small temporary hash table filled with large mappings, 149 * a small temporary hash table filled with large mappings,
diff --git a/arch/powerpc/kernel/sys_ppc32.c b/arch/powerpc/kernel/sys_ppc32.c
index b1b6043a56c4..4e5bf1edc0f2 100644
--- a/arch/powerpc/kernel/sys_ppc32.c
+++ b/arch/powerpc/kernel/sys_ppc32.c
@@ -23,7 +23,6 @@
23#include <linux/resource.h> 23#include <linux/resource.h>
24#include <linux/times.h> 24#include <linux/times.h>
25#include <linux/smp.h> 25#include <linux/smp.h>
26#include <linux/smp_lock.h>
27#include <linux/sem.h> 26#include <linux/sem.h>
28#include <linux/msg.h> 27#include <linux/msg.h>
29#include <linux/shm.h> 28#include <linux/shm.h>
diff --git a/arch/powerpc/kernel/sysfs.c b/arch/powerpc/kernel/sysfs.c
index c0d8c2006bf4..f0f2199e64e1 100644
--- a/arch/powerpc/kernel/sysfs.c
+++ b/arch/powerpc/kernel/sysfs.c
@@ -182,6 +182,41 @@ static SYSDEV_ATTR(mmcra, 0600, show_mmcra, store_mmcra);
182static SYSDEV_ATTR(spurr, 0600, show_spurr, NULL); 182static SYSDEV_ATTR(spurr, 0600, show_spurr, NULL);
183static SYSDEV_ATTR(dscr, 0600, show_dscr, store_dscr); 183static SYSDEV_ATTR(dscr, 0600, show_dscr, store_dscr);
184static SYSDEV_ATTR(purr, 0600, show_purr, store_purr); 184static SYSDEV_ATTR(purr, 0600, show_purr, store_purr);
185
186unsigned long dscr_default = 0;
187EXPORT_SYMBOL(dscr_default);
188
189static ssize_t show_dscr_default(struct sysdev_class *class,
190 struct sysdev_class_attribute *attr, char *buf)
191{
192 return sprintf(buf, "%lx\n", dscr_default);
193}
194
195static ssize_t __used store_dscr_default(struct sysdev_class *class,
196 struct sysdev_class_attribute *attr, const char *buf,
197 size_t count)
198{
199 unsigned long val;
200 int ret = 0;
201
202 ret = sscanf(buf, "%lx", &val);
203 if (ret != 1)
204 return -EINVAL;
205 dscr_default = val;
206
207 return count;
208}
209
210static SYSDEV_CLASS_ATTR(dscr_default, 0600,
211 show_dscr_default, store_dscr_default);
212
213static void sysfs_create_dscr_default(void)
214{
215 int err = 0;
216 if (cpu_has_feature(CPU_FTR_DSCR))
217 err = sysfs_create_file(&cpu_sysdev_class.kset.kobj,
218 &attr_dscr_default.attr);
219}
185#endif /* CONFIG_PPC64 */ 220#endif /* CONFIG_PPC64 */
186 221
187#ifdef HAS_PPC_PMC_PA6T 222#ifdef HAS_PPC_PMC_PA6T
@@ -617,6 +652,9 @@ static int __init topology_init(void)
617 if (cpu_online(cpu)) 652 if (cpu_online(cpu))
618 register_cpu_online(cpu); 653 register_cpu_online(cpu);
619 } 654 }
655#ifdef CONFIG_PPC64
656 sysfs_create_dscr_default();
657#endif /* CONFIG_PPC64 */
620 658
621 return 0; 659 return 0;
622} 660}
diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index 8533b3b83f5d..f33acfd872ad 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -53,7 +53,7 @@
53#include <linux/posix-timers.h> 53#include <linux/posix-timers.h>
54#include <linux/irq.h> 54#include <linux/irq.h>
55#include <linux/delay.h> 55#include <linux/delay.h>
56#include <linux/perf_event.h> 56#include <linux/irq_work.h>
57#include <asm/trace.h> 57#include <asm/trace.h>
58 58
59#include <asm/io.h> 59#include <asm/io.h>
@@ -155,16 +155,15 @@ EXPORT_SYMBOL_GPL(rtc_lock);
155 155
156static u64 tb_to_ns_scale __read_mostly; 156static u64 tb_to_ns_scale __read_mostly;
157static unsigned tb_to_ns_shift __read_mostly; 157static unsigned tb_to_ns_shift __read_mostly;
158static unsigned long boot_tb __read_mostly; 158static u64 boot_tb __read_mostly;
159 159
160extern struct timezone sys_tz; 160extern struct timezone sys_tz;
161static long timezone_offset; 161static long timezone_offset;
162 162
163unsigned long ppc_proc_freq; 163unsigned long ppc_proc_freq;
164EXPORT_SYMBOL(ppc_proc_freq); 164EXPORT_SYMBOL_GPL(ppc_proc_freq);
165unsigned long ppc_tb_freq; 165unsigned long ppc_tb_freq;
166 166EXPORT_SYMBOL_GPL(ppc_tb_freq);
167static DEFINE_PER_CPU(u64, last_jiffy);
168 167
169#ifdef CONFIG_VIRT_CPU_ACCOUNTING 168#ifdef CONFIG_VIRT_CPU_ACCOUNTING
170/* 169/*
@@ -185,6 +184,8 @@ DEFINE_PER_CPU(unsigned long, cputime_scaled_last_delta);
185 184
186cputime_t cputime_one_jiffy; 185cputime_t cputime_one_jiffy;
187 186
187void (*dtl_consumer)(struct dtl_entry *, u64);
188
188static void calc_cputime_factors(void) 189static void calc_cputime_factors(void)
189{ 190{
190 struct div_result res; 191 struct div_result res;
@@ -200,62 +201,171 @@ static void calc_cputime_factors(void)
200} 201}
201 202
202/* 203/*
203 * Read the PURR on systems that have it, otherwise the timebase. 204 * Read the SPURR on systems that have it, otherwise the PURR,
205 * or if that doesn't exist return the timebase value passed in.
204 */ 206 */
205static u64 read_purr(void) 207static u64 read_spurr(u64 tb)
206{ 208{
209 if (cpu_has_feature(CPU_FTR_SPURR))
210 return mfspr(SPRN_SPURR);
207 if (cpu_has_feature(CPU_FTR_PURR)) 211 if (cpu_has_feature(CPU_FTR_PURR))
208 return mfspr(SPRN_PURR); 212 return mfspr(SPRN_PURR);
209 return mftb(); 213 return tb;
210} 214}
211 215
216#ifdef CONFIG_PPC_SPLPAR
217
212/* 218/*
213 * Read the SPURR on systems that have it, otherwise the purr 219 * Scan the dispatch trace log and count up the stolen time.
220 * Should be called with interrupts disabled.
