aboutsummaryrefslogtreecommitdiffstats
path: root/arch/i386/kernel
diff options
context:
space:
mode:
Diffstat (limited to 'arch/i386/kernel')
-rw-r--r--arch/i386/kernel/Makefile5
-rw-r--r--arch/i386/kernel/acpi/cstate.c6
-rw-r--r--arch/i386/kernel/acpi/earlyquirk.c21
-rw-r--r--arch/i386/kernel/alternative.c63
-rw-r--r--arch/i386/kernel/apic.c22
-rw-r--r--arch/i386/kernel/apm.c3
-rw-r--r--arch/i386/kernel/asm-offsets.c39
-rw-r--r--arch/i386/kernel/cpu/amd.c5
-rw-r--r--arch/i386/kernel/cpu/common.c249
-rw-r--r--arch/i386/kernel/cpu/intel.c12
-rw-r--r--arch/i386/kernel/cpu/intel_cacheinfo.c11
-rw-r--r--arch/i386/kernel/cpu/mcheck/therm_throt.c2
-rw-r--r--arch/i386/kernel/cpu/mtrr/Makefile4
-rw-r--r--arch/i386/kernel/cpu/mtrr/amd.c2
-rw-r--r--arch/i386/kernel/cpu/mtrr/centaur.c9
-rw-r--r--arch/i386/kernel/cpu/mtrr/cyrix.c25
-rw-r--r--arch/i386/kernel/cpu/mtrr/generic.c78
-rw-r--r--arch/i386/kernel/cpu/mtrr/if.c31
-rw-r--r--arch/i386/kernel/cpu/mtrr/main.c71
-rw-r--r--arch/i386/kernel/cpu/mtrr/mtrr.h25
-rw-r--r--arch/i386/kernel/cpu/proc.c3
-rw-r--r--arch/i386/kernel/cpuid.c3
-rw-r--r--arch/i386/kernel/crash.c66
-rw-r--r--arch/i386/kernel/e820.c894
-rw-r--r--arch/i386/kernel/efi.c17
-rw-r--r--arch/i386/kernel/entry.S331
-rw-r--r--arch/i386/kernel/head.S66
-rw-r--r--arch/i386/kernel/hpet.c7
-rw-r--r--arch/i386/kernel/i8259.c5
-rw-r--r--arch/i386/kernel/io_apic.c66
-rw-r--r--arch/i386/kernel/kprobes.c4
-rw-r--r--arch/i386/kernel/ldt.c4
-rw-r--r--arch/i386/kernel/mca.c13
-rw-r--r--arch/i386/kernel/microcode.c2
-rw-r--r--arch/i386/kernel/module.c11
-rw-r--r--arch/i386/kernel/mpparse.c2
-rw-r--r--arch/i386/kernel/msr.c5
-rw-r--r--arch/i386/kernel/nmi.c42
-rw-r--r--arch/i386/kernel/paravirt.c569
-rw-r--r--arch/i386/kernel/pci-dma.c6
-rw-r--r--arch/i386/kernel/process.c81
-rw-r--r--arch/i386/kernel/ptrace.c18
-rw-r--r--arch/i386/kernel/quirks.c46
-rw-r--r--arch/i386/kernel/reboot.c1
-rw-r--r--arch/i386/kernel/setup.c855
-rw-r--r--arch/i386/kernel/signal.c6
-rw-r--r--arch/i386/kernel/smp.c10
-rw-r--r--arch/i386/kernel/smpboot.c69
-rw-r--r--arch/i386/kernel/sysenter.c6
-rw-r--r--arch/i386/kernel/time.c15
-rw-r--r--arch/i386/kernel/time_hpet.c15
-rw-r--r--arch/i386/kernel/topology.c8
-rw-r--r--arch/i386/kernel/traps.c120
-rw-r--r--arch/i386/kernel/tsc.c1
-rw-r--r--arch/i386/kernel/vm86.c121
-rw-r--r--arch/i386/kernel/vmlinux.lds.S147
56 files changed, 2709 insertions, 1609 deletions
diff --git a/arch/i386/kernel/Makefile b/arch/i386/kernel/Makefile
index 1a884b6e6e5c..1e8988e558c5 100644
--- a/arch/i386/kernel/Makefile
+++ b/arch/i386/kernel/Makefile
@@ -6,7 +6,7 @@ extra-y := head.o init_task.o vmlinux.lds
6 6
7obj-y := process.o signal.o entry.o traps.o irq.o \ 7obj-y := process.o signal.o entry.o traps.o irq.o \
8 ptrace.o time.o ioport.o ldt.o setup.o i8259.o sys_i386.o \ 8 ptrace.o time.o ioport.o ldt.o setup.o i8259.o sys_i386.o \
9 pci-dma.o i386_ksyms.o i387.o bootflag.o \ 9 pci-dma.o i386_ksyms.o i387.o bootflag.o e820.o\
10 quirks.o i8237.o topology.o alternative.o i8253.o tsc.o 10 quirks.o i8237.o topology.o alternative.o i8253.o tsc.o
11 11
12obj-$(CONFIG_STACKTRACE) += stacktrace.o 12obj-$(CONFIG_STACKTRACE) += stacktrace.o
@@ -40,6 +40,9 @@ obj-$(CONFIG_EARLY_PRINTK) += early_printk.o
40obj-$(CONFIG_HPET_TIMER) += hpet.o 40obj-$(CONFIG_HPET_TIMER) += hpet.o
41obj-$(CONFIG_K8_NB) += k8.o 41obj-$(CONFIG_K8_NB) += k8.o
42 42
43# Make sure this is linked after any other paravirt_ops structs: see head.S
44obj-$(CONFIG_PARAVIRT) += paravirt.o
45
43EXTRA_AFLAGS := -traditional 46EXTRA_AFLAGS := -traditional
44 47
45obj-$(CONFIG_SCx200) += scx200.o 48obj-$(CONFIG_SCx200) += scx200.o
diff --git a/arch/i386/kernel/acpi/cstate.c b/arch/i386/kernel/acpi/cstate.c
index 4664b55f623e..12e937c1ce4b 100644
--- a/arch/i386/kernel/acpi/cstate.c
+++ b/arch/i386/kernel/acpi/cstate.c
@@ -156,10 +156,8 @@ static int __init ffh_cstate_init(void)
156 156
157static void __exit ffh_cstate_exit(void) 157static void __exit ffh_cstate_exit(void)
158{ 158{
159 if (cpu_cstate_entry) { 159 free_percpu(cpu_cstate_entry);
160 free_percpu(cpu_cstate_entry); 160 cpu_cstate_entry = NULL;
161 cpu_cstate_entry = NULL;
162 }
163} 161}
164 162
165arch_initcall(ffh_cstate_init); 163arch_initcall(ffh_cstate_init);
diff --git a/arch/i386/kernel/acpi/earlyquirk.c b/arch/i386/kernel/acpi/earlyquirk.c
index c9841692bb7c..4b60af7f91dd 100644
--- a/arch/i386/kernel/acpi/earlyquirk.c
+++ b/arch/i386/kernel/acpi/earlyquirk.c
@@ -10,6 +10,7 @@
10#include <asm/pci-direct.h> 10#include <asm/pci-direct.h>
11#include <asm/acpi.h> 11#include <asm/acpi.h>
12#include <asm/apic.h> 12#include <asm/apic.h>
13#include <asm/irq.h>
13 14
14#ifdef CONFIG_ACPI 15#ifdef CONFIG_ACPI
15 16
@@ -49,6 +50,24 @@ static int __init check_bridge(int vendor, int device)
49 return 0; 50 return 0;
50} 51}
51 52
53static void check_intel(void)
54{
55 u16 vendor, device;
56
57 vendor = read_pci_config_16(0, 0, 0, PCI_VENDOR_ID);
58
59 if (vendor != PCI_VENDOR_ID_INTEL)
60 return;
61
62 device = read_pci_config_16(0, 0, 0, PCI_DEVICE_ID);
63#ifdef CONFIG_SMP
64 if (device == PCI_DEVICE_ID_INTEL_E7320_MCH ||
65 device == PCI_DEVICE_ID_INTEL_E7520_MCH ||
66 device == PCI_DEVICE_ID_INTEL_E7525_MCH)
67 quirk_intel_irqbalance();
68#endif
69}
70
52void __init check_acpi_pci(void) 71void __init check_acpi_pci(void)
53{ 72{
54 int num, slot, func; 73 int num, slot, func;
@@ -60,6 +79,8 @@ void __init check_acpi_pci(void)
60 if (!early_pci_allowed()) 79 if (!early_pci_allowed())
61 return; 80 return;
62 81
82 check_intel();
83
63 /* Poor man's PCI discovery */ 84 /* Poor man's PCI discovery */
64 for (num = 0; num < 32; num++) { 85 for (num = 0; num < 32; num++) {
65 for (slot = 0; slot < 32; slot++) { 86 for (slot = 0; slot < 32; slot++) {
diff --git a/arch/i386/kernel/alternative.c b/arch/i386/kernel/alternative.c
index 535f9794fba1..9eca21b49f6b 100644
--- a/arch/i386/kernel/alternative.c
+++ b/arch/i386/kernel/alternative.c
@@ -124,6 +124,20 @@ static unsigned char** find_nop_table(void)
124 124
125#endif /* CONFIG_X86_64 */ 125#endif /* CONFIG_X86_64 */
126 126
127static void nop_out(void *insns, unsigned int len)
128{
129 unsigned char **noptable = find_nop_table();
130
131 while (len > 0) {
132 unsigned int noplen = len;
133 if (noplen > ASM_NOP_MAX)
134 noplen = ASM_NOP_MAX;
135 memcpy(insns, noptable[noplen], noplen);
136 insns += noplen;
137 len -= noplen;
138 }
139}
140
127extern struct alt_instr __alt_instructions[], __alt_instructions_end[]; 141extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
128extern struct alt_instr __smp_alt_instructions[], __smp_alt_instructions_end[]; 142extern struct alt_instr __smp_alt_instructions[], __smp_alt_instructions_end[];
129extern u8 *__smp_locks[], *__smp_locks_end[]; 143extern u8 *__smp_locks[], *__smp_locks_end[];
@@ -138,10 +152,9 @@ extern u8 __smp_alt_begin[], __smp_alt_end[];
138 152
139void apply_alternatives(struct alt_instr *start, struct alt_instr *end) 153void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
140{ 154{
141 unsigned char **noptable = find_nop_table();
142 struct alt_instr *a; 155 struct alt_instr *a;
143 u8 *instr; 156 u8 *instr;
144 int diff, i, k; 157 int diff;
145 158
146 DPRINTK("%s: alt table %p -> %p\n", __FUNCTION__, start, end); 159 DPRINTK("%s: alt table %p -> %p\n", __FUNCTION__, start, end);
147 for (a = start; a < end; a++) { 160 for (a = start; a < end; a++) {
@@ -159,13 +172,7 @@ void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
159#endif 172#endif
160 memcpy(instr, a->replacement, a->replacementlen); 173 memcpy(instr, a->replacement, a->replacementlen);
161 diff = a->instrlen - a->replacementlen; 174 diff = a->instrlen - a->replacementlen;
162 /* Pad the rest with nops */ 175 nop_out(instr + a->replacementlen, diff);
163 for (i = a->replacementlen; diff > 0; diff -= k, i += k) {
164 k = diff;
165 if (k > ASM_NOP_MAX)
166 k = ASM_NOP_MAX;
167 memcpy(a->instr + i, noptable[k], k);
168 }
169 } 176 }
170} 177}
171 178
@@ -209,7 +216,6 @@ static void alternatives_smp_lock(u8 **start, u8 **end, u8 *text, u8 *text_end)
209 216
210static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end) 217static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end)
211{ 218{
212 unsigned char **noptable = find_nop_table();
213 u8 **ptr; 219 u8 **ptr;
214 220
215 for (ptr = start; ptr < end; ptr++) { 221 for (ptr = start; ptr < end; ptr++) {
@@ -217,7 +223,7 @@ static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end
217 continue; 223 continue;
218 if (*ptr > text_end) 224 if (*ptr > text_end)
219 continue; 225 continue;
220 **ptr = noptable[1][0]; 226 nop_out(*ptr, 1);
221 }; 227 };
222} 228}
223 229
@@ -343,6 +349,40 @@ void alternatives_smp_switch(int smp)
343 349
344#endif 350#endif
345 351
352#ifdef CONFIG_PARAVIRT
353void apply_paravirt(struct paravirt_patch *start, struct paravirt_patch *end)
354{
355 struct paravirt_patch *p;
356
357 for (p = start; p < end; p++) {
358 unsigned int used;
359
360 used = paravirt_ops.patch(p->instrtype, p->clobbers, p->instr,
361 p->len);
362#ifdef CONFIG_DEBUG_PARAVIRT
363 {
364 int i;
365 /* Deliberately clobber regs using "not %reg" to find bugs. */
366 for (i = 0; i < 3; i++) {
367 if (p->len - used >= 2 && (p->clobbers & (1 << i))) {
368 memcpy(p->instr + used, "\xf7\xd0", 2);
369 p->instr[used+1] |= i;
370 used += 2;
371 }
372 }
373 }
374#endif
375 /* Pad the rest with nops */
376 nop_out(p->instr + used, p->len - used);
377 }
378
379 /* Sync to be conservative, in case we patched following instructions */
380 sync_core();
381}
382extern struct paravirt_patch __start_parainstructions[],
383 __stop_parainstructions[];
384#endif /* CONFIG_PARAVIRT */
385
346void __init alternative_instructions(void) 386void __init alternative_instructions(void)
347{ 387{
348 unsigned long flags; 388 unsigned long flags;
@@ -390,5 +430,6 @@ void __init alternative_instructions(void)
390 alternatives_smp_switch(0); 430 alternatives_smp_switch(0);
391 } 431 }
392#endif 432#endif
433 apply_paravirt(__start_parainstructions, __stop_parainstructions);
393 local_irq_restore(flags); 434 local_irq_restore(flags);
394} 435}
diff --git a/arch/i386/kernel/apic.c b/arch/i386/kernel/apic.c
index 2fd4b7d927c2..776d9be26af9 100644
--- a/arch/i386/kernel/apic.c
+++ b/arch/i386/kernel/apic.c
@@ -647,23 +647,30 @@ static struct {
647static int lapic_suspend(struct sys_device *dev, pm_message_t state) 647static int lapic_suspend(struct sys_device *dev, pm_message_t state)
648{ 648{
649 unsigned long flags; 649 unsigned long flags;
650 int maxlvt;
650 651
651 if (!apic_pm_state.active) 652 if (!apic_pm_state.active)
652 return 0; 653 return 0;
653 654
655 maxlvt = get_maxlvt();
656
654 apic_pm_state.apic_id = apic_read(APIC_ID); 657 apic_pm_state.apic_id = apic_read(APIC_ID);
655 apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI); 658 apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI);
656 apic_pm_state.apic_ldr = apic_read(APIC_LDR); 659 apic_pm_state.apic_ldr = apic_read(APIC_LDR);
657 apic_pm_state.apic_dfr = apic_read(APIC_DFR); 660 apic_pm_state.apic_dfr = apic_read(APIC_DFR);
658 apic_pm_state.apic_spiv = apic_read(APIC_SPIV); 661 apic_pm_state.apic_spiv = apic_read(APIC_SPIV);
659 apic_pm_state.apic_lvtt = apic_read(APIC_LVTT); 662 apic_pm_state.apic_lvtt = apic_read(APIC_LVTT);
660 apic_pm_state.apic_lvtpc = apic_read(APIC_LVTPC); 663 if (maxlvt >= 4)
664 apic_pm_state.apic_lvtpc = apic_read(APIC_LVTPC);
661 apic_pm_state.apic_lvt0 = apic_read(APIC_LVT0); 665 apic_pm_state.apic_lvt0 = apic_read(APIC_LVT0);
662 apic_pm_state.apic_lvt1 = apic_read(APIC_LVT1); 666 apic_pm_state.apic_lvt1 = apic_read(APIC_LVT1);
663 apic_pm_state.apic_lvterr = apic_read(APIC_LVTERR); 667 apic_pm_state.apic_lvterr = apic_read(APIC_LVTERR);
664 apic_pm_state.apic_tmict = apic_read(APIC_TMICT); 668 apic_pm_state.apic_tmict = apic_read(APIC_TMICT);
665 apic_pm_state.apic_tdcr = apic_read(APIC_TDCR); 669 apic_pm_state.apic_tdcr = apic_read(APIC_TDCR);
666 apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR); 670#ifdef CONFIG_X86_MCE_P4THERMAL
671 if (maxlvt >= 5)
672 apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR);
673#endif
667 674
668 local_irq_save(flags); 675 local_irq_save(flags);
669 disable_local_APIC(); 676 disable_local_APIC();
@@ -675,10 +682,13 @@ static int lapic_resume(struct sys_device *dev)
675{ 682{
676 unsigned int l, h; 683 unsigned int l, h;
677 unsigned long flags; 684 unsigned long flags;
685 int maxlvt;
678 686
679 if (!apic_pm_state.active) 687 if (!apic_pm_state.active)
680 return 0; 688 return 0;
681 689
690 maxlvt = get_maxlvt();
691
682 local_irq_save(flags); 692 local_irq_save(flags);
683 693
684 /* 694 /*
@@ -700,8 +710,12 @@ static int lapic_resume(struct sys_device *dev)
700 apic_write(APIC_SPIV, apic_pm_state.apic_spiv); 710 apic_write(APIC_SPIV, apic_pm_state.apic_spiv);
701 apic_write(APIC_LVT0, apic_pm_state.apic_lvt0); 711 apic_write(APIC_LVT0, apic_pm_state.apic_lvt0);
702 apic_write(APIC_LVT1, apic_pm_state.apic_lvt1); 712 apic_write(APIC_LVT1, apic_pm_state.apic_lvt1);
703 apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr); 713#ifdef CONFIG_X86_MCE_P4THERMAL
704 apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc); 714 if (maxlvt >= 5)
715 apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr);
716#endif
717 if (maxlvt >= 4)
718 apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc);
705 apic_write(APIC_LVTT, apic_pm_state.apic_lvtt); 719 apic_write(APIC_LVTT, apic_pm_state.apic_lvtt);
706 apic_write(APIC_TDCR, apic_pm_state.apic_tdcr); 720 apic_write(APIC_TDCR, apic_pm_state.apic_tdcr);
707 apic_write(APIC_TMICT, apic_pm_state.apic_tmict); 721 apic_write(APIC_TMICT, apic_pm_state.apic_tmict);
diff --git a/arch/i386/kernel/apm.c b/arch/i386/kernel/apm.c
index a60358fe9a49..a97847da9ed5 100644
--- a/arch/i386/kernel/apm.c
+++ b/arch/i386/kernel/apm.c
@@ -231,6 +231,7 @@
231#include <asm/uaccess.h> 231#include <asm/uaccess.h>
232#include <asm/desc.h> 232#include <asm/desc.h>
233#include <asm/i8253.h> 233#include <asm/i8253.h>
234#include <asm/paravirt.h>
234 235
235#include "io_ports.h" 236#include "io_ports.h"
236 237
@@ -2235,7 +2236,7 @@ static int __init apm_init(void)
2235 2236
2236 dmi_check_system(apm_dmi_table); 2237 dmi_check_system(apm_dmi_table);
2237 2238
2238 if (apm_info.bios.version == 0) { 2239 if (apm_info.bios.version == 0 || paravirt_enabled()) {
2239 printk(KERN_INFO "apm: BIOS not found.\n"); 2240 printk(KERN_INFO "apm: BIOS not found.\n");
2240 return -ENODEV; 2241 return -ENODEV;
2241 } 2242 }
diff --git a/arch/i386/kernel/asm-offsets.c b/arch/i386/kernel/asm-offsets.c
index c80271f8f084..1b2f3cd33270 100644
--- a/arch/i386/kernel/asm-offsets.c
+++ b/arch/i386/kernel/asm-offsets.c
@@ -15,6 +15,7 @@
15#include <asm/processor.h> 15#include <asm/processor.h>
16#include <asm/thread_info.h> 16#include <asm/thread_info.h>
17#include <asm/elf.h> 17#include <asm/elf.h>
18#include <asm/pda.h>
18 19
19#define DEFINE(sym, val) \ 20#define DEFINE(sym, val) \
20 asm volatile("\n->" #sym " %0 " #val : : "i" (val)) 21 asm volatile("\n->" #sym " %0 " #val : : "i" (val))
@@ -51,13 +52,35 @@ void foo(void)
51 OFFSET(TI_exec_domain, thread_info, exec_domain); 52 OFFSET(TI_exec_domain, thread_info, exec_domain);
52 OFFSET(TI_flags, thread_info, flags); 53 OFFSET(TI_flags, thread_info, flags);
53 OFFSET(TI_status, thread_info, status); 54 OFFSET(TI_status, thread_info, status);
54 OFFSET(TI_cpu, thread_info, cpu);
55 OFFSET(TI_preempt_count, thread_info, preempt_count); 55 OFFSET(TI_preempt_count, thread_info, preempt_count);
56 OFFSET(TI_addr_limit, thread_info, addr_limit); 56 OFFSET(TI_addr_limit, thread_info, addr_limit);
57 OFFSET(TI_restart_block, thread_info, restart_block); 57 OFFSET(TI_restart_block, thread_info, restart_block);
58 OFFSET(TI_sysenter_return, thread_info, sysenter_return); 58 OFFSET(TI_sysenter_return, thread_info, sysenter_return);
59 BLANK(); 59 BLANK();
60 60
61 OFFSET(GDS_size, Xgt_desc_struct, size);
62 OFFSET(GDS_address, Xgt_desc_struct, address);
63 OFFSET(GDS_pad, Xgt_desc_struct, pad);
64 BLANK();
65
66 OFFSET(PT_EBX, pt_regs, ebx);
67 OFFSET(PT_ECX, pt_regs, ecx);
68 OFFSET(PT_EDX, pt_regs, edx);
69 OFFSET(PT_ESI, pt_regs, esi);
70 OFFSET(PT_EDI, pt_regs, edi);
71 OFFSET(PT_EBP, pt_regs, ebp);
72 OFFSET(PT_EAX, pt_regs, eax);
73 OFFSET(PT_DS, pt_regs, xds);
74 OFFSET(PT_ES, pt_regs, xes);
75 OFFSET(PT_GS, pt_regs, xgs);
76 OFFSET(PT_ORIG_EAX, pt_regs, orig_eax);
77 OFFSET(PT_EIP, pt_regs, eip);
78 OFFSET(PT_CS, pt_regs, xcs);
79 OFFSET(PT_EFLAGS, pt_regs, eflags);
80 OFFSET(PT_OLDESP, pt_regs, esp);
81 OFFSET(PT_OLDSS, pt_regs, xss);
82 BLANK();
83
61 OFFSET(EXEC_DOMAIN_handler, exec_domain, handler); 84 OFFSET(EXEC_DOMAIN_handler, exec_domain, handler);
62 OFFSET(RT_SIGFRAME_sigcontext, rt_sigframe, uc.uc_mcontext); 85 OFFSET(RT_SIGFRAME_sigcontext, rt_sigframe, uc.uc_mcontext);
63 BLANK(); 86 BLANK();
@@ -74,4 +97,18 @@ void foo(void)
74 DEFINE(VDSO_PRELINK, VDSO_PRELINK); 97 DEFINE(VDSO_PRELINK, VDSO_PRELINK);
75 98
76 OFFSET(crypto_tfm_ctx_offset, crypto_tfm, __crt_ctx); 99 OFFSET(crypto_tfm_ctx_offset, crypto_tfm, __crt_ctx);
100
101 BLANK();
102 OFFSET(PDA_cpu, i386_pda, cpu_number);
103 OFFSET(PDA_pcurrent, i386_pda, pcurrent);
104
105#ifdef CONFIG_PARAVIRT
106 BLANK();
107 OFFSET(PARAVIRT_enabled, paravirt_ops, paravirt_enabled);
108 OFFSET(PARAVIRT_irq_disable, paravirt_ops, irq_disable);
109 OFFSET(PARAVIRT_irq_enable, paravirt_ops, irq_enable);
110 OFFSET(PARAVIRT_irq_enable_sysexit, paravirt_ops, irq_enable_sysexit);
111 OFFSET(PARAVIRT_iret, paravirt_ops, iret);
112 OFFSET(PARAVIRT_read_cr0, paravirt_ops, read_cr0);
113#endif
77} 114}
diff --git a/arch/i386/kernel/cpu/amd.c b/arch/i386/kernel/cpu/amd.c
index e4758095d87a..41cfea57232b 100644
--- a/arch/i386/kernel/cpu/amd.c
+++ b/arch/i386/kernel/cpu/amd.c
@@ -104,10 +104,7 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
104 f_vide(); 104 f_vide();
105 rdtscl(d2); 105 rdtscl(d2);
106 d = d2-d; 106 d = d2-d;
107 107
108 /* Knock these two lines out if it debugs out ok */
109 printk(KERN_INFO "AMD K6 stepping B detected - ");
110 /* -- cut here -- */
111 if (d > 20*K6_BUG_LOOP) 108 if (d > 20*K6_BUG_LOOP)
112 printk("system stability may be impaired when more than 32 MB are used.\n"); 109 printk("system stability may be impaired when more than 32 MB are used.\n");
113 else 110 else
diff --git a/arch/i386/kernel/cpu/common.c b/arch/i386/kernel/cpu/common.c
index d9f3e3c31f05..1b34c56f8123 100644
--- a/arch/i386/kernel/cpu/common.c
+++ b/arch/i386/kernel/cpu/common.c
@@ -18,14 +18,15 @@
18#include <asm/apic.h> 18#include <asm/apic.h>
19#include <mach_apic.h> 19#include <mach_apic.h>
20#endif 20#endif
21#include <asm/pda.h>
21 22
22#include "cpu.h" 23#include "cpu.h"
23 24
24DEFINE_PER_CPU(struct Xgt_desc_struct, cpu_gdt_descr); 25DEFINE_PER_CPU(struct Xgt_desc_struct, cpu_gdt_descr);
25EXPORT_PER_CPU_SYMBOL(cpu_gdt_descr); 26EXPORT_PER_CPU_SYMBOL(cpu_gdt_descr);
26 27
27DEFINE_PER_CPU(unsigned char, cpu_16bit_stack[CPU_16BIT_STACK_SIZE]); 28struct i386_pda *_cpu_pda[NR_CPUS] __read_mostly;
28EXPORT_PER_CPU_SYMBOL(cpu_16bit_stack); 29EXPORT_SYMBOL(_cpu_pda);
29 30
30static int cachesize_override __cpuinitdata = -1; 31static int cachesize_override __cpuinitdata = -1;
31static int disable_x86_fxsr __cpuinitdata; 32static int disable_x86_fxsr __cpuinitdata;
@@ -235,29 +236,14 @@ static int __cpuinit have_cpuid_p(void)
235 return flag_is_changeable_p(X86_EFLAGS_ID); 236 return flag_is_changeable_p(X86_EFLAGS_ID);
236} 237}
237 238
238/* Do minimum CPU detection early. 239void __init cpu_detect(struct cpuinfo_x86 *c)
239 Fields really needed: vendor, cpuid_level, family, model, mask, cache alignment.
240 The others are not touched to avoid unwanted side effects.
241
242 WARNING: this function is only called on the BP. Don't add code here
243 that is supposed to run on all CPUs. */
244static void __init early_cpu_detect(void)
245{ 240{
246 struct cpuinfo_x86 *c = &boot_cpu_data;
247
248 c->x86_cache_alignment = 32;
249
250 if (!have_cpuid_p())
251 return;
252
253 /* Get vendor name */ 241 /* Get vendor name */
254 cpuid(0x00000000, &c->cpuid_level, 242 cpuid(0x00000000, &c->cpuid_level,
255 (int *)&c->x86_vendor_id[0], 243 (int *)&c->x86_vendor_id[0],
256 (int *)&c->x86_vendor_id[8], 244 (int *)&c->x86_vendor_id[8],
257 (int *)&c->x86_vendor_id[4]); 245 (int *)&c->x86_vendor_id[4]);
258 246
259 get_cpu_vendor(c, 1);
260
261 c->x86 = 4; 247 c->x86 = 4;
262 if (c->cpuid_level >= 0x00000001) { 248 if (c->cpuid_level >= 0x00000001) {
263 u32 junk, tfms, cap0, misc; 249 u32 junk, tfms, cap0, misc;
@@ -274,6 +260,26 @@ static void __init early_cpu_detect(void)
274 } 260 }
275} 261}
276 262
263/* Do minimum CPU detection early.
264 Fields really needed: vendor, cpuid_level, family, model, mask, cache alignment.
265 The others are not touched to avoid unwanted side effects.
266
267 WARNING: this function is only called on the BP. Don't add code here
268 that is supposed to run on all CPUs. */
269static void __init early_cpu_detect(void)
270{
271 struct cpuinfo_x86 *c = &boot_cpu_data;
272
273 c->x86_cache_alignment = 32;
274
275 if (!have_cpuid_p())
276 return;
277
278 cpu_detect(c);
279
280 get_cpu_vendor(c, 1);
281}
282
277static void __cpuinit generic_identify(struct cpuinfo_x86 * c) 283static void __cpuinit generic_identify(struct cpuinfo_x86 * c)
278{ 284{
279 u32 tfms, xlvl; 285 u32 tfms, xlvl;
@@ -308,6 +314,8 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 * c)
308#else 314#else
309 c->apicid = (ebx >> 24) & 0xFF; 315 c->apicid = (ebx >> 24) & 0xFF;
310#endif 316#endif
317 if (c->x86_capability[0] & (1<<19))
318 c->x86_clflush_size = ((ebx >> 8) & 0xff) * 8;
311 } else { 319 } else {
312 /* Have CPUID level 0 only - unheard of */ 320 /* Have CPUID level 0 only - unheard of */
313 c->x86 = 4; 321 c->x86 = 4;
@@ -372,6 +380,7 @@ void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
372 c->x86_vendor_id[0] = '\0'; /* Unset */ 380 c->x86_vendor_id[0] = '\0'; /* Unset */
373 c->x86_model_id[0] = '\0'; /* Unset */ 381 c->x86_model_id[0] = '\0'; /* Unset */
374 c->x86_max_cores = 1; 382 c->x86_max_cores = 1;
383 c->x86_clflush_size = 32;
375 memset(&c->x86_capability, 0, sizeof c->x86_capability); 384 memset(&c->x86_capability, 0, sizeof c->x86_capability);
376 385
377 if (!have_cpuid_p()) { 386 if (!have_cpuid_p()) {
@@ -591,42 +600,24 @@ void __init early_cpu_init(void)
591 disable_pse = 1; 600 disable_pse = 1;
592#endif 601#endif
593} 602}
594/* 603
595 * cpu_init() initializes state that is per-CPU. Some data is already 604/* Make sure %gs is initialized properly in idle threads */
596 * initialized (naturally) in the bootstrap process, such as the GDT 605struct pt_regs * __devinit idle_regs(struct pt_regs *regs)
597 * and IDT. We reload them nevertheless, this function acts as a
598 * 'CPU state barrier', nothing should get across.