214 */ 221 */
215static u64 read_spurr(u64 purr) 222static u64 scan_dispatch_log(u64 stop_tb)
216{ 223{
217 /* 224 u64 i = local_paca->dtl_ridx;
218 * cpus without PURR won't have a SPURR 225 struct dtl_entry *dtl = local_paca->dtl_curr;
219 * We already know the former when we use this, so tell gcc 226 struct dtl_entry *dtl_end = local_paca->dispatch_log_end;
227 struct lppaca *vpa = local_paca->lppaca_ptr;
228 u64 tb_delta;
229 u64 stolen = 0;
230 u64 dtb;
231
232 if (!dtl)
233 return 0;
234
235 if (i == vpa->dtl_idx)
236 return 0;
237 while (i < vpa->dtl_idx) {
238 if (dtl_consumer)
239 dtl_consumer(dtl, i);
240 dtb = dtl->timebase;
241 tb_delta = dtl->enqueue_to_dispatch_time +
242 dtl->ready_to_enqueue_time;
243 barrier();
244 if (i + N_DISPATCH_LOG < vpa->dtl_idx) {
245 /* buffer has overflowed */
246 i = vpa->dtl_idx - N_DISPATCH_LOG;
247 dtl = local_paca->dispatch_log + (i % N_DISPATCH_LOG);
248 continue;
249 }
250 if (dtb > stop_tb)
251 break;
252 stolen += tb_delta;
253 ++i;
254 ++dtl;
255 if (dtl == dtl_end)
256 dtl = local_paca->dispatch_log;
257 }
258 local_paca->dtl_ridx = i;
259 local_paca->dtl_curr = dtl;
260 return stolen;
261}
262
263/*
264 * Accumulate stolen time by scanning the dispatch trace log.
265 * Called on entry from user mode.
266 */
267void accumulate_stolen_time(void)
268{
269 u64 sst, ust;
270
271 u8 save_soft_enabled = local_paca->soft_enabled;
272 u8 save_hard_enabled = local_paca->hard_enabled;
273
274 /* We are called early in the exception entry, before
275 * soft/hard_enabled are sync'ed to the expected state
276 * for the exception. We are hard disabled but the PACA
277 * needs to reflect that so various debug stuff doesn't
278 * complain
220 */ 279 */
221 if (cpu_has_feature(CPU_FTR_PURR) && cpu_has_feature(CPU_FTR_SPURR)) 280 local_paca->soft_enabled = 0;
222 return mfspr(SPRN_SPURR); 281 local_paca->hard_enabled = 0;
223 return purr; 282
283 sst = scan_dispatch_log(local_paca->starttime_user);
284 ust = scan_dispatch_log(local_paca->starttime);
285 local_paca->system_time -= sst;
286 local_paca->user_time -= ust;
287 local_paca->stolen_time += ust + sst;
288
289 local_paca->soft_enabled = save_soft_enabled;
290 local_paca->hard_enabled = save_hard_enabled;
291}
292
293static inline u64 calculate_stolen_time(u64 stop_tb)
294{
295 u64 stolen = 0;
296
297 if (get_paca()->dtl_ridx != get_paca()->lppaca_ptr->dtl_idx) {
298 stolen = scan_dispatch_log(stop_tb);
299 get_paca()->system_time -= stolen;
300 }
301
302 stolen += get_paca()->stolen_time;
303 get_paca()->stolen_time = 0;
304 return stolen;
224} 305}
225 306
307#else /* CONFIG_PPC_SPLPAR */
308static inline u64 calculate_stolen_time(u64 stop_tb)
309{
310 return 0;
311}
312
313#endif /* CONFIG_PPC_SPLPAR */
314
226/* 315/*
227 * Account time for a transition between system, hard irq 316 * Account time for a transition between system, hard irq
228 * or soft irq state. 317 * or soft irq state.
229 */ 318 */
230void account_system_vtime(struct task_struct *tsk) 319void account_system_vtime(struct task_struct *tsk)
231{ 320{
232 u64 now, nowscaled, delta, deltascaled, sys_time; 321 u64 now, nowscaled, delta, deltascaled;
233 unsigned long flags; 322 unsigned long flags;
323 u64 stolen, udelta, sys_scaled, user_scaled;
234 324
235 local_irq_save(flags); 325 local_irq_save(flags);
236 now = read_purr(); 326 now = mftb();
237 nowscaled = read_spurr(now); 327 nowscaled = read_spurr(now);
238 delta = now - get_paca()->startpurr; 328 get_paca()->system_time += now - get_paca()->starttime;
329 get_paca()->starttime = now;
239 deltascaled = nowscaled - get_paca()->startspurr; 330 deltascaled = nowscaled - get_paca()->startspurr;
240 get_paca()->startpurr = now;
241 get_paca()->startspurr = nowscaled; 331 get_paca()->startspurr = nowscaled;
242 if (!in_interrupt()) { 332
243 /* deltascaled includes both user and system time. 333 stolen = calculate_stolen_time(now);
244 * Hence scale it based on the purr ratio to estimate 334
245 * the system time */ 335 delta = get_paca()->system_time;
246 sys_time = get_paca()->system_time; 336 get_paca()->system_time = 0;
247 if (get_paca()->user_time) 337 udelta = get_paca()->user_time - get_paca()->utime_sspurr;
248 deltascaled = deltascaled * sys_time / 338 get_paca()->utime_sspurr = get_paca()->user_time;
249 (sys_time + get_paca()->user_time); 339
250 delta += sys_time; 340 /*
251 get_paca()->system_time = 0; 341 * Because we don't read the SPURR on every kernel entry/exit,
342 * deltascaled includes both user and system SPURR ticks.
343 * Apportion these ticks to system SPURR ticks and user
344 * SPURR ticks in the same ratio as the system time (delta)
345 * and user time (udelta) values obtained from the timebase
346 * over the same interval. The system ticks get accounted here;
347 * the user ticks get saved up in paca->user_time_scaled to be
348 * used by account_process_tick.
349 */
350 sys_scaled = delta;
351 user_scaled = udelta;
352 if (deltascaled != delta + udelta) {
353 if (udelta) {
354 sys_scaled = deltascaled * delta / (delta + udelta);
355 user_scaled = deltascaled - sys_scaled;
356 } else {
357 sys_scaled = deltascaled;
358 }
359 }
360 get_paca()->user_time_scaled += user_scaled;
361
362 if (in_interrupt() || idle_task(smp_processor_id()) != tsk) {
363 account_system_time(tsk, 0, delta, sys_scaled);
364 if (stolen)
365 account_steal_time(stolen);
366 } else {
367 account_idle_time(delta + stolen);
252 } 368 }
253 if (in_irq() || idle_task(smp_processor_id()) != tsk)
254 account_system_time(tsk, 0, delta, deltascaled);
255 else
256 account_idle_time(delta);
257 __get_cpu_var(cputime_last_delta) = delta;
258 __get_cpu_var(cputime_scaled_last_delta) = deltascaled;
259 local_irq_restore(flags); 369 local_irq_restore(flags);
260} 370}
261EXPORT_SYMBOL_GPL(account_system_vtime); 371EXPORT_SYMBOL_GPL(account_system_vtime);
@@ -265,125 +375,26 @@ EXPORT_SYMBOL_GPL(account_system_vtime);
265 * by the exception entry and exit code to the generic process 375 * by the exception entry and exit code to the generic process
266 * user and system time records. 376 * user and system time records.
267 * Must be called with interrupts disabled. 377 * Must be called with interrupts disabled.
378 * Assumes that account_system_vtime() has been called recently
379 * (i.e. since the last entry from usermode) so that
380 * get_paca()->user_time_scaled is up to date.
268 */ 381 */
269void account_process_tick(struct task_struct *tsk, int user_tick) 382void account_process_tick(struct task_struct *tsk, int user_tick)
270{ 383{
271 cputime_t utime, utimescaled; 384 cputime_t utime, utimescaled;
272 385
273 utime = get_paca()->user_time; 386 utime = get_paca()->user_time;
387 utimescaled = get_paca()->user_time_scaled;
274 get_paca()->user_time = 0; 388 get_paca()->user_time = 0;
275 utimescaled = cputime_to_scaled(utime); 389 get_paca()->user_time_scaled = 0;
390 get_paca()->utime_sspurr = 0;
276 account_user_time(tsk, utime, utimescaled); 391 account_user_time(tsk, utime, utimescaled);
277} 392}
278 393
279/*
280 * Stuff for accounting stolen time.
281 */
282struct cpu_purr_data {
283 int initialized; /* thread is running */
284 u64 tb; /* last TB value read */
285 u64 purr; /* last PURR value read */
286 u64 spurr; /* last SPURR value read */
287};
288
289/*
290 * Each entry in the cpu_purr_data array is manipulated only by its
291 * "owner" cpu -- usually in the timer interrupt but also occasionally
292 * in process context for cpu online. As long as cpus do not touch
293 * each others' cpu_purr_data, disabling local interrupts is
294 * sufficient to serialize accesses.
295 */
296static DEFINE_PER_CPU(struct cpu_purr_data, cpu_purr_data);
297
298static void snapshot_tb_and_purr(void *data)
299{
300 unsigned long flags;
301 struct cpu_purr_data *p = &__get_cpu_var(cpu_purr_data);
302
303 local_irq_save(flags);
304 p->tb = get_tb_or_rtc();
305 p->purr = mfspr(SPRN_PURR);
306 wmb();
307 p->initialized = 1;
308 local_irq_restore(flags);
309}
310
311/*
312 * Called during boot when all cpus have come up.
313 */
314void snapshot_timebases(void)
315{
316 if (!cpu_has_feature(CPU_FTR_PURR))
317 return;
318 on_each_cpu(snapshot_tb_and_purr, NULL, 1);
319}
320
321/*
322 * Must be called with interrupts disabled.
323 */
324void calculate_steal_time(void)
325{
326 u64 tb, purr;
327 s64 stolen;
328 struct cpu_purr_data *pme;
329
330 pme = &__get_cpu_var(cpu_purr_data);
331 if (!pme->initialized)
332 return; /* !CPU_FTR_PURR or early in early boot */
333 tb = mftb();
334 purr = mfspr(SPRN_PURR);
335 stolen = (tb - pme->tb) - (purr - pme->purr);
336 if (stolen > 0) {
337 if (idle_task(smp_processor_id()) != current)
338 account_steal_time(stolen);
339 else
340 account_idle_time(stolen);
341 }
342 pme->tb = tb;
343 pme->purr = purr;
344}
345
346#ifdef CONFIG_PPC_SPLPAR
347/*
348 * Must be called before the cpu is added to the online map when
349 * a cpu is being brought up at runtime.
350 */
351static void snapshot_purr(void)
352{
353 struct cpu_purr_data *pme;
354 unsigned long flags;
355
356 if (!cpu_has_feature(CPU_FTR_PURR))
357 return;
358 local_irq_save(flags);
359 pme = &__get_cpu_var(cpu_purr_data);
360 pme->tb = mftb();
361 pme->purr = mfspr(SPRN_PURR);
362 pme->initialized = 1;
363 local_irq_restore(flags);
364}
365
366#endif /* CONFIG_PPC_SPLPAR */
367
368#else /* ! CONFIG_VIRT_CPU_ACCOUNTING */ 394#else /* ! CONFIG_VIRT_CPU_ACCOUNTING */
369#define calc_cputime_factors() 395#define calc_cputime_factors()
370#define calculate_steal_time() do { } while (0)
371#endif
372
373#if !(defined(CONFIG_VIRT_CPU_ACCOUNTING) && defined(CONFIG_PPC_SPLPAR))
374#define snapshot_purr() do { } while (0)
375#endif 396#endif
376 397
377/*
378 * Called when a cpu comes up after the system has finished booting,
379 * i.e. as a result of a hotplug cpu action.
380 */
381void snapshot_timebase(void)
382{
383 __get_cpu_var(last_jiffy) = get_tb_or_rtc();
384 snapshot_purr();
385}
386
387void __delay(unsigned long loops) 398void __delay(unsigned long loops)
388{ 399{
389 unsigned long start; 400 unsigned long start;
@@ -493,60 +504,60 @@ void __init iSeries_time_init_early(void)
493} 504}
494#endif /* CONFIG_PPC_ISERIES */ 505#endif /* CONFIG_PPC_ISERIES */
495 506
496#ifdef CONFIG_PERF_EVENTS 507#ifdef CONFIG_IRQ_WORK
497 508
498/* 509/*
499 * 64-bit uses a byte in the PACA, 32-bit uses a per-cpu variable... 510 * 64-bit uses a byte in the PACA, 32-bit uses a per-cpu variable...