599 */
600void __cpuinit cpu_init(void)
601{ 606{
602 int cpu = smp_processor_id(); 607 memset(regs, 0, sizeof(struct pt_regs));
603 struct tss_struct * t = &per_cpu(init_tss, cpu); 608 regs->xgs = __KERNEL_PDA;
604 struct thread_struct *thread = &current->thread; 609 return regs;
605 struct desc_struct *gdt; 610}
606 __u32 stk16_off = (__u32)&per_cpu(cpu_16bit_stack, cpu);
607 struct Xgt_desc_struct *cpu_gdt_descr = &per_cpu(cpu_gdt_descr, cpu);
608 611
609 if (cpu_test_and_set(cpu, cpu_initialized)) { 612static __cpuinit int alloc_gdt(int cpu)
610 printk(KERN_WARNING "CPU#%d already initialized!\n", cpu); 613{
611 for (;;) local_irq_enable(); 614 struct Xgt_desc_struct *cpu_gdt_descr = &per_cpu(cpu_gdt_descr, cpu);
612 } 615 struct desc_struct *gdt;
613 printk(KERN_INFO "Initializing CPU#%d\n", cpu); 616 struct i386_pda *pda;
614 617
615 if (cpu_has_vme || cpu_has_tsc || cpu_has_de) 618 gdt = (struct desc_struct *)cpu_gdt_descr->address;
616 clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE); 619 pda = cpu_pda(cpu);
617 if (tsc_disable && cpu_has_tsc) {
618 printk(KERN_NOTICE "Disabling TSC...\n");
619 /**** FIX-HPA: DOES THIS REALLY BELONG HERE? ****/
620 clear_bit(X86_FEATURE_TSC, boot_cpu_data.x86_capability);
621 set_in_cr4(X86_CR4_TSD);
622 }
623 620
624 /* The CPU hotplug case */
625 if (cpu_gdt_descr->address) {
626 gdt = (struct desc_struct *)cpu_gdt_descr->address;
627 memset(gdt, 0, PAGE_SIZE);
628 goto old_gdt;
629 }
630 /* 621 /*
631 * This is a horrible hack to allocate the GDT. The problem 622 * This is a horrible hack to allocate the GDT. The problem
632 * is that cpu_init() is called really early for the boot CPU 623 * is that cpu_init() is called really early for the boot CPU
@@ -634,43 +625,130 @@ void __cpuinit cpu_init(void)
634 * CPUs, when bootmem will have gone away 625 * CPUs, when bootmem will have gone away
635 */ 626 */
636 if (NODE_DATA(0)->bdata->node_bootmem_map) { 627 if (NODE_DATA(0)->bdata->node_bootmem_map) {
637 gdt = (struct desc_struct *)alloc_bootmem_pages(PAGE_SIZE); 628 BUG_ON(gdt != NULL || pda != NULL);
638 /* alloc_bootmem_pages panics on failure, so no check */ 629
630 gdt = alloc_bootmem_pages(PAGE_SIZE);
631 pda = alloc_bootmem(sizeof(*pda));
632 /* alloc_bootmem(_pages) panics on failure, so no check */
633
639 memset(gdt, 0, PAGE_SIZE); 634 memset(gdt, 0, PAGE_SIZE);
635 memset(pda, 0, sizeof(*pda));
640 } else { 636 } else {
641 gdt = (struct desc_struct *)get_zeroed_page(GFP_KERNEL); 637 /* GDT and PDA might already have been allocated if
642 if (unlikely(!gdt)) { 638 this is a CPU hotplug re-insertion. */
643 printk(KERN_CRIT "CPU%d failed to allocate GDT\n", cpu); 639 if (gdt == NULL)
644 for (;;) 640 gdt = (struct desc_struct *)get_zeroed_page(GFP_KERNEL);
645 local_irq_enable(); 641
642 if (pda == NULL)
643 pda = kmalloc_node(sizeof(*pda), GFP_KERNEL, cpu_to_node(cpu));
644
645 if (unlikely(!gdt || !pda)) {
646 free_pages((unsigned long)gdt, 0);
647 kfree(pda);
648 return 0;
646 } 649 }
647 } 650 }
648old_gdt: 651
652 cpu_gdt_descr->address = (unsigned long)gdt;
653 cpu_pda(cpu) = pda;
654
655 return 1;
656}
657
658/* Initial PDA used by boot CPU */
659struct i386_pda boot_pda = {
660 ._pda = &boot_pda,
661 .cpu_number = 0,
662 .pcurrent = &init_task,
663};
664
665static inline void set_kernel_gs(void)
666{
667 /* Set %gs for this CPU's PDA. Memory clobber is to create a
668 barrier with respect to any PDA operations, so the compiler
669 doesn't move any before here. */
670 asm volatile ("mov %0, %%gs" : : "r" (__KERNEL_PDA) : "memory");
671}
672
673/* Initialize the CPU's GDT and PDA. The boot CPU does this for
674 itself, but secondaries find this done for them. */
675__cpuinit int init_gdt(int cpu, struct task_struct *idle)
676{
677 struct Xgt_desc_struct *cpu_gdt_descr = &per_cpu(cpu_gdt_descr, cpu);
678 struct desc_struct *gdt;
679 struct i386_pda *pda;
680
681 /* For non-boot CPUs, the GDT and PDA should already have been
682 allocated. */
683 if (!alloc_gdt(cpu)) {
684 printk(KERN_CRIT "CPU%d failed to allocate GDT or PDA\n", cpu);
685 return 0;
686 }
687
688 gdt = (struct desc_struct *)cpu_gdt_descr->address;
689 pda = cpu_pda(cpu);
690
691 BUG_ON(gdt == NULL || pda == NULL);
692
649 /* 693 /*
650 * Initialize the per-CPU GDT with the boot GDT, 694 * Initialize the per-CPU GDT with the boot GDT,
651 * and set up the GDT descriptor: 695 * and set up the GDT descriptor:
652 */ 696 */
653 memcpy(gdt, cpu_gdt_table, GDT_SIZE); 697 memcpy(gdt, cpu_gdt_table, GDT_SIZE);
698 cpu_gdt_descr->size = GDT_SIZE - 1;
654 699
655 /* Set up GDT entry for 16bit stack */ 700 pack_descriptor((u32 *)&gdt[GDT_ENTRY_PDA].a,
656 *(__u64 *)(&gdt[GDT_ENTRY_ESPFIX_SS]) |= 701 (u32 *)&gdt[GDT_ENTRY_PDA].b,
657 ((((__u64)stk16_off) << 16) & 0x000000ffffff0000ULL) | 702 (unsigned long)pda, sizeof(*pda) - 1,
658 ((((__u64)stk16_off) << 32) & 0xff00000000000000ULL) | 703 0x80 | DESCTYPE_S | 0x2, 0); /* present read-write data segment */
659 (CPU_16BIT_STACK_SIZE - 1);
660 704
661 cpu_gdt_descr->size = GDT_SIZE - 1; 705 memset(pda, 0, sizeof(*pda));
662 cpu_gdt_descr->address = (unsigned long)gdt; 706 pda->_pda = pda;
707 pda->cpu_number = cpu;
708 pda->pcurrent = idle;
709
710 return 1;
711}
712
713/* Common CPU init for both boot and secondary CPUs */
714static void __cpuinit _cpu_init(int cpu, struct task_struct *curr)
715{
716 struct tss_struct * t = &per_cpu(init_tss, cpu);
717 struct thread_struct *thread = &curr->thread;
718 struct Xgt_desc_struct *cpu_gdt_descr = &per_cpu(cpu_gdt_descr, cpu);
663 719
720 /* Reinit these anyway, even if they've already been done (on
721 the boot CPU, this will transition from the boot gdt+pda to
722 the real ones). */
664 load_gdt(cpu_gdt_descr); 723 load_gdt(cpu_gdt_descr);
724 set_kernel_gs();
725
726 if (cpu_test_and_set(cpu, cpu_initialized)) {
727 printk(KERN_WARNING "CPU#%d already initialized!\n", cpu);
728 for (;;) local_irq_enable();
729 }
730
731 printk(KERN_INFO "Initializing CPU#%d\n", cpu);
732
733 if (cpu_has_vme || cpu_has_tsc || cpu_has_de)
734 clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
735 if (tsc_disable && cpu_has_tsc) {
736 printk(KERN_NOTICE "Disabling TSC...\n");
737 /**** FIX-HPA: DOES THIS REALLY BELONG HERE? ****/
738 clear_bit(X86_FEATURE_TSC, boot_cpu_data.x86_capability);
739 set_in_cr4(X86_CR4_TSD);
740 }
741
665 load_idt(&idt_descr); 742 load_idt(&idt_descr);
666 743
667 /* 744 /*
668 * Set up and load the per-CPU TSS and LDT 745 * Set up and load the per-CPU TSS and LDT
669 */ 746 */
670 atomic_inc(&init_mm.mm_count); 747 atomic_inc(&init_mm.mm_count);
671 current->active_mm = &init_mm; 748 curr->active_mm = &init_mm;
672 BUG_ON(current->mm); 749 if (curr->mm)
673 enter_lazy_tlb(&init_mm, current); 750 BUG();
751 enter_lazy_tlb(&init_mm, curr);
674 752
675 load_esp0(t, thread); 753 load_esp0(t, thread);
676 set_tss_desc(cpu,t); 754 set_tss_desc(cpu,t);
@@ -682,8 +760,8 @@ old_gdt:
682 __set_tss_desc(cpu, GDT_ENTRY_DOUBLEFAULT_TSS, &doublefault_tss); 760 __set_tss_desc(cpu, GDT_ENTRY_DOUBLEFAULT_TSS, &doublefault_tss);
683#endif 761#endif
684 762
685 /* Clear %fs and %gs. */ 763 /* Clear %fs. */
686 asm volatile ("movl %0, %%fs; movl %0, %%gs" : : "r" (0)); 764 asm volatile ("mov %0, %%fs" : : "r" (0));
687 765
688 /* Clear all 6 debug registers: */ 766 /* Clear all 6 debug registers: */
689 set_debugreg(0, 0); 767 set_debugreg(0, 0);
@@ -701,6 +779,37 @@ old_gdt:
701 mxcsr_feature_mask_init(); 779 mxcsr_feature_mask_init();
702} 780}
703 781
782/* Entrypoint to initialize secondary CPU */
783void __cpuinit secondary_cpu_init(void)
784{
785 int cpu = smp_processor_id();
786 struct task_struct *curr = current;
787
788 _cpu_init(cpu, curr);
789}
790
791/*
792 * cpu_init() initializes state that is per-CPU. Some data is already
793 * initialized (naturally) in the bootstrap process, such as the GDT
794 * and IDT. We reload them nevertheless, this function acts as a
795 * 'CPU state barrier', nothing should get across.
796 */
797void __cpuinit cpu_init(void)
798{
799 int cpu = smp_processor_id();
800 struct task_struct *curr = current;
801
802 /* Set up the real GDT and PDA, so we can transition from the
803 boot versions. */
804 if (!init_gdt(cpu, curr)) {
805 /* failed to allocate something; not much we can do... */
806 for (;;)
807 local_irq_enable();
808 }
809
810 _cpu_init(cpu, curr);
811}
812
704#ifdef CONFIG_HOTPLUG_CPU 813#ifdef CONFIG_HOTPLUG_CPU
705void __cpuinit cpu_uninit(void) 814void __cpuinit cpu_uninit(void)
706{ 815{
diff --git a/arch/i386/kernel/cpu/intel.c b/arch/i386/kernel/cpu/intel.c
index 94a95aa5227e..56fe26584957 100644
--- a/arch/i386/kernel/cpu/intel.c
+++ b/arch/i386/kernel/cpu/intel.c
@@ -107,7 +107,7 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
107 * Note that the workaround only should be initialized once... 107 * Note that the workaround only should be initialized once...
108 */ 108 */
109 c->f00f_bug = 0; 109 c->f00f_bug = 0;
110 if ( c->x86 == 5 ) { 110 if (!paravirt_enabled() && c->x86 == 5) {
111 static int f00f_workaround_enabled = 0; 111 static int f00f_workaround_enabled = 0;
112 112
113 c->f00f_bug = 1; 113 c->f00f_bug = 1;
@@ -195,8 +195,16 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
195 if ((c->x86 == 0xf && c->x86_model >= 0x03) || 195 if ((c->x86 == 0xf && c->x86_model >= 0x03) ||
196 (c->x86 == 0x6 && c->x86_model >= 0x0e)) 196 (c->x86 == 0x6 && c->x86_model >= 0x0e))
197 set_bit(X86_FEATURE_CONSTANT_TSC, c->x86_capability); 197 set_bit(X86_FEATURE_CONSTANT_TSC, c->x86_capability);
198}
199 198
199 if (cpu_has_ds) {
200 unsigned int l1;
201 rdmsr(MSR_IA32_MISC_ENABLE, l1, l2);
202 if (!(l1 & (1<<11)))
203 set_bit(X86_FEATURE_BTS, c->x86_capability);
204 if (!(l1 & (1<<12)))
205 set_bit(X86_FEATURE_PEBS, c->x86_capability);
206 }
207}
200 208
201static unsigned int __cpuinit intel_size_cache(struct cpuinfo_x86 * c, unsigned int size) 209static unsigned int __cpuinit intel_size_cache(struct cpuinfo_x86 * c, unsigned int size)
202{ 210{
diff --git a/arch/i386/kernel/cpu/intel_cacheinfo.c b/arch/i386/kernel/cpu/intel_cacheinfo.c
index 5c43be47587f..80b4c5d421b1 100644
--- a/arch/i386/kernel/cpu/intel_cacheinfo.c
+++ b/arch/i386/kernel/cpu/intel_cacheinfo.c
@@ -480,12 +480,10 @@ static int __cpuinit detect_cache_attributes(unsigned int cpu)
480 if (num_cache_leaves == 0) 480 if (num_cache_leaves == 0)
481 return -ENOENT; 481 return -ENOENT;
482 482
483 cpuid4_info[cpu] = kmalloc( 483 cpuid4_info[cpu] = kzalloc(
484 sizeof(struct _cpuid4_info) * num_cache_leaves, GFP_KERNEL); 484 sizeof(struct _cpuid4_info) * num_cache_leaves, GFP_KERNEL);
485 if (unlikely(cpuid4_info[cpu] == NULL)) 485 if (unlikely(cpuid4_info[cpu] == NULL))
486 return -ENOMEM; 486 return -ENOMEM;
487 memset(cpuid4_info[cpu], 0,
488 sizeof(struct _cpuid4_info) * num_cache_leaves);
489 487
490 oldmask = current->cpus_allowed; 488 oldmask = current->cpus_allowed;
491 retval = set_cpus_allowed(current, cpumask_of_cpu(cpu)); 489 retval = set_cpus_allowed(current, cpumask_of_cpu(cpu));
@@ -658,17 +656,14 @@ static int __cpuinit cpuid4_cache_sysfs_init(unsigned int cpu)
658 return -ENOENT; 656 return -ENOENT;
659 657
660 /* Allocate all required memory */ 658 /* Allocate all required memory */
661 cache_kobject[cpu] = kmalloc(sizeof(struct kobject), GFP_KERNEL); 659 cache_kobject[cpu] = kzalloc(sizeof(struct kobject), GFP_KERNEL);
662 if (unlikely(cache_kobject[cpu] == NULL)) 660 if (unlikely(cache_kobject[cpu] == NULL))
663 goto err_out; 661 goto err_out;
664 memset(cache_kobject[cpu], 0, sizeof(struct kobject));
665 662
666 index_kobject[cpu] = kmalloc( 663 index_kobject[cpu] = kzalloc(
667 sizeof(struct _index_kobject ) * num_cache_leaves, GFP_KERNEL); 664 sizeof(struct _index_kobject ) * num_cache_leaves, GFP_KERNEL);
668 if (unlikely(index_kobject[cpu] == NULL)) 665 if (unlikely(index_kobject[cpu] == NULL))
669 goto err_out; 666 goto err_out;
670 memset(index_kobject[cpu], 0,
671 sizeof(struct _index_kobject) * num_cache_leaves);
672 667
673 return 0; 668 return 0;
674 669
diff --git a/arch/i386/kernel/cpu/mcheck/therm_throt.c b/arch/i386/kernel/cpu/mcheck/therm_throt.c
index bad8b4420709..065005c3f168 100644
--- a/arch/i386/kernel/cpu/mcheck/therm_throt.c
+++ b/arch/i386/kernel/cpu/mcheck/therm_throt.c
@@ -116,7 +116,6 @@ static __cpuinit int thermal_throttle_add_dev(struct sys_device *sys_dev)
116 return sysfs_create_group(&sys_dev->kobj, &thermal_throttle_attr_group); 116 return sysfs_create_group(&sys_dev->kobj, &thermal_throttle_attr_group);
117} 117}
118 118
119#ifdef CONFIG_HOTPLUG_CPU
120static __cpuinit void thermal_throttle_remove_dev(struct sys_device *sys_dev) 119static __cpuinit void thermal_throttle_remove_dev(struct sys_device *sys_dev)
121{ 120{
122 return sysfs_remove_group(&sys_dev->kobj, &thermal_throttle_attr_group); 121 return sysfs_remove_group(&sys_dev->kobj, &thermal_throttle_attr_group);
@@ -153,7 +152,6 @@ static struct notifier_block thermal_throttle_cpu_notifier =
153{ 152{
154 .notifier_call = thermal_throttle_cpu_callback, 153 .notifier_call = thermal_throttle_cpu_callback,
155}; 154};
156#endif /* CONFIG_HOTPLUG_CPU */
157 155
158static __init int thermal_throttle_init_device(void) 156static __init int thermal_throttle_init_device(void)
159{ 157{
diff --git a/arch/i386/kernel/cpu/mtrr/Makefile b/arch/i386/kernel/cpu/mtrr/Makefile
index a25b701ab84e..191fc0533649 100644
--- a/arch/i386/kernel/cpu/mtrr/Makefile
+++ b/arch/i386/kernel/cpu/mtrr/Makefile
@@ -1,5 +1,3 @@
1obj-y := main.o if.o generic.o state.o 1obj-y := main.o if.o generic.o state.o
2obj-y += amd.o 2obj-$(CONFIG_X86_32) += amd.o cyrix.o centaur.o
3obj-y += cyrix.o
4obj-y += centaur.o
5 3
diff --git a/arch/i386/kernel/cpu/mtrr/amd.c b/arch/i386/kernel/cpu/mtrr/amd.c
index 1a1e04b6fd00..0949cdbf848a 100644
--- a/arch/i386/kernel/cpu/mtrr/amd.c
+++ b/arch/i386/kernel/cpu/mtrr/amd.c
@@ -7,7 +7,7 @@
7 7
8static void 8static void
9amd_get_mtrr(unsigned int reg, unsigned long *base, 9amd_get_mtrr(unsigned int reg, unsigned long *base,
10 unsigned int *size, mtrr_type * type) 10 unsigned long *size, mtrr_type * type)
11{ 11{
12 unsigned long low, high; 12 unsigned long low, high;
13 13
diff --git a/arch/i386/kernel/cpu/mtrr/centaur.c b/arch/i386/kernel/cpu/mtrr/centaur.c
index 33f00ac314ef..cb9aa3a7a7ab 100644
--- a/arch/i386/kernel/cpu/mtrr/centaur.c
+++ b/arch/i386/kernel/cpu/mtrr/centaur.c
@@ -17,7 +17,7 @@ static u8 centaur_mcr_type; /* 0 for winchip, 1 for winchip2 */
17 */ 17 */
18 18
19static int 19static int
20centaur_get_free_region(unsigned long base, unsigned long size) 20centaur_get_free_region(unsigned long base, unsigned long size, int replace_reg)
21/* [SUMMARY] Get a free MTRR. 21/* [SUMMARY] Get a free MTRR.
22 <base> The starting (base) address of the region. 22 <base> The starting (base) address of the region.
23 <size> The size (in bytes) of the region. 23 <size> The size (in bytes) of the region.
@@ -26,10 +26,11 @@ centaur_get_free_region(unsigned long base, unsigned long size)
26{ 26{
27 int i, max; 27 int i, max;
28 mtrr_type ltype; 28 mtrr_type ltype;
29 unsigned long lbase; 29 unsigned long lbase, lsize;
30 unsigned int lsize;
31 30
32 max = num_var_ranges; 31 max = num_var_ranges;
32 if (replace_reg >= 0 && replace_reg < max)
33 return replace_reg;
33 for (i = 0; i < max; ++i) { 34 for (i = 0; i < max; ++i) {
34 if (centaur_mcr_reserved & (1 << i)) 35 if (centaur_mcr_reserved & (1 << i))
35 continue; 36 continue;
@@ -49,7 +50,7 @@ mtrr_centaur_report_mcr(int mcr, u32 lo, u32 hi)
49 50
50static void 51static void
51centaur_get_mcr(unsigned int reg, unsigned long *base, 52centaur_get_mcr(unsigned int reg, unsigned long *base,
52 unsigned int *size, mtrr_type * type) 53 unsigned long *size, mtrr_type * type)
53{ 54{
54 *base = centaur_mcr[reg].high >> PAGE_SHIFT; 55 *base = centaur_mcr[reg].high >> PAGE_SHIFT;
55 *size = -(centaur_mcr[reg].low & 0xfffff000) >> PAGE_SHIFT; 56 *size = -(centaur_mcr[reg].low & 0xfffff000) >> PAGE_SHIFT;
diff --git a/arch/i386/kernel/cpu/mtrr/cyrix.c b/arch/i386/kernel/cpu/mtrr/cyrix.c
index 9027a987006b..0737a596db43 100644
--- a/arch/i386/kernel/cpu/mtrr/cyrix.c
+++ b/arch/i386/kernel/cpu/mtrr/cyrix.c
@@ -9,7 +9,7 @@ int arr3_protected;
9 9
10static void 10static void
11cyrix_get_arr(unsigned int reg, unsigned long *base, 11cyrix_get_arr(unsigned int reg, unsigned long *base,
12 unsigned int *size, mtrr_type * type) 12 unsigned long *size, mtrr_type * type)
13{ 13{
14 unsigned long flags; 14 unsigned long flags;
15 unsigned char arr, ccr3, rcr, shift; 15 unsigned char arr, ccr3, rcr, shift;
@@ -77,7 +77,7 @@ cyrix_get_arr(unsigned int reg, unsigned long *base,
77} 77}
78 78
79static int 79static int
80cyrix_get_free_region(unsigned long base, unsigned long size) 80cyrix_get_free_region(unsigned long base, unsigned long size, int replace_reg)
81/* [SUMMARY] Get a free ARR. 81/* [SUMMARY] Get a free ARR.
82 <base> The starting (base) address of the region. 82 <base> The starting (base) address of the region.
83 <size> The size (in bytes) of the region. 83 <size> The size (in bytes) of the region.
@@ -86,9 +86,24 @@ cyrix_get_free_region(unsigned long base, unsigned long size)
86{ 86{
87 int i; 87 int i;
88 mtrr_type ltype; 88 mtrr_type ltype;
89 unsigned long lbase; 89 unsigned long lbase, lsize;
90 unsigned int lsize;
91 90
91 switch (replace_reg) {
92 case 7:
93 if (size < 0x40)
94 break;
95 case 6:
96 case 5:
97 case 4:
98 return replace_reg;
99 case 3:
100 if (arr3_protected)
101 break;
102 case 2:
103 case 1:
104 case 0:
105 return replace_reg;
106 }
92 /* If we are to set up a region >32M then look at ARR7 immediately */ 107 /* If we are to set up a region >32M then look at ARR7 immediately */
93 if (size > 0x2000) { 108 if (size > 0x2000) {
94 cyrix_get_arr(7, &lbase, &lsize, &ltype); 109 cyrix_get_arr(7, &lbase, &lsize, &ltype);
@@ -214,7 +229,7 @@ static void cyrix_set_arr(unsigned int reg, unsigned long base,
214 229
215typedef struct { 230typedef struct {
216 unsigned long base; 231 unsigned long base;
217 unsigned int size; 232 unsigned long size;
218 mtrr_type type; 233 mtrr_type type;
219} arr_state_t; 234} arr_state_t;
220 235
diff --git a/arch/i386/kernel/cpu/mtrr/generic.c b/arch/i386/kernel/cpu/mtrr/generic.c
index 0b61eed8bbd8..f77fc53db654 100644
--- a/arch/i386/kernel/cpu/mtrr/generic.c
+++ b/arch/i386/kernel/cpu/mtrr/generic.c
@@ -3,6 +3,7 @@
3#include <linux/init.h> 3#include <linux/init.h>
4#include <linux/slab.h> 4#include <linux/slab.h>
5#include <linux/mm.h> 5#include <linux/mm.h>
6#include <linux/module.h>
6#include <asm/io.h> 7#include <asm/io.h>
7#include <asm/mtrr.h> 8#include <asm/mtrr.h>
8#include <asm/msr.h> 9#include <asm/msr.h>
@@ -15,12 +16,19 @@ struct mtrr_state {
15 struct mtrr_var_range *var_ranges; 16 struct mtrr_var_range *var_ranges;
16 mtrr_type fixed_ranges[NUM_FIXED_RANGES]; 17 mtrr_type fixed_ranges[NUM_FIXED_RANGES];
17 unsigned char enabled; 18 unsigned char enabled;
19 unsigned char have_fixed;
18 mtrr_type def_type; 20 mtrr_type def_type;
19}; 21};
20 22
21static unsigned long smp_changes_mask; 23static unsigned long smp_changes_mask;
22static struct mtrr_state mtrr_state = {}; 24static struct mtrr_state mtrr_state = {};
23 25
26#undef MODULE_PARAM_PREFIX
27#define MODULE_PARAM_PREFIX "mtrr."
28
29static __initdata int mtrr_show;
30module_param_named(show, mtrr_show, bool, 0);
31
24/* Get the MSR pair relating to a var range */ 32/* Get the MSR pair relating to a var range */
25static void __init 33static void __init
26get_mtrr_var_range(unsigned int index, struct mtrr_var_range *vr) 34get_mtrr_var_range(unsigned int index, struct mtrr_var_range *vr)
@@ -43,6 +51,14 @@ get_fixed_ranges(mtrr_type * frs)
43 rdmsr(MTRRfix4K_C0000_MSR + i, p[6 + i * 2], p[7 + i * 2]); 51 rdmsr(MTRRfix4K_C0000_MSR + i, p[6 + i * 2], p[7 + i * 2]);
44} 52}
45 53
54static void __init print_fixed(unsigned base, unsigned step, const mtrr_type*types)
55{
56 unsigned i;
57
58 for (i = 0; i < 8; ++i, ++types, base += step)
59 printk(KERN_INFO "MTRR %05X-%05X %s\n", base, base + step - 1, mtrr_attrib_to_str(*types));
60}
61
46/* Grab all of the MTRR state for this CPU into *state */ 62/* Grab all of the MTRR state for this CPU into *state */
47void __init get_mtrr_state(void) 63void __init get_mtrr_state(void)
48{ 64{
@@ -58,13 +74,49 @@ void __init get_mtrr_state(void)
58 } 74 }
59 vrs = mtrr_state.var_ranges; 75 vrs = mtrr_state.var_ranges;
60 76
77 rdmsr(MTRRcap_MSR, lo, dummy);
78 mtrr_state.have_fixed = (lo >> 8) & 1;
79
61 for (i = 0; i < num_var_ranges; i++) 80 for (i = 0; i < num_var_ranges; i++)
62 get_mtrr_var_range(i, &vrs[i]); 81 get_mtrr_var_range(i, &vrs[i]);
63 get_fixed_ranges(mtrr_state.fixed_ranges); 82 if (mtrr_state.have_fixed)
83 get_fixed_ranges(mtrr_state.fixed_ranges);
64 84
65 rdmsr(MTRRdefType_MSR, lo, dummy); 85 rdmsr(MTRRdefType_MSR, lo, dummy);
66 mtrr_state.def_type = (lo & 0xff); 86 mtrr_state.def_type = (lo & 0xff);
67 mtrr_state.enabled = (lo & 0xc00) >> 10; 87 mtrr_state.enabled = (lo & 0xc00) >> 10;
88
89 if (mtrr_show) {
90 int high_width;
91
92 printk(KERN_INFO "MTRR default type: %s\n", mtrr_attrib_to_str(mtrr_state.def_type));
93 if (mtrr_state.have_fixed) {
94 printk(KERN_INFO "MTRR fixed ranges %sabled:\n",
95 mtrr_state.enabled & 1 ? "en" : "dis");
96 print_fixed(0x00000, 0x10000, mtrr_state.fixed_ranges + 0);
97 for (i = 0; i < 2; ++i)
98 print_fixed(0x80000 + i * 0x20000, 0x04000, mtrr_state.fixed_ranges + (i + 1) * 8);
99 for (i = 0; i < 8; ++i)
100 print_fixed(0xC0000 + i * 0x08000, 0x01000, mtrr_state.fixed_ranges + (i + 3) * 8);
101 }
102 printk(KERN_INFO "MTRR variable ranges %sabled:\n",
103 mtrr_state.enabled & 2 ? "en" : "dis");
104 high_width = ((size_or_mask ? ffs(size_or_mask) - 1 : 32) - (32 - PAGE_SHIFT) + 3) / 4;
105 for (i = 0; i < num_var_ranges; ++i) {
106 if (mtrr_state.var_ranges[i].mask_lo & (1 << 11))
107 printk(KERN_INFO "MTRR %u base %0*X%05X000 mask %0*X%05X000 %s\n",
108 i,
109 high_width,
110 mtrr_state.var_ranges[i].base_hi,
111 mtrr_state.var_ranges[i].base_lo >> 12,
112 high_width,
113 mtrr_state.var_ranges[i].mask_hi,
114 mtrr_state.var_ranges[i].mask_lo >> 12,
115 mtrr_attrib_to_str(mtrr_state.var_ranges[i].base_lo & 0xff));
116 else
117 printk(KERN_INFO "MTRR %u disabled\n", i);
118 }
119 }
68} 120}
69 121
70/* Some BIOS's are fucked and don't set all MTRRs the same! */ 122/* Some BIOS's are fucked and don't set all MTRRs the same! */
@@ -95,7 +147,7 @@ void mtrr_wrmsr(unsigned msr, unsigned a, unsigned b)
95 smp_processor_id(), msr, a, b); 147 smp_processor_id(), msr, a, b);
96} 148}
97 149
98int generic_get_free_region(unsigned long base, unsigned long size) 150int generic_get_free_region(unsigned long base, unsigned long size, int replace_reg)
99/* [SUMMARY] Get a free MTRR. 151/* [SUMMARY] Get a free MTRR.
100 <base> The starting (base) address of the region. 152 <base> The starting (base) address of the region.
101 <size> The size (in bytes) of the region. 153 <size> The size (in bytes) of the region.
@@ -104,10 +156,11 @@ int generic_get_free_region(unsigned long base, unsigned long size)
104{ 156{
105 int i, max; 157 int i, max;
106 mtrr_type ltype; 158 mtrr_type ltype;
107 unsigned long lbase; 159 unsigned long lbase, lsize;
108 unsigned lsize;
109 160
110 max = num_var_ranges; 161 max = num_var_ranges;
162 if (replace_reg >= 0 && replace_reg < max)
163 return replace_reg;
111 for (i = 0; i < max; ++i) { 164 for (i = 0; i < max; ++i) {
112 mtrr_if->get(i, &lbase, &lsize, &ltype); 165 mtrr_if->get(i, &lbase, &lsize, &ltype);
113 if (lsize == 0) 166 if (lsize == 0)
@@ -117,7 +170,7 @@ int generic_get_free_region(unsigned long base, unsigned long size)
117} 170}
118 171
119static void generic_get_mtrr(unsigned int reg, unsigned long *base, 172static void generic_get_mtrr(unsigned int reg, unsigned long *base,
120 unsigned int *size, mtrr_type * type) 173 unsigned long *size, mtrr_type *type)
121{ 174{
122 unsigned int mask_lo, mask_hi, base_lo, base_hi; 175 unsigned int mask_lo, mask_hi, base_lo, base_hi;
123 176
@@ -202,7 +255,9 @@ static int set_mtrr_var_ranges(unsigned int index, struct mtrr_var_range *vr)
202 return changed; 255 return changed;
203} 256}
204 257
205static unsigned long set_mtrr_state(u32 deftype_lo, u32 deftype_hi) 258static u32 deftype_lo, deftype_hi;
259
260static unsigned long set_mtrr_state(void)
206/* [SUMMARY] Set the MTRR state for this CPU. 261/* [SUMMARY] Set the MTRR state for this CPU.
207 <state> The MTRR state information to read. 262 <state> The MTRR state information to read.
208 <ctxt> Some relevant CPU context. 263 <ctxt> Some relevant CPU context.