500 */ 511 */
501#ifdef CONFIG_PPC64 512#ifdef CONFIG_PPC64
502static inline unsigned long test_perf_event_pending(void) 513static inline unsigned long test_irq_work_pending(void)
503{ 514{
504 unsigned long x; 515 unsigned long x;
505 516
506 asm volatile("lbz %0,%1(13)" 517 asm volatile("lbz %0,%1(13)"
507 : "=r" (x) 518 : "=r" (x)
508 : "i" (offsetof(struct paca_struct, perf_event_pending))); 519 : "i" (offsetof(struct paca_struct, irq_work_pending)));
509 return x; 520 return x;
510} 521}
511 522
512static inline void set_perf_event_pending_flag(void) 523static inline void set_irq_work_pending_flag(void)
513{ 524{
514 asm volatile("stb %0,%1(13)" : : 525 asm volatile("stb %0,%1(13)" : :
515 "r" (1), 526 "r" (1),
516 "i" (offsetof(struct paca_struct, perf_event_pending))); 527 "i" (offsetof(struct paca_struct, irq_work_pending)));
517} 528}
518 529
519static inline void clear_perf_event_pending(void) 530static inline void clear_irq_work_pending(void)
520{ 531{
521 asm volatile("stb %0,%1(13)" : : 532 asm volatile("stb %0,%1(13)" : :
522 "r" (0), 533 "r" (0),
523 "i" (offsetof(struct paca_struct, perf_event_pending))); 534 "i" (offsetof(struct paca_struct, irq_work_pending)));
524} 535}
525 536
526#else /* 32-bit */ 537#else /* 32-bit */
527 538
528DEFINE_PER_CPU(u8, perf_event_pending); 539DEFINE_PER_CPU(u8, irq_work_pending);
529 540
530#define set_perf_event_pending_flag() __get_cpu_var(perf_event_pending) = 1 541#define set_irq_work_pending_flag() __get_cpu_var(irq_work_pending) = 1
531#define test_perf_event_pending() __get_cpu_var(perf_event_pending) 542#define test_irq_work_pending() __get_cpu_var(irq_work_pending)
532#define clear_perf_event_pending() __get_cpu_var(perf_event_pending) = 0 543#define clear_irq_work_pending() __get_cpu_var(irq_work_pending) = 0
533 544
534#endif /* 32 vs 64 bit */ 545#endif /* 32 vs 64 bit */
535 546
536void set_perf_event_pending(void) 547void set_irq_work_pending(void)
537{ 548{
538 preempt_disable(); 549 preempt_disable();
539 set_perf_event_pending_flag(); 550 set_irq_work_pending_flag();
540 set_dec(1); 551 set_dec(1);
541 preempt_enable(); 552 preempt_enable();
542} 553}
543 554
544#else /* CONFIG_PERF_EVENTS */ 555#else /* CONFIG_IRQ_WORK */
545 556
546#define test_perf_event_pending() 0 557#define test_irq_work_pending() 0
547#define clear_perf_event_pending() 558#define clear_irq_work_pending()
548 559
549#endif /* CONFIG_PERF_EVENTS */ 560#endif /* CONFIG_IRQ_WORK */
550 561
551/* 562/*
552 * For iSeries shared processors, we have to let the hypervisor 563 * For iSeries shared processors, we have to let the hypervisor
@@ -569,14 +580,21 @@ void timer_interrupt(struct pt_regs * regs)
569 struct clock_event_device *evt = &decrementer->event; 580 struct clock_event_device *evt = &decrementer->event;
570 u64 now; 581 u64 now;
571 582
583 /* Ensure a positive value is written to the decrementer, or else
584 * some CPUs will continue to take decrementer exceptions.
585 */
586 set_dec(DECREMENTER_MAX);
587
588 /* Some implementations of hotplug will get timer interrupts while
589 * offline, just ignore these
590 */
591 if (!cpu_online(smp_processor_id()))
592 return;
593
572 trace_timer_interrupt_entry(regs); 594 trace_timer_interrupt_entry(regs);
573 595
574 __get_cpu_var(irq_stat).timer_irqs++; 596 __get_cpu_var(irq_stat).timer_irqs++;
575 597
576 /* Ensure a positive value is written to the decrementer, or else
577 * some CPUs will continuue to take decrementer exceptions */
578 set_dec(DECREMENTER_MAX);
579
580#if defined(CONFIG_PPC32) && defined(CONFIG_PMAC) 598#if defined(CONFIG_PPC32) && defined(CONFIG_PMAC)
581 if (atomic_read(&ppc_n_lost_interrupts) != 0) 599 if (atomic_read(&ppc_n_lost_interrupts) != 0)
582 do_IRQ(regs); 600 do_IRQ(regs);
@@ -585,11 +603,9 @@ void timer_interrupt(struct pt_regs * regs)
585 old_regs = set_irq_regs(regs); 603 old_regs = set_irq_regs(regs);
586 irq_enter(); 604 irq_enter();
587 605
588 calculate_steal_time(); 606 if (test_irq_work_pending()) {
589 607 clear_irq_work_pending();
590 if (test_perf_event_pending()) { 608 irq_work_run();
591 clear_perf_event_pending();
592 perf_event_do_pending();
593 } 609 }
594 610
595#ifdef CONFIG_PPC_ISERIES 611#ifdef CONFIG_PPC_ISERIES
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index a45a63c3a0c7..1a0141426cda 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -34,6 +34,7 @@
34#include <linux/bug.h> 34#include <linux/bug.h>
35#include <linux/kdebug.h> 35#include <linux/kdebug.h>
36#include <linux/debugfs.h> 36#include <linux/debugfs.h>
37#include <linux/ratelimit.h>
37 38
38#include <asm/emulated_ops.h> 39#include <asm/emulated_ops.h>
39#include <asm/pgtable.h> 40#include <asm/pgtable.h>
@@ -55,6 +56,7 @@
55#endif 56#endif
56#include <asm/kexec.h> 57#include <asm/kexec.h>
57#include <asm/ppc-opcode.h> 58#include <asm/ppc-opcode.h>
59#include <asm/rio.h>
58 60
59#if defined(CONFIG_DEBUGGER) || defined(CONFIG_KEXEC) 61#if defined(CONFIG_DEBUGGER) || defined(CONFIG_KEXEC)
60int (*__debugger)(struct pt_regs *regs) __read_mostly; 62int (*__debugger)(struct pt_regs *regs) __read_mostly;
@@ -143,7 +145,6 @@ int die(const char *str, struct pt_regs *regs, long err)
143#endif 145#endif
144 printk("%s\n", ppc_md.name ? ppc_md.name : ""); 146 printk("%s\n", ppc_md.name ? ppc_md.name : "");
145 147
146 sysfs_printk_last_file();
147 if (notify_die(DIE_OOPS, str, regs, err, 255, 148 if (notify_die(DIE_OOPS, str, regs, err, 255,
148 SIGSEGV) == NOTIFY_STOP) 149 SIGSEGV) == NOTIFY_STOP)
149 return 1; 150 return 1;
@@ -197,12 +198,11 @@ void _exception(int signr, struct pt_regs *regs, int code, unsigned long addr)
197 if (die("Exception in kernel mode", regs, signr)) 198 if (die("Exception in kernel mode", regs, signr))
198 return; 199 return;
199 } else if (show_unhandled_signals && 200 } else if (show_unhandled_signals &&
200 unhandled_signal(current, signr) && 201 unhandled_signal(current, signr)) {
201 printk_ratelimit()) { 202 printk_ratelimited(regs->msr & MSR_64BIT ? fmt64 : fmt32,
202 printk(regs->msr & MSR_SF ? fmt64 : fmt32, 203 current->comm, current->pid, signr,
203 current->comm, current->pid, signr, 204 addr, regs->nip, regs->link, code);
204 addr, regs->nip, regs->link, code); 205 }
205 }
206 206
207 memset(&info, 0, sizeof(info)); 207 memset(&info, 0, sizeof(info));
208 info.si_signo = signr; 208 info.si_signo = signr;
@@ -221,7 +221,7 @@ void system_reset_exception(struct pt_regs *regs)
221 } 221 }
222 222
223#ifdef CONFIG_KEXEC 223#ifdef CONFIG_KEXEC
224 cpu_set(smp_processor_id(), cpus_in_sr); 224 cpumask_set_cpu(smp_processor_id(), &cpus_in_sr);
225#endif 225#endif
226 226
227 die("System Reset", regs, SIGABRT); 227 die("System Reset", regs, SIGABRT);
@@ -425,6 +425,12 @@ int machine_check_e500mc(struct pt_regs *regs)
425 unsigned long reason = mcsr; 425 unsigned long reason = mcsr;
426 int recoverable = 1; 426 int recoverable = 1;
427 427
428 if (reason & MCSR_LD) {
429 recoverable = fsl_rio_mcheck_exception(regs);
430 if (recoverable == 1)
431 goto silent_out;
432 }
433
428 printk("Machine check in kernel mode.\n"); 434 printk("Machine check in kernel mode.\n");
429 printk("Caused by (from MCSR=%lx): ", reason); 435 printk("Caused by (from MCSR=%lx): ", reason);
430 436
@@ -500,6 +506,7 @@ int machine_check_e500mc(struct pt_regs *regs)
500 reason & MCSR_MEA ? "Effective" : "Physical", addr); 506 reason & MCSR_MEA ? "Effective" : "Physical", addr);
501 } 507 }
502 508
509silent_out:
503 mtspr(SPRN_MCSR, mcsr); 510 mtspr(SPRN_MCSR, mcsr);
504 return mfspr(SPRN_MCSR) == 0 && recoverable; 511 return mfspr(SPRN_MCSR) == 0 && recoverable;
505} 512}
@@ -508,6 +515,11 @@ int machine_check_e500(struct pt_regs *regs)
508{ 515{
509 unsigned long reason = get_mc_reason(regs); 516 unsigned long reason = get_mc_reason(regs);
510 517
518 if (reason & MCSR_BUS_RBERR) {
519 if (fsl_rio_mcheck_exception(regs))
520 return 1;
521 }
522
511 printk("Machine check in kernel mode.\n"); 523 printk("Machine check in kernel mode.\n");
512 printk("Caused by (from MCSR=%lx): ", reason); 524 printk("Caused by (from MCSR=%lx): ", reason);
513 525
@@ -538,6 +550,11 @@ int machine_check_e500(struct pt_regs *regs)
538 550
539 return 0; 551 return 0;
540} 552}
553
554int machine_check_generic(struct pt_regs *regs)
555{
556 return 0;
557}
541#elif defined(CONFIG_E200) 558#elif defined(CONFIG_E200)
542int machine_check_e200(struct pt_regs *regs) 559int machine_check_e200(struct pt_regs *regs)
543{ 560{
@@ -621,12 +638,6 @@ void machine_check_exception(struct pt_regs *regs)
621 if (recover > 0) 638 if (recover > 0)
622 return; 639 return;
623 640
624 if (user_mode(regs)) {
625 regs->msr |= MSR_RI;
626 _exception(SIGBUS, regs, BUS_ADRERR, regs->nip);
627 return;
628 }
629
630#if defined(CONFIG_8xx) && defined(CONFIG_PCI) 641#if defined(CONFIG_8xx) && defined(CONFIG_PCI)
631 /* the qspan pci read routines can cause machine checks -- Cort 642 /* the qspan pci read routines can cause machine checks -- Cort
632 * 643 *
@@ -638,16 +649,12 @@ void machine_check_exception(struct pt_regs *regs)
638 return; 649 return;
639#endif 650#endif
640 651
641 if (debugger_fault_handler(regs)) { 652 if (debugger_fault_handler(regs))
642 regs->msr |= MSR_RI;
643 return; 653 return;
644 }
645 654
646 if (check_io_access(regs)) 655 if (check_io_access(regs))
647 return; 656 return;
648 657
649 if (debugger_fault_handler(regs))
650 return;
651 die("Machine check", regs, SIGBUS); 658 die("Machine check", regs, SIGBUS);
652 659
653 /* Must die if the interrupt is not recoverable */ 660 /* Must die if the interrupt is not recoverable */
@@ -914,6 +921,26 @@ static int emulate_instruction(struct pt_regs *regs)
914 return emulate_isel(regs, instword); 921 return emulate_isel(regs, instword);
915 } 922 }
916 923
924#ifdef CONFIG_PPC64
925 /* Emulate the mfspr rD, DSCR. */
926 if (((instword & PPC_INST_MFSPR_DSCR_MASK) == PPC_INST_MFSPR_DSCR) &&
927 cpu_has_feature(CPU_FTR_DSCR)) {
928 PPC_WARN_EMULATED(mfdscr, regs);
929 rd = (instword >> 21) & 0x1f;
930 regs->gpr[rd] = mfspr(SPRN_DSCR);
931 return 0;
932 }
933 /* Emulate the mtspr DSCR, rD. */
934 if (((instword & PPC_INST_MTSPR_DSCR_MASK) == PPC_INST_MTSPR_DSCR) &&
935 cpu_has_feature(CPU_FTR_DSCR)) {
936 PPC_WARN_EMULATED(mtdscr, regs);
937 rd = (instword >> 21) & 0x1f;
938 mtspr(SPRN_DSCR, regs->gpr[rd]);
939 current->thread.dscr_inherit = 1;
940 return 0;
941 }
942#endif
943
917 return -EINVAL; 944 return -EINVAL;
918} 945}
919 946
@@ -964,7 +991,7 @@ void __kprobes program_check_exception(struct pt_regs *regs)
964 * ESR_DST (!?) or 0. In the process of chasing this with the 991 * ESR_DST (!?) or 0. In the process of chasing this with the
965 * hardware people - not sure if it can happen on any illegal 992 * hardware people - not sure if it can happen on any illegal
966 * instruction or only on FP instructions, whether there is a 993 * instruction or only on FP instructions, whether there is a
967 * pattern to occurences etc. -dgibson 31/Mar/2003 */ 994 * pattern to occurrences etc. -dgibson 31/Mar/2003 */
968 switch (do_mathemu(regs)) { 995 switch (do_mathemu(regs)) {
969 case 0: 996 case 0:
970 emulate_single_step(regs); 997 emulate_single_step(regs);