@@ -217,14 +272,14 @@ static unsigned long set_mtrr_state(u32 deftype_lo, u32 deftype_hi)
217 if (set_mtrr_var_ranges(i, &mtrr_state.var_ranges[i])) 272 if (set_mtrr_var_ranges(i, &mtrr_state.var_ranges[i]))
218 change_mask |= MTRR_CHANGE_MASK_VARIABLE; 273 change_mask |= MTRR_CHANGE_MASK_VARIABLE;
219 274
220 if (set_fixed_ranges(mtrr_state.fixed_ranges)) 275 if (mtrr_state.have_fixed && set_fixed_ranges(mtrr_state.fixed_ranges))
221 change_mask |= MTRR_CHANGE_MASK_FIXED; 276 change_mask |= MTRR_CHANGE_MASK_FIXED;
222 277
223 /* Set_mtrr_restore restores the old value of MTRRdefType, 278 /* Set_mtrr_restore restores the old value of MTRRdefType,
224 so to set it we fiddle with the saved value */ 279 so to set it we fiddle with the saved value */
225 if ((deftype_lo & 0xff) != mtrr_state.def_type 280 if ((deftype_lo & 0xff) != mtrr_state.def_type
226 || ((deftype_lo & 0xc00) >> 10) != mtrr_state.enabled) { 281 || ((deftype_lo & 0xc00) >> 10) != mtrr_state.enabled) {
227 deftype_lo |= (mtrr_state.def_type | mtrr_state.enabled << 10); 282 deftype_lo = (deftype_lo & ~0xcff) | mtrr_state.def_type | (mtrr_state.enabled << 10);
228 change_mask |= MTRR_CHANGE_MASK_DEFTYPE; 283 change_mask |= MTRR_CHANGE_MASK_DEFTYPE;
229 } 284 }
230 285
@@ -233,7 +288,6 @@ static unsigned long set_mtrr_state(u32 deftype_lo, u32 deftype_hi)
233 288
234 289
235static unsigned long cr4 = 0; 290static unsigned long cr4 = 0;
236static u32 deftype_lo, deftype_hi;
237static DEFINE_SPINLOCK(set_atomicity_lock); 291static DEFINE_SPINLOCK(set_atomicity_lock);
238 292
239/* 293/*
@@ -271,7 +325,7 @@ static void prepare_set(void) __acquires(set_atomicity_lock)
271 rdmsr(MTRRdefType_MSR, deftype_lo, deftype_hi); 325 rdmsr(MTRRdefType_MSR, deftype_lo, deftype_hi);
272 326
273 /* Disable MTRRs, and set the default type to uncached */ 327 /* Disable MTRRs, and set the default type to uncached */
274 mtrr_wrmsr(MTRRdefType_MSR, deftype_lo & 0xf300UL, deftype_hi); 328 mtrr_wrmsr(MTRRdefType_MSR, deftype_lo & ~0xcff, deftype_hi);
275} 329}
276 330
277static void post_set(void) __releases(set_atomicity_lock) 331static void post_set(void) __releases(set_atomicity_lock)
@@ -300,7 +354,7 @@ static void generic_set_all(void)
300 prepare_set(); 354 prepare_set();
301 355
302 /* Actually set the state */ 356 /* Actually set the state */
303 mask = set_mtrr_state(deftype_lo,deftype_hi); 357 mask = set_mtrr_state();
304 358
305 post_set(); 359 post_set();
306 local_irq_restore(flags); 360 local_irq_restore(flags);
@@ -366,7 +420,7 @@ int generic_validate_add_page(unsigned long base, unsigned long size, unsigned i
366 printk(KERN_WARNING "mtrr: base(0x%lx000) is not 4 MiB aligned\n", base); 420 printk(KERN_WARNING "mtrr: base(0x%lx000) is not 4 MiB aligned\n", base);
367 return -EINVAL; 421 return -EINVAL;
368 } 422 }
369 if (!(base + size < 0x70000000 || base > 0x7003FFFF) && 423 if (!(base + size < 0x70000 || base > 0x7003F) &&
370 (type == MTRR_TYPE_WRCOMB 424 (type == MTRR_TYPE_WRCOMB
371 || type == MTRR_TYPE_WRBACK)) { 425 || type == MTRR_TYPE_WRBACK)) {
372 printk(KERN_WARNING "mtrr: writable mtrr between 0x70000000 and 0x7003FFFF may hang the CPU.\n"); 426 printk(KERN_WARNING "mtrr: writable mtrr between 0x70000000 and 0x7003FFFF may hang the CPU.\n");
diff --git a/arch/i386/kernel/cpu/mtrr/if.c b/arch/i386/kernel/cpu/mtrr/if.c
index 5ac051bb9d55..5ae1705eafa6 100644
--- a/arch/i386/kernel/cpu/mtrr/if.c
+++ b/arch/i386/kernel/cpu/mtrr/if.c
@@ -17,7 +17,7 @@ extern unsigned int *usage_table;
17 17
18#define FILE_FCOUNT(f) (((struct seq_file *)((f)->private_data))->private) 18#define FILE_FCOUNT(f) (((struct seq_file *)((f)->private_data))->private)
19 19
20static char *mtrr_strings[MTRR_NUM_TYPES] = 20static const char *const mtrr_strings[MTRR_NUM_TYPES] =
21{ 21{
22 "uncachable", /* 0 */ 22 "uncachable", /* 0 */
23 "write-combining", /* 1 */ 23 "write-combining", /* 1 */
@@ -28,7 +28,7 @@ static char *mtrr_strings[MTRR_NUM_TYPES] =
28 "write-back", /* 6 */ 28 "write-back", /* 6 */
29}; 29};
30 30
31char *mtrr_attrib_to_str(int x) 31const char *mtrr_attrib_to_str(int x)
32{ 32{
33 return (x <= 6) ? mtrr_strings[x] : "?"; 33 return (x <= 6) ? mtrr_strings[x] : "?";
34} 34}
@@ -44,10 +44,9 @@ mtrr_file_add(unsigned long base, unsigned long size,
44 44
45 max = num_var_ranges; 45 max = num_var_ranges;
46 if (fcount == NULL) { 46 if (fcount == NULL) {
47 fcount = kmalloc(max * sizeof *fcount, GFP_KERNEL); 47 fcount = kzalloc(max * sizeof *fcount, GFP_KERNEL);
48 if (!fcount) 48 if (!fcount)
49 return -ENOMEM; 49 return -ENOMEM;
50 memset(fcount, 0, max * sizeof *fcount);
51 FILE_FCOUNT(file) = fcount; 50 FILE_FCOUNT(file) = fcount;
52 } 51 }
53 if (!page) { 52 if (!page) {
@@ -155,6 +154,7 @@ mtrr_ioctl(struct file *file, unsigned int cmd, unsigned long __arg)
155{ 154{
156 int err = 0; 155 int err = 0;
157 mtrr_type type; 156 mtrr_type type;
157 unsigned long size;
158 struct mtrr_sentry sentry; 158 struct mtrr_sentry sentry;
159 struct mtrr_gentry gentry; 159 struct mtrr_gentry gentry;
160 void __user *arg = (void __user *) __arg; 160 void __user *arg = (void __user *) __arg;
@@ -235,15 +235,15 @@ mtrr_ioctl(struct file *file, unsigned int cmd, unsigned long __arg)
235 case MTRRIOC_GET_ENTRY: 235 case MTRRIOC_GET_ENTRY:
236 if (gentry.regnum >= num_var_ranges) 236 if (gentry.regnum >= num_var_ranges)
237 return -EINVAL; 237 return -EINVAL;
238 mtrr_if->get(gentry.regnum, &gentry.base, &gentry.size, &type); 238 mtrr_if->get(gentry.regnum, &gentry.base, &size, &type);
239 239
240 /* Hide entries that go above 4GB */ 240 /* Hide entries that go above 4GB */
241 if (gentry.base + gentry.size > 0x100000 241 if (gentry.base + size - 1 >= (1UL << (8 * sizeof(gentry.size) - PAGE_SHIFT))
242 || gentry.size == 0x100000) 242 || size >= (1UL << (8 * sizeof(gentry.size) - PAGE_SHIFT)))
243 gentry.base = gentry.size = gentry.type = 0; 243 gentry.base = gentry.size = gentry.type = 0;
244 else { 244 else {
245 gentry.base <<= PAGE_SHIFT; 245 gentry.base <<= PAGE_SHIFT;
246 gentry.size <<= PAGE_SHIFT; 246 gentry.size = size << PAGE_SHIFT;
247 gentry.type = type; 247 gentry.type = type;
248 } 248 }
249 249
@@ -273,8 +273,14 @@ mtrr_ioctl(struct file *file, unsigned int cmd, unsigned long __arg)
273 case MTRRIOC_GET_PAGE_ENTRY: 273 case MTRRIOC_GET_PAGE_ENTRY:
274 if (gentry.regnum >= num_var_ranges) 274 if (gentry.regnum >= num_var_ranges)
275 return -EINVAL; 275 return -EINVAL;
276 mtrr_if->get(gentry.regnum, &gentry.base, &gentry.size, &type); 276 mtrr_if->get(gentry.regnum, &gentry.base, &size, &type);
277 gentry.type = type; 277 /* Hide entries that would overflow */
278 if (size != (__typeof__(gentry.size))size)
279 gentry.base = gentry.size = gentry.type = 0;
280 else {
281 gentry.size = size;
282 gentry.type = type;
283 }
278 break; 284 break;
279 } 285 }
280 286
@@ -353,8 +359,7 @@ static int mtrr_seq_show(struct seq_file *seq, void *offset)
353 char factor; 359 char factor;
354 int i, max, len; 360 int i, max, len;
355 mtrr_type type; 361 mtrr_type type;
356 unsigned long base; 362 unsigned long base, size;
357 unsigned int size;
358 363
359 len = 0; 364 len = 0;
360 max = num_var_ranges; 365 max = num_var_ranges;
@@ -373,7 +378,7 @@ static int mtrr_seq_show(struct seq_file *seq, void *offset)
373 } 378 }
374 /* RED-PEN: base can be > 32bit */ 379 /* RED-PEN: base can be > 32bit */
375 len += seq_printf(seq, 380 len += seq_printf(seq,
376 "reg%02i: base=0x%05lx000 (%4liMB), size=%4i%cB: %s, count=%d\n", 381 "reg%02i: base=0x%05lx000 (%4luMB), size=%4lu%cB: %s, count=%d\n",
377 i, base, base >> (20 - PAGE_SHIFT), size, factor, 382 i, base, base >> (20 - PAGE_SHIFT), size, factor,
378 mtrr_attrib_to_str(type), usage_table[i]); 383 mtrr_attrib_to_str(type), usage_table[i]);
379 } 384 }
diff --git a/arch/i386/kernel/cpu/mtrr/main.c b/arch/i386/kernel/cpu/mtrr/main.c
index fff90bda4733..16bb7ea87145 100644
--- a/arch/i386/kernel/cpu/mtrr/main.c
+++ b/arch/i386/kernel/cpu/mtrr/main.c
@@ -59,7 +59,11 @@ struct mtrr_ops * mtrr_if = NULL;
59static void set_mtrr(unsigned int reg, unsigned long base, 59static void set_mtrr(unsigned int reg, unsigned long base,
60 unsigned long size, mtrr_type type); 60 unsigned long size, mtrr_type type);
61 61
62#ifndef CONFIG_X86_64
62extern int arr3_protected; 63extern int arr3_protected;
64#else
65#define arr3_protected 0
66#endif
63 67
64void set_mtrr_ops(struct mtrr_ops * ops) 68void set_mtrr_ops(struct mtrr_ops * ops)
65{ 69{
@@ -168,6 +172,13 @@ static void ipi_handler(void *info)
168 172
169#endif 173#endif
170 174
175static inline int types_compatible(mtrr_type type1, mtrr_type type2) {
176 return type1 == MTRR_TYPE_UNCACHABLE ||
177 type2 == MTRR_TYPE_UNCACHABLE ||
178 (type1 == MTRR_TYPE_WRTHROUGH && type2 == MTRR_TYPE_WRBACK) ||
179 (type1 == MTRR_TYPE_WRBACK && type2 == MTRR_TYPE_WRTHROUGH);
180}
181
171/** 182/**
172 * set_mtrr - update mtrrs on all processors 183 * set_mtrr - update mtrrs on all processors
173 * @reg: mtrr in question 184 * @reg: mtrr in question
@@ -263,8 +274,8 @@ static void set_mtrr(unsigned int reg, unsigned long base,
263 274
264/** 275/**
265 * mtrr_add_page - Add a memory type region 276 * mtrr_add_page - Add a memory type region
266 * @base: Physical base address of region in pages (4 KB) 277 * @base: Physical base address of region in pages (in units of 4 kB!)
267 * @size: Physical size of region in pages (4 KB) 278 * @size: Physical size of region in pages (4 kB)
268 * @type: Type of MTRR desired 279 * @type: Type of MTRR desired
269 * @increment: If this is true do usage counting on the region 280 * @increment: If this is true do usage counting on the region
270 * 281 *
@@ -300,11 +311,9 @@ static void set_mtrr(unsigned int reg, unsigned long base,
300int mtrr_add_page(unsigned long base, unsigned long size, 311int mtrr_add_page(unsigned long base, unsigned long size,
301 unsigned int type, char increment) 312 unsigned int type, char increment)
302{ 313{
303 int i; 314 int i, replace, error;
304 mtrr_type ltype; 315 mtrr_type ltype;
305 unsigned long lbase; 316 unsigned long lbase, lsize;
306 unsigned int lsize;
307 int error;
308 317
309 if (!mtrr_if) 318 if (!mtrr_if)
310 return -ENXIO; 319 return -ENXIO;
@@ -324,12 +333,18 @@ int mtrr_add_page(unsigned long base, unsigned long size,
324 return -ENOSYS; 333 return -ENOSYS;
325 } 334 }
326 335
336 if (!size) {
337 printk(KERN_WARNING "mtrr: zero sized request\n");
338 return -EINVAL;
339 }
340
327 if (base & size_or_mask || size & size_or_mask) { 341 if (base & size_or_mask || size & size_or_mask) {
328 printk(KERN_WARNING "mtrr: base or size exceeds the MTRR width\n"); 342 printk(KERN_WARNING "mtrr: base or size exceeds the MTRR width\n");
329 return -EINVAL; 343 return -EINVAL;
330 } 344 }
331 345
332 error = -EINVAL; 346 error = -EINVAL;
347 replace = -1;
333 348
334 /* No CPU hotplug when we change MTRR entries */ 349 /* No CPU hotplug when we change MTRR entries */
335 lock_cpu_hotplug(); 350 lock_cpu_hotplug();
@@ -337,21 +352,28 @@ int mtrr_add_page(unsigned long base, unsigned long size,
337 mutex_lock(&mtrr_mutex); 352 mutex_lock(&mtrr_mutex);
338 for (i = 0; i < num_var_ranges; ++i) { 353 for (i = 0; i < num_var_ranges; ++i) {
339 mtrr_if->get(i, &lbase, &lsize, &ltype); 354 mtrr_if->get(i, &lbase, &lsize, &ltype);
340 if (base >= lbase + lsize) 355 if (!lsize || base > lbase + lsize - 1 || base + size - 1 < lbase)
341 continue;
342 if ((base < lbase) && (base + size <= lbase))
343 continue; 356 continue;
344 /* At this point we know there is some kind of overlap/enclosure */ 357 /* At this point we know there is some kind of overlap/enclosure */
345 if ((base < lbase) || (base + size > lbase + lsize)) { 358 if (base < lbase || base + size - 1 > lbase + lsize - 1) {
359 if (base <= lbase && base + size - 1 >= lbase + lsize - 1) {
360 /* New region encloses an existing region */
361 if (type == ltype) {
362 replace = replace == -1 ? i : -2;
363 continue;
364 }
365 else if (types_compatible(type, ltype))
366 continue;
367 }
346 printk(KERN_WARNING 368 printk(KERN_WARNING
347 "mtrr: 0x%lx000,0x%lx000 overlaps existing" 369 "mtrr: 0x%lx000,0x%lx000 overlaps existing"
348 " 0x%lx000,0x%x000\n", base, size, lbase, 370 " 0x%lx000,0x%lx000\n", base, size, lbase,
349 lsize); 371 lsize);
350 goto out; 372 goto out;
351 } 373 }
352 /* New region is enclosed by an existing region */ 374 /* New region is enclosed by an existing region */
353 if (ltype != type) { 375 if (ltype != type) {
354 if (type == MTRR_TYPE_UNCACHABLE) 376 if (types_compatible(type, ltype))
355 continue; 377 continue;
356 printk (KERN_WARNING "mtrr: type mismatch for %lx000,%lx000 old: %s new: %s\n", 378 printk (KERN_WARNING "mtrr: type mismatch for %lx000,%lx000 old: %s new: %s\n",
357 base, size, mtrr_attrib_to_str(ltype), 379 base, size, mtrr_attrib_to_str(ltype),
@@ -364,10 +386,18 @@ int mtrr_add_page(unsigned long base, unsigned long size,
364 goto out; 386 goto out;
365 } 387 }
366 /* Search for an empty MTRR */ 388 /* Search for an empty MTRR */
367 i = mtrr_if->get_free_region(base, size); 389 i = mtrr_if->get_free_region(base, size, replace);
368 if (i >= 0) { 390 if (i >= 0) {
369 set_mtrr(i, base, size, type); 391 set_mtrr(i, base, size, type);
370 usage_table[i] = 1; 392 if (likely(replace < 0))
393 usage_table[i] = 1;
394 else {
395 usage_table[i] = usage_table[replace] + !!increment;
396 if (unlikely(replace != i)) {
397 set_mtrr(replace, 0, 0, 0);
398 usage_table[replace] = 0;
399 }
400 }
371 } else 401 } else
372 printk(KERN_INFO "mtrr: no more MTRRs available\n"); 402 printk(KERN_INFO "mtrr: no more MTRRs available\n");
373 error = i; 403 error = i;
@@ -455,8 +485,7 @@ int mtrr_del_page(int reg, unsigned long base, unsigned long size)
455{ 485{
456 int i, max; 486 int i, max;
457 mtrr_type ltype; 487 mtrr_type ltype;
458 unsigned long lbase; 488 unsigned long lbase, lsize;
459 unsigned int lsize;
460 int error = -EINVAL; 489 int error = -EINVAL;
461 490
462 if (!mtrr_if) 491 if (!mtrr_if)
@@ -544,9 +573,11 @@ extern void centaur_init_mtrr(void);
544 573
545static void __init init_ifs(void) 574static void __init init_ifs(void)
546{ 575{
576#ifndef CONFIG_X86_64
547 amd_init_mtrr(); 577 amd_init_mtrr();
548 cyrix_init_mtrr(); 578 cyrix_init_mtrr();
549 centaur_init_mtrr(); 579 centaur_init_mtrr();
580#endif
550} 581}
551 582
552/* The suspend/resume methods are only for CPU without MTRR. CPU using generic 583/* The suspend/resume methods are only for CPU without MTRR. CPU using generic
@@ -555,7 +586,7 @@ static void __init init_ifs(void)
555struct mtrr_value { 586struct mtrr_value {
556 mtrr_type ltype; 587 mtrr_type ltype;
557 unsigned long lbase; 588 unsigned long lbase;
558 unsigned int lsize; 589 unsigned long lsize;
559}; 590};
560 591
561static struct mtrr_value * mtrr_state; 592static struct mtrr_value * mtrr_state;
@@ -565,10 +596,8 @@ static int mtrr_save(struct sys_device * sysdev, pm_message_t state)
565 int i; 596 int i;
566 int size = num_var_ranges * sizeof(struct mtrr_value); 597 int size = num_var_ranges * sizeof(struct mtrr_value);
567 598
568 mtrr_state = kmalloc(size,GFP_ATOMIC); 599 mtrr_state = kzalloc(size,GFP_ATOMIC);
569 if (mtrr_state) 600 if (!mtrr_state)
570 memset(mtrr_state,0,size);
571 else
572 return -ENOMEM; 601 return -ENOMEM;
573 602
574 for (i = 0; i < num_var_ranges; i++) { 603 for (i = 0; i < num_var_ranges; i++) {
diff --git a/arch/i386/kernel/cpu/mtrr/mtrr.h b/arch/i386/kernel/cpu/mtrr/mtrr.h
index 99c9f2682041..d61ea9db6cfe 100644
--- a/arch/i386/kernel/cpu/mtrr/mtrr.h
+++ b/arch/i386/kernel/cpu/mtrr/mtrr.h
@@ -43,15 +43,16 @@ struct mtrr_ops {
43 void (*set_all)(void); 43 void (*set_all)(void);
44 44
45 void (*get)(unsigned int reg, unsigned long *base, 45 void (*get)(unsigned int reg, unsigned long *base,
46 unsigned int *size, mtrr_type * type); 46 unsigned long *size, mtrr_type * type);
47 int (*get_free_region) (unsigned long base, unsigned long size); 47 int (*get_free_region)(unsigned long base, unsigned long size,
48 48 int replace_reg);
49 int (*validate_add_page)(unsigned long base, unsigned long size, 49 int (*validate_add_page)(unsigned long base, unsigned long size,
50 unsigned int type); 50 unsigned int type);
51 int (*have_wrcomb)(void); 51 int (*have_wrcomb)(void);
52}; 52};
53 53
54extern int generic_get_free_region(unsigned long base, unsigned long size); 54extern int generic_get_free_region(unsigned long base, unsigned long size,
55 int replace_reg);
55extern int generic_validate_add_page(unsigned long base, unsigned long size, 56extern int generic_validate_add_page(unsigned long base, unsigned long size,
56 unsigned int type); 57 unsigned int type);
57 58
@@ -62,17 +63,17 @@ extern int positive_have_wrcomb(void);
62/* library functions for processor-specific routines */ 63/* library functions for processor-specific routines */
63struct set_mtrr_context { 64struct set_mtrr_context {
64 unsigned long flags; 65 unsigned long flags;
65 unsigned long deftype_lo;
66 unsigned long deftype_hi;
67 unsigned long cr4val; 66 unsigned long cr4val;
68 unsigned long ccr3; 67 u32 deftype_lo;
68 u32 deftype_hi;
69 u32 ccr3;
69}; 70};
70 71
71struct mtrr_var_range { 72struct mtrr_var_range {
72 unsigned long base_lo; 73 u32 base_lo;
73 unsigned long base_hi; 74 u32 base_hi;
74 unsigned long mask_lo; 75 u32 mask_lo;
75 unsigned long mask_hi; 76 u32 mask_hi;
76}; 77};
77 78
78void set_mtrr_done(struct set_mtrr_context *ctxt); 79void set_mtrr_done(struct set_mtrr_context *ctxt);
@@ -92,6 +93,6 @@ extern struct mtrr_ops * mtrr_if;
92extern unsigned int num_var_ranges; 93extern unsigned int num_var_ranges;
93 94
94void mtrr_state_warn(void); 95void mtrr_state_warn(void);
95char *mtrr_attrib_to_str(int x); 96const char *mtrr_attrib_to_str(int x);
96void mtrr_wrmsr(unsigned, unsigned, unsigned); 97void mtrr_wrmsr(unsigned, unsigned, unsigned);
97 98
diff --git a/arch/i386/kernel/cpu/proc.c b/arch/i386/kernel/cpu/proc.c
index 76aac088a323..6624d8583c42 100644
--- a/arch/i386/kernel/cpu/proc.c
+++ b/arch/i386/kernel/cpu/proc.c
@@ -152,9 +152,10 @@ static int show_cpuinfo(struct seq_file *m, void *v)
152 seq_printf(m, " [%d]", i); 152 seq_printf(m, " [%d]", i);
153 } 153 }
154 154
155 seq_printf(m, "\nbogomips\t: %lu.%02lu\n\n", 155 seq_printf(m, "\nbogomips\t: %lu.%02lu\n",
156 c->loops_per_jiffy/(500000/HZ), 156 c->loops_per_jiffy/(500000/HZ),
157 (c->loops_per_jiffy/(5000/HZ)) % 100); 157 (c->loops_per_jiffy/(5000/HZ)) % 100);
158 seq_printf(m, "clflush size\t: %u\n\n", c->x86_clflush_size);
158 159
159 return 0; 160 return 0;
160} 161}
diff --git a/arch/i386/kernel/cpuid.c b/arch/i386/kernel/cpuid.c
index ab0c327e79dc..db6dd20c3589 100644
--- a/arch/i386/kernel/cpuid.c
+++ b/arch/i386/kernel/cpuid.c
@@ -34,7 +34,6 @@
34#include <linux/major.h> 34#include <linux/major.h>
35#include <linux/fs.h> 35#include <linux/fs.h>
36#include <linux/smp_lock.h> 36#include <linux/smp_lock.h>
37#include <linux/fs.h>
38#include <linux/device.h> 37#include <linux/device.h>
39#include <linux/cpu.h> 38#include <linux/cpu.h>
40#include <linux/notifier.h> 39#include <linux/notifier.h>
@@ -167,7 +166,6 @@ static int cpuid_device_create(int i)
167 return err; 166 return err;
168} 167}
169 168
170#ifdef CONFIG_HOTPLUG_CPU
171static int cpuid_class_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) 169static int cpuid_class_cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu)
172{ 170{
173 unsigned int cpu = (unsigned long)hcpu; 171 unsigned int cpu = (unsigned long)hcpu;
@@ -187,7 +185,6 @@ static struct notifier_block __cpuinitdata cpuid_class_cpu_notifier =
187{ 185{
188 .notifier_call = cpuid_class_cpu_callback, 186 .notifier_call = cpuid_class_cpu_callback,
189}; 187};
190#endif /* !CONFIG_HOTPLUG_CPU */
191 188
192static int __init cpuid_init(void) 189static int __init cpuid_init(void)
193{ 190{
diff --git a/arch/i386/kernel/crash.c b/arch/i386/kernel/crash.c
index 144b43288965..a5e0e990ea95 100644
--- a/arch/i386/kernel/crash.c
+++ b/arch/i386/kernel/crash.c
@@ -31,68 +31,6 @@
31/* This keeps a track of which one is crashing cpu. */ 31/* This keeps a track of which one is crashing cpu. */
32static int crashing_cpu; 32static int crashing_cpu;
33 33
34static u32 *append_elf_note(u32 *buf, char *name, unsigned type, void *data,
35 size_t data_len)
36{
37 struct elf_note note;
38
39 note.n_namesz = strlen(name) + 1;
40 note.n_descsz = data_len;
41 note.n_type = type;
42 memcpy(buf, &note, sizeof(note));
43 buf += (sizeof(note) +3)/4;
44 memcpy(buf, name, note.n_namesz);
45 buf += (note.n_namesz + 3)/4;
46 memcpy(buf, data, note.n_descsz);
47 buf += (note.n_descsz + 3)/4;
48
49 return buf;
50}
51
52static void final_note(u32 *buf)
53{
54 struct elf_note note;
55
56 note.n_namesz = 0;
57 note.n_descsz = 0;
58 note.n_type = 0;
59 memcpy(buf, &note, sizeof(note));
60}
61
62static void crash_save_this_cpu(struct pt_regs *regs, int cpu)
63{
64 struct elf_prstatus prstatus;
65 u32 *buf;
66
67 if ((cpu < 0) || (cpu >= NR_CPUS))
68 return;
69
70 /* Using ELF notes here is opportunistic.
71 * I need a well defined structure format
72 * for the data I pass, and I need tags
73 * on the data to indicate what information I have
74 * squirrelled away. ELF notes happen to provide
75 * all of that, so there is no need to invent something new.
76 */
77 buf = (u32*)per_cpu_ptr(crash_notes, cpu);
78 if (!buf)
79 return;
80 memset(&prstatus, 0, sizeof(prstatus));
81 prstatus.pr_pid = current->pid;
82 elf_core_copy_regs(&prstatus.pr_reg, regs);
83 buf = append_elf_note(buf, "CORE", NT_PRSTATUS, &prstatus,
84 sizeof(prstatus));
85 final_note(buf);
86}
87
88static void crash_save_self(struct pt_regs *regs)
89{
90 int cpu;
91
92 cpu = safe_smp_processor_id();
93 crash_save_this_cpu(regs, cpu);
94}
95
96#if defined(CONFIG_SMP) && defined(CONFIG_X86_LOCAL_APIC) 34#if defined(CONFIG_SMP) && defined(CONFIG_X86_LOCAL_APIC)
97static atomic_t waiting_for_crash_ipi; 35static atomic_t waiting_for_crash_ipi;
98 36
@@ -121,7 +59,7 @@ static int crash_nmi_callback(struct notifier_block *self,
121 crash_fixup_ss_esp(&fixed_regs, regs); 59 crash_fixup_ss_esp(&fixed_regs, regs);
122 regs = &fixed_regs; 60 regs = &fixed_regs;
123 } 61 }
124 crash_save_this_cpu(regs, cpu); 62 crash_save_cpu(regs, cpu);
125 disable_local_APIC(); 63 disable_local_APIC();
126 atomic_dec(&waiting_for_crash_ipi); 64 atomic_dec(&waiting_for_crash_ipi);
127 /* Assume hlt works */ 65 /* Assume hlt works */
@@ -195,5 +133,5 @@ void machine_crash_shutdown(struct pt_regs *regs)
195#if defined(CONFIG_X86_IO_APIC) 133#if defined(CONFIG_X86_IO_APIC)
196 disable_IO_APIC(); 134 disable_IO_APIC();
197#endif 135#endif
198 crash_save_self(regs); 136 crash_save_cpu(regs, safe_smp_processor_id());
199} 137}
diff --git a/arch/i386/kernel/e820.c b/arch/i386/kernel/e820.c
new file mode 100644
index 000000000000..2f7d0a92fd7c
--- /dev/null
+++ b/arch/i386/kernel/e820.c
@@ -0,0 +1,894 @@
1#include <linux/kernel.h>
2#include <linux/types.h>
3#include <linux/init.h>
4#include <linux/bootmem.h>
5#include <linux/ioport.h>
6#include <linux/string.h>
7#include <linux/kexec.h>
8#include <linux/module.h>
9#include <linux/mm.h>
10#include <linux/efi.h>
11#include <linux/pfn.h>
12#include <linux/uaccess.h>
13
14#include <asm/pgtable.h>
15#include <asm/page.h>
16#include <asm/e820.h>
17
18#ifdef CONFIG_EFI
19int efi_enabled = 0;
20EXPORT_SYMBOL(efi_enabled);
21#endif
22
23struct e820map e820;
24struct change_member {
25 struct e820entry *pbios; /* pointer to original bios entry */
26 unsigned long long addr; /* address for this change point */
27};
28static struct change_member change_point_list[2*E820MAX] __initdata;
29static struct change_member *change_point[2*E820MAX] __initdata;
30static struct e820entry *overlap_list[E820MAX] __initdata;
31static struct e820entry new_bios[E820MAX] __initdata;
32/* For PCI or other memory-mapped resources */
33unsigned long pci_mem_start = 0x10000000;
34#ifdef CONFIG_PCI
35EXPORT_SYMBOL(pci_mem_start);
36#endif
37extern int user_defined_memmap;
38struct resource data_resource = {
39 .name = "Kernel data",
40 .start = 0,
41 .end = 0,
42 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
43};
44
45struct resource code_resource = {
46 .name = "Kernel code",
47 .start = 0,
48 .end = 0,
49 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
50};
51
52static struct resource system_rom_resource = {
53 .name = "System ROM",
54 .start = 0xf0000,
55 .end = 0xfffff,
56 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
57};
58
59static struct resource extension_rom_resource = {
60 .name = "Extension ROM",
61 .start = 0xe0000,
62 .end = 0xeffff,
63 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
64};
65
66static struct resource adapter_rom_resources[] = { {
67 .name = "Adapter ROM",
68 .start = 0xc8000,
69 .end = 0,
70 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
71}, {
72 .name = "Adapter ROM",
73 .start = 0,
74 .end = 0,
75 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
76}, {
77 .name = "Adapter ROM",
78 .start = 0,
79 .end = 0,
80 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
81}, {
82 .name = "Adapter ROM",
83 .start = 0,
84 .end = 0,
85 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
86}, {
87 .name = "Adapter ROM",
88 .start = 0,
89 .end = 0,
90 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
91}, {
92 .name = "Adapter ROM",
93 .start = 0,
94 .end = 0,
95 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
96} };
97
98static struct resource video_rom_resource = {
99 .name = "Video ROM",
100 .start = 0xc0000,
101 .end = 0xc7fff,
102 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
103};
104
105static struct resource video_ram_resource = {
106 .name = "Video RAM area",
107 .start = 0xa0000,
108 .end = 0xbffff,
109 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
110};
111
112static struct resource standard_io_resources[] = { {
113 .name = "dma1",
114 .start = 0x0000,
115 .end = 0x001f,
116 .flags = IORESOURCE_BUSY | IORESOURCE_IO
117}, {
118 .name = "pic1",
119 .start = 0x0020,
120 .end = 0x0021,
121 .flags = IORESOURCE_BUSY | IORESOURCE_IO
122}, {
123 .name = "timer0",
124 .start = 0x0040,
125 .end = 0x0043,
126 .flags = IORESOURCE_BUSY | IORESOURCE_IO
127}, {
128 .name = "timer1",
129 .start = 0x0050,
130 .end = 0x0053,
131 .flags = IORESOURCE_BUSY | IORESOURCE_IO
132}, {
133 .name = "keyboard",
134 .start = 0x0060,
135 .end = 0x006f,
136 .flags = IORESOURCE_BUSY | IORESOURCE_IO
137}, {
138 .name = "dma page reg",
139 .start = 0x0080,
140 .end = 0x008f,
141 .flags = IORESOURCE_BUSY | IORESOURCE_IO
142}, {
143 .name = "pic2",
144 .start = 0x00a0,
145 .end = 0x00a1,
146 .flags = IORESOURCE_BUSY | IORESOURCE_IO
147}, {
148 .name = "dma2",
149 .start = 0x00c0,
150 .end = 0x00df,
151 .flags = IORESOURCE_BUSY | IORESOURCE_IO
152}, {
153 .name = "fpu",
154 .start = 0x00f0,
155 .end = 0x00ff,
156 .flags = IORESOURCE_BUSY | IORESOURCE_IO
157} };
158
159static int romsignature(const unsigned char *x)
160{
161 unsigned short sig;
162 int ret = 0;
163 if (probe_kernel_address((const unsigned short *)x, sig) == 0)
164 ret = (sig == 0xaa55);
165 return ret;
166}
167
168static int __init romchecksum(unsigned char *rom, unsigned long length)
169{
170 unsigned char *p, sum = 0;
171
172 for (p = rom; p < rom + length; p++)
173 sum += *p;
174 return sum == 0;
175}
176
177static void __init probe_roms(void)
178{
179 unsigned long start, length, upper;
180 unsigned char *rom;
181 int i;
182
183 /* video rom */
184 upper = adapter_rom_resources[0].start;
185 for (start = video_rom_resource.start; start < upper; start += 2048) {
186 rom = isa_bus_to_virt(start);
187 if (!romsignature(rom))
188 continue;
189
190 video_rom_resource.start = start;
191
192 /* 0 < length <= 0x7f * 512, historically */
193 length = rom[2] * 512;
194
195 /* if checksum okay, trust length byte */
196 if (length && romchecksum(rom, length))
197 video_rom_resource.end = start + length - 1;
198
199 request_resource(&iomem_resource, &video_rom_resource);
200 break;
201 }
202
203 start = (video_rom_resource.end + 1 + 2047) & ~2047UL;
204 if (start < upper)
205 start = upper;
206
207 /* system rom */
208 request_resource(&iomem_resource, &system_rom_resource);
209 upper = system_rom_resource.start;
210
211 /* check for extension rom (ignore length byte!) */
212 rom = isa_bus_to_virt(extension_rom_resource.start);
213 if (romsignature(rom)) {
214 length = extension_rom_resource.end - extension_rom_resource.start + 1;
215 if (romchecksum(rom, length)) {
216 request_resource(&iomem_resource, &extension_rom_resource);
217 upper = extension_rom_resource.start;
218 }
219 }
220
221 /* check for adapter roms on 2k boundaries */
222 for (i = 0; i < ARRAY_SIZE(adapter_rom_resources) && start < upper; start += 2048) {
223 rom = isa_bus_to_virt(start);
224 if (!romsignature(rom))
225 continue;
226
227 /* 0 < length <= 0x7f * 512, historically */
228 length = rom[2] * 512;
229
230 /* but accept any length that fits if checksum okay */
231 if (!length || start + length > upper || !romchecksum(rom, length))
232 continue;
233
234 adapter_rom_resources[i].start = start;
235 adapter_rom_resources[i].end = start + length - 1;
236 request_resource(&iomem_resource, &adapter_rom_resources[i]);
237
238 start = adapter_rom_resources[i++].end & ~2047UL;
239 }
240}
241
242/*
243 * Request address space for all standard RAM and ROM resources
244 * and also for regions reported as reserved by the e820.