@@ -1315,9 +1342,8 @@ void altivec_assist_exception(struct pt_regs *regs)
1315 } else { 1342 } else {
1316 /* didn't recognize the instruction */ 1343 /* didn't recognize the instruction */
1317 /* XXX quick hack for now: set the non-Java bit in the VSCR */ 1344 /* XXX quick hack for now: set the non-Java bit in the VSCR */
1318 if (printk_ratelimit()) 1345 printk_ratelimited(KERN_ERR "Unrecognized altivec instruction "
1319 printk(KERN_ERR "Unrecognized altivec instruction " 1346 "in %s at %lx\n", current->comm, regs->nip);
1320 "in %s at %lx\n", current->comm, regs->nip);
1321 current->thread.vscr.u[3] |= 0x10000; 1347 current->thread.vscr.u[3] |= 0x10000;
1322 } 1348 }
1323} 1349}
@@ -1511,15 +1537,18 @@ struct ppc_emulated ppc_emulated = {
1511#ifdef CONFIG_VSX 1537#ifdef CONFIG_VSX
1512 WARN_EMULATED_SETUP(vsx), 1538 WARN_EMULATED_SETUP(vsx),
1513#endif 1539#endif
1540#ifdef CONFIG_PPC64
1541 WARN_EMULATED_SETUP(mfdscr),
1542 WARN_EMULATED_SETUP(mtdscr),
1543#endif
1514}; 1544};
1515 1545
1516u32 ppc_warn_emulated; 1546u32 ppc_warn_emulated;
1517 1547
1518void ppc_warn_emulated_print(const char *type) 1548void ppc_warn_emulated_print(const char *type)
1519{ 1549{
1520 if (printk_ratelimit()) 1550 pr_warn_ratelimited("%s used emulated %s instruction\n", current->comm,
1521 pr_warning("%s used emulated %s instruction\n", current->comm, 1551 type);
1522 type);
1523} 1552}
1524 1553
1525static int __init ppc_warn_emulated_init(void) 1554static int __init ppc_warn_emulated_init(void)
diff --git a/arch/powerpc/kernel/udbg.c b/arch/powerpc/kernel/udbg.c
index e39cad83c884..23d65abbedce 100644
--- a/arch/powerpc/kernel/udbg.c
+++ b/arch/powerpc/kernel/udbg.c
@@ -62,6 +62,8 @@ void __init udbg_early_init(void)
62 udbg_init_cpm(); 62 udbg_init_cpm();
63#elif defined(CONFIG_PPC_EARLY_DEBUG_USBGECKO) 63#elif defined(CONFIG_PPC_EARLY_DEBUG_USBGECKO)
64 udbg_init_usbgecko(); 64 udbg_init_usbgecko();
65#elif defined(CONFIG_PPC_EARLY_DEBUG_WSP)
66 udbg_init_wsp();
65#endif 67#endif
66 68
67#ifdef CONFIG_PPC_EARLY_DEBUG 69#ifdef CONFIG_PPC_EARLY_DEBUG
diff --git a/arch/powerpc/kernel/udbg_16550.c b/arch/powerpc/kernel/udbg_16550.c
index b4b167b33643..6837f839ab78 100644
--- a/arch/powerpc/kernel/udbg_16550.c
+++ b/arch/powerpc/kernel/udbg_16550.c
@@ -1,5 +1,5 @@
1/* 1/*
2 * udbg for NS16550 compatable serial ports 2 * udbg for NS16550 compatible serial ports
3 * 3 *
4 * Copyright (C) 2001-2005 PPC 64 Team, IBM Corp 4 * Copyright (C) 2001-2005 PPC 64 Team, IBM Corp
5 * 5 *
@@ -11,6 +11,7 @@
11#include <linux/types.h> 11#include <linux/types.h>
12#include <asm/udbg.h> 12#include <asm/udbg.h>
13#include <asm/io.h> 13#include <asm/io.h>
14#include <asm/reg_a2.h>
14 15
15extern u8 real_readb(volatile u8 __iomem *addr); 16extern u8 real_readb(volatile u8 __iomem *addr);
16extern void real_writeb(u8 data, volatile u8 __iomem *addr); 17extern void real_writeb(u8 data, volatile u8 __iomem *addr);
@@ -298,3 +299,53 @@ void __init udbg_init_40x_realmode(void)
298 udbg_getc_poll = NULL; 299 udbg_getc_poll = NULL;
299} 300}
300#endif /* CONFIG_PPC_EARLY_DEBUG_40x */ 301#endif /* CONFIG_PPC_EARLY_DEBUG_40x */
302
303#ifdef CONFIG_PPC_EARLY_DEBUG_WSP
304static void udbg_wsp_flush(void)
305{
306 if (udbg_comport) {
307 while ((readb(&udbg_comport->lsr) & LSR_THRE) == 0)
308 /* wait for idle */;
309 }
310}
311
312static void udbg_wsp_putc(char c)
313{
314 if (udbg_comport) {
315 if (c == '\n')
316 udbg_wsp_putc('\r');
317 udbg_wsp_flush();
318 writeb(c, &udbg_comport->thr); eieio();
319 }
320}
321
322static int udbg_wsp_getc(void)
323{
324 if (udbg_comport) {
325 while ((readb(&udbg_comport->lsr) & LSR_DR) == 0)
326 ; /* wait for char */
327 return readb(&udbg_comport->rbr);
328 }
329 return -1;
330}
331
332static int udbg_wsp_getc_poll(void)
333{
334 if (udbg_comport)
335 if (readb(&udbg_comport->lsr) & LSR_DR)
336 return readb(&udbg_comport->rbr);
337 return -1;
338}
339
340void __init udbg_init_wsp(void)
341{
342 udbg_comport = (struct NS16550 __iomem *)WSP_UART_VIRT;
343
344 udbg_init_uart(udbg_comport, 57600, 50000000);
345
346 udbg_putc = udbg_wsp_putc;
347 udbg_flush = udbg_wsp_flush;
348 udbg_getc = udbg_wsp_getc;
349 udbg_getc_poll = udbg_wsp_getc_poll;
350}
351#endif /* CONFIG_PPC_EARLY_DEBUG_WSP */
diff --git a/arch/powerpc/kernel/vdso.c b/arch/powerpc/kernel/vdso.c
index 13002fe206e7..142ab1008c3b 100644
--- a/arch/powerpc/kernel/vdso.c
+++ b/arch/powerpc/kernel/vdso.c
@@ -159,7 +159,7 @@ static void dump_vdso_pages(struct vm_area_struct * vma)
159{ 159{
160 int i; 160 int i;
161 161
162 if (!vma || test_thread_flag(TIF_32BIT)) { 162 if (!vma || is_32bit_task()) {
163 printk("vDSO32 @ %016lx:\n", (unsigned long)vdso32_kbase); 163 printk("vDSO32 @ %016lx:\n", (unsigned long)vdso32_kbase);
164 for (i=0; i<vdso32_pages; i++) { 164 for (i=0; i<vdso32_pages; i++) {
165 struct page *pg = virt_to_page(vdso32_kbase + 165 struct page *pg = virt_to_page(vdso32_kbase +
@@ -170,7 +170,7 @@ static void dump_vdso_pages(struct vm_area_struct * vma)
170 dump_one_vdso_page(pg, upg); 170 dump_one_vdso_page(pg, upg);
171 } 171 }
172 } 172 }
173 if (!vma || !test_thread_flag(TIF_32BIT)) { 173 if (!vma || !is_32bit_task()) {
174 printk("vDSO64 @ %016lx:\n", (unsigned long)vdso64_kbase); 174 printk("vDSO64 @ %016lx:\n", (unsigned long)vdso64_kbase);
175 for (i=0; i<vdso64_pages; i++) { 175 for (i=0; i<vdso64_pages; i++) {
176 struct page *pg = virt_to_page(vdso64_kbase + 176 struct page *pg = virt_to_page(vdso64_kbase +
@@ -200,7 +200,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
200 return 0; 200 return 0;
201 201
202#ifdef CONFIG_PPC64 202#ifdef CONFIG_PPC64