245 */
246static void __init
247legacy_init_iomem_resources(struct resource *code_resource, struct resource *data_resource)
248{
249 int i;
250
251 probe_roms();
252 for (i = 0; i < e820.nr_map; i++) {
253 struct resource *res;
254#ifndef CONFIG_RESOURCES_64BIT
255 if (e820.map[i].addr + e820.map[i].size > 0x100000000ULL)
256 continue;
257#endif
258 res = kzalloc(sizeof(struct resource), GFP_ATOMIC);
259 switch (e820.map[i].type) {
260 case E820_RAM: res->name = "System RAM"; break;
261 case E820_ACPI: res->name = "ACPI Tables"; break;
262 case E820_NVS: res->name = "ACPI Non-volatile Storage"; break;
263 default: res->name = "reserved";
264 }
265 res->start = e820.map[i].addr;
266 res->end = res->start + e820.map[i].size - 1;
267 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
268 if (request_resource(&iomem_resource, res)) {
269 kfree(res);
270 continue;
271 }
272 if (e820.map[i].type == E820_RAM) {
273 /*
274 * We don't know which RAM region contains kernel data,
275 * so we try it repeatedly and let the resource manager
276 * test it.
277 */
278 request_resource(res, code_resource);
279 request_resource(res, data_resource);
280#ifdef CONFIG_KEXEC
281 request_resource(res, &crashk_res);
282#endif
283 }
284 }
285}
286
287/*
288 * Request address space for all standard resources
289 *
290 * This is called just before pcibios_init(), which is also a
291 * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
292 */
293static int __init request_standard_resources(void)
294{
295 int i;
296
297 printk("Setting up standard PCI resources\n");
298 if (efi_enabled)
299 efi_initialize_iomem_resources(&code_resource, &data_resource);
300 else
301 legacy_init_iomem_resources(&code_resource, &data_resource);
302
303 /* EFI systems may still have VGA */
304 request_resource(&iomem_resource, &video_ram_resource);
305
306 /* request I/O space for devices used on all i[345]86 PCs */
307 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
308 request_resource(&ioport_resource, &standard_io_resources[i]);
309 return 0;
310}
311
312subsys_initcall(request_standard_resources);
313
314void __init add_memory_region(unsigned long long start,
315 unsigned long long size, int type)
316{
317 int x;
318
319 if (!efi_enabled) {
320 x = e820.nr_map;
321
322 if (x == E820MAX) {
323 printk(KERN_ERR "Ooops! Too many entries in the memory map!\n");
324 return;
325 }
326
327 e820.map[x].addr = start;
328 e820.map[x].size = size;
329 e820.map[x].type = type;
330 e820.nr_map++;
331 }
332} /* add_memory_region */
333
334/*
335 * Sanitize the BIOS e820 map.
336 *
337 * Some e820 responses include overlapping entries. The following
338 * replaces the original e820 map with a new one, removing overlaps.
339 *
340 */
341int __init sanitize_e820_map(struct e820entry * biosmap, char * pnr_map)
342{
343 struct change_member *change_tmp;
344 unsigned long current_type, last_type;
345 unsigned long long last_addr;
346 int chgidx, still_changing;
347 int overlap_entries;
348 int new_bios_entry;
349 int old_nr, new_nr, chg_nr;
350 int i;
351
352 /*
353 Visually we're performing the following (1,2,3,4 = memory types)...
354
355 Sample memory map (w/overlaps):
356 ____22__________________
357 ______________________4_
358 ____1111________________
359 _44_____________________
360 11111111________________
361 ____________________33__
362 ___________44___________
363 __________33333_________
364 ______________22________
365 ___________________2222_
366 _________111111111______
367 _____________________11_
368 _________________4______
369
370 Sanitized equivalent (no overlap):
371 1_______________________
372 _44_____________________
373 ___1____________________
374 ____22__________________
375 ______11________________
376 _________1______________
377 __________3_____________
378 ___________44___________
379 _____________33_________
380 _______________2________
381 ________________1_______
382 _________________4______
383 ___________________2____
384 ____________________33__
385 ______________________4_
386 */
387 printk("sanitize start\n");
388 /* if there's only one memory region, don't bother */
389 if (*pnr_map < 2) {
390 printk("sanitize bail 0\n");
391 return -1;
392 }
393
394 old_nr = *pnr_map;
395
396 /* bail out if we find any unreasonable addresses in bios map */
397 for (i=0; i<old_nr; i++)
398 if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr) {
399 printk("sanitize bail 1\n");
400 return -1;
401 }
402
403 /* create pointers for initial change-point information (for sorting) */
404 for (i=0; i < 2*old_nr; i++)
405 change_point[i] = &change_point_list[i];
406
407 /* record all known change-points (starting and ending addresses),
408 omitting those that are for empty memory regions */
409 chgidx = 0;
410 for (i=0; i < old_nr; i++) {
411 if (biosmap[i].size != 0) {
412 change_point[chgidx]->addr = biosmap[i].addr;
413 change_point[chgidx++]->pbios = &biosmap[i];
414 change_point[chgidx]->addr = biosmap[i].addr + biosmap[i].size;
415 change_point[chgidx++]->pbios = &biosmap[i];
416 }
417 }
418 chg_nr = chgidx; /* true number of change-points */
419
420 /* sort change-point list by memory addresses (low -> high) */
421 still_changing = 1;
422 while (still_changing) {
423 still_changing = 0;
424 for (i=1; i < chg_nr; i++) {
425 /* if <current_addr> > <last_addr>, swap */
426 /* or, if current=<start_addr> & last=<end_addr>, swap */
427 if ((change_point[i]->addr < change_point[i-1]->addr) ||
428 ((change_point[i]->addr == change_point[i-1]->addr) &&
429 (change_point[i]->addr == change_point[i]->pbios->addr) &&
430 (change_point[i-1]->addr != change_point[i-1]->pbios->addr))
431 )
432 {
433 change_tmp = change_point[i];
434 change_point[i] = change_point[i-1];
435 change_point[i-1] = change_tmp;
436 still_changing=1;
437 }
438 }
439 }
440
441 /* create a new bios memory map, removing overlaps */
442 overlap_entries=0; /* number of entries in the overlap table */
443 new_bios_entry=0; /* index for creating new bios map entries */
444 last_type = 0; /* start with undefined memory type */
445 last_addr = 0; /* start with 0 as last starting address */
446 /* loop through change-points, determining affect on the new bios map */
447 for (chgidx=0; chgidx < chg_nr; chgidx++)
448 {
449 /* keep track of all overlapping bios entries */
450 if (change_point[chgidx]->addr == change_point[chgidx]->pbios->addr)
451 {
452 /* add map entry to overlap list (> 1 entry implies an overlap) */
453 overlap_list[overlap_entries++]=change_point[chgidx]->pbios;
454 }
455 else
456 {
457 /* remove entry from list (order independent, so swap with last) */
458 for (i=0; i<overlap_entries; i++)
459 {
460 if (overlap_list[i] == change_point[chgidx]->pbios)
461 overlap_list[i] = overlap_list[overlap_entries-1];
462 }
463 overlap_entries--;
464 }
465 /* if there are overlapping entries, decide which "type" to use */
466 /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */
467 current_type = 0;
468 for (i=0; i<overlap_entries; i++)
469 if (overlap_list[i]->type > current_type)
470 current_type = overlap_list[i]->type;
471 /* continue building up new bios map based on this information */
472 if (current_type != last_type) {
473 if (last_type != 0) {
474 new_bios[new_bios_entry].size =
475 change_point[chgidx]->addr - last_addr;
476 /* move forward only if the new size was non-zero */
477 if (new_bios[new_bios_entry].size != 0)
478 if (++new_bios_entry >= E820MAX)
479 break; /* no more space left for new bios entries */
480 }
481 if (current_type != 0) {
482 new_bios[new_bios_entry].addr = change_point[chgidx]->addr;
483 new_bios[new_bios_entry].type = current_type;
484 last_addr=change_point[chgidx]->addr;
485 }
486 last_type = current_type;
487 }
488 }
489 new_nr = new_bios_entry; /* retain count for new bios entries */
490
491 /* copy new bios mapping into original location */
492 memcpy(biosmap, new_bios, new_nr*sizeof(struct e820entry));
493 *pnr_map = new_nr;
494
495 printk("sanitize end\n");
496 return 0;
497}
498
499/*
500 * Copy the BIOS e820 map into a safe place.
501 *
502 * Sanity-check it while we're at it..
503 *
504 * If we're lucky and live on a modern system, the setup code
505 * will have given us a memory map that we can use to properly
506 * set up memory. If we aren't, we'll fake a memory map.
507 *
508 * We check to see that the memory map contains at least 2 elements
509 * before we'll use it, because the detection code in setup.S may
510 * not be perfect and most every PC known to man has two memory
511 * regions: one from 0 to 640k, and one from 1mb up. (The IBM
512 * thinkpad 560x, for example, does not cooperate with the memory
513 * detection code.)
514 */
515int __init copy_e820_map(struct e820entry * biosmap, int nr_map)
516{
517 /* Only one memory region (or negative)? Ignore it */
518 if (nr_map < 2)
519 return -1;
520
521 do {
522 unsigned long long start = biosmap->addr;
523 unsigned long long size = biosmap->size;
524 unsigned long long end = start + size;
525 unsigned long type = biosmap->type;
526 printk("copy_e820_map() start: %016Lx size: %016Lx end: %016Lx type: %ld\n", start, size, end, type);
527
528 /* Overflow in 64 bits? Ignore the memory map. */
529 if (start > end)
530 return -1;
531
532 /*
533 * Some BIOSes claim RAM in the 640k - 1M region.
534 * Not right. Fix it up.
535 */
536 if (type == E820_RAM) {
537 printk("copy_e820_map() type is E820_RAM\n");
538 if (start < 0x100000ULL && end > 0xA0000ULL) {
539 printk("copy_e820_map() lies in range...\n");
540 if (start < 0xA0000ULL) {
541 printk("copy_e820_map() start < 0xA0000ULL\n");
542 add_memory_region(start, 0xA0000ULL-start, type);
543 }
544 if (end <= 0x100000ULL) {
545 printk("copy_e820_map() end <= 0x100000ULL\n");
546 continue;
547 }
548 start = 0x100000ULL;
549 size = end - start;
550 }
551 }
552 add_memory_region(start, size, type);
553 } while (biosmap++,--nr_map);
554 return 0;
555}
556
557/*
558 * Callback for efi_memory_walk.
559 */
560static int __init
561efi_find_max_pfn(unsigned long start, unsigned long end, void *arg)
562{
563 unsigned long *max_pfn = arg, pfn;
564
565 if (start < end) {
566 pfn = PFN_UP(end -1);
567 if (pfn > *max_pfn)
568 *max_pfn = pfn;
569 }
570 return 0;
571}
572
573static int __init
574efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg)
575{
576 memory_present(0, PFN_UP(start), PFN_DOWN(end));
577 return 0;
578}
579
580/*
581 * Find the highest page frame number we have available
582 */
583void __init find_max_pfn(void)
584{
585 int i;
586
587 max_pfn = 0;
588 if (efi_enabled) {
589 efi_memmap_walk(efi_find_max_pfn, &max_pfn);
590 efi_memmap_walk(efi_memory_present_wrapper, NULL);
591 return;
592 }
593
594 for (i = 0; i < e820.nr_map; i++) {
595 unsigned long start, end;
596 /* RAM? */
597 if (e820.map[i].type != E820_RAM)
598 continue;
599 start = PFN_UP(e820.map[i].addr);
600 end = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
601 if (start >= end)
602 continue;
603 if (end > max_pfn)
604 max_pfn = end;
605 memory_present(0, start, end);
606 }
607}
608
609/*
610 * Free all available memory for boot time allocation. Used
611 * as a callback function by efi_memory_walk()
612 */
613
614static int __init
615free_available_memory(unsigned long start, unsigned long end, void *arg)
616{
617 /* check max_low_pfn */
618 if (start >= (max_low_pfn << PAGE_SHIFT))
619 return 0;
620 if (end >= (max_low_pfn << PAGE_SHIFT))
621 end = max_low_pfn << PAGE_SHIFT;
622 if (start < end)
623 free_bootmem(start, end - start);
624
625 return 0;
626}
627/*
628 * Register fully available low RAM pages with the bootmem allocator.
629 */
630void __init register_bootmem_low_pages(unsigned long max_low_pfn)
631{
632 int i;
633
634 if (efi_enabled) {
635 efi_memmap_walk(free_available_memory, NULL);
636 return;
637 }
638 for (i = 0; i < e820.nr_map; i++) {
639 unsigned long curr_pfn, last_pfn, size;
640 /*
641 * Reserve usable low memory
642 */
643 if (e820.map[i].type != E820_RAM)
644 continue;
645 /*
646 * We are rounding up the start address of usable memory:
647 */
648 curr_pfn = PFN_UP(e820.map[i].addr);
649 if (curr_pfn >= max_low_pfn)
650 continue;
651 /*
652 * ... and at the end of the usable range downwards:
653 */
654 last_pfn = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
655
656 if (last_pfn > max_low_pfn)
657 last_pfn = max_low_pfn;
658
659 /*
660 * .. finally, did all the rounding and playing
661 * around just make the area go away?
662 */
663 if (last_pfn <= curr_pfn)
664 continue;
665
666 size = last_pfn - curr_pfn;
667 free_bootmem(PFN_PHYS(curr_pfn), PFN_PHYS(size));
668 }
669}
670
671void __init register_memory(void)
672{
673 unsigned long gapstart, gapsize, round;
674 unsigned long long last;
675 int i;
676
677 /*
678 * Search for the bigest gap in the low 32 bits of the e820
679 * memory space.
680 */
681 last = 0x100000000ull;
682 gapstart = 0x10000000;
683 gapsize = 0x400000;
684 i = e820.nr_map;
685 while (--i >= 0) {
686 unsigned long long start = e820.map[i].addr;
687 unsigned long long end = start + e820.map[i].size;
688
689 /*
690 * Since "last" is at most 4GB, we know we'll
691 * fit in 32 bits if this condition is true
692 */
693 if (last > end) {
694 unsigned long gap = last - end;
695
696 if (gap > gapsize) {
697 gapsize = gap;
698 gapstart = end;
699 }
700 }
701 if (start < last)
702 last = start;
703 }
704
705 /*
706 * See how much we want to round up: start off with
707 * rounding to the next 1MB area.
708 */
709 round = 0x100000;
710 while ((gapsize >> 4) > round)
711 round += round;
712 /* Fun with two's complement */
713 pci_mem_start = (gapstart + round) & -round;
714
715 printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n",
716 pci_mem_start, gapstart, gapsize);
717}
718
719void __init print_memory_map(char *who)
720{
721 int i;
722
723 for (i = 0; i < e820.nr_map; i++) {
724 printk(" %s: %016Lx - %016Lx ", who,
725 e820.map[i].addr,
726 e820.map[i].addr + e820.map[i].size);
727 switch (e820.map[i].type) {
728 case E820_RAM: printk("(usable)\n");
729 break;
730 case E820_RESERVED:
731 printk("(reserved)\n");
732 break;
733 case E820_ACPI:
734 printk("(ACPI data)\n");
735 break;
736 case E820_NVS:
737 printk("(ACPI NVS)\n");
738 break;
739 default: printk("type %lu\n", e820.map[i].type);
740 break;
741 }
742 }
743}
744
745static __init __always_inline void efi_limit_regions(unsigned long long size)
746{
747 unsigned long long current_addr = 0;
748 efi_memory_desc_t *md, *next_md;
749 void *p, *p1;
750 int i, j;
751
752 j = 0;
753 p1 = memmap.map;
754 for (p = p1, i = 0; p < memmap.map_end; p += memmap.desc_size, i++) {
755 md = p;
756 next_md = p1;
757 current_addr = md->phys_addr +
758 PFN_PHYS(md->num_pages);
759 if (is_available_memory(md)) {
760 if (md->phys_addr >= size) continue;
761 memcpy(next_md, md, memmap.desc_size);
762 if (current_addr >= size) {
763 next_md->num_pages -=
764 PFN_UP(current_addr-size);
765 }
766 p1 += memmap.desc_size;
767 next_md = p1;
768 j++;
769 } else if ((md->attribute & EFI_MEMORY_RUNTIME) ==
770 EFI_MEMORY_RUNTIME) {
771 /* In order to make runtime services
772 * available we have to include runtime
773 * memory regions in memory map */
774 memcpy(next_md, md, memmap.desc_size);
775 p1 += memmap.desc_size;
776 next_md = p1;
777 j++;
778 }
779 }
780 memmap.nr_map = j;
781 memmap.map_end = memmap.map +
782 (memmap.nr_map * memmap.desc_size);
783}
784
785void __init limit_regions(unsigned long long size)
786{
787 unsigned long long current_addr;
788 int i;
789
790 print_memory_map("limit_regions start");
791 if (efi_enabled) {
792 efi_limit_regions(size);
793 return;
794 }
795 for (i = 0; i < e820.nr_map; i++) {
796 current_addr = e820.map[i].addr + e820.map[i].size;
797 if (current_addr < size)
798 continue;
799
800 if (e820.map[i].type != E820_RAM)
801 continue;
802
803 if (e820.map[i].addr >= size) {
804 /*
805 * This region starts past the end of the
806 * requested size, skip it completely.
807 */
808 e820.nr_map = i;
809 } else {
810 e820.nr_map = i + 1;
811 e820.map[i].size -= current_addr - size;
812 }
813 print_memory_map("limit_regions endfor");
814 return;
815 }
816 print_memory_map("limit_regions endfunc");
817}
818
819 /*
820 * This function checks if the entire range <start,end> is mapped with type.
821 *
822 * Note: this function only works correct if the e820 table is sorted and
823 * not-overlapping, which is the case
824 */
825int __init
826e820_all_mapped(unsigned long s, unsigned long e, unsigned type)
827{
828 u64 start = s;
829 u64 end = e;
830 int i;
831 for (i = 0; i < e820.nr_map; i++) {
832 struct e820entry *ei = &e820.map[i];
833 if (type && ei->type != type)
834 continue;
835 /* is the region (part) in overlap with the current region ?*/
836 if (ei->addr >= end || ei->addr + ei->size <= start)
837 continue;
838 /* if the region is at the beginning of <start,end> we move
839 * start to the end of the region since it's ok until there
840 */
841 if (ei->addr <= start)
842 start = ei->addr + ei->size;
843 /* if start is now at or beyond end, we're done, full
844 * coverage */
845 if (start >= end)
846 return 1; /* we're done */
847 }
848 return 0;
849}
850
851static int __init parse_memmap(char *arg)
852{
853 if (!arg)
854 return -EINVAL;
855
856 if (strcmp(arg, "exactmap") == 0) {
857#ifdef CONFIG_CRASH_DUMP
858 /* If we are doing a crash dump, we
859 * still need to know the real mem
860 * size before original memory map is
861 * reset.
862 */
863 find_max_pfn();
864 saved_max_pfn = max_pfn;
865#endif
866 e820.nr_map = 0;
867 user_defined_memmap = 1;
868 } else {
869 /* If the user specifies memory size, we
870 * limit the BIOS-provided memory map to
871 * that size. exactmap can be used to specify
872 * the exact map. mem=number can be used to
873 * trim the existing memory map.
874 */
875 unsigned long long start_at, mem_size;
876
877 mem_size = memparse(arg, &arg);
878 if (*arg == '@') {
879 start_at = memparse(arg+1, &arg);
880 add_memory_region(start_at, mem_size, E820_RAM);
881 } else if (*arg == '#') {
882 start_at = memparse(arg+1, &arg);
883 add_memory_region(start_at, mem_size, E820_ACPI);
884 } else if (*arg == '$') {
885 start_at = memparse(arg+1, &arg);
886 add_memory_region(start_at, mem_size, E820_RESERVED);
887 } else {
888 limit_regions(mem_size);
889 user_defined_memmap = 1;
890 }
891 }
892 return 0;
893}
894early_param("memmap", parse_memmap);
diff --git a/arch/i386/kernel/efi.c b/arch/i386/kernel/efi.c
index 8b40648d0ef0..b92c7f0a358a 100644
--- a/arch/i386/kernel/efi.c
+++ b/arch/i386/kernel/efi.c
@@ -194,17 +194,24 @@ inline int efi_set_rtc_mmss(unsigned long nowtime)
194 return 0; 194 return 0;
195} 195}
196/* 196/*
197 * This should only be used during kernel init and before runtime 197 * This is used during kernel init before runtime
198 * services have been remapped, therefore, we'll need to call in physical 198 * services have been remapped and also during suspend, therefore,
199 * mode. Note, this call isn't used later, so mark it __init. 199 * we'll need to call both in physical and virtual modes.
200 */ 200 */
201inline unsigned long __init efi_get_time(void) 201inline unsigned long efi_get_time(void)
202{ 202{
203 efi_status_t status; 203 efi_status_t status;
204 efi_time_t eft; 204 efi_time_t eft;
205 efi_time_cap_t cap; 205 efi_time_cap_t cap;
206 206
207 status = phys_efi_get_time(&eft, &cap); 207 if (efi.get_time) {
208 /* if we are in virtual mode use remapped function */
209 status = efi.get_time(&eft, &cap);
210 } else {
211 /* we are in physical mode */
212 status = phys_efi_get_time(&eft, &cap);
213 }
214
208 if (status != EFI_SUCCESS) 215 if (status != EFI_SUCCESS)
209 printk("Oops: efitime: can't read time status: 0x%lx\n",status); 216 printk("Oops: efitime: can't read time status: 0x%lx\n",status);
210 217
diff --git a/arch/i386/kernel/entry.S b/arch/i386/kernel/entry.S
index 5a63d6fdb70e..de34b7fed3c1 100644
--- a/arch/i386/kernel/entry.S
+++ b/arch/i386/kernel/entry.S
@@ -30,12 +30,13 @@
30 * 18(%esp) - %eax 30 * 18(%esp) - %eax
31 * 1C(%esp) - %ds 31 * 1C(%esp) - %ds
32 * 20(%esp) - %es 32 * 20(%esp) - %es
33 * 24(%esp) - orig_eax 33 * 24(%esp) - %gs
34 * 28(%esp) - %eip 34 * 28(%esp) - orig_eax
35 * 2C(%esp) - %cs 35 * 2C(%esp) - %eip
36 * 30(%esp) - %eflags 36 * 30(%esp) - %cs
37 * 34(%esp) - %oldesp 37 * 34(%esp) - %eflags
38 * 38(%esp) - %oldss 38 * 38(%esp) - %oldesp
39 * 3C(%esp) - %oldss
39 * 40 *
40 * "current" is in register %ebx during any slow entries. 41 * "current" is in register %ebx during any slow entries.
41 */ 42 */
@@ -48,26 +49,24 @@
48#include <asm/smp.h> 49#include <asm/smp.h>
49#include <asm/page.h> 50#include <asm/page.h>
50#include <asm/desc.h> 51#include <asm/desc.h>
52#include <asm/percpu.h>
51#include <asm/dwarf2.h> 53#include <asm/dwarf2.h>
52#include "irq_vectors.h" 54#include "irq_vectors.h"
53 55
54#define nr_syscalls ((syscall_table_size)/4) 56/*
57 * We use macros for low-level operations which need to be overridden
58 * for paravirtualization. The following will never clobber any registers:
59 * INTERRUPT_RETURN (aka. "iret")
60 * GET_CR0_INTO_EAX (aka. "movl %cr0, %eax")
61 * ENABLE_INTERRUPTS_SYSEXIT (aka "sti; sysexit").
62 *
63 * For DISABLE_INTERRUPTS/ENABLE_INTERRUPTS (aka "cli"/"sti"), you must
64 * specify what registers can be overwritten (CLBR_NONE, CLBR_EAX/EDX/ECX/ANY).
65 * Allowing a register to be clobbered can shrink the paravirt replacement
66 * enough to patch inline, increasing performance.
67 */
55 68
56EBX = 0x00 69#define nr_syscalls ((syscall_table_size)/4)
57ECX = 0x04
58EDX = 0x08
59ESI = 0x0C
60EDI = 0x10
61EBP = 0x14
62EAX = 0x18
63DS = 0x1C
64ES = 0x20
65ORIG_EAX = 0x24
66EIP = 0x28
67CS = 0x2C
68EFLAGS = 0x30
69OLDESP = 0x34
70OLDSS = 0x38
71 70
72CF_MASK = 0x00000001 71CF_MASK = 0x00000001
73TF_MASK = 0x00000100 72TF_MASK = 0x00000100
@@ -76,23 +75,16 @@ DF_MASK = 0x00000400
76NT_MASK = 0x00004000 75NT_MASK = 0x00004000
77VM_MASK = 0x00020000 76VM_MASK = 0x00020000
78 77
79/* These are replaces for paravirtualization */
80#define DISABLE_INTERRUPTS cli
81#define ENABLE_INTERRUPTS sti
82#define ENABLE_INTERRUPTS_SYSEXIT sti; sysexit
83#define INTERRUPT_RETURN iret
84#define GET_CR0_INTO_EAX movl %cr0, %eax
85
86#ifdef CONFIG_PREEMPT 78#ifdef CONFIG_PREEMPT
87#define preempt_stop DISABLE_INTERRUPTS; TRACE_IRQS_OFF 79#define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF
88#else 80#else
89#define preempt_stop 81#define preempt_stop(clobbers)
90#define resume_kernel restore_nocheck 82#define resume_kernel restore_nocheck
91#endif 83#endif
92 84
93.macro TRACE_IRQS_IRET 85.macro TRACE_IRQS_IRET
94#ifdef CONFIG_TRACE_IRQFLAGS 86#ifdef CONFIG_TRACE_IRQFLAGS
95 testl $IF_MASK,EFLAGS(%esp) # interrupts off? 87 testl $IF_MASK,PT_EFLAGS(%esp) # interrupts off?
96 jz 1f 88 jz 1f
97 TRACE_IRQS_ON 89 TRACE_IRQS_ON
981: 901:
@@ -107,6 +99,9 @@ VM_MASK = 0x00020000
107 99
108#define SAVE_ALL \ 100#define SAVE_ALL \
109 cld; \ 101 cld; \
102 pushl %gs; \
103 CFI_ADJUST_CFA_OFFSET 4;\
104 /*CFI_REL_OFFSET gs, 0;*/\
110 pushl %es; \ 105 pushl %es; \
111 CFI_ADJUST_CFA_OFFSET 4;\ 106 CFI_ADJUST_CFA_OFFSET 4;\
112 /*CFI_REL_OFFSET es, 0;*/\ 107 /*CFI_REL_OFFSET es, 0;*/\
@@ -136,7 +131,9 @@ VM_MASK = 0x00020000
136 CFI_REL_OFFSET ebx, 0;\ 131 CFI_REL_OFFSET ebx, 0;\
137 movl $(__USER_DS), %edx; \ 132 movl $(__USER_DS), %edx; \
138 movl %edx, %ds; \ 133 movl %edx, %ds; \
139 movl %edx, %es; 134 movl %edx, %es; \
135 movl $(__KERNEL_PDA), %edx; \
136 movl %edx, %gs
140 137
141#define RESTORE_INT_REGS \ 138#define RESTORE_INT_REGS \
142 popl %ebx; \ 139 popl %ebx; \
@@ -169,17 +166,22 @@ VM_MASK = 0x00020000
1692: popl %es; \ 1662: popl %es; \
170 CFI_ADJUST_CFA_OFFSET -4;\ 167 CFI_ADJUST_CFA_OFFSET -4;\
171 /*CFI_RESTORE es;*/\ 168 /*CFI_RESTORE es;*/\
172.section .fixup,"ax"; \ 1693: popl %gs; \
1733: movl $0,(%esp); \ 170 CFI_ADJUST_CFA_OFFSET -4;\
174 jmp 1b; \ 171 /*CFI_RESTORE gs;*/\
172.pushsection .fixup,"ax"; \
1754: movl $0,(%esp); \ 1734: movl $0,(%esp); \
174 jmp 1b; \
1755: movl $0,(%esp); \
176 jmp 2b; \ 176 jmp 2b; \
177.previous; \ 1776: movl $0,(%esp); \
178 jmp 3b; \
178.section __ex_table,"a";\ 179.section __ex_table,"a";\
179 .align 4; \ 180 .align 4; \
180 .long 1b,3b; \ 181 .long 1b,4b; \
181 .long 2b,4b; \ 182 .long 2b,5b; \
182.previous 183 .long 3b,6b; \
184.popsection
183 185
184#define RING0_INT_FRAME \ 186#define RING0_INT_FRAME \
185 CFI_STARTPROC simple;\ 187 CFI_STARTPROC simple;\
@@ -198,18 +200,18 @@ VM_MASK = 0x00020000
198#define RING0_PTREGS_FRAME \ 200#define RING0_PTREGS_FRAME \
199 CFI_STARTPROC simple;\ 201 CFI_STARTPROC simple;\
200 CFI_SIGNAL_FRAME;\ 202 CFI_SIGNAL_FRAME;\
201 CFI_DEF_CFA esp, OLDESP-EBX;\ 203 CFI_DEF_CFA esp, PT_OLDESP-PT_EBX;\
202 /*CFI_OFFSET cs, CS-OLDESP;*/\ 204 /*CFI_OFFSET cs, PT_CS-PT_OLDESP;*/\
203 CFI_OFFSET eip, EIP-OLDESP;\ 205 CFI_OFFSET eip, PT_EIP-PT_OLDESP;\
204 /*CFI_OFFSET es, ES-OLDESP;*/\ 206 /*CFI_OFFSET es, PT_ES-PT_OLDESP;*/\
205 /*CFI_OFFSET ds, DS-OLDESP;*/\ 207 /*CFI_OFFSET ds, PT_DS-PT_OLDESP;*/\
206 CFI_OFFSET eax, EAX-OLDESP;\ 208 CFI_OFFSET eax, PT_EAX-PT_OLDESP;\
207 CFI_OFFSET ebp, EBP-OLDESP;\ 209 CFI_OFFSET ebp, PT_EBP-PT_OLDESP;\
208 CFI_OFFSET edi, EDI-OLDESP;\ 210 CFI_OFFSET edi, PT_EDI-PT_OLDESP;\
209 CFI_OFFSET esi, ESI-OLDESP;\ 211 CFI_OFFSET esi, PT_ESI-PT_OLDESP;\
210 CFI_OFFSET edx, EDX-OLDESP;\ 212 CFI_OFFSET edx, PT_EDX-PT_OLDESP;\
211 CFI_OFFSET ecx, ECX-OLDESP;\ 213 CFI_OFFSET ecx, PT_ECX-PT_OLDESP;\
212 CFI_OFFSET ebx, EBX-OLDESP 214 CFI_OFFSET ebx, PT_EBX-PT_OLDESP
213 215
214ENTRY(ret_from_fork) 216ENTRY(ret_from_fork)
215 CFI_STARTPROC 217 CFI_STARTPROC
@@ -237,17 +239,18 @@ ENTRY(ret_from_fork)
237 ALIGN 239 ALIGN
238 RING0_PTREGS_FRAME 240 RING0_PTREGS_FRAME
239ret_from_exception: 241ret_from_exception:
240 preempt_stop 242 preempt_stop(CLBR_ANY)
241ret_from_intr: 243ret_from_intr:
242 GET_THREAD_INFO(%ebp) 244 GET_THREAD_INFO(%ebp)
243check_userspace: 245check_userspace:
244 movl EFLAGS(%esp), %eax # mix EFLAGS and CS 246 movl PT_EFLAGS(%esp), %eax # mix EFLAGS and CS
245 movb CS(%esp), %al 247 movb PT_CS(%esp), %al
246 andl $(VM_MASK | SEGMENT_RPL_MASK), %eax 248 andl $(VM_MASK | SEGMENT_RPL_MASK), %eax
247 cmpl $USER_RPL, %eax 249 cmpl $USER_RPL, %eax
248 jb resume_kernel # not returning to v8086 or userspace 250 jb resume_kernel # not returning to v8086 or userspace
251
249ENTRY(resume_userspace) 252ENTRY(resume_userspace)
250 DISABLE_INTERRUPTS # make sure we don't miss an interrupt 253 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
251 # setting need_resched or sigpending 254 # setting need_resched or sigpending
252 # between sampling and the iret 255 # between sampling and the iret
253 movl TI_flags(%ebp), %ecx 256 movl TI_flags(%ebp), %ecx
@@ -258,14 +261,14 @@ ENTRY(resume_userspace)
258 261
259#ifdef CONFIG_PREEMPT 262#ifdef CONFIG_PREEMPT
260ENTRY(resume_kernel) 263ENTRY(resume_kernel)
261 DISABLE_INTERRUPTS 264 DISABLE_INTERRUPTS(CLBR_ANY)
262 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ? 265 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ?
263 jnz restore_nocheck 266 jnz restore_nocheck
264need_resched: 267need_resched:
265 movl TI_flags(%ebp), %ecx # need_resched set ? 268 movl TI_flags(%ebp), %ecx # need_resched set ?
266 testb $_TIF_NEED_RESCHED, %cl 269 testb $_TIF_NEED_RESCHED, %cl
267 jz restore_all 270 jz restore_all
268 testl $IF_MASK,EFLAGS(%esp) # interrupts off (exception path) ? 271 testl $IF_MASK,PT_EFLAGS(%esp) # interrupts off (exception path) ?
269 jz restore_all 272 jz restore_all
270 call preempt_schedule_irq 273 call preempt_schedule_irq
271 jmp need_resched 274 jmp need_resched
@@ -287,7 +290,7 @@ sysenter_past_esp:
287 * No need to follow this irqs on/off section: the syscall 290 * No need to follow this irqs on/off section: the syscall
288 * disabled irqs and here we enable it straight after entry: 291 * disabled irqs and here we enable it straight after entry:
289 */ 292 */
290 ENABLE_INTERRUPTS 293 ENABLE_INTERRUPTS(CLBR_NONE)
291 pushl $(__USER_DS) 294 pushl $(__USER_DS)
292 CFI_ADJUST_CFA_OFFSET 4 295 CFI_ADJUST_CFA_OFFSET 4
293 /*CFI_REL_OFFSET ss, 0*/ 296 /*CFI_REL_OFFSET ss, 0*/
@@ -331,20 +334,27 @@ sysenter_past_esp:
331 cmpl $(nr_syscalls), %eax 334 cmpl $(nr_syscalls), %eax
332 jae syscall_badsys 335 jae syscall_badsys
333 call *sys_call_table(,%eax,4) 336 call *sys_call_table(,%eax,4)
334 movl %eax,EAX(%esp) 337 movl %eax,PT_EAX(%esp)
335 DISABLE_INTERRUPTS 338 DISABLE_INTERRUPTS(CLBR_ECX|CLBR_EDX)
336 TRACE_IRQS_OFF 339 TRACE_IRQS_OFF
337 movl TI_flags(%ebp), %ecx 340 movl TI_flags(%ebp), %ecx
338 testw $_TIF_ALLWORK_MASK, %cx 341 testw $_TIF_ALLWORK_MASK, %cx
339 jne syscall_exit_work 342 jne syscall_exit_work
340/* if something modifies registers it must also disable sysexit */ 343/* if something modifies registers it must also disable sysexit */
341 movl EIP(%esp), %edx 344 movl PT_EIP(%esp), %edx
342 movl OLDESP(%esp), %ecx 345 movl PT_OLDESP(%esp), %ecx
343 xorl %ebp,%ebp 346 xorl %ebp,%ebp
344 TRACE_IRQS_ON 347 TRACE_IRQS_ON
3481: mov PT_GS(%esp), %gs
345 ENABLE_INTERRUPTS_SYSEXIT 349 ENABLE_INTERRUPTS_SYSEXIT
346 CFI_ENDPROC 350 CFI_ENDPROC
347 351.pushsection .fixup,"ax"
3522: movl $0,PT_GS(%esp)
353 jmp 1b
354.section __ex_table,"a"
355 .align 4
356 .long 1b,2b
357.popsection
348 358
349 # system call handler stub 359 # system call handler stub
350ENTRY(system_call) 360ENTRY(system_call)
@@ -353,7 +363,7 @@ ENTRY(system_call)
353 CFI_ADJUST_CFA_OFFSET 4 363 CFI_ADJUST_CFA_OFFSET 4
354 SAVE_ALL 364 SAVE_ALL
355 GET_THREAD_INFO(%ebp) 365 GET_THREAD_INFO(%ebp)
356 testl $TF_MASK,EFLAGS(%esp) 366 testl $TF_MASK,PT_EFLAGS(%esp)
357 jz no_singlestep 367 jz no_singlestep
358 orl $_TIF_SINGLESTEP,TI_flags(%ebp) 368 orl $_TIF_SINGLESTEP,TI_flags(%ebp)
359no_singlestep: 369no_singlestep:
@@ -365,9 +375,9 @@ no_singlestep:
365 jae syscall_badsys 375 jae syscall_badsys
366syscall_call: 376syscall_call:
367 call *sys_call_table(,%eax,4) 377 call *sys_call_table(,%eax,4)
368 movl %eax,EAX(%esp) # store the return value 378 movl %eax,PT_EAX(%esp) # store the return value
369syscall_exit: 379syscall_exit:
370 DISABLE_INTERRUPTS # make sure we don't miss an interrupt 380 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
371 # setting need_resched or sigpending 381 # setting need_resched or sigpending
372 # between sampling and the iret 382 # between sampling and the iret
373 TRACE_IRQS_OFF 383 TRACE_IRQS_OFF
@@ -376,12 +386,12 @@ syscall_exit:
376 jne syscall_exit_work 386 jne syscall_exit_work
377 387
378restore_all: 388restore_all:
379 movl EFLAGS(%esp), %eax # mix EFLAGS, SS and CS 389 movl PT_EFLAGS(%esp), %eax # mix EFLAGS, SS and CS
380 # Warning: OLDSS(%esp) contains the wrong/random values if we 390 # Warning: PT_OLDSS(%esp) contains the wrong/random values if we
381 # are returning to the kernel. 391 # are returning to the kernel.
382 # See comments in process.c:copy_thread() for details. 392 # See comments in process.c:copy_thread() for details.
383 movb OLDSS(%esp), %ah 393 movb PT_OLDSS(%esp), %ah
384 movb CS(%esp), %al 394 movb PT_CS(%esp), %al
385 andl $(VM_MASK | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax 395 andl $(VM_MASK | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax
386 cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax 396 cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax
387 CFI_REMEMBER_STATE 397 CFI_REMEMBER_STATE
@@ -390,13 +400,13 @@ restore_nocheck:
390 TRACE_IRQS_IRET 400 TRACE_IRQS_IRET
391restore_nocheck_notrace: 401restore_nocheck_notrace:
392 RESTORE_REGS 402 RESTORE_REGS
393 addl $4, %esp 403 addl $4, %esp # skip orig_eax/error_code
394 CFI_ADJUST_CFA_OFFSET -4 404 CFI_ADJUST_CFA_OFFSET -4
3951: INTERRUPT_RETURN 4051: INTERRUPT_RETURN
396.section .fixup,"ax" 406.section .fixup,"ax"
397iret_exc: 407iret_exc:
398 TRACE_IRQS_ON 408 TRACE_IRQS_ON
399 ENABLE_INTERRUPTS 409 ENABLE_INTERRUPTS(CLBR_NONE)
400 pushl $0 # no error code 410 pushl $0 # no error code
401 pushl $do_iret_error 411 pushl $do_iret_error
402 jmp error_code 412 jmp error_code
@@ -408,33 +418,42 @@ iret_exc:
408 418
409 CFI_RESTORE_STATE 419 CFI_RESTORE_STATE
410ldt_ss: 420ldt_ss:
411 larl OLDSS(%esp), %eax 421 larl PT_OLDSS(%esp), %eax
412 jnz restore_nocheck 422 jnz restore_nocheck
413 testl $0x00400000, %eax # returning to 32bit stack? 423 testl $0x00400000, %eax # returning to 32bit stack?
414 jnz restore_nocheck # allright, normal return 424 jnz restore_nocheck # allright, normal return
425
426#ifdef CONFIG_PARAVIRT
427 /*
428 * The kernel can't run on a non-flat stack if paravirt mode
429 * is active. Rather than try to fixup the high bits of
430 * ESP, bypass this code entirely. This may break DOSemu
431 * and/or Wine support in a paravirt VM, although the option
432 * is still available to implement the setting of the high
433 * 16-bits in the INTERRUPT_RETURN paravirt-op.
434 */
435 cmpl $0, paravirt_ops+PARAVIRT_enabled
436 jne restore_nocheck
437#endif
438
415 /* If returning to userspace with 16bit stack, 439 /* If returning to userspace with 16bit stack,
416 * try to fix the higher word of ESP, as the CPU 440 * try to fix the higher word of ESP, as the CPU
417 * won't restore it. 441 * won't restore it.
418 * This is an "official" bug of all the x86-compatible 442 * This is an "official" bug of all the x86-compatible
419 * CPUs, which we can try to work around to make 443 * CPUs, which we can try to work around to make
420 * dosemu and wine happy. */ 444 * dosemu and wine happy. */
421 subl $8, %esp # reserve space for switch16 pointer 445 movl PT_OLDESP(%esp), %eax
422 CFI_ADJUST_CFA_OFFSET 8 446 movl %esp, %edx
423 DISABLE_INTERRUPTS 447 call patch_espfix_desc
448 pushl $__ESPFIX_SS
449 CFI_ADJUST_CFA_OFFSET 4
450 pushl %eax
451 CFI_ADJUST_CFA_OFFSET 4
452 DISABLE_INTERRUPTS(CLBR_EAX)
424 TRACE_IRQS_OFF 453 TRACE_IRQS_OFF
425 movl %esp, %eax 454 lss (%esp), %esp
426 /* Set up the 16bit stack frame with switch32 pointer on top, 455 CFI_ADJUST_CFA_OFFSET -8
427 * and a switch16 pointer on top of the current frame. */ 456 jmp restore_nocheck
428 call setup_x86_bogus_stack
429 CFI_ADJUST_CFA_OFFSET -8 # frame has moved
430 TRACE_IRQS_IRET
431 RESTORE_REGS
432 lss 20+4(%esp), %esp # switch to 16bit stack
4331: INTERRUPT_RETURN
434.section __ex_table,"a"
435 .align 4
436 .long 1b,iret_exc
437.previous
438 CFI_ENDPROC 457 CFI_ENDPROC
439 458
440 # perform work that needs to be done immediately before resumption 459 # perform work that needs to be done immediately before resumption
@@ -445,7 +464,7 @@ work_pending:
445 jz work_notifysig 464 jz work_notifysig
446work_resched: 465work_resched:
447 call schedule 466 call schedule
448 DISABLE_INTERRUPTS # make sure we don't miss an interrupt 467 DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
449 # setting need_resched or sigpending 468 # setting need_resched or sigpending
450 # between sampling and the iret 469 # between sampling and the iret
451 TRACE_IRQS_OFF 470 TRACE_IRQS_OFF
@@ -458,7 +477,8 @@ work_resched:
458 477
459work_notifysig: # deal with pending signals and 478work_notifysig: # deal with pending signals and
460 # notify-resume requests 479 # notify-resume requests
461 testl $VM_MASK, EFLAGS(%esp) 480#ifdef CONFIG_VM86
481 testl $VM_MASK, PT_EFLAGS(%esp)
462 movl %esp, %eax 482 movl %esp, %eax
463 jne work_notifysig_v86 # returning to kernel-space or 483 jne work_notifysig_v86 # returning to kernel-space or
464 # vm86-space 484 # vm86-space
@@ -468,29 +488,30 @@ work_notifysig: # deal with pending signals and
468 488
469 ALIGN 489 ALIGN
470work_notifysig_v86: 490work_notifysig_v86:
471#ifdef CONFIG_VM86
472 pushl %ecx # save ti_flags for do_notify_resume 491 pushl %ecx # save ti_flags for do_notify_resume
473 CFI_ADJUST_CFA_OFFSET 4 492 CFI_ADJUST_CFA_OFFSET 4
474 call save_v86_state # %eax contains pt_regs pointer 493 call save_v86_state # %eax contains pt_regs pointer
475 popl %ecx 494 popl %ecx
476 CFI_ADJUST_CFA_OFFSET -4 495 CFI_ADJUST_CFA_OFFSET -4
477 movl %eax, %esp 496 movl %eax, %esp
497#else
498 movl %esp, %eax
499#endif
478 xorl %edx, %edx 500 xorl %edx, %edx
479 call do_notify_resume 501 call do_notify_resume
480 jmp resume_userspace_sig 502 jmp resume_userspace_sig
481#endif
482 503
483 # perform syscall exit tracing 504 # perform syscall exit tracing
484 ALIGN 505 ALIGN
485syscall_trace_entry: 506syscall_trace_entry:
486 movl $-ENOSYS,EAX(%esp) 507 movl $-ENOSYS,PT_EAX(%esp)
487 movl %esp, %eax 508 movl %esp, %eax
488 xorl %edx,%edx 509 xorl %edx,%edx
489 call do_syscall_trace 510 call do_syscall_trace
490 cmpl $0, %eax 511 cmpl $0, %eax
491 jne resume_userspace # ret != 0 -> running under PTRACE_SYSEMU, 512 jne resume_userspace # ret != 0 -> running under PTRACE_SYSEMU,
492 # so must skip actual syscall 513 # so must skip actual syscall
493 movl ORIG_EAX(%esp), %eax 514 movl PT_ORIG_EAX(%esp), %eax
494 cmpl $(nr_syscalls), %eax 515 cmpl $(nr_syscalls), %eax
495 jnae syscall_call 516 jnae syscall_call
496 jmp syscall_exit 517 jmp syscall_exit
@@ -501,7 +522,7 @@ syscall_exit_work:
501 testb $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP), %cl 522 testb $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP), %cl
502 jz work_pending 523 jz work_pending
503 TRACE_IRQS_ON 524 TRACE_IRQS_ON
504 ENABLE_INTERRUPTS # could let do_syscall_trace() call 525 ENABLE_INTERRUPTS(CLBR_ANY) # could let do_syscall_trace() call
505 # schedule() instead 526 # schedule() instead
506 movl %esp, %eax 527 movl %esp, %eax
507 movl $1, %edx 528 movl $1, %edx
@@ -515,39 +536,38 @@ syscall_fault:
515 CFI_ADJUST_CFA_OFFSET 4 536 CFI_ADJUST_CFA_OFFSET 4
516 SAVE_ALL 537 SAVE_ALL
517 GET_THREAD_INFO(%ebp) 538 GET_THREAD_INFO(%ebp)
518 movl $-EFAULT,EAX(%esp) 539 movl $-EFAULT,PT_EAX(%esp)
519 jmp resume_userspace 540 jmp resume_userspace
520 541
521syscall_badsys: 542syscall_badsys:
522 movl $-ENOSYS,EAX(%esp) 543 movl $-ENOSYS,PT_EAX(%esp)
523 jmp resume_userspace 544 jmp resume_userspace
524 CFI_ENDPROC 545 CFI_ENDPROC
525 546
526#define FIXUP_ESPFIX_STACK \ 547#define FIXUP_ESPFIX_STACK \
527 movl %esp, %eax; \ 548 /* since we are on a wrong stack, we cant make it a C code :( */ \
528 /* switch to 32bit stack using the pointer on top of 16bit stack */ \ 549 movl %gs:PDA_cpu, %ebx; \
529 lss %ss:CPU_16BIT_STACK_SIZE-8, %esp; \ 550 PER_CPU(cpu_gdt_descr, %ebx); \
530 /* copy data from 16bit stack to 32bit stack */ \ 551 movl GDS_address(%ebx), %ebx; \
531 call fixup_x86_bogus_stack; \ 552 GET_DESC_BASE(GDT_ENTRY_ESPFIX_SS, %ebx, %eax, %ax, %al, %ah); \
532 /* put ESP to the proper location */ \ 553 addl %esp, %eax; \
533 movl %eax, %esp; 554 pushl $__KERNEL_DS; \
534#define UNWIND_ESPFIX_STACK \ 555 CFI_ADJUST_CFA_OFFSET 4; \
535 pushl %eax; \ 556 pushl %eax; \
536 CFI_ADJUST_CFA_OFFSET 4; \ 557 CFI_ADJUST_CFA_OFFSET 4; \
558 lss (%esp), %esp; \
559 CFI_ADJUST_CFA_OFFSET -8;
560#define UNWIND_ESPFIX_STACK \
537 movl %ss, %eax; \ 561 movl %ss, %eax; \
538 /* see if on 16bit stack */ \ 562 /* see if on espfix stack */ \
539 cmpw $__ESPFIX_SS, %ax; \ 563 cmpw $__ESPFIX_SS, %ax; \
540 je 28f; \ 564 jne 27f; \
54127: popl %eax; \ 565 movl $__KERNEL_DS, %eax; \
542 CFI_ADJUST_CFA_OFFSET -4; \
543.section .fixup,"ax"; \
54428: movl $__KERNEL_DS, %eax; \
545 movl %eax, %ds; \ 566 movl %eax, %ds; \
546 movl %eax, %es; \ 567 movl %eax, %es; \
547 /* switch to 32bit stack */ \ 568 /* switch to normal stack */ \
548 FIXUP_ESPFIX_STACK; \ 569 FIXUP_ESPFIX_STACK; \
549 jmp 27b; \ 57027:;
550.previous
551 571
552/* 572/*
553 * Build the entry stubs and pointer table with 573 * Build the entry stubs and pointer table with
@@ -608,13 +628,16 @@ KPROBE_ENTRY(page_fault)
608 CFI_ADJUST_CFA_OFFSET 4 628 CFI_ADJUST_CFA_OFFSET 4
609 ALIGN 629 ALIGN
610error_code: 630error_code:
631 /* the function address is in %gs's slot on the stack */
632 pushl %es
633 CFI_ADJUST_CFA_OFFSET 4
634 /*CFI_REL_OFFSET es, 0*/
611 pushl %ds 635 pushl %ds
612 CFI_ADJUST_CFA_OFFSET 4 636 CFI_ADJUST_CFA_OFFSET 4
613 /*CFI_REL_OFFSET ds, 0*/ 637 /*CFI_REL_OFFSET ds, 0*/
614 pushl %eax 638 pushl %eax
615 CFI_ADJUST_CFA_OFFSET 4 639 CFI_ADJUST_CFA_OFFSET 4
616 CFI_REL_OFFSET eax, 0 640 CFI_REL_OFFSET eax, 0
617 xorl %eax, %eax
618 pushl %ebp 641 pushl %ebp
619 CFI_ADJUST_CFA_OFFSET 4 642 CFI_ADJUST_CFA_OFFSET 4
620 CFI_REL_OFFSET ebp, 0 643 CFI_REL_OFFSET ebp, 0
@@ -627,7 +650,6 @@ error_code:
627 pushl %edx 650 pushl %edx
628 CFI_ADJUST_CFA_OFFSET 4 651 CFI_ADJUST_CFA_OFFSET 4
629 CFI_REL_OFFSET edx, 0 652 CFI_REL_OFFSET edx, 0
630 decl %eax # eax = -1
631 pushl %ecx 653 pushl %ecx
632 CFI_ADJUST_CFA_OFFSET 4 654 CFI_ADJUST_CFA_OFFSET 4
633 CFI_REL_OFFSET ecx, 0 655 CFI_REL_OFFSET ecx, 0
@@ -635,18 +657,20 @@ error_code:
635 CFI_ADJUST_CFA_OFFSET 4 657 CFI_ADJUST_CFA_OFFSET 4
636 CFI_REL_OFFSET ebx, 0 658 CFI_REL_OFFSET ebx, 0
637 cld 659 cld
638 pushl %es 660 pushl %gs
639 CFI_ADJUST_CFA_OFFSET 4 661 CFI_ADJUST_CFA_OFFSET 4
640 /*CFI_REL_OFFSET es, 0*/ 662 /*CFI_REL_OFFSET gs, 0*/
663 movl $(__KERNEL_PDA), %ecx
664 movl %ecx, %gs
641 UNWIND_ESPFIX_STACK 665 UNWIND_ESPFIX_STACK
642 popl %ecx 666 popl %ecx
643 CFI_ADJUST_CFA_OFFSET -4 667 CFI_ADJUST_CFA_OFFSET -4
644 /*CFI_REGISTER es, ecx*/ 668 /*CFI_REGISTER es, ecx*/
645 movl ES(%esp), %edi # get the function address 669 movl PT_GS(%esp), %edi # get the function address
646 movl ORIG_EAX(%esp), %edx # get the error code 670 movl PT_ORIG_EAX(%esp), %edx # get the error code
647 movl %eax, ORIG_EAX(%esp) 671 movl $-1, PT_ORIG_EAX(%esp) # no syscall to restart
648 movl %ecx, ES(%esp) 672 mov %ecx, PT_GS(%esp)
649 /*CFI_REL_OFFSET es, ES*/ 673 /*CFI_REL_OFFSET gs, ES*/
650 movl $(__USER_DS), %ecx 674 movl $(__USER_DS), %ecx
651 movl %ecx, %ds 675 movl %ecx, %ds
652 movl %ecx, %es 676 movl %ecx, %es
@@ -682,7 +706,7 @@ ENTRY(device_not_available)
682 GET_CR0_INTO_EAX 706 GET_CR0_INTO_EAX
683 testl $0x4, %eax # EM (math emulation bit) 707 testl $0x4, %eax # EM (math emulation bit)
684 jne device_not_available_emulate 708 jne device_not_available_emulate
685 preempt_stop 709 preempt_stop(CLBR_ANY)
686 call math_state_restore 710 call math_state_restore
687 jmp ret_from_exception 711 jmp ret_from_exception
688device_not_available_emulate: 712device_not_available_emulate:
@@ -754,7 +778,7 @@ KPROBE_ENTRY(nmi)
754 cmpw $__ESPFIX_SS, %ax 778 cmpw $__ESPFIX_SS, %ax
755 popl %eax 779 popl %eax
756 CFI_ADJUST_CFA_OFFSET -4 780 CFI_ADJUST_CFA_OFFSET -4
757 je nmi_16bit_stack 781 je nmi_espfix_stack
758 cmpl $sysenter_entry,(%esp) 782 cmpl $sysenter_entry,(%esp)
759 je nmi_stack_fixup 783 je nmi_stack_fixup
760 pushl %eax 784 pushl %eax
@@ -797,7 +821,7 @@ nmi_debug_stack_check:
797 FIX_STACK(24,nmi_stack_correct, 1) 821 FIX_STACK(24,nmi_stack_correct, 1)
798 jmp nmi_stack_correct 822 jmp nmi_stack_correct
799 823
800nmi_16bit_stack: 824nmi_espfix_stack:
801 /* We have a RING0_INT_FRAME here. 825 /* We have a RING0_INT_FRAME here.
802 * 826 *
803 * create the pointer to lss back 827 * create the pointer to lss back
@@ -806,7 +830,6 @@ nmi_16bit_stack:
806 CFI_ADJUST_CFA_OFFSET 4 830 CFI_ADJUST_CFA_OFFSET 4
807 pushl %esp 831 pushl %esp
808 CFI_ADJUST_CFA_OFFSET 4 832 CFI_ADJUST_CFA_OFFSET 4
809 movzwl %sp, %esp
810 addw $4, (%esp) 833 addw $4, (%esp)
811 /* copy the iret frame of 12 bytes */ 834 /* copy the iret frame of 12 bytes */
812 .rept 3 835 .rept 3
@@ -817,11 +840,11 @@ nmi_16bit_stack:
817 CFI_ADJUST_CFA_OFFSET 4 840 CFI_ADJUST_CFA_OFFSET 4
818 SAVE_ALL 841 SAVE_ALL
819 FIXUP_ESPFIX_STACK # %eax == %esp 842 FIXUP_ESPFIX_STACK # %eax == %esp
820 CFI_ADJUST_CFA_OFFSET -20 # the frame has now moved
821 xorl %edx,%edx # zero error code 843 xorl %edx,%edx # zero error code
822 call do_nmi 844 call do_nmi
823 RESTORE_REGS 845 RESTORE_REGS
824 lss 12+4(%esp), %esp # back to 16bit stack 846 lss 12+4(%esp), %esp # back to espfix stack
847 CFI_ADJUST_CFA_OFFSET -24
8251: INTERRUPT_RETURN 8481: INTERRUPT_RETURN
826 CFI_ENDPROC 849 CFI_ENDPROC
827.section __ex_table,"a" 850.section __ex_table,"a"
@@ -830,6 +853,19 @@ nmi_16bit_stack:
830.previous 853.previous
831KPROBE_END(nmi) 854KPROBE_END(nmi)
832 855
856#ifdef CONFIG_PARAVIRT
857ENTRY(native_iret)
8581: iret
859.section __ex_table,"a"
860 .align 4
861 .long 1b,iret_exc
862.previous
863
864ENTRY(native_irq_enable_sysexit)
865 sti
866 sysexit
867#endif
868
833KPROBE_ENTRY(int3) 869KPROBE_ENTRY(int3)
834 RING0_INT_FRAME 870 RING0_INT_FRAME
835 pushl $-1 # mark this as an int 871 pushl $-1 # mark this as an int
@@ -949,26 +985,27 @@ ENTRY(arch_unwind_init_running)
949 movl 4(%esp), %edx 985 movl 4(%esp), %edx
950 movl (%esp), %ecx 986 movl (%esp), %ecx
951 leal 4(%esp), %eax 987 leal 4(%esp), %eax
952 movl %ebx, EBX(%edx) 988 movl %ebx, PT_EBX(%edx)
953 xorl %ebx, %ebx 989 xorl %ebx, %ebx
954 movl %ebx, ECX(%edx) 990 movl %ebx, PT_ECX(%edx)
955 movl %ebx, EDX(%edx) 991 movl %ebx, PT_EDX(%edx)
956 movl %esi, ESI(%edx) 992 movl %esi, PT_ESI(%edx)
957 movl %edi, EDI(%edx) 993 movl %edi, PT_EDI(%edx)
958 movl %ebp, EBP(%edx) 994 movl %ebp, PT_EBP(%edx)
959 movl %ebx, EAX(%edx) 995 movl %ebx, PT_EAX(%edx)
960 movl $__USER_DS, DS(%edx) 996 movl $__USER_DS, PT_DS(%edx)
961 movl $__USER_DS, ES(%edx) 997 movl $__USER_DS, PT_ES(%edx)
962 movl %ebx, ORIG_EAX(%edx) 998 movl $0, PT_GS(%edx)
963 movl %ecx, EIP(%edx) 999 movl %ebx, PT_ORIG_EAX(%edx)
1000 movl %ecx, PT_EIP(%edx)
964 movl 12(%esp), %ecx 1001 movl 12(%esp), %ecx
965 movl $__KERNEL_CS, CS(%edx) 1002 movl $__KERNEL_CS, PT_CS(%edx)
966 movl %ebx, EFLAGS(%edx) 1003 movl %ebx, PT_EFLAGS(%edx)
967 movl %eax, OLDESP(%edx) 1004 movl %eax, PT_OLDESP(%edx)
968 movl 8(%esp), %eax 1005 movl 8(%esp), %eax
969 movl %ecx, 8(%esp) 1006 movl %ecx, 8(%esp)
970 movl EBX(%edx), %ebx 1007 movl PT_EBX(%edx), %ebx
971 movl $__KERNEL_DS, OLDSS(%edx) 1008 movl $__KERNEL_DS, PT_OLDSS(%edx)
972 jmpl *%eax 1009 jmpl *%eax
973 CFI_ENDPROC 1010 CFI_ENDPROC
974ENDPROC(arch_unwind_init_running) 1011ENDPROC(arch_unwind_init_running)
diff --git a/arch/i386/kernel/head.S b/arch/i386/kernel/head.S
index ca31f18d277c..edef5084ce17 100644
--- a/arch/i386/kernel/head.S
+++ b/arch/i386/kernel/head.S
@@ -55,6 +55,12 @@
55 */ 55 */
56ENTRY(startup_32) 56ENTRY(startup_32)
57 57
58#ifdef CONFIG_PARAVIRT
59 movl %cs, %eax
60 testl $0x3, %eax
61 jnz startup_paravirt
62#endif
63
58/* 64/*
59 * Set segments to known values. 65 * Set segments to known values.
60 */ 66 */
@@ -302,6 +308,7 @@ is386: movl $2,%ecx # set MP
302 movl %eax,%cr0 308 movl %eax,%cr0
303 309
304 call check_x87 310 call check_x87
311 call setup_pda
305 lgdt cpu_gdt_descr 312 lgdt cpu_gdt_descr
306 lidt idt_descr 313 lidt idt_descr
307 ljmp $(__KERNEL_CS),$1f 314 ljmp $(__KERNEL_CS),$1f
@@ -312,10 +319,13 @@ is386: movl $2,%ecx # set MP
312 movl %eax,%ds 319 movl %eax,%ds
313 movl %eax,%es 320 movl %eax,%es
314 321
315 xorl %eax,%eax # Clear FS/GS and LDT 322 xorl %eax,%eax # Clear FS and LDT
316 movl %eax,%fs 323 movl %eax,%fs
317 movl %eax,%gs
318 lldt %ax 324 lldt %ax
325
326 movl $(__KERNEL_PDA),%eax
327 mov %eax,%gs
328
319 cld # gcc2 wants the direction flag cleared at all times 329 cld # gcc2 wants the direction flag cleared at all times
320 pushl $0 # fake return address for unwinder 330 pushl $0 # fake return address for unwinder
321#ifdef CONFIG_SMP 331#ifdef CONFIG_SMP
@@ -346,6 +356,23 @@ check_x87:
346 ret 356 ret
347 357
348/* 358/*
359 * Point the GDT at this CPU's PDA. On boot this will be
360 * cpu_gdt_table and boot_pda; for secondary CPUs, these will be
361 * that CPU's GDT and PDA.
362 */
363setup_pda:
364 /* get the PDA pointer */
365 movl start_pda, %eax
366
367 /* slot the PDA address into the GDT */
368 mov cpu_gdt_descr+2, %ecx
369 mov %ax, (__KERNEL_PDA+0+2)(%ecx) /* base & 0x0000ffff */
370 shr $16, %eax
371 mov %al, (__KERNEL_PDA+4+0)(%ecx) /* base & 0x00ff0000 */
372 mov %ah, (__KERNEL_PDA+4+3)(%ecx) /* base & 0xff000000 */
373 ret
374
375/*
349 * setup_idt 376 * setup_idt
350 * 377 *
351 * sets up a idt with 256 entries pointing to 378 * sets up a idt with 256 entries pointing to
@@ -465,6 +492,33 @@ ignore_int:
465#endif 492#endif
466 iret 493 iret
467 494
495#ifdef CONFIG_PARAVIRT
496startup_paravirt:
497 cld
498 movl $(init_thread_union+THREAD_SIZE),%esp
499
500 /* We take pains to preserve all the regs. */
501 pushl %edx
502 pushl %ecx
503 pushl %eax
504
505 /* paravirt.o is last in link, and that probe fn never returns */
506 pushl $__start_paravirtprobe
5071:
508 movl 0(%esp), %eax
509 pushl (%eax)
510 movl 8(%esp), %eax
511 call *(%esp)
512 popl %eax
513
514 movl 4(%esp), %eax
515 movl 8(%esp), %ecx
516 movl 12(%esp), %edx
517
518 addl $4, (%esp)
519 jmp 1b
520#endif
521
468/* 522/*
469 * Real beginning of normal "text" segment 523 * Real beginning of normal "text" segment
470 */ 524 */
@@ -484,6 +538,8 @@ ENTRY(empty_zero_page)
484 * This starts the data section. 538 * This starts the data section.