203 if (test_thread_flag(TIF_32BIT)) { 203 if (is_32bit_task()) {
204 vdso_pagelist = vdso32_pagelist; 204 vdso_pagelist = vdso32_pagelist;
205 vdso_pages = vdso32_pages; 205 vdso_pages = vdso32_pages;
206 vdso_base = VDSO32_MBASE; 206 vdso_base = VDSO32_MBASE;
@@ -820,17 +820,17 @@ static int __init vdso_init(void)
820} 820}
821arch_initcall(vdso_init); 821arch_initcall(vdso_init);
822 822
823int in_gate_area_no_task(unsigned long addr) 823int in_gate_area_no_mm(unsigned long addr)
824{ 824{
825 return 0; 825 return 0;
826} 826}
827 827
828int in_gate_area(struct task_struct *task, unsigned long addr) 828int in_gate_area(struct mm_struct *mm, unsigned long addr)
829{ 829{
830 return 0; 830 return 0;
831} 831}
832 832
833struct vm_area_struct *get_gate_vma(struct task_struct *tsk) 833struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
834{ 834{
835 return NULL; 835 return NULL;
836} 836}
diff --git a/arch/powerpc/kernel/vdso32/Makefile b/arch/powerpc/kernel/vdso32/Makefile
index 51ead52141bd..9a7946c41738 100644
--- a/arch/powerpc/kernel/vdso32/Makefile
+++ b/arch/powerpc/kernel/vdso32/Makefile
@@ -14,10 +14,10 @@ obj-vdso32 := $(addprefix $(obj)/, $(obj-vdso32))
14 14
15GCOV_PROFILE := n 15GCOV_PROFILE := n
16 16
17EXTRA_CFLAGS := -shared -fno-common -fno-builtin 17ccflags-y := -shared -fno-common -fno-builtin
18EXTRA_CFLAGS += -nostdlib -Wl,-soname=linux-vdso32.so.1 \ 18ccflags-y += -nostdlib -Wl,-soname=linux-vdso32.so.1 \
19 $(call cc-ldoption, -Wl$(comma)--hash-style=sysv) 19 $(call cc-ldoption, -Wl$(comma)--hash-style=sysv)
20EXTRA_AFLAGS := -D__VDSO32__ -s 20asflags-y := -D__VDSO32__ -s
21 21
22obj-y += vdso32_wrapper.o 22obj-y += vdso32_wrapper.o
23extra-y += vdso32.lds 23extra-y += vdso32.lds
diff --git a/arch/powerpc/kernel/vdso32/sigtramp.S b/arch/powerpc/kernel/vdso32/sigtramp.S
index 68d49dd71dcc..cf0c9c9c24f9 100644
--- a/arch/powerpc/kernel/vdso32/sigtramp.S
+++ b/arch/powerpc/kernel/vdso32/sigtramp.S
@@ -19,7 +19,7 @@
19 19
20/* The nop here is a hack. The dwarf2 unwind routines subtract 1 from 20/* The nop here is a hack. The dwarf2 unwind routines subtract 1 from
21 the return address to get an address in the middle of the presumed 21 the return address to get an address in the middle of the presumed
22 call instruction. Since we don't have a call here, we artifically 22 call instruction. Since we don't have a call here, we artificially
23 extend the range covered by the unwind info by adding a nop before 23 extend the range covered by the unwind info by adding a nop before
24 the real start. */ 24 the real start. */
25 nop 25 nop
diff --git a/arch/powerpc/kernel/vdso64/Makefile b/arch/powerpc/kernel/vdso64/Makefile
index 79da65d44a2a..8c500d8622e4 100644
--- a/arch/powerpc/kernel/vdso64/Makefile
+++ b/arch/powerpc/kernel/vdso64/Makefile
@@ -9,10 +9,10 @@ obj-vdso64 := $(addprefix $(obj)/, $(obj-vdso64))
9 9
10GCOV_PROFILE := n 10GCOV_PROFILE := n
11 11
12EXTRA_CFLAGS := -shared -fno-common -fno-builtin 12ccflags-y := -shared -fno-common -fno-builtin
13EXTRA_CFLAGS += -nostdlib -Wl,-soname=linux-vdso64.so.1 \ 13ccflags-y += -nostdlib -Wl,-soname=linux-vdso64.so.1 \
14 $(call cc-ldoption, -Wl$(comma)--hash-style=sysv) 14 $(call cc-ldoption, -Wl$(comma)--hash-style=sysv)
15EXTRA_AFLAGS := -D__VDSO64__ -s 15asflags-y := -D__VDSO64__ -s
16 16
17obj-y += vdso64_wrapper.o 17obj-y += vdso64_wrapper.o
18extra-y += vdso64.lds 18extra-y += vdso64.lds
diff --git a/arch/powerpc/kernel/vdso64/sigtramp.S b/arch/powerpc/kernel/vdso64/sigtramp.S
index 59eb59bb4082..45ea281e9a21 100644
--- a/arch/powerpc/kernel/vdso64/sigtramp.S
+++ b/arch/powerpc/kernel/vdso64/sigtramp.S
@@ -20,7 +20,7 @@
20 20
21/* The nop here is a hack. The dwarf2 unwind routines subtract 1 from 21/* The nop here is a hack. The dwarf2 unwind routines subtract 1 from
22 the return address to get an address in the middle of the presumed 22 the return address to get an address in the middle of the presumed
23 call instruction. Since we don't have a call here, we artifically 23 call instruction. Since we don't have a call here, we artificially
24 extend the range covered by the unwind info by padding before the 24 extend the range covered by the unwind info by padding before the
25 real start. */ 25 real start. */
26 nop 26 nop
diff --git a/arch/powerpc/kernel/vector.S b/arch/powerpc/kernel/vector.S
index fe460482fa68..4d5a3edff49e 100644
--- a/arch/powerpc/kernel/vector.S
+++ b/arch/powerpc/kernel/vector.S
@@ -5,6 +5,7 @@
5#include <asm/cputable.h> 5#include <asm/cputable.h>
6#include <asm/thread_info.h> 6#include <asm/thread_info.h>
7#include <asm/page.h> 7#include <asm/page.h>
8#include <asm/ptrace.h>
8 9
9/* 10/*
10 * load_up_altivec(unused, unused, tsk) 11 * load_up_altivec(unused, unused, tsk)
@@ -101,7 +102,7 @@ _GLOBAL(giveup_altivec)
101 MTMSRD(r5) /* enable use of VMX now */ 102 MTMSRD(r5) /* enable use of VMX now */
102 isync 103 isync
103 PPC_LCMPI 0,r3,0 104 PPC_LCMPI 0,r3,0
104 beqlr- /* if no previous owner, done */ 105 beqlr /* if no previous owner, done */
105 addi r3,r3,THREAD /* want THREAD of task */ 106 addi r3,r3,THREAD /* want THREAD of task */
106 PPC_LL r5,PT_REGS(r3) 107 PPC_LL r5,PT_REGS(r3)
107 PPC_LCMPI 0,r5,0 108 PPC_LCMPI 0,r5,0
diff --git a/arch/powerpc/kernel/vio.c b/arch/powerpc/kernel/vio.c
index fa3469ddaef8..1b695fdc362b 100644
--- a/arch/powerpc/kernel/vio.c
+++ b/arch/powerpc/kernel/vio.c
@@ -238,9 +238,7 @@ static inline void vio_cmo_dealloc(struct vio_dev *viodev, size_t size)
238 * memory in this pool does not change. 238 * memory in this pool does not change.