485 */ 539 */
486.data 540.data
541ENTRY(start_pda)
542 .long boot_pda
487 543
488ENTRY(stack_start) 544ENTRY(stack_start)
489 .long init_thread_union+THREAD_SIZE 545 .long init_thread_union+THREAD_SIZE
@@ -525,7 +581,7 @@ idt_descr:
525 581
526# boot GDT descriptor (later on used by CPU#0): 582# boot GDT descriptor (later on used by CPU#0):
527 .word 0 # 32 bit align gdt_desc.address 583 .word 0 # 32 bit align gdt_desc.address
528cpu_gdt_descr: 584ENTRY(cpu_gdt_descr)
529 .word GDT_ENTRIES*8-1 585 .word GDT_ENTRIES*8-1
530 .long cpu_gdt_table 586 .long cpu_gdt_table
531 587
@@ -584,8 +640,8 @@ ENTRY(cpu_gdt_table)
584 .quad 0x00009a000000ffff /* 0xc0 APM CS 16 code (16 bit) */ 640 .quad 0x00009a000000ffff /* 0xc0 APM CS 16 code (16 bit) */
585 .quad 0x004092000000ffff /* 0xc8 APM DS data */ 641 .quad 0x004092000000ffff /* 0xc8 APM DS data */
586 642
587 .quad 0x0000920000000000 /* 0xd0 - ESPFIX 16-bit SS */ 643 .quad 0x00c0920000000000 /* 0xd0 - ESPFIX SS */
588 .quad 0x0000000000000000 /* 0xd8 - unused */ 644 .quad 0x00cf92000000ffff /* 0xd8 - PDA */
589 .quad 0x0000000000000000 /* 0xe0 - unused */ 645 .quad 0x0000000000000000 /* 0xe0 - unused */
590 .quad 0x0000000000000000 /* 0xe8 - unused */ 646 .quad 0x0000000000000000 /* 0xe8 - unused */
591 .quad 0x0000000000000000 /* 0xf0 - unused */ 647 .quad 0x0000000000000000 /* 0xf0 - unused */
diff --git a/arch/i386/kernel/hpet.c b/arch/i386/kernel/hpet.c
index 17647a530b2f..45a8685bb60b 100644
--- a/arch/i386/kernel/hpet.c
+++ b/arch/i386/kernel/hpet.c
@@ -34,6 +34,7 @@ static int __init init_hpet_clocksource(void)
34 unsigned long hpet_period; 34 unsigned long hpet_period;
35 void __iomem* hpet_base; 35 void __iomem* hpet_base;
36 u64 tmp; 36 u64 tmp;
37 int err;
37 38
38 if (!is_hpet_enabled()) 39 if (!is_hpet_enabled())
39 return -ENODEV; 40 return -ENODEV;
@@ -61,7 +62,11 @@ static int __init init_hpet_clocksource(void)
61 do_div(tmp, FSEC_PER_NSEC); 62 do_div(tmp, FSEC_PER_NSEC);
62 clocksource_hpet.mult = (u32)tmp; 63 clocksource_hpet.mult = (u32)tmp;
63 64
64 return clocksource_register(&clocksource_hpet); 65 err = clocksource_register(&clocksource_hpet);
66 if (err)
67 iounmap(hpet_base);
68
69 return err;
65} 70}
66 71
67module_init(init_hpet_clocksource); 72module_init(init_hpet_clocksource);
diff --git a/arch/i386/kernel/i8259.c b/arch/i386/kernel/i8259.c
index 62996cd17084..c8d45821c788 100644
--- a/arch/i386/kernel/i8259.c
+++ b/arch/i386/kernel/i8259.c
@@ -381,7 +381,10 @@ void __init init_ISA_irqs (void)
381 } 381 }
382} 382}
383 383
384void __init init_IRQ(void) 384/* Overridden in paravirt.c */
385void init_IRQ(void) __attribute__((weak, alias("native_init_IRQ")));
386
387void __init native_init_IRQ(void)
385{ 388{
386 int i; 389 int i;
387 390
diff --git a/arch/i386/kernel/io_apic.c b/arch/i386/kernel/io_apic.c
index 3b7a63e0ed1a..e21dcde0790e 100644
--- a/arch/i386/kernel/io_apic.c
+++ b/arch/i386/kernel/io_apic.c
@@ -34,6 +34,7 @@
34#include <linux/pci.h> 34#include <linux/pci.h>
35#include <linux/msi.h> 35#include <linux/msi.h>
36#include <linux/htirq.h> 36#include <linux/htirq.h>
37#include <linux/freezer.h>
37 38
38#include <asm/io.h> 39#include <asm/io.h>
39#include <asm/smp.h> 40#include <asm/smp.h>
@@ -153,14 +154,20 @@ static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin)
153 * the interrupt, and we need to make sure the entry is fully populated 154 * the interrupt, and we need to make sure the entry is fully populated
154 * before that happens. 155 * before that happens.
155 */ 156 */
156static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e) 157static void
158__ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
157{ 159{
158 unsigned long flags;
159 union entry_union eu; 160 union entry_union eu;
160 eu.entry = e; 161 eu.entry = e;
161 spin_lock_irqsave(&ioapic_lock, flags);
162 io_apic_write(apic, 0x11 + 2*pin, eu.w2); 162 io_apic_write(apic, 0x11 + 2*pin, eu.w2);
163 io_apic_write(apic, 0x10 + 2*pin, eu.w1); 163 io_apic_write(apic, 0x10 + 2*pin, eu.w1);
164}
165
166static void ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
167{
168 unsigned long flags;
169 spin_lock_irqsave(&ioapic_lock, flags);
170 __ioapic_write_entry(apic, pin, e);
164 spin_unlock_irqrestore(&ioapic_lock, flags); 171 spin_unlock_irqrestore(&ioapic_lock, flags);
165} 172}
166 173
@@ -836,8 +843,7 @@ static int __init find_isa_irq_pin(int irq, int type)
836 843
837 if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA || 844 if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA ||
838 mp_bus_id_to_type[lbus] == MP_BUS_EISA || 845 mp_bus_id_to_type[lbus] == MP_BUS_EISA ||
839 mp_bus_id_to_type[lbus] == MP_BUS_MCA || 846 mp_bus_id_to_type[lbus] == MP_BUS_MCA
840 mp_bus_id_to_type[lbus] == MP_BUS_NEC98
841 ) && 847 ) &&
842 (mp_irqs[i].mpc_irqtype == type) && 848 (mp_irqs[i].mpc_irqtype == type) &&
843 (mp_irqs[i].mpc_srcbusirq == irq)) 849 (mp_irqs[i].mpc_srcbusirq == irq))
@@ -856,8 +862,7 @@ static int __init find_isa_irq_apic(int irq, int type)
856 862
857 if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA || 863 if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA ||
858 mp_bus_id_to_type[lbus] == MP_BUS_EISA || 864 mp_bus_id_to_type[lbus] == MP_BUS_EISA ||
859 mp_bus_id_to_type[lbus] == MP_BUS_MCA || 865 mp_bus_id_to_type[lbus] == MP_BUS_MCA
860 mp_bus_id_to_type[lbus] == MP_BUS_NEC98
861 ) && 866 ) &&
862 (mp_irqs[i].mpc_irqtype == type) && 867 (mp_irqs[i].mpc_irqtype == type) &&
863 (mp_irqs[i].mpc_srcbusirq == irq)) 868 (mp_irqs[i].mpc_srcbusirq == irq))
@@ -987,12 +992,6 @@ static int EISA_ELCR(unsigned int irq)
987#define default_MCA_trigger(idx) (1) 992#define default_MCA_trigger(idx) (1)
988#define default_MCA_polarity(idx) (0) 993#define default_MCA_polarity(idx) (0)
989 994
990/* NEC98 interrupts are always polarity zero edge triggered,
991 * when listed as conforming in the MP table. */
992
993#define default_NEC98_trigger(idx) (0)
994#define default_NEC98_polarity(idx) (0)
995
996static int __init MPBIOS_polarity(int idx) 995static int __init MPBIOS_polarity(int idx)
997{ 996{
998 int bus = mp_irqs[idx].mpc_srcbus; 997 int bus = mp_irqs[idx].mpc_srcbus;
@@ -1027,11 +1026,6 @@ static int __init MPBIOS_polarity(int idx)
1027 polarity = default_MCA_polarity(idx); 1026 polarity = default_MCA_polarity(idx);
1028 break; 1027 break;
1029 } 1028 }
1030 case MP_BUS_NEC98: /* NEC 98 pin */
1031 {
1032 polarity = default_NEC98_polarity(idx);
1033 break;
1034 }
1035 default: 1029 default:
1036 { 1030 {
1037 printk(KERN_WARNING "broken BIOS!!\n"); 1031 printk(KERN_WARNING "broken BIOS!!\n");
@@ -1101,11 +1095,6 @@ static int MPBIOS_trigger(int idx)
1101 trigger = default_MCA_trigger(idx); 1095 trigger = default_MCA_trigger(idx);
1102 break; 1096 break;
1103 } 1097 }
1104 case MP_BUS_NEC98: /* NEC 98 pin */
1105 {
1106 trigger = default_NEC98_trigger(idx);
1107 break;
1108 }
1109 default: 1098 default:
1110 { 1099 {
1111 printk(KERN_WARNING "broken BIOS!!\n"); 1100 printk(KERN_WARNING "broken BIOS!!\n");
@@ -1167,7 +1156,6 @@ static int pin_2_irq(int idx, int apic, int pin)
1167 case MP_BUS_ISA: /* ISA pin */ 1156 case MP_BUS_ISA: /* ISA pin */
1168 case MP_BUS_EISA: 1157 case MP_BUS_EISA:
1169 case MP_BUS_MCA: 1158 case MP_BUS_MCA:
1170 case MP_BUS_NEC98:
1171 { 1159 {
1172 irq = mp_irqs[idx].mpc_srcbusirq; 1160 irq = mp_irqs[idx].mpc_srcbusirq;
1173 break; 1161 break;
@@ -1235,7 +1223,7 @@ static inline int IO_APIC_irq_trigger(int irq)
1235} 1223}
1236 1224
1237/* irq_vectors is indexed by the sum of all RTEs in all I/O APICs. */ 1225/* irq_vectors is indexed by the sum of all RTEs in all I/O APICs. */
1238u8 irq_vector[NR_IRQ_VECTORS] __read_mostly = { FIRST_DEVICE_VECTOR , 0 }; 1226static u8 irq_vector[NR_IRQ_VECTORS] __read_mostly = { FIRST_DEVICE_VECTOR , 0 };
1239 1227
1240static int __assign_irq_vector(int irq) 1228static int __assign_irq_vector(int irq)
1241{ 1229{
@@ -1360,8 +1348,8 @@ static void __init setup_IO_APIC_irqs(void)
1360 if (!apic && (irq < 16)) 1348 if (!apic && (irq < 16))
1361 disable_8259A_irq(irq); 1349 disable_8259A_irq(irq);
1362 } 1350 }
1363 ioapic_write_entry(apic, pin, entry);
1364 spin_lock_irqsave(&ioapic_lock, flags); 1351 spin_lock_irqsave(&ioapic_lock, flags);
1352 __ioapic_write_entry(apic, pin, entry);
1365 set_native_irq_info(irq, TARGET_CPUS); 1353 set_native_irq_info(irq, TARGET_CPUS);
1366 spin_unlock_irqrestore(&ioapic_lock, flags); 1354 spin_unlock_irqrestore(&ioapic_lock, flags);
1367 } 1355 }
@@ -1926,6 +1914,15 @@ static void __init setup_ioapic_ids_from_mpc(void)
1926static void __init setup_ioapic_ids_from_mpc(void) { } 1914static void __init setup_ioapic_ids_from_mpc(void) { }
1927#endif 1915#endif
1928 1916
1917static int no_timer_check __initdata;
1918
1919static int __init notimercheck(char *s)
1920{
1921 no_timer_check = 1;
1922 return 1;
1923}
1924__setup("no_timer_check", notimercheck);
1925
1929/* 1926/*
1930 * There is a nasty bug in some older SMP boards, their mptable lies 1927 * There is a nasty bug in some older SMP boards, their mptable lies
1931 * about the timer IRQ. We do the following to work around the situation: 1928 * about the timer IRQ. We do the following to work around the situation:
@@ -1934,10 +1931,13 @@ static void __init setup_ioapic_ids_from_mpc(void) { }
1934 * - if this function detects that timer IRQs are defunct, then we fall 1931 * - if this function detects that timer IRQs are defunct, then we fall
1935 * back to ISA timer IRQs 1932 * back to ISA timer IRQs
1936 */ 1933 */
1937static int __init timer_irq_works(void) 1934int __init timer_irq_works(void)
1938{ 1935{
1939 unsigned long t1 = jiffies; 1936 unsigned long t1 = jiffies;
1940 1937
1938 if (no_timer_check)
1939 return 1;
1940
1941 local_irq_enable(); 1941 local_irq_enable();
1942 /* Let ten ticks pass... */ 1942 /* Let ten ticks pass... */
1943 mdelay((10 * 1000) / HZ); 1943 mdelay((10 * 1000) / HZ);
@@ -2161,9 +2161,15 @@ static inline void unlock_ExtINT_logic(void)
2161 unsigned char save_control, save_freq_select; 2161 unsigned char save_control, save_freq_select;
2162 2162
2163 pin = find_isa_irq_pin(8, mp_INT); 2163 pin = find_isa_irq_pin(8, mp_INT);
2164 if (pin == -1) {
2165 WARN_ON_ONCE(1);
2166 return;
2167 }
2164 apic = find_isa_irq_apic(8, mp_INT); 2168 apic = find_isa_irq_apic(8, mp_INT);
2165 if (pin == -1) 2169 if (apic == -1) {
2170 WARN_ON_ONCE(1);
2166 return; 2171 return;
2172 }
2167 2173
2168 entry0 = ioapic_read_entry(apic, pin); 2174 entry0 = ioapic_read_entry(apic, pin);
2169 clear_IO_APIC_pin(apic, pin); 2175 clear_IO_APIC_pin(apic, pin);
@@ -2208,7 +2214,7 @@ int timer_uses_ioapic_pin_0;
2208 * is so screwy. Thanks to Brian Perkins for testing/hacking this beast 2214 * is so screwy. Thanks to Brian Perkins for testing/hacking this beast
2209 * fanatically on his truly buggy board. 2215 * fanatically on his truly buggy board.
2210 */ 2216 */
2211static inline void check_timer(void) 2217static inline void __init check_timer(void)
2212{ 2218{
2213 int apic1, pin1, apic2, pin2; 2219 int apic1, pin1, apic2, pin2;
2214 int vector; 2220 int vector;
@@ -2856,8 +2862,8 @@ int io_apic_set_pci_routing (int ioapic, int pin, int irq, int edge_level, int a
2856 if (!ioapic && (irq < 16)) 2862 if (!ioapic && (irq < 16))
2857 disable_8259A_irq(irq); 2863 disable_8259A_irq(irq);
2858 2864
2859 ioapic_write_entry(ioapic, pin, entry);
2860 spin_lock_irqsave(&ioapic_lock, flags); 2865 spin_lock_irqsave(&ioapic_lock, flags);
2866 __ioapic_write_entry(ioapic, pin, entry);
2861 set_native_irq_info(irq, TARGET_CPUS); 2867 set_native_irq_info(irq, TARGET_CPUS);
2862 spin_unlock_irqrestore(&ioapic_lock, flags); 2868 spin_unlock_irqrestore(&ioapic_lock, flags);
2863 2869
diff --git a/arch/i386/kernel/kprobes.c b/arch/i386/kernel/kprobes.c
index fc79e1e859c4..af1d53344993 100644
--- a/arch/i386/kernel/kprobes.c
+++ b/arch/i386/kernel/kprobes.c
@@ -184,7 +184,7 @@ void __kprobes arch_disarm_kprobe(struct kprobe *p)
184void __kprobes arch_remove_kprobe(struct kprobe *p) 184void __kprobes arch_remove_kprobe(struct kprobe *p)
185{ 185{
186 mutex_lock(&kprobe_mutex); 186 mutex_lock(&kprobe_mutex);
187 free_insn_slot(p->ainsn.insn); 187 free_insn_slot(p->ainsn.insn, (p->ainsn.boostable == 1));
188 mutex_unlock(&kprobe_mutex); 188 mutex_unlock(&kprobe_mutex);
189} 189}
190 190
@@ -333,7 +333,7 @@ static int __kprobes kprobe_handler(struct pt_regs *regs)
333 return 1; 333 return 1;
334 334
335ss_probe: 335ss_probe:
336#ifndef CONFIG_PREEMPT 336#if !defined(CONFIG_PREEMPT) || defined(CONFIG_PM)
337 if (p->ainsn.boostable == 1 && !p->post_handler){ 337 if (p->ainsn.boostable == 1 && !p->post_handler){
338 /* Boost up -- we can execute copied instructions directly */ 338 /* Boost up -- we can execute copied instructions directly */
339 reset_current_kprobe(); 339 reset_current_kprobe();
diff --git a/arch/i386/kernel/ldt.c b/arch/i386/kernel/ldt.c
index 445211eb2d57..b410e5fb034f 100644
--- a/arch/i386/kernel/ldt.c
+++ b/arch/i386/kernel/ldt.c
@@ -160,16 +160,14 @@ static int read_default_ldt(void __user * ptr, unsigned long bytecount)
160{ 160{
161 int err; 161 int err;
162 unsigned long size; 162 unsigned long size;
163 void *address;
164 163
165 err = 0; 164 err = 0;
166 address = &default_ldt[0];
167 size = 5*sizeof(struct desc_struct); 165 size = 5*sizeof(struct desc_struct);
168 if (size > bytecount) 166 if (size > bytecount)
169 size = bytecount; 167 size = bytecount;
170 168
171 err = size; 169 err = size;
172 if (copy_to_user(ptr, address, size)) 170 if (clear_user(ptr, size))
173 err = -EFAULT; 171 err = -EFAULT;
174 172
175 return err; 173 return err;
diff --git a/arch/i386/kernel/mca.c b/arch/i386/kernel/mca.c
index eb57a851789d..b83672b89527 100644
--- a/arch/i386/kernel/mca.c
+++ b/arch/i386/kernel/mca.c
@@ -283,10 +283,9 @@ static int __init mca_init(void)
283 bus->f.mca_transform_memory = mca_dummy_transform_memory; 283 bus->f.mca_transform_memory = mca_dummy_transform_memory;
284 284
285 /* get the motherboard device */ 285 /* get the motherboard device */
286 mca_dev = kmalloc(sizeof(struct mca_device), GFP_KERNEL); 286 mca_dev = kzalloc(sizeof(struct mca_device), GFP_KERNEL);
287 if(unlikely(!mca_dev)) 287 if(unlikely(!mca_dev))
288 goto out_nomem; 288 goto out_nomem;
289 memset(mca_dev, 0, sizeof(struct mca_device));
290 289
291 /* 290 /*
292 * We do not expect many MCA interrupts during initialization, 291 * We do not expect many MCA interrupts during initialization,
@@ -310,11 +309,9 @@ static int __init mca_init(void)
310 mca_dev->slot = MCA_MOTHERBOARD; 309 mca_dev->slot = MCA_MOTHERBOARD;
311 mca_register_device(MCA_PRIMARY_BUS, mca_dev); 310 mca_register_device(MCA_PRIMARY_BUS, mca_dev);
312 311
313 mca_dev = kmalloc(sizeof(struct mca_device), GFP_ATOMIC); 312 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC);
314 if(unlikely(!mca_dev)) 313 if(unlikely(!mca_dev))
315 goto out_unlock_nomem; 314 goto out_unlock_nomem;
316 memset(mca_dev, 0, sizeof(struct mca_device));
317
318 315
319 /* Put motherboard into video setup mode, read integrated video 316 /* Put motherboard into video setup mode, read integrated video
320 * POS registers, and turn motherboard setup off. 317 * POS registers, and turn motherboard setup off.
@@ -349,10 +346,9 @@ static int __init mca_init(void)
349 } 346 }
350 if(which_scsi) { 347 if(which_scsi) {
351 /* found a scsi card */ 348 /* found a scsi card */
352 mca_dev = kmalloc(sizeof(struct mca_device), GFP_ATOMIC); 349 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC);
353 if(unlikely(!mca_dev)) 350 if(unlikely(!mca_dev))
354 goto out_unlock_nomem; 351 goto out_unlock_nomem;
355 memset(mca_dev, 0, sizeof(struct mca_device));
356 352
357 for(j = 0; j < 8; j++) 353 for(j = 0; j < 8; j++)
358 mca_dev->pos[j] = pos[j]; 354 mca_dev->pos[j] = pos[j];
@@ -378,10 +374,9 @@ static int __init mca_init(void)
378 if(!mca_read_and_store_pos(pos)) 374 if(!mca_read_and_store_pos(pos))
379 continue; 375 continue;
380 376
381 mca_dev = kmalloc(sizeof(struct mca_device), GFP_ATOMIC); 377 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC);
382 if(unlikely(!mca_dev)) 378 if(unlikely(!mca_dev))
383 goto out_unlock_nomem; 379 goto out_unlock_nomem;
384 memset(mca_dev, 0, sizeof(struct mca_device));
385 380
386 for(j=0; j<8; j++) 381 for(j=0; j<8; j++)
387 mca_dev->pos[j]=pos[j]; 382 mca_dev->pos[j]=pos[j];
diff --git a/arch/i386/kernel/microcode.c b/arch/i386/kernel/microcode.c
index 23f5984d0654..972346604f9d 100644
--- a/arch/i386/kernel/microcode.c
+++ b/arch/i386/kernel/microcode.c
@@ -703,7 +703,6 @@ static struct sysdev_driver mc_sysdev_driver = {
703 .resume = mc_sysdev_resume, 703 .resume = mc_sysdev_resume,
704}; 704};
705 705
706#ifdef CONFIG_HOTPLUG_CPU
707static __cpuinit int 706static __cpuinit int
708mc_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu) 707mc_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu)
709{ 708{
@@ -726,7 +725,6 @@ mc_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu)
726static struct notifier_block mc_cpu_notifier = { 725static struct notifier_block mc_cpu_notifier = {
727 .notifier_call = mc_cpu_callback, 726 .notifier_call = mc_cpu_callback,
728}; 727};
729#endif
730 728
731static int __init microcode_init (void) 729static int __init microcode_init (void)
732{ 730{
diff --git a/arch/i386/kernel/module.c b/arch/i386/kernel/module.c
index 470cf97e7cd3..d7d9c8b23f72 100644
--- a/arch/i386/kernel/module.c
+++ b/arch/i386/kernel/module.c
@@ -108,7 +108,8 @@ int module_finalize(const Elf_Ehdr *hdr,
108 const Elf_Shdr *sechdrs, 108 const Elf_Shdr *sechdrs,
109 struct module *me) 109 struct module *me)
110{ 110{
111 const Elf_Shdr *s, *text = NULL, *alt = NULL, *locks = NULL; 111 const Elf_Shdr *s, *text = NULL, *alt = NULL, *locks = NULL,
112 *para = NULL;
112 char *secstrings = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset; 113 char *secstrings = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset;
113 114
114 for (s = sechdrs; s < sechdrs + hdr->e_shnum; s++) { 115 for (s = sechdrs; s < sechdrs + hdr->e_shnum; s++) {
@@ -118,6 +119,8 @@ int module_finalize(const Elf_Ehdr *hdr,
118 alt = s; 119 alt = s;
119 if (!strcmp(".smp_locks", secstrings + s->sh_name)) 120 if (!strcmp(".smp_locks", secstrings + s->sh_name))
120 locks= s; 121 locks= s;
122 if (!strcmp(".parainstructions", secstrings + s->sh_name))
123 para = s;
121 } 124 }
122 125
123 if (alt) { 126 if (alt) {
@@ -132,6 +135,12 @@ int module_finalize(const Elf_Ehdr *hdr,
132 lseg, lseg + locks->sh_size, 135 lseg, lseg + locks->sh_size,
133 tseg, tseg + text->sh_size); 136 tseg, tseg + text->sh_size);
134 } 137 }
138
139 if (para) {
140 void *pseg = (void *)para->sh_addr;
141 apply_paravirt(pseg, pseg + para->sh_size);
142 }
143
135 return 0; 144 return 0;
136} 145}
137 146
diff --git a/arch/i386/kernel/mpparse.c b/arch/i386/kernel/mpparse.c
index 442aaf8c77eb..2ce67228dff8 100644
--- a/arch/i386/kernel/mpparse.c
+++ b/arch/i386/kernel/mpparse.c
@@ -249,8 +249,6 @@ static void __init MP_bus_info (struct mpc_config_bus *m)
249 mp_current_pci_id++; 249 mp_current_pci_id++;
250 } else if (strncmp(str, BUSTYPE_MCA, sizeof(BUSTYPE_MCA)-1) == 0) { 250 } else if (strncmp(str, BUSTYPE_MCA, sizeof(BUSTYPE_MCA)-1) == 0) {
251 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_MCA; 251 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_MCA;
252 } else if (strncmp(str, BUSTYPE_NEC98, sizeof(BUSTYPE_NEC98)-1) == 0) {
253 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_NEC98;
254 } else { 252 } else {
255 printk(KERN_WARNING "Unknown bustype %s - ignoring\n", str); 253 printk(KERN_WARNING "Unknown bustype %s - ignoring\n", str);
256 } 254 }
diff --git a/arch/i386/kernel/msr.c b/arch/i386/kernel/msr.c
index a773f776c9ea..1d1a56cae340 100644
--- a/arch/i386/kernel/msr.c
+++ b/arch/i386/kernel/msr.c
@@ -195,7 +195,6 @@ static ssize_t msr_write(struct file *file, const char __user *buf,
195{ 195{
196 const u32 __user *tmp = (const u32 __user *)buf; 196 const u32 __user *tmp = (const u32 __user *)buf;
197 u32 data[2]; 197 u32 data[2];
198 size_t rv;
199 u32 reg = *ppos; 198 u32 reg = *ppos;
200 int cpu = iminor(file->f_dentry->d_inode); 199 int cpu = iminor(file->f_dentry->d_inode);
201 int err; 200 int err;
@@ -203,7 +202,7 @@ static ssize_t msr_write(struct file *file, const char __user *buf,
203 if (count % 8) 202 if (count % 8)
204 return -EINVAL; /* Invalid chunk size */ 203 return -EINVAL; /* Invalid chunk size */
205 204
206 for (rv = 0; count; count -= 8) { 205 for (; count; count -= 8) {
207 if (copy_from_user(&data, tmp, 8)) 206 if (copy_from_user(&data, tmp, 8))
208 return -EFAULT; 207 return -EFAULT;
209 err = do_wrmsr(cpu, reg, data[0], data[1]); 208 err = do_wrmsr(cpu, reg, data[0], data[1]);
@@ -250,7 +249,6 @@ static int msr_device_create(int i)
250 return err; 249 return err;
251} 250}
252 251
253#ifdef CONFIG_HOTPLUG_CPU
254static int msr_class_cpu_callback(struct notifier_block *nfb, 252static int msr_class_cpu_callback(struct notifier_block *nfb,
255 unsigned long action, void *hcpu) 253 unsigned long action, void *hcpu)
256{ 254{
@@ -271,7 +269,6 @@ static struct notifier_block __cpuinitdata msr_class_cpu_notifier =
271{ 269{
272 .notifier_call = msr_class_cpu_callback, 270 .notifier_call = msr_class_cpu_callback,
273}; 271};
274#endif
275 272
276static int __init msr_init(void) 273static int __init msr_init(void)
277{ 274{
diff --git a/arch/i386/kernel/nmi.c b/arch/i386/kernel/nmi.c
index eaafe233a5da..f5bc7e1be801 100644
--- a/arch/i386/kernel/nmi.c
+++ b/arch/i386/kernel/nmi.c
@@ -22,6 +22,7 @@
22#include <linux/percpu.h> 22#include <linux/percpu.h>
23#include <linux/dmi.h> 23#include <linux/dmi.h>
24#include <linux/kprobes.h> 24#include <linux/kprobes.h>
25#include <linux/cpumask.h>
25 26
26#include <asm/smp.h> 27#include <asm/smp.h>
27#include <asm/nmi.h> 28#include <asm/nmi.h>
@@ -42,6 +43,8 @@ int nmi_watchdog_enabled;
42static DEFINE_PER_CPU(unsigned long, perfctr_nmi_owner); 43static DEFINE_PER_CPU(unsigned long, perfctr_nmi_owner);
43static DEFINE_PER_CPU(unsigned long, evntsel_nmi_owner[3]); 44static DEFINE_PER_CPU(unsigned long, evntsel_nmi_owner[3]);
44 45
46static cpumask_t backtrace_mask = CPU_MASK_NONE;
47
45/* this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's 48/* this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's
46 * offset from MSR_P4_BSU_ESCR0. It will be the max for all platforms (for now) 49 * offset from MSR_P4_BSU_ESCR0. It will be the max for all platforms (for now)
47 */ 50 */
@@ -867,14 +870,16 @@ static unsigned int
867 870
868void touch_nmi_watchdog (void) 871void touch_nmi_watchdog (void)
869{ 872{
870 int i; 873 if (nmi_watchdog > 0) {
874 unsigned cpu;
871 875
872 /* 876 /*
873 * Just reset the alert counters, (other CPUs might be 877 * Just reset the alert counters, (other CPUs might be
874 * spinning on locks we hold): 878 * spinning on locks we hold):
875 */ 879 */
876 for_each_possible_cpu(i) 880 for_each_present_cpu (cpu)
877 alert_counter[i] = 0; 881 alert_counter[cpu] = 0;
882 }
878 883
879 /* 884 /*
880 * Tickle the softlockup detector too: 885 * Tickle the softlockup detector too:
@@ -907,6 +912,16 @@ __kprobes int nmi_watchdog_tick(struct pt_regs * regs, unsigned reason)
907 touched = 1; 912 touched = 1;
908 } 913 }
909 914
915 if (cpu_isset(cpu, backtrace_mask)) {
916 static DEFINE_SPINLOCK(lock); /* Serialise the printks */
917
918 spin_lock(&lock);
919 printk("NMI backtrace for cpu %d\n", cpu);
920 dump_stack();
921 spin_unlock(&lock);
922 cpu_clear(cpu, backtrace_mask);
923 }
924
910 sum = per_cpu(irq_stat, cpu).apic_timer_irqs; 925 sum = per_cpu(irq_stat, cpu).apic_timer_irqs;
911 926
912 /* if the apic timer isn't firing, this cpu isn't doing much */ 927 /* if the apic timer isn't firing, this cpu isn't doing much */
@@ -1033,6 +1048,19 @@ int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file,
1033 1048
1034#endif 1049#endif
1035 1050
1051void __trigger_all_cpu_backtrace(void)
1052{
1053 int i;
1054
1055 backtrace_mask = cpu_online_map;
1056 /* Wait for up to 10 seconds for all CPUs to do the backtrace */
1057 for (i = 0; i < 10 * 1000; i++) {
1058 if (cpus_empty(backtrace_mask))
1059 break;
1060 mdelay(1);
1061 }
1062}
1063
1036EXPORT_SYMBOL(nmi_active); 1064EXPORT_SYMBOL(nmi_active);
1037EXPORT_SYMBOL(nmi_watchdog); 1065EXPORT_SYMBOL(nmi_watchdog);
1038EXPORT_SYMBOL(avail_to_resrv_perfctr_nmi); 1066EXPORT_SYMBOL(avail_to_resrv_perfctr_nmi);
diff --git a/arch/i386/kernel/paravirt.c b/arch/i386/kernel/paravirt.c
new file mode 100644
index 000000000000..3dceab5828f1
--- /dev/null
+++ b/arch/i386/kernel/paravirt.c
@@ -0,0 +1,569 @@
1/* Paravirtualization interfaces
2 Copyright (C) 2006 Rusty Russell IBM Corporation
3
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2 of the License, or
7 (at your option) any later version.
8
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
13
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17*/
18#include <linux/errno.h>
19#include <linux/module.h>
20#include <linux/efi.h>
21#include <linux/bcd.h>
22#include <linux/start_kernel.h>
23
24#include <asm/bug.h>
25#include <asm/paravirt.h>
26#include <asm/desc.h>
27#include <asm/setup.h>
28#include <asm/arch_hooks.h>
29#include <asm/time.h>
30#include <asm/irq.h>
31#include <asm/delay.h>
32#include <asm/fixmap.h>
33#include <asm/apic.h>
34#include <asm/tlbflush.h>
35
36/* nop stub */
37static void native_nop(void)
38{
39}
40
41static void __init default_banner(void)
42{
43 printk(KERN_INFO "Booting paravirtualized kernel on %s\n",
44 paravirt_ops.name);
45}
46
47char *memory_setup(void)
48{
49 return paravirt_ops.memory_setup();
50}
51
52/* Simple instruction patching code. */
53#define DEF_NATIVE(name, code) \
54 extern const char start_##name[], end_##name[]; \
55 asm("start_" #name ": " code "; end_" #name ":")
56DEF_NATIVE(cli, "cli");
57DEF_NATIVE(sti, "sti");
58DEF_NATIVE(popf, "push %eax; popf");
59DEF_NATIVE(pushf, "pushf; pop %eax");
60DEF_NATIVE(pushf_cli, "pushf; pop %eax; cli");
61DEF_NATIVE(iret, "iret");
62DEF_NATIVE(sti_sysexit, "sti; sysexit");
63
64static const struct native_insns
65{
66 const char *start, *end;
67} native_insns[] = {
68 [PARAVIRT_IRQ_DISABLE] = { start_cli, end_cli },
69 [PARAVIRT_IRQ_ENABLE] = { start_sti, end_sti },
70 [PARAVIRT_RESTORE_FLAGS] = { start_popf, end_popf },
71 [PARAVIRT_SAVE_FLAGS] = { start_pushf, end_pushf },
72 [PARAVIRT_SAVE_FLAGS_IRQ_DISABLE] = { start_pushf_cli, end_pushf_cli },
73 [PARAVIRT_INTERRUPT_RETURN] = { start_iret, end_iret },
74 [PARAVIRT_STI_SYSEXIT] = { start_sti_sysexit, end_sti_sysexit },
75};
76
77static unsigned native_patch(u8 type, u16 clobbers, void *insns, unsigned len)
78{
79 unsigned int insn_len;
80
81 /* Don't touch it if we don't have a replacement */
82 if (type >= ARRAY_SIZE(native_insns) || !native_insns[type].start)
83 return len;
84
85 insn_len = native_insns[type].end - native_insns[type].start;
86
87 /* Similarly if we can't fit replacement. */
88 if (len < insn_len)
89 return len;
90
91 memcpy(insns, native_insns[type].start, insn_len);
92 return insn_len;
93}
94
95static fastcall unsigned long native_get_debugreg(int regno)
96{
97 unsigned long val = 0; /* Damn you, gcc! */
98
99 switch (regno) {
100 case 0:
101 asm("movl %%db0, %0" :"=r" (val)); break;
102 case 1:
103 asm("movl %%db1, %0" :"=r" (val)); break;
104 case 2:
105 asm("movl %%db2, %0" :"=r" (val)); break;
106 case 3:
107 asm("movl %%db3, %0" :"=r" (val)); break;
108 case 6:
109 asm("movl %%db6, %0" :"=r" (val)); break;
110 case 7:
111 asm("movl %%db7, %0" :"=r" (val)); break;
112 default:
113 BUG();
114 }
115 return val;
116}
117
118static fastcall void native_set_debugreg(int regno, unsigned long value)
119{
120 switch (regno) {
121 case 0:
122 asm("movl %0,%%db0" : /* no output */ :"r" (value));
123 break;
124 case 1:
125 asm("movl %0,%%db1" : /* no output */ :"r" (value));
126 break;
127 case 2:
128 asm("movl %0,%%db2" : /* no output */ :"r" (value));
129 break;
130 case 3:
131 asm("movl %0,%%db3" : /* no output */ :"r" (value));
132 break;
133 case 6:
134 asm("movl %0,%%db6" : /* no output */ :"r" (value));
135 break;
136 case 7:
137 asm("movl %0,%%db7" : /* no output */ :"r" (value));
138 break;
139 default:
140 BUG();
141 }
142}
143
144void init_IRQ(void)
145{
146 paravirt_ops.init_IRQ();
147}
148
149static fastcall void native_clts(void)
150{
151 asm volatile ("clts");
152}
153
154static fastcall unsigned long native_read_cr0(void)
155{
156 unsigned long val;
157 asm volatile("movl %%cr0,%0\n\t" :"=r" (val));
158 return val;
159}
160
161static fastcall void native_write_cr0(unsigned long val)
162{
163 asm volatile("movl %0,%%cr0": :"r" (val));
164}
165
166static fastcall unsigned long native_read_cr2(void)
167{
168 unsigned long val;
169 asm volatile("movl %%cr2,%0\n\t" :"=r" (val));
170 return val;
171}
172
173static fastcall void native_write_cr2(unsigned long val)
174{
175 asm volatile("movl %0,%%cr2": :"r" (val));
176}
177
178static fastcall unsigned long native_read_cr3(void)
179{
180 unsigned long val;
181 asm volatile("movl %%cr3,%0\n\t" :"=r" (val));
182 return val;
183}
184
185static fastcall void native_write_cr3(unsigned long val)
186{
187 asm volatile("movl %0,%%cr3": :"r" (val));
188}
189
190static fastcall unsigned long native_read_cr4(void)
191{
192 unsigned long val;
193 asm volatile("movl %%cr4,%0\n\t" :"=r" (val));
194 return val;
195}
196
197static fastcall unsigned long native_read_cr4_safe(void)
198{
199 unsigned long val;
200 /* This could fault if %cr4 does not exist */
201 asm("1: movl %%cr4, %0 \n"
202 "2: \n"
203 ".section __ex_table,\"a\" \n"
204 ".long 1b,2b \n"
205 ".previous \n"
206 : "=r" (val): "0" (0));
207 return val;
208}
209
210static fastcall void native_write_cr4(unsigned long val)
211{
212 asm volatile("movl %0,%%cr4": :"r" (val));
213}
214
215static fastcall unsigned long native_save_fl(void)
216{
217 unsigned long f;
218 asm volatile("pushfl ; popl %0":"=g" (f): /* no input */);
219 return f;
220}
221
222static fastcall void native_restore_fl(unsigned long f)
223{
224 asm volatile("pushl %0 ; popfl": /* no output */
225 :"g" (f)
226 :"memory", "cc");
227}
228
229static fastcall void native_irq_disable(void)
230{
231 asm volatile("cli": : :"memory");
232}
233
234static fastcall void native_irq_enable(void)
235{
236 asm volatile("sti": : :"memory");
237}
238
239static fastcall void native_safe_halt(void)
240{
241 asm volatile("sti; hlt": : :"memory");
242}
243
244static fastcall void native_halt(void)
245{
246 asm volatile("hlt": : :"memory");
247}
248
249static fastcall void native_wbinvd(void)
250{
251 asm volatile("wbinvd": : :"memory");
252}
253
254static fastcall unsigned long long native_read_msr(unsigned int msr, int *err)
255{
256 unsigned long long val;
257
258 asm volatile("2: rdmsr ; xorl %0,%0\n"
259 "1:\n\t"
260 ".section .fixup,\"ax\"\n\t"
261 "3: movl %3,%0 ; jmp 1b\n\t"
262 ".previous\n\t"
263 ".section __ex_table,\"a\"\n"
264 " .align 4\n\t"
265 " .long 2b,3b\n\t"
266 ".previous"
267 : "=r" (*err), "=A" (val)
268 : "c" (msr), "i" (-EFAULT));
269
270 return val;
271}
272
273static fastcall int native_write_msr(unsigned int msr, unsigned long long val)
274{
275 int err;
276 asm volatile("2: wrmsr ; xorl %0,%0\n"
277 "1:\n\t"
278 ".section .fixup,\"ax\"\n\t"
279 "3: movl %4,%0 ; jmp 1b\n\t"
280 ".previous\n\t"
281 ".section __ex_table,\"a\"\n"
282 " .align 4\n\t"
283 " .long 2b,3b\n\t"
284 ".previous"
285 : "=a" (err)
286 : "c" (msr), "0" ((u32)val), "d" ((u32)(val>>32)),
287 "i" (-EFAULT));
288 return err;
289}
290
291static fastcall unsigned long long native_read_tsc(void)
292{
293 unsigned long long val;
294 asm volatile("rdtsc" : "=A" (val));
295 return val;
296}
297
298static fastcall unsigned long long native_read_pmc(void)
299{
300 unsigned long long val;
301 asm volatile("rdpmc" : "=A" (val));
302 return val;
303}
304
305static fastcall void native_load_tr_desc(void)
306{
307 asm volatile("ltr %w0"::"q" (GDT_ENTRY_TSS*8));
308}
309
310static fastcall void native_load_gdt(const struct Xgt_desc_struct *dtr)
311{
312 asm volatile("lgdt %0"::"m" (*dtr));
313}
314
315static fastcall void native_load_idt(const struct Xgt_desc_struct *dtr)
316{
317 asm volatile("lidt %0"::"m" (*dtr));
318}
319
320static fastcall void native_store_gdt(struct Xgt_desc_struct *dtr)
321{
322 asm ("sgdt %0":"=m" (*dtr));
323}
324
325static fastcall void native_store_idt(struct Xgt_desc_struct *dtr)
326{
327 asm ("sidt %0":"=m" (*dtr));
328}
329
330static fastcall unsigned long native_store_tr(void)
331{
332 unsigned long tr;
333 asm ("str %0":"=r" (tr));
334 return tr;
335}
336
337static fastcall void native_load_tls(struct thread_struct *t, unsigned int cpu)
338{
339#define C(i) get_cpu_gdt_table(cpu)[GDT_ENTRY_TLS_MIN + i] = t->tls_array[i]
340 C(0); C(1); C(2);
341#undef C
342}
343
344static inline void native_write_dt_entry(void *dt, int entry, u32 entry_low, u32 entry_high)
345{
346 u32 *lp = (u32 *)((char *)dt + entry*8);
347 lp[0] = entry_low;
348 lp[1] = entry_high;
349}
350
351static fastcall void native_write_ldt_entry(void *dt, int entrynum, u32 low, u32 high)
352{
353 native_write_dt_entry(dt, entrynum, low, high);
354}
355
356static fastcall void native_write_gdt_entry(void *dt, int entrynum, u32 low, u32 high)
357{
358 native_write_dt_entry(dt, entrynum, low, high);
359}
360
361static fastcall void native_write_idt_entry(void *dt, int entrynum, u32 low, u32 high)
362{
363 native_write_dt_entry(dt, entrynum, low, high);
364}
365
366static fastcall void native_load_esp0(struct tss_struct *tss,
367 struct thread_struct *thread)
368{
369 tss->esp0 = thread->esp0;
370
371 /* This can only happen when SEP is enabled, no need to test "SEP"arately */
372 if (unlikely(tss->ss1 != thread->sysenter_cs)) {
373 tss->ss1 = thread->sysenter_cs;
374 wrmsr(MSR_IA32_SYSENTER_CS, thread->sysenter_cs, 0);
375 }
376}
377
378static fastcall void native_io_delay(void)
379{
380 asm volatile("outb %al,$0x80");
381}
382
383static fastcall void native_flush_tlb(void)
384{
385 __native_flush_tlb();
386}
387
388/*
389 * Global pages have to be flushed a bit differently. Not a real
390 * performance problem because this does not happen often.
391 */
392static fastcall void native_flush_tlb_global(void)
393{
394 __native_flush_tlb_global();
395}
396
397static fastcall void native_flush_tlb_single(u32 addr)
398{
399 __native_flush_tlb_single(addr);
400}
401
402#ifndef CONFIG_X86_PAE
403static fastcall void native_set_pte(pte_t *ptep, pte_t pteval)
404{
405 *ptep = pteval;
406}
407
408static fastcall void native_set_pte_at(struct mm_struct *mm, u32 addr, pte_t *ptep, pte_t pteval)
409{
410 *ptep = pteval;
411}
412
413static fastcall void native_set_pmd(pmd_t *pmdp, pmd_t pmdval)
414{
415 *pmdp = pmdval;
416}
417
418#else /* CONFIG_X86_PAE */
419
420static fastcall void native_set_pte(pte_t *ptep, pte_t pte)
421{
422 ptep->pte_high = pte.pte_high;
423 smp_wmb();
424 ptep->pte_low = pte.pte_low;
425}
426
427static fastcall void native_set_pte_at(struct mm_struct *mm, u32 addr, pte_t *ptep, pte_t pte)
428{
429 ptep->pte_high = pte.pte_high;
430 smp_wmb();
431 ptep->pte_low = pte.pte_low;
432}
433
434static fastcall void native_set_pte_present(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
435{
436 ptep->pte_low = 0;
437 smp_wmb();
438 ptep->pte_high = pte.pte_high;
439 smp_wmb();
440 ptep->pte_low = pte.pte_low;
441}
442
443static fastcall void native_set_pte_atomic(pte_t *ptep, pte_t pteval)
444{
445 set_64bit((unsigned long long *)ptep,pte_val(pteval));
446}
447
448static fastcall void native_set_pmd(pmd_t *pmdp, pmd_t pmdval)
449{
450 set_64bit((unsigned long long *)pmdp,pmd_val(pmdval));
451}
452
453static fastcall void native_set_pud(pud_t *pudp, pud_t pudval)
454{
455 *pudp = pudval;
456}
457
458static fastcall void native_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
459{
460 ptep->pte_low = 0;
461 smp_wmb();
462 ptep->pte_high = 0;
463}
464
465static fastcall void native_pmd_clear(pmd_t *pmd)
466{
467 u32 *tmp = (u32 *)pmd;
468 *tmp = 0;
469 smp_wmb();
470 *(tmp + 1) = 0;
471}
472#endif /* CONFIG_X86_PAE */
473
474/* These are in entry.S */
475extern fastcall void native_iret(void);
476extern fastcall void native_irq_enable_sysexit(void);
477
478static int __init print_banner(void)
479{
480 paravirt_ops.banner();
481 return 0;
482}
483core_initcall(print_banner);
484
485/* We simply declare start_kernel to be the paravirt probe of last resort. */
486paravirt_probe(start_kernel);
487
488struct paravirt_ops paravirt_ops = {
489 .name = "bare hardware",
490 .paravirt_enabled = 0,
491 .kernel_rpl = 0,
492
493 .patch = native_patch,
494 .banner = default_banner,
495 .arch_setup = native_nop,
496 .memory_setup = machine_specific_memory_setup,
497 .get_wallclock = native_get_wallclock,
498 .set_wallclock = native_set_wallclock,
499 .time_init = time_init_hook,
500 .init_IRQ = native_init_IRQ,
501
502 .cpuid = native_cpuid,
503 .get_debugreg = native_get_debugreg,
504 .set_debugreg = native_set_debugreg,
505 .clts = native_clts,
506 .read_cr0 = native_read_cr0,
507 .write_cr0 = native_write_cr0,
508 .read_cr2 = native_read_cr2,
509 .write_cr2 = native_write_cr2,
510 .read_cr3 = native_read_cr3,
511 .write_cr3 = native_write_cr3,
512 .read_cr4 = native_read_cr4,
513 .read_cr4_safe = native_read_cr4_safe,
514 .write_cr4 = native_write_cr4,
515 .save_fl = native_save_fl,
516 .restore_fl = native_restore_fl,
517 .irq_disable = native_irq_disable,
518 .irq_enable = native_irq_enable,
519 .safe_halt = native_safe_halt,
520 .halt = native_halt,
521 .wbinvd = native_wbinvd,
522 .read_msr = native_read_msr,
523 .write_msr = native_write_msr,
524 .read_tsc = native_read_tsc,
525 .read_pmc = native_read_pmc,
526 .load_tr_desc = native_load_tr_desc,
527 .set_ldt = native_set_ldt,
528 .load_gdt = native_load_gdt,
529 .load_idt = native_load_idt,
530 .store_gdt = native_store_gdt,
531 .store_idt = native_store_idt,
532 .store_tr = native_store_tr,
533 .load_tls = native_load_tls,
534 .write_ldt_entry = native_write_ldt_entry,
535 .write_gdt_entry = native_write_gdt_entry,
536 .write_idt_entry = native_write_idt_entry,
537 .load_esp0 = native_load_esp0,
538
539 .set_iopl_mask = native_set_iopl_mask,
540 .io_delay = native_io_delay,
541 .const_udelay = __const_udelay,
542
543#ifdef CONFIG_X86_LOCAL_APIC
544 .apic_write = native_apic_write,
545 .apic_write_atomic = native_apic_write_atomic,
546 .apic_read = native_apic_read,
547#endif
548
549 .flush_tlb_user = native_flush_tlb,
550 .flush_tlb_kernel = native_flush_tlb_global,
551 .flush_tlb_single = native_flush_tlb_single,
552
553 .set_pte = native_set_pte,
554 .set_pte_at = native_set_pte_at,
555 .set_pmd = native_set_pmd,
556 .pte_update = (void *)native_nop,
557 .pte_update_defer = (void *)native_nop,
558#ifdef CONFIG_X86_PAE
559 .set_pte_atomic = native_set_pte_atomic,
560 .set_pte_present = native_set_pte_present,
561 .set_pud = native_set_pud,
562 .pte_clear = native_pte_clear,
563 .pmd_clear = native_pmd_clear,
564#endif
565
566 .irq_enable_sysexit = native_irq_enable_sysexit,
567 .iret = native_iret,
568};
569EXPORT_SYMBOL(paravirt_ops);
diff --git a/arch/i386/kernel/pci-dma.c b/arch/i386/kernel/pci-dma.c
index 5c8c6ef1fc5e..41af692c1584 100644
--- a/arch/i386/kernel/pci-dma.c
+++ b/arch/i386/kernel/pci-dma.c
@@ -92,14 +92,12 @@ int dma_declare_coherent_memory(struct device *dev, dma_addr_t bus_addr,
92 if (!mem_base) 92 if (!mem_base)
93 goto out; 93 goto out;
94 94
95 dev->dma_mem = kmalloc(sizeof(struct dma_coherent_mem), GFP_KERNEL); 95 dev->dma_mem = kzalloc(sizeof(struct dma_coherent_mem), GFP_KERNEL);
96 if (!dev->dma_mem) 96 if (!dev->dma_mem)
97 goto out; 97 goto out;
98 memset(dev->dma_mem, 0, sizeof(struct dma_coherent_mem)); 98 dev->dma_mem->bitmap = kzalloc(bitmap_size, GFP_KERNEL);
99 dev->dma_mem->bitmap = kmalloc(bitmap_size, GFP_KERNEL);
100 if (!dev->dma_mem->bitmap) 99 if (!dev->dma_mem->bitmap)
101 goto free1_out; 100 goto free1_out;
102 memset(dev->dma_mem->bitmap, 0, bitmap_size);
103 101
104 dev->dma_mem->virt_base = mem_base; 102 dev->dma_mem->virt_base = mem_base;
105 dev->dma_mem->device_base = device_addr; 103 dev->dma_mem->device_base = device_addr;
diff --git a/arch/i386/kernel/process.c b/arch/i386/kernel/process.c
index dd53c58f64f1..99308510a17c 100644
--- a/arch/i386/kernel/process.c
+++ b/arch/i386/kernel/process.c
@@ -56,6 +56,7 @@
56 56
57#include <asm/tlbflush.h> 57#include <asm/tlbflush.h>
58#include <asm/cpu.h> 58#include <asm/cpu.h>
59#include <asm/pda.h>
59 60
60asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); 61asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
61 62
@@ -99,22 +100,18 @@ EXPORT_SYMBOL(enable_hlt);
99 */ 100 */
100void default_idle(void) 101void default_idle(void)
101{ 102{
102 local_irq_enable();
103
104 if (!hlt_counter && boot_cpu_data.hlt_works_ok) { 103 if (!hlt_counter && boot_cpu_data.hlt_works_ok) {
105 current_thread_info()->status &= ~TS_POLLING; 104 current_thread_info()->status &= ~TS_POLLING;
106 smp_mb__after_clear_bit(); 105 smp_mb__after_clear_bit();
107 while (!need_resched()) { 106 local_irq_disable();
108 local_irq_disable(); 107 if (!need_resched())
109 if (!need_resched()) 108 safe_halt(); /* enables interrupts racelessly */
110 safe_halt(); 109 else
111 else 110 local_irq_enable();
112 local_irq_enable();
113 }
114 current_thread_info()->status |= TS_POLLING; 111 current_thread_info()->status |= TS_POLLING;
115 } else { 112 } else {
116 while (!need_resched()) 113 /* loop is done by the caller */
117 cpu_relax(); 114 cpu_relax();
118 } 115 }
119} 116}
120#ifdef CONFIG_APM_MODULE 117#ifdef CONFIG_APM_MODULE
@@ -128,14 +125,7 @@ EXPORT_SYMBOL(default_idle);
128 */ 125 */
129static void poll_idle (void) 126static void poll_idle (void)
130{ 127{
131 local_irq_enable(); 128 cpu_relax();
132
133 asm volatile(
134 "2:"
135 "testl %0, %1;"
136 "rep; nop;"
137 "je 2b;"
138 : : "i"(_TIF_NEED_RESCHED), "m" (current_thread_info()->flags));
139} 129}
140 130
141#ifdef CONFIG_HOTPLUG_CPU 131#ifdef CONFIG_HOTPLUG_CPU
@@ -256,8 +246,7 @@ void mwait_idle_with_hints(unsigned long eax, unsigned long ecx)
256static void mwait_idle(void) 246static void mwait_idle(void)
257{ 247{
258 local_irq_enable(); 248 local_irq_enable();
259 while (!need_resched()) 249 mwait_idle_with_hints(0, 0);
260 mwait_idle_with_hints(0, 0);
261} 250}
262 251
263void __devinit select_idle_routine(const struct cpuinfo_x86 *c) 252void __devinit select_idle_routine(const struct cpuinfo_x86 *c)
@@ -314,8 +303,8 @@ void show_regs(struct pt_regs * regs)
314 regs->eax,regs->ebx,regs->ecx,regs->edx); 303 regs->eax,regs->ebx,regs->ecx,regs->edx);
315 printk("ESI: %08lx EDI: %08lx EBP: %08lx", 304 printk("ESI: %08lx EDI: %08lx EBP: %08lx",
316 regs->esi, regs->edi, regs->ebp); 305 regs->esi, regs->edi, regs->ebp);
317 printk(" DS: %04x ES: %04x\n", 306 printk(" DS: %04x ES: %04x GS: %04x\n",
318 0xffff & regs->xds,0xffff & regs->xes); 307 0xffff & regs->xds,0xffff & regs->xes, 0xffff & regs->xgs);
319 308
320 cr0 = read_cr0(); 309 cr0 = read_cr0();
321 cr2 = read_cr2(); 310 cr2 = read_cr2();
@@ -346,6 +335,7 @@ int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
346 335
347 regs.xds = __USER_DS; 336 regs.xds = __USER_DS;
348 regs.xes = __USER_DS; 337 regs.xes = __USER_DS;
338 regs.xgs = __KERNEL_PDA;
349 regs.orig_eax = -1; 339 regs.orig_eax = -1;
350 regs.eip = (unsigned long) kernel_thread_helper; 340 regs.eip = (unsigned long) kernel_thread_helper;
351 regs.xcs = __KERNEL_CS | get_kernel_rpl(); 341 regs.xcs = __KERNEL_CS | get_kernel_rpl();
@@ -431,7 +421,6 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
431 p->thread.eip = (unsigned long) ret_from_fork; 421 p->thread.eip = (unsigned long) ret_from_fork;
432 422
433 savesegment(fs,p->thread.fs); 423 savesegment(fs,p->thread.fs);
434 savesegment(gs,p->thread.gs);
435 424
436 tsk = current; 425 tsk = current;
437 if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { 426 if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
@@ -508,7 +497,7 @@ void dump_thread(struct pt_regs * regs, struct user * dump)
508 dump->regs.ds = regs->xds; 497 dump->regs.ds = regs->xds;
509 dump->regs.es = regs->xes; 498 dump->regs.es = regs->xes;
510 savesegment(fs,dump->regs.fs); 499 savesegment(fs,dump->regs.fs);
511 savesegment(gs,dump->regs.gs); 500 dump->regs.gs = regs->xgs;
512 dump->regs.orig_eax = regs->orig_eax; 501 dump->regs.orig_eax = regs->orig_eax;
513 dump->regs.eip = regs->eip; 502 dump->regs.eip = regs->eip;
514 dump->regs.cs = regs->xcs; 503 dump->regs.cs = regs->xcs;
@@ -648,22 +637,27 @@ struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct tas
648 637
649 __unlazy_fpu(prev_p); 638 __unlazy_fpu(prev_p);
650 639
640
641 /* we're going to use this soon, after a few expensive things */
642 if (next_p->fpu_counter > 5)
643 prefetch(&next->i387.fxsave);
644
651 /* 645 /*
652 * Reload esp0. 646 * Reload esp0.
653 */ 647 */
654 load_esp0(tss, next); 648 load_esp0(tss, next);
655 649
656 /* 650 /*
657 * Save away %fs and %gs. No need to save %es and %ds, as 651 * Save away %fs. No need to save %gs, as it was saved on the
658 * those are always kernel segments while inside the kernel. 652 * stack on entry. No need to save %es and %ds, as those are
659 * Doing this before setting the new TLS descriptors avoids 653 * always kernel segments while inside the kernel. Doing this
660 * the situation where we temporarily have non-reloadable 654 * before setting the new TLS descriptors avoids the situation
661 * segments in %fs and %gs. This could be an issue if the 655 * where we temporarily have non-reloadable segments in %fs
662 * NMI handler ever used %fs or %gs (it does not today), or 656 * and %gs. This could be an issue if the NMI handler ever
663 * if the kernel is running inside of a hypervisor layer. 657 * used %fs or %gs (it does not today), or if the kernel is
658 * running inside of a hypervisor layer.
664 */ 659 */
665 savesegment(fs, prev->fs); 660 savesegment(fs, prev->fs);
666 savesegment(gs, prev->gs);
667 661
668 /* 662 /*
669 * Load the per-thread Thread-Local Storage descriptor. 663 * Load the per-thread Thread-Local Storage descriptor.
@@ -671,22 +665,14 @@ struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct tas
671 load_TLS(next, cpu); 665 load_TLS(next, cpu);
672 666
673 /* 667 /*
674 * Restore %fs and %gs if needed. 668 * Restore %fs if needed.
675 * 669 *
676 * Glibc normally makes %fs be zero, and %gs is one of 670 * Glibc normally makes %fs be zero.
677 * the TLS segments.
678 */ 671 */
679 if (unlikely(prev->fs | next->fs)) 672 if (unlikely(prev->fs | next->fs))
680 loadsegment(fs, next->fs); 673 loadsegment(fs, next->fs);
681 674
682 if (prev->gs | next->gs) 675 write_pda(pcurrent, next_p);
683 loadsegment(gs, next->gs);
684
685 /*
686 * Restore IOPL if needed.
687 */
688 if (unlikely(prev->iopl != next->iopl))
689 set_iopl_mask(next->iopl);
690 676
691 /* 677 /*
692 * Now maybe handle debug registers and/or IO bitmaps 678 * Now maybe handle debug registers and/or IO bitmaps
@@ -697,6 +683,13 @@ struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct tas
697 683
698 disable_tsc(prev_p, next_p); 684 disable_tsc(prev_p, next_p);
699 685
686 /* If the task has used fpu the last 5 timeslices, just do a full
687 * restore of the math state immediately to avoid the trap; the
688 * chances of needing FPU soon are obviously high now
689 */
690 if (next_p->fpu_counter > 5)
691 math_state_restore();
692
700 return prev_p; 693 return prev_p;
701} 694}
702 695
diff --git a/arch/i386/kernel/ptrace.c b/arch/i386/kernel/ptrace.c
index 775f50e9395b..f3f94ac5736a 100644
--- a/arch/i386/kernel/ptrace.c
+++ b/arch/i386/kernel/ptrace.c
@@ -94,13 +94,9 @@ static int putreg(struct task_struct *child,
94 return -EIO; 94 return -EIO;
95 child->thread.fs = value; 95 child->thread.fs = value;
96 return 0; 96 return 0;
97 case GS:
98 if (value && (value & 3) != 3)
99 return -EIO;
100 child->thread.gs = value;
101 return 0;
102 case DS: 97 case DS:
103 case ES: 98 case ES:
99 case GS:
104 if (value && (value & 3) != 3) 100 if (value && (value & 3) != 3)
105 return -EIO; 101 return -EIO;
106 value &= 0xffff; 102 value &= 0xffff;
@@ -116,8 +112,8 @@ static int putreg(struct task_struct *child,
116 value |= get_stack_long(child, EFL_OFFSET) & ~FLAG_MASK; 112 value |= get_stack_long(child, EFL_OFFSET) & ~FLAG_MASK;
117 break; 113 break;
118 } 114 }
119 if (regno > GS*4) 115 if (regno > ES*4)
120 regno -= 2*4; 116 regno -= 1*4;
121 put_stack_long(child, regno - sizeof(struct pt_regs), value); 117 put_stack_long(child, regno - sizeof(struct pt_regs), value);
122 return 0; 118 return 0;
123} 119}
@@ -131,18 +127,16 @@ static unsigned long getreg(struct task_struct *child,
131 case FS: 127 case FS:
132 retval = child->thread.fs; 128 retval = child->thread.fs;
133 break; 129 break;
134 case GS:
135 retval = child->thread.gs;
136 break;
137 case DS: 130 case DS:
138 case ES: 131 case ES:
132 case GS:
139 case SS: 133 case SS:
140 case CS: 134 case CS:
141 retval = 0xffff; 135 retval = 0xffff;
142 /* fall through */ 136 /* fall through */
143 default: 137 default:
144 if (regno > GS*4) 138 if (regno > ES*4)
145 regno -= 2*4; 139 regno -= 1*4;
146 regno = regno - sizeof(struct pt_regs); 140 regno = regno - sizeof(struct pt_regs);
147 retval &= get_stack_long(child, regno); 141 retval &= get_stack_long(child, regno);
148 } 142 }
diff --git a/arch/i386/kernel/quirks.c b/arch/i386/kernel/quirks.c
index 9f6ab1789bb0..a01320a7b636 100644
--- a/arch/i386/kernel/quirks.c
+++ b/arch/i386/kernel/quirks.c
@@ -3,10 +3,23 @@
3 */ 3 */
4#include <linux/pci.h> 4#include <linux/pci.h>
5#include <linux/irq.h> 5#include <linux/irq.h>
6#include <asm/pci-direct.h>
7#include <asm/genapic.h>
8#include <asm/cpu.h>
6 9
7#if defined(CONFIG_X86_IO_APIC) && defined(CONFIG_SMP) && defined(CONFIG_PCI) 10#if defined(CONFIG_X86_IO_APIC) && defined(CONFIG_SMP) && defined(CONFIG_PCI)
11static void __devinit verify_quirk_intel_irqbalance(struct pci_dev *dev)
12{
13#ifdef CONFIG_X86_64
14 if (genapic != &apic_flat)
15 panic("APIC mode must be flat on this system\n");
16#elif defined(CONFIG_X86_GENERICARCH)
17 if (genapic != &apic_default)
18 panic("APIC mode must be default(flat) on this system. Use apic=default\n");
19#endif
20}
8 21
9static void __devinit quirk_intel_irqbalance(struct pci_dev *dev) 22void __init quirk_intel_irqbalance(void)
10{ 23{
11 u8 config, rev; 24 u8 config, rev;
12 u32 word; 25 u32 word;
@@ -16,18 +29,18 @@ static void __devinit quirk_intel_irqbalance(struct pci_dev *dev)
16 * based platforms. 29 * based platforms.
17 * Disable SW irqbalance/affinity on those platforms. 30 * Disable SW irqbalance/affinity on those platforms.
18 */ 31 */
19 pci_read_config_byte(dev, PCI_CLASS_REVISION, &rev); 32 rev = read_pci_config_byte(0, 0, 0, PCI_CLASS_REVISION);
20 if (rev > 0x9) 33 if (rev > 0x9)
21 return; 34 return;
22 35
23 printk(KERN_INFO "Intel E7520/7320/7525 detected."); 36 printk(KERN_INFO "Intel E7520/7320/7525 detected.");
24 37
25 /* enable access to config space*/ 38 /* enable access to config space */
26 pci_read_config_byte(dev, 0xf4, &config); 39 config = read_pci_config_byte(0, 0, 0, 0xf4);
27 pci_write_config_byte(dev, 0xf4, config|0x2); 40 write_pci_config_byte(0, 0, 0, 0xf4, config|0x2);
28 41
29 /* read xTPR register */ 42 /* read xTPR register */
30 raw_pci_ops->read(0, 0, 0x40, 0x4c, 2, &word); 43 word = read_pci_config_16(0, 0, 0x40, 0x4c);
31 44
32 if (!(word & (1 << 13))) { 45 if (!(word & (1 << 13))) {
33 printk(KERN_INFO "Disabling irq balancing and affinity\n"); 46 printk(KERN_INFO "Disabling irq balancing and affinity\n");
@@ -38,13 +51,24 @@ static void __devinit quirk_intel_irqbalance(struct pci_dev *dev)
38#ifdef CONFIG_PROC_FS 51#ifdef CONFIG_PROC_FS
39 no_irq_affinity = 1; 52 no_irq_affinity = 1;
40#endif 53#endif
54#ifdef CONFIG_HOTPLUG_CPU
55 printk(KERN_INFO "Disabling cpu hotplug control\n");
56 enable_cpu_hotplug = 0;
57#endif
58#ifdef CONFIG_X86_64
59 /* force the genapic selection to flat mode so that
60 * interrupts can be redirected to more than one CPU.