239 */ 239 */
240 if (spare_needed && reserve_freed) { 240 if (spare_needed && reserve_freed) {
241 tmp = min(spare_needed, min(reserve_freed, 241 tmp = min3(spare_needed, reserve_freed, (viodev->cmo.entitled - VIO_CMO_MIN_ENT));
242 (viodev->cmo.entitled -
243 VIO_CMO_MIN_ENT)));
244 242
245 vio_cmo.spare += tmp; 243 vio_cmo.spare += tmp;
246 viodev->cmo.entitled -= tmp; 244 viodev->cmo.entitled -= tmp;
@@ -602,6 +600,11 @@ static void vio_dma_iommu_unmap_sg(struct device *dev,
602 vio_cmo_dealloc(viodev, alloc_size); 600 vio_cmo_dealloc(viodev, alloc_size);
603} 601}
604 602
603static int vio_dma_iommu_dma_supported(struct device *dev, u64 mask)
604{
605 return dma_iommu_ops.dma_supported(dev, mask);
606}
607
605struct dma_map_ops vio_dma_mapping_ops = { 608struct dma_map_ops vio_dma_mapping_ops = {
606 .alloc_coherent = vio_dma_iommu_alloc_coherent, 609 .alloc_coherent = vio_dma_iommu_alloc_coherent,
607 .free_coherent = vio_dma_iommu_free_coherent, 610 .free_coherent = vio_dma_iommu_free_coherent,
@@ -609,6 +612,7 @@ struct dma_map_ops vio_dma_mapping_ops = {
609 .unmap_sg = vio_dma_iommu_unmap_sg, 612 .unmap_sg = vio_dma_iommu_unmap_sg,
610 .map_page = vio_dma_iommu_map_page, 613 .map_page = vio_dma_iommu_map_page,
611 .unmap_page = vio_dma_iommu_unmap_page, 614 .unmap_page = vio_dma_iommu_unmap_page,
615 .dma_supported = vio_dma_iommu_dma_supported,
612 616
613}; 617};
614 618
@@ -860,8 +864,7 @@ static void vio_cmo_bus_remove(struct vio_dev *viodev)
860 864
861static void vio_cmo_set_dma_ops(struct vio_dev *viodev) 865static void vio_cmo_set_dma_ops(struct vio_dev *viodev)
862{ 866{
863 vio_dma_mapping_ops.dma_supported = dma_iommu_ops.dma_supported; 867 set_dma_ops(&viodev->dev, &vio_dma_mapping_ops);
864 viodev->dev.archdata.dma_ops = &vio_dma_mapping_ops;
865} 868}
866 869
867/** 870/**
@@ -1184,7 +1187,12 @@ EXPORT_SYMBOL(vio_unregister_driver);
1184/* vio_dev refcount hit 0 */ 1187/* vio_dev refcount hit 0 */
1185static void __devinit vio_dev_release(struct device *dev) 1188static void __devinit vio_dev_release(struct device *dev)
1186{ 1189{
1187 /* XXX should free TCE table */ 1190 struct iommu_table *tbl = get_iommu_table_base(dev);
1191
1192 /* iSeries uses a common table for all vio devices */
1193 if (!firmware_has_feature(FW_FEATURE_ISERIES) && tbl)
1194 iommu_free_table(tbl, dev->of_node ?
1195 dev->of_node->full_name : dev_name(dev));
1188 of_node_put(dev->of_node); 1196 of_node_put(dev->of_node);
1189 kfree(to_vio_dev(dev)); 1197 kfree(to_vio_dev(dev));
1190} 1198}
@@ -1241,7 +1249,7 @@ struct vio_dev *vio_register_device_node(struct device_node *of_node)
1241 if (firmware_has_feature(FW_FEATURE_CMO)) 1249 if (firmware_has_feature(FW_FEATURE_CMO))
1242 vio_cmo_set_dma_ops(viodev); 1250 vio_cmo_set_dma_ops(viodev);
1243 else 1251 else
1244 viodev->dev.archdata.dma_ops = &dma_iommu_ops; 1252 set_dma_ops(&viodev->dev, &dma_iommu_ops);
1245 set_iommu_table_base(&viodev->dev, vio_build_iommu_table(viodev)); 1253 set_iommu_table_base(&viodev->dev, vio_build_iommu_table(viodev));
1246 set_dev_node(&viodev->dev, of_node_to_nid(of_node)); 1254 set_dev_node(&viodev->dev, of_node_to_nid(of_node));
1247 1255
@@ -1249,13 +1257,16 @@ struct vio_dev *vio_register_device_node(struct device_node *of_node)
1249 viodev->dev.parent = &vio_bus_device.dev; 1257 viodev->dev.parent = &vio_bus_device.dev;
1250 viodev->dev.bus = &vio_bus_type; 1258 viodev->dev.bus = &vio_bus_type;
1251 viodev->dev.release = vio_dev_release; 1259 viodev->dev.release = vio_dev_release;
1260 /* needed to ensure proper operation of coherent allocations
1261 * later, in case driver doesn't set it explicitly */
1262 dma_set_mask(&viodev->dev, DMA_BIT_MASK(64));
1263 dma_set_coherent_mask(&viodev->dev, DMA_BIT_MASK(64));
1252 1264
1253 /* register with generic device framework */ 1265 /* register with generic device framework */
1254 if (device_register(&viodev->dev)) { 1266 if (device_register(&viodev->dev)) {
1255 printk(KERN_ERR "%s: failed to register device %s\n", 1267 printk(KERN_ERR "%s: failed to register device %s\n",
1256 __func__, dev_name(&viodev->dev)); 1268 __func__, dev_name(&viodev->dev));
1257 /* XXX free TCE table */ 1269 put_device(&viodev->dev);
1258 kfree(viodev);
1259 return NULL; 1270 return NULL;
1260 } 1271 }
1261 1272
diff --git a/arch/powerpc/kernel/vmlinux.lds.S b/arch/powerpc/kernel/vmlinux.lds.S
index 8a0deefac08d..920276c0f6a1 100644
--- a/arch/powerpc/kernel/vmlinux.lds.S
+++ b/arch/powerpc/kernel/vmlinux.lds.S
@@ -160,7 +160,7 @@ SECTIONS
160 INIT_RAM_FS 160 INIT_RAM_FS
161 } 161 }
162 162
163 PERCPU(PAGE_SIZE) 163 PERCPU_SECTION(L1_CACHE_BYTES)
164 164
165 . = ALIGN(8); 165 . = ALIGN(8);
166 .machine.desc : AT(ADDR(.machine.desc) - LOAD_OFFSET) { 166 .machine.desc : AT(ADDR(.machine.desc) - LOAD_OFFSET) {
generated by cgit v1.2.2 (git 2.25.0) at 2025-03-13 23:27:23 -0400