61 */
62 genapic_force = &apic_flat;
63#endif
41 } 64 }
42 65
43 /* put back the original value for config space*/ 66 /* put back the original value for config space */
44 if (!(config & 0x2)) 67 if (!(config & 0x2))
45 pci_write_config_byte(dev, 0xf4, config); 68 write_pci_config_byte(0, 0, 0, 0xf4, config);
46} 69}
47DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_E7320_MCH, quirk_intel_irqbalance); 70DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_E7320_MCH, verify_quirk_intel_irqbalance);
48DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_E7525_MCH, quirk_intel_irqbalance); 71DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_E7525_MCH, verify_quirk_intel_irqbalance);
49DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_E7520_MCH, quirk_intel_irqbalance); 72DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_E7520_MCH, verify_quirk_intel_irqbalance);
73
50#endif 74#endif
diff --git a/arch/i386/kernel/reboot.c b/arch/i386/kernel/reboot.c
index 84278e0093a2..3514b4153f7f 100644
--- a/arch/i386/kernel/reboot.c
+++ b/arch/i386/kernel/reboot.c
@@ -12,6 +12,7 @@
12#include <linux/dmi.h> 12#include <linux/dmi.h>
13#include <linux/ctype.h> 13#include <linux/ctype.h>
14#include <linux/pm.h> 14#include <linux/pm.h>
15#include <linux/reboot.h>
15#include <asm/uaccess.h> 16#include <asm/uaccess.h>
16#include <asm/apic.h> 17#include <asm/apic.h>
17#include <asm/desc.h> 18#include <asm/desc.h>
diff --git a/arch/i386/kernel/setup.c b/arch/i386/kernel/setup.c
index 141041dde74d..79df6e612dbd 100644
--- a/arch/i386/kernel/setup.c
+++ b/arch/i386/kernel/setup.c
@@ -63,9 +63,6 @@
63#include <setup_arch.h> 63#include <setup_arch.h>
64#include <bios_ebda.h> 64#include <bios_ebda.h>
65 65
66/* Forward Declaration. */
67void __init find_max_pfn(void);
68
69/* This value is set up by the early boot code to point to the value 66/* This value is set up by the early boot code to point to the value
70 immediately after the boot time page tables. It contains a *physical* 67 immediately after the boot time page tables. It contains a *physical*
71 address, and must not be in the .bss segment! */ 68 address, and must not be in the .bss segment! */
@@ -76,11 +73,8 @@ int disable_pse __devinitdata = 0;
76/* 73/*
77 * Machine setup.. 74 * Machine setup..
78 */ 75 */
79 76extern struct resource code_resource;
80#ifdef CONFIG_EFI 77extern struct resource data_resource;
81int efi_enabled = 0;
82EXPORT_SYMBOL(efi_enabled);
83#endif
84 78
85/* cpu data as detected by the assembly code in head.S */ 79/* cpu data as detected by the assembly code in head.S */
86struct cpuinfo_x86 new_cpu_data __initdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; 80struct cpuinfo_x86 new_cpu_data __initdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
@@ -99,12 +93,6 @@ unsigned int machine_submodel_id;
99unsigned int BIOS_revision; 93unsigned int BIOS_revision;
100unsigned int mca_pentium_flag; 94unsigned int mca_pentium_flag;
101 95
102/* For PCI or other memory-mapped resources */
103unsigned long pci_mem_start = 0x10000000;
104#ifdef CONFIG_PCI
105EXPORT_SYMBOL(pci_mem_start);
106#endif
107
108/* Boot loader ID as an integer, for the benefit of proc_dointvec */ 96/* Boot loader ID as an integer, for the benefit of proc_dointvec */
109int bootloader_type; 97int bootloader_type;
110 98
@@ -134,7 +122,6 @@ struct ist_info ist_info;
134 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) 122 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
135EXPORT_SYMBOL(ist_info); 123EXPORT_SYMBOL(ist_info);
136#endif 124#endif
137struct e820map e820;
138 125
139extern void early_cpu_init(void); 126extern void early_cpu_init(void);
140extern int root_mountflags; 127extern int root_mountflags;
@@ -149,516 +136,6 @@ static char command_line[COMMAND_LINE_SIZE];
149 136
150unsigned char __initdata boot_params[PARAM_SIZE]; 137unsigned char __initdata boot_params[PARAM_SIZE];
151 138
152static struct resource data_resource = {
153 .name = "Kernel data",
154 .start = 0,
155 .end = 0,
156 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
157};
158
159static struct resource code_resource = {
160 .name = "Kernel code",
161 .start = 0,
162 .end = 0,
163 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
164};
165
166static struct resource system_rom_resource = {
167 .name = "System ROM",
168 .start = 0xf0000,
169 .end = 0xfffff,
170 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
171};
172
173static struct resource extension_rom_resource = {
174 .name = "Extension ROM",
175 .start = 0xe0000,
176 .end = 0xeffff,
177 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
178};
179
180static struct resource adapter_rom_resources[] = { {
181 .name = "Adapter ROM",
182 .start = 0xc8000,
183 .end = 0,
184 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
185}, {
186 .name = "Adapter ROM",
187 .start = 0,
188 .end = 0,
189 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
190}, {
191 .name = "Adapter ROM",
192 .start = 0,
193 .end = 0,
194 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
195}, {
196 .name = "Adapter ROM",
197 .start = 0,
198 .end = 0,
199 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
200}, {
201 .name = "Adapter ROM",
202 .start = 0,
203 .end = 0,
204 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
205}, {
206 .name = "Adapter ROM",
207 .start = 0,
208 .end = 0,
209 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
210} };
211
212static struct resource video_rom_resource = {
213 .name = "Video ROM",
214 .start = 0xc0000,
215 .end = 0xc7fff,
216 .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM
217};
218
219static struct resource video_ram_resource = {
220 .name = "Video RAM area",
221 .start = 0xa0000,
222 .end = 0xbffff,
223 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
224};
225
226static struct resource standard_io_resources[] = { {
227 .name = "dma1",
228 .start = 0x0000,
229 .end = 0x001f,
230 .flags = IORESOURCE_BUSY | IORESOURCE_IO
231}, {
232 .name = "pic1",
233 .start = 0x0020,
234 .end = 0x0021,
235 .flags = IORESOURCE_BUSY | IORESOURCE_IO
236}, {
237 .name = "timer0",
238 .start = 0x0040,
239 .end = 0x0043,
240 .flags = IORESOURCE_BUSY | IORESOURCE_IO
241}, {
242 .name = "timer1",
243 .start = 0x0050,
244 .end = 0x0053,
245 .flags = IORESOURCE_BUSY | IORESOURCE_IO
246}, {
247 .name = "keyboard",
248 .start = 0x0060,
249 .end = 0x006f,
250 .flags = IORESOURCE_BUSY | IORESOURCE_IO
251}, {
252 .name = "dma page reg",
253 .start = 0x0080,
254 .end = 0x008f,
255 .flags = IORESOURCE_BUSY | IORESOURCE_IO
256}, {
257 .name = "pic2",
258 .start = 0x00a0,
259 .end = 0x00a1,
260 .flags = IORESOURCE_BUSY | IORESOURCE_IO
261}, {
262 .name = "dma2",
263 .start = 0x00c0,
264 .end = 0x00df,
265 .flags = IORESOURCE_BUSY | IORESOURCE_IO
266}, {
267 .name = "fpu",
268 .start = 0x00f0,
269 .end = 0x00ff,
270 .flags = IORESOURCE_BUSY | IORESOURCE_IO
271} };
272
273#define romsignature(x) (*(unsigned short *)(x) == 0xaa55)
274
275static int __init romchecksum(unsigned char *rom, unsigned long length)
276{
277 unsigned char *p, sum = 0;
278
279 for (p = rom; p < rom + length; p++)
280 sum += *p;
281 return sum == 0;
282}
283
284static void __init probe_roms(void)
285{
286 unsigned long start, length, upper;
287 unsigned char *rom;
288 int i;
289
290 /* video rom */
291 upper = adapter_rom_resources[0].start;
292 for (start = video_rom_resource.start; start < upper; start += 2048) {
293 rom = isa_bus_to_virt(start);
294 if (!romsignature(rom))
295 continue;
296
297 video_rom_resource.start = start;
298
299 /* 0 < length <= 0x7f * 512, historically */
300 length = rom[2] * 512;
301
302 /* if checksum okay, trust length byte */
303 if (length && romchecksum(rom, length))
304 video_rom_resource.end = start + length - 1;
305
306 request_resource(&iomem_resource, &video_rom_resource);
307 break;
308 }
309
310 start = (video_rom_resource.end + 1 + 2047) & ~2047UL;
311 if (start < upper)
312 start = upper;
313
314 /* system rom */
315 request_resource(&iomem_resource, &system_rom_resource);
316 upper = system_rom_resource.start;
317
318 /* check for extension rom (ignore length byte!) */
319 rom = isa_bus_to_virt(extension_rom_resource.start);
320 if (romsignature(rom)) {
321 length = extension_rom_resource.end - extension_rom_resource.start + 1;
322 if (romchecksum(rom, length)) {
323 request_resource(&iomem_resource, &extension_rom_resource);
324 upper = extension_rom_resource.start;
325 }
326 }
327
328 /* check for adapter roms on 2k boundaries */
329 for (i = 0; i < ARRAY_SIZE(adapter_rom_resources) && start < upper; start += 2048) {
330 rom = isa_bus_to_virt(start);
331 if (!romsignature(rom))
332 continue;
333
334 /* 0 < length <= 0x7f * 512, historically */
335 length = rom[2] * 512;
336
337 /* but accept any length that fits if checksum okay */
338 if (!length || start + length > upper || !romchecksum(rom, length))
339 continue;
340
341 adapter_rom_resources[i].start = start;
342 adapter_rom_resources[i].end = start + length - 1;
343 request_resource(&iomem_resource, &adapter_rom_resources[i]);
344
345 start = adapter_rom_resources[i++].end & ~2047UL;
346 }
347}
348
349static void __init limit_regions(unsigned long long size)
350{
351 unsigned long long current_addr = 0;
352 int i;
353
354 if (efi_enabled) {
355 efi_memory_desc_t *md;
356 void *p;
357
358 for (p = memmap.map, i = 0; p < memmap.map_end;
359 p += memmap.desc_size, i++) {
360 md = p;
361 current_addr = md->phys_addr + (md->num_pages << 12);
362 if (md->type == EFI_CONVENTIONAL_MEMORY) {
363 if (current_addr >= size) {
364 md->num_pages -=
365 (((current_addr-size) + PAGE_SIZE-1) >> PAGE_SHIFT);
366 memmap.nr_map = i + 1;
367 return;
368 }
369 }
370 }
371 }
372 for (i = 0; i < e820.nr_map; i++) {
373 current_addr = e820.map[i].addr + e820.map[i].size;
374 if (current_addr < size)
375 continue;
376
377 if (e820.map[i].type != E820_RAM)
378 continue;
379
380 if (e820.map[i].addr >= size) {
381 /*
382 * This region starts past the end of the
383 * requested size, skip it completely.
384 */
385 e820.nr_map = i;
386 } else {
387 e820.nr_map = i + 1;
388 e820.map[i].size -= current_addr - size;
389 }
390 return;
391 }
392}
393
394void __init add_memory_region(unsigned long long start,
395 unsigned long long size, int type)
396{
397 int x;
398
399 if (!efi_enabled) {
400 x = e820.nr_map;
401
402 if (x == E820MAX) {
403 printk(KERN_ERR "Ooops! Too many entries in the memory map!\n");
404 return;
405 }
406
407 e820.map[x].addr = start;
408 e820.map[x].size = size;
409 e820.map[x].type = type;
410 e820.nr_map++;
411 }
412} /* add_memory_region */
413
414#define E820_DEBUG 1
415
416static void __init print_memory_map(char *who)
417{
418 int i;
419
420 for (i = 0; i < e820.nr_map; i++) {
421 printk(" %s: %016Lx - %016Lx ", who,
422 e820.map[i].addr,
423 e820.map[i].addr + e820.map[i].size);
424 switch (e820.map[i].type) {
425 case E820_RAM: printk("(usable)\n");
426 break;
427 case E820_RESERVED:
428 printk("(reserved)\n");
429 break;
430 case E820_ACPI:
431 printk("(ACPI data)\n");
432 break;
433 case E820_NVS:
434 printk("(ACPI NVS)\n");
435 break;
436 default: printk("type %lu\n", e820.map[i].type);
437 break;
438 }
439 }
440}
441
442/*
443 * Sanitize the BIOS e820 map.
444 *
445 * Some e820 responses include overlapping entries. The following
446 * replaces the original e820 map with a new one, removing overlaps.
447 *
448 */
449struct change_member {
450 struct e820entry *pbios; /* pointer to original bios entry */
451 unsigned long long addr; /* address for this change point */
452};
453static struct change_member change_point_list[2*E820MAX] __initdata;
454static struct change_member *change_point[2*E820MAX] __initdata;
455static struct e820entry *overlap_list[E820MAX] __initdata;
456static struct e820entry new_bios[E820MAX] __initdata;
457
458int __init sanitize_e820_map(struct e820entry * biosmap, char * pnr_map)
459{
460 struct change_member *change_tmp;
461 unsigned long current_type, last_type;
462 unsigned long long last_addr;
463 int chgidx, still_changing;
464 int overlap_entries;
465 int new_bios_entry;
466 int old_nr, new_nr, chg_nr;
467 int i;
468
469 /*
470 Visually we're performing the following (1,2,3,4 = memory types)...
471
472 Sample memory map (w/overlaps):
473 ____22__________________
474 ______________________4_
475 ____1111________________
476 _44_____________________
477 11111111________________
478 ____________________33__
479 ___________44___________
480 __________33333_________
481 ______________22________
482 ___________________2222_
483 _________111111111______
484 _____________________11_
485 _________________4______
486
487 Sanitized equivalent (no overlap):
488 1_______________________
489 _44_____________________
490 ___1____________________
491 ____22__________________
492 ______11________________
493 _________1______________
494 __________3_____________
495 ___________44___________
496 _____________33_________
497 _______________2________
498 ________________1_______
499 _________________4______
500 ___________________2____
501 ____________________33__
502 ______________________4_
503 */
504
505 /* if there's only one memory region, don't bother */
506 if (*pnr_map < 2)
507 return -1;
508
509 old_nr = *pnr_map;
510
511 /* bail out if we find any unreasonable addresses in bios map */
512 for (i=0; i<old_nr; i++)
513 if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr)
514 return -1;
515
516 /* create pointers for initial change-point information (for sorting) */
517 for (i=0; i < 2*old_nr; i++)
518 change_point[i] = &change_point_list[i];
519
520 /* record all known change-points (starting and ending addresses),
521 omitting those that are for empty memory regions */
522 chgidx = 0;
523 for (i=0; i < old_nr; i++) {
524 if (biosmap[i].size != 0) {
525 change_point[chgidx]->addr = biosmap[i].addr;
526 change_point[chgidx++]->pbios = &biosmap[i];
527 change_point[chgidx]->addr = biosmap[i].addr + biosmap[i].size;
528 change_point[chgidx++]->pbios = &biosmap[i];
529 }
530 }
531 chg_nr = chgidx; /* true number of change-points */
532
533 /* sort change-point list by memory addresses (low -> high) */
534 still_changing = 1;
535 while (still_changing) {
536 still_changing = 0;
537 for (i=1; i < chg_nr; i++) {
538 /* if <current_addr> > <last_addr>, swap */
539 /* or, if current=<start_addr> & last=<end_addr>, swap */
540 if ((change_point[i]->addr < change_point[i-1]->addr) ||
541 ((change_point[i]->addr == change_point[i-1]->addr) &&
542 (change_point[i]->addr == change_point[i]->pbios->addr) &&
543 (change_point[i-1]->addr != change_point[i-1]->pbios->addr))
544 )
545 {
546 change_tmp = change_point[i];
547 change_point[i] = change_point[i-1];
548 change_point[i-1] = change_tmp;
549 still_changing=1;
550 }
551 }
552 }
553
554 /* create a new bios memory map, removing overlaps */
555 overlap_entries=0; /* number of entries in the overlap table */
556 new_bios_entry=0; /* index for creating new bios map entries */
557 last_type = 0; /* start with undefined memory type */
558 last_addr = 0; /* start with 0 as last starting address */
559 /* loop through change-points, determining affect on the new bios map */
560 for (chgidx=0; chgidx < chg_nr; chgidx++)
561 {
562 /* keep track of all overlapping bios entries */
563 if (change_point[chgidx]->addr == change_point[chgidx]->pbios->addr)
564 {
565 /* add map entry to overlap list (> 1 entry implies an overlap) */
566 overlap_list[overlap_entries++]=change_point[chgidx]->pbios;
567 }
568 else
569 {
570 /* remove entry from list (order independent, so swap with last) */
571 for (i=0; i<overlap_entries; i++)
572 {
573 if (overlap_list[i] == change_point[chgidx]->pbios)
574 overlap_list[i] = overlap_list[overlap_entries-1];
575 }
576 overlap_entries--;
577 }
578 /* if there are overlapping entries, decide which "type" to use */
579 /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */
580 current_type = 0;
581 for (i=0; i<overlap_entries; i++)
582 if (overlap_list[i]->type > current_type)
583 current_type = overlap_list[i]->type;
584 /* continue building up new bios map based on this information */
585 if (current_type != last_type) {
586 if (last_type != 0) {
587 new_bios[new_bios_entry].size =
588 change_point[chgidx]->addr - last_addr;
589 /* move forward only if the new size was non-zero */
590 if (new_bios[new_bios_entry].size != 0)
591 if (++new_bios_entry >= E820MAX)
592 break; /* no more space left for new bios entries */
593 }
594 if (current_type != 0) {
595 new_bios[new_bios_entry].addr = change_point[chgidx]->addr;
596 new_bios[new_bios_entry].type = current_type;
597 last_addr=change_point[chgidx]->addr;
598 }
599 last_type = current_type;
600 }
601 }
602 new_nr = new_bios_entry; /* retain count for new bios entries */
603
604 /* copy new bios mapping into original location */
605 memcpy(biosmap, new_bios, new_nr*sizeof(struct e820entry));
606 *pnr_map = new_nr;
607
608 return 0;
609}
610
611/*
612 * Copy the BIOS e820 map into a safe place.
613 *
614 * Sanity-check it while we're at it..
615 *
616 * If we're lucky and live on a modern system, the setup code
617 * will have given us a memory map that we can use to properly
618 * set up memory. If we aren't, we'll fake a memory map.
619 *
620 * We check to see that the memory map contains at least 2 elements
621 * before we'll use it, because the detection code in setup.S may
622 * not be perfect and most every PC known to man has two memory
623 * regions: one from 0 to 640k, and one from 1mb up. (The IBM
624 * thinkpad 560x, for example, does not cooperate with the memory
625 * detection code.)
626 */
627int __init copy_e820_map(struct e820entry * biosmap, int nr_map)
628{
629 /* Only one memory region (or negative)? Ignore it */
630 if (nr_map < 2)
631 return -1;
632
633 do {
634 unsigned long long start = biosmap->addr;
635 unsigned long long size = biosmap->size;
636 unsigned long long end = start + size;
637 unsigned long type = biosmap->type;
638
639 /* Overflow in 64 bits? Ignore the memory map. */
640 if (start > end)
641 return -1;
642
643 /*
644 * Some BIOSes claim RAM in the 640k - 1M region.
645 * Not right. Fix it up.
646 */
647 if (type == E820_RAM) {
648 if (start < 0x100000ULL && end > 0xA0000ULL) {
649 if (start < 0xA0000ULL)
650 add_memory_region(start, 0xA0000ULL-start, type);
651 if (end <= 0x100000ULL)
652 continue;
653 start = 0x100000ULL;
654 size = end - start;
655 }
656 }
657 add_memory_region(start, size, type);
658 } while (biosmap++,--nr_map);
659 return 0;
660}
661
662#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) 139#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
663struct edd edd; 140struct edd edd;
664#ifdef CONFIG_EDD_MODULE 141#ifdef CONFIG_EDD_MODULE
@@ -682,7 +159,7 @@ static inline void copy_edd(void)
682} 159}
683#endif 160#endif
684 161
685static int __initdata user_defined_memmap = 0; 162int __initdata user_defined_memmap = 0;
686 163
687/* 164/*
688 * "mem=nopentium" disables the 4MB page tables. 165 * "mem=nopentium" disables the 4MB page tables.
@@ -719,51 +196,6 @@ static int __init parse_mem(char *arg)
719} 196}
720early_param("mem", parse_mem); 197early_param("mem", parse_mem);
721 198
722static int __init parse_memmap(char *arg)
723{
724 if (!arg)
725 return -EINVAL;
726
727 if (strcmp(arg, "exactmap") == 0) {
728#ifdef CONFIG_CRASH_DUMP
729 /* If we are doing a crash dump, we
730 * still need to know the real mem
731 * size before original memory map is
732 * reset.
733 */
734 find_max_pfn();
735 saved_max_pfn = max_pfn;
736#endif
737 e820.nr_map = 0;
738 user_defined_memmap = 1;
739 } else {
740 /* If the user specifies memory size, we
741 * limit the BIOS-provided memory map to
742 * that size. exactmap can be used to specify
743 * the exact map. mem=number can be used to
744 * trim the existing memory map.
745 */
746 unsigned long long start_at, mem_size;
747
748 mem_size = memparse(arg, &arg);
749 if (*arg == '@') {
750 start_at = memparse(arg+1, &arg);
751 add_memory_region(start_at, mem_size, E820_RAM);
752 } else if (*arg == '#') {
753 start_at = memparse(arg+1, &arg);
754 add_memory_region(start_at, mem_size, E820_ACPI);
755 } else if (*arg == '$') {
756 start_at = memparse(arg+1, &arg);
757 add_memory_region(start_at, mem_size, E820_RESERVED);
758 } else {
759 limit_regions(mem_size);
760 user_defined_memmap = 1;
761 }
762 }
763 return 0;
764}
765early_param("memmap", parse_memmap);
766
767#ifdef CONFIG_PROC_VMCORE 199#ifdef CONFIG_PROC_VMCORE
768/* elfcorehdr= specifies the location of elf core header 200/* elfcorehdr= specifies the location of elf core header
769 * stored by the crashed kernel. 201 * stored by the crashed kernel.
@@ -828,90 +260,6 @@ static int __init parse_reservetop(char *arg)
828early_param("reservetop", parse_reservetop); 260early_param("reservetop", parse_reservetop);
829 261
830/* 262/*
831 * Callback for efi_memory_walk.
832 */
833static int __init
834efi_find_max_pfn(unsigned long start, unsigned long end, void *arg)
835{
836 unsigned long *max_pfn = arg, pfn;
837
838 if (start < end) {
839 pfn = PFN_UP(end -1);
840 if (pfn > *max_pfn)
841 *max_pfn = pfn;
842 }
843 return 0;
844}
845
846static int __init
847efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg)
848{
849 memory_present(0, PFN_UP(start), PFN_DOWN(end));
850 return 0;
851}
852
853 /*
854 * This function checks if the entire range <start,end> is mapped with type.
855 *
856 * Note: this function only works correct if the e820 table is sorted and
857 * not-overlapping, which is the case
858 */
859int __init
860e820_all_mapped(unsigned long s, unsigned long e, unsigned type)
861{
862 u64 start = s;
863 u64 end = e;
864 int i;
865 for (i = 0; i < e820.nr_map; i++) {
866 struct e820entry *ei = &e820.map[i];
867 if (type && ei->type != type)
868 continue;
869 /* is the region (part) in overlap with the current region ?*/
870 if (ei->addr >= end || ei->addr + ei->size <= start)
871 continue;
872 /* if the region is at the beginning of <start,end> we move
873 * start to the end of the region since it's ok until there
874 */
875 if (ei->addr <= start)
876 start = ei->addr + ei->size;
877 /* if start is now at or beyond end, we're done, full
878 * coverage */
879 if (start >= end)
880 return 1; /* we're done */
881 }
882 return 0;
883}
884
885/*
886 * Find the highest page frame number we have available
887 */
888void __init find_max_pfn(void)
889{
890 int i;
891
892 max_pfn = 0;
893 if (efi_enabled) {
894 efi_memmap_walk(efi_find_max_pfn, &max_pfn);
895 efi_memmap_walk(efi_memory_present_wrapper, NULL);
896 return;
897 }
898
899 for (i = 0; i < e820.nr_map; i++) {
900 unsigned long start, end;
901 /* RAM? */
902 if (e820.map[i].type != E820_RAM)
903 continue;
904 start = PFN_UP(e820.map[i].addr);
905 end = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
906 if (start >= end)
907 continue;
908 if (end > max_pfn)
909 max_pfn = end;
910 memory_present(0, start, end);
911 }
912}
913
914/*
915 * Determine low and high memory ranges: 263 * Determine low and high memory ranges:
916 */ 264 */
917unsigned long __init find_max_low_pfn(void) 265unsigned long __init find_max_low_pfn(void)
@@ -971,68 +319,6 @@ unsigned long __init find_max_low_pfn(void)
971} 319}
972 320
973/* 321/*
974 * Free all available memory for boot time allocation. Used
975 * as a callback function by efi_memory_walk()
976 */
977
978static int __init
979free_available_memory(unsigned long start, unsigned long end, void *arg)
980{
981 /* check max_low_pfn */
982 if (start >= (max_low_pfn << PAGE_SHIFT))
983 return 0;
984 if (end >= (max_low_pfn << PAGE_SHIFT))
985 end = max_low_pfn << PAGE_SHIFT;
986 if (start < end)
987 free_bootmem(start, end - start);
988
989 return 0;
990}
991/*
992 * Register fully available low RAM pages with the bootmem allocator.
993 */
994static void __init register_bootmem_low_pages(unsigned long max_low_pfn)
995{
996 int i;
997
998 if (efi_enabled) {
999 efi_memmap_walk(free_available_memory, NULL);
1000 return;
1001 }
1002 for (i = 0; i < e820.nr_map; i++) {
1003 unsigned long curr_pfn, last_pfn, size;
1004 /*
1005 * Reserve usable low memory
1006 */
1007 if (e820.map[i].type != E820_RAM)
1008 continue;
1009 /*
1010 * We are rounding up the start address of usable memory:
1011 */
1012 curr_pfn = PFN_UP(e820.map[i].addr);
1013 if (curr_pfn >= max_low_pfn)
1014 continue;
1015 /*
1016 * ... and at the end of the usable range downwards:
1017 */
1018 last_pfn = PFN_DOWN(e820.map[i].addr + e820.map[i].size);
1019
1020 if (last_pfn > max_low_pfn)
1021 last_pfn = max_low_pfn;
1022
1023 /*
1024 * .. finally, did all the rounding and playing
1025 * around just make the area go away?
1026 */
1027 if (last_pfn <= curr_pfn)
1028 continue;
1029
1030 size = last_pfn - curr_pfn;
1031 free_bootmem(PFN_PHYS(curr_pfn), PFN_PHYS(size));
1032 }
1033}
1034
1035/*
1036 * workaround for Dell systems that neglect to reserve EBDA 322 * workaround for Dell systems that neglect to reserve EBDA
1037 */ 323 */
1038static void __init reserve_ebda_region(void) 324static void __init reserve_ebda_region(void)
@@ -1118,8 +404,8 @@ void __init setup_bootmem_allocator(void)
1118 * the (very unlikely) case of us accidentally initializing the 404 * the (very unlikely) case of us accidentally initializing the
1119 * bootmem allocator with an invalid RAM area. 405 * bootmem allocator with an invalid RAM area.
1120 */ 406 */
1121 reserve_bootmem(__PHYSICAL_START, (PFN_PHYS(min_low_pfn) + 407 reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) +
1122 bootmap_size + PAGE_SIZE-1) - (__PHYSICAL_START)); 408 bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text));
1123 409
1124 /* 410 /*
1125 * reserve physical page 0 - it's a special BIOS page on many boxes, 411 * reserve physical page 0 - it's a special BIOS page on many boxes,
@@ -1162,8 +448,7 @@ void __init setup_bootmem_allocator(void)
1162 if (LOADER_TYPE && INITRD_START) { 448 if (LOADER_TYPE && INITRD_START) {
1163 if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) { 449 if (INITRD_START + INITRD_SIZE <= (max_low_pfn << PAGE_SHIFT)) {
1164 reserve_bootmem(INITRD_START, INITRD_SIZE); 450 reserve_bootmem(INITRD_START, INITRD_SIZE);
1165 initrd_start = 451 initrd_start = INITRD_START + PAGE_OFFSET;
1166 INITRD_START ? INITRD_START + PAGE_OFFSET : 0;
1167 initrd_end = initrd_start+INITRD_SIZE; 452 initrd_end = initrd_start+INITRD_SIZE;
1168 } 453 }
1169 else { 454 else {
@@ -1200,126 +485,6 @@ void __init remapped_pgdat_init(void)
1200 } 485 }
1201} 486}
1202 487
1203/*
1204 * Request address space for all standard RAM and ROM resources
1205 * and also for regions reported as reserved by the e820.
1206 */
1207static void __init
1208legacy_init_iomem_resources(struct resource *code_resource, struct resource *data_resource)
1209{
1210 int i;
1211
1212 probe_roms();
1213 for (i = 0; i < e820.nr_map; i++) {
1214 struct resource *res;
1215#ifndef CONFIG_RESOURCES_64BIT
1216 if (e820.map[i].addr + e820.map[i].size > 0x100000000ULL)
1217 continue;
1218#endif
1219 res = kzalloc(sizeof(struct resource), GFP_ATOMIC);
1220 switch (e820.map[i].type) {
1221 case E820_RAM: res->name = "System RAM"; break;
1222 case E820_ACPI: res->name = "ACPI Tables"; break;
1223 case E820_NVS: res->name = "ACPI Non-volatile Storage"; break;
1224 default: res->name = "reserved";
1225 }
1226 res->start = e820.map[i].addr;
1227 res->end = res->start + e820.map[i].size - 1;
1228 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
1229 if (request_resource(&iomem_resource, res)) {
1230 kfree(res);
1231 continue;
1232 }
1233 if (e820.map[i].type == E820_RAM) {
1234 /*
1235 * We don't know which RAM region contains kernel data,
1236 * so we try it repeatedly and let the resource manager
1237 * test it.
1238 */
1239 request_resource(res, code_resource);
1240 request_resource(res, data_resource);
1241#ifdef CONFIG_KEXEC
1242 request_resource(res, &crashk_res);
1243#endif
1244 }
1245 }
1246}
1247
1248/*
1249 * Request address space for all standard resources
1250 *
1251 * This is called just before pcibios_init(), which is also a
1252 * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
1253 */
1254static int __init request_standard_resources(void)
1255{
1256 int i;
1257
1258 printk("Setting up standard PCI resources\n");
1259 if (efi_enabled)
1260 efi_initialize_iomem_resources(&code_resource, &data_resource);
1261 else
1262 legacy_init_iomem_resources(&code_resource, &data_resource);
1263
1264 /* EFI systems may still have VGA */
1265 request_resource(&iomem_resource, &video_ram_resource);
1266
1267 /* request I/O space for devices used on all i[345]86 PCs */
1268 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
1269 request_resource(&ioport_resource, &standard_io_resources[i]);
1270 return 0;
1271}
1272
1273subsys_initcall(request_standard_resources);
1274
1275static void __init register_memory(void)
1276{
1277 unsigned long gapstart, gapsize, round;
1278 unsigned long long last;
1279 int i;
1280
1281 /*
1282 * Search for the bigest gap in the low 32 bits of the e820
1283 * memory space.
1284 */
1285 last = 0x100000000ull;
1286 gapstart = 0x10000000;
1287 gapsize = 0x400000;
1288 i = e820.nr_map;
1289 while (--i >= 0) {
1290 unsigned long long start = e820.map[i].addr;
1291 unsigned long long end = start + e820.map[i].size;
1292
1293 /*
1294 * Since "last" is at most 4GB, we know we'll
1295 * fit in 32 bits if this condition is true
1296 */
1297 if (last > end) {
1298 unsigned long gap = last - end;
1299
1300 if (gap > gapsize) {
1301 gapsize = gap;
1302 gapstart = end;
1303 }
1304 }
1305 if (start < last)
1306 last = start;
1307 }
1308
1309 /*
1310 * See how much we want to round up: start off with
1311 * rounding to the next 1MB area.
1312 */
1313 round = 0x100000;
1314 while ((gapsize >> 4) > round)
1315 round += round;
1316 /* Fun with two's complement */
1317 pci_mem_start = (gapstart + round) & -round;
1318
1319 printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n",
1320 pci_mem_start, gapstart, gapsize);
1321}
1322
1323#ifdef CONFIG_MCA 488#ifdef CONFIG_MCA
1324static void set_mca_bus(int x) 489static void set_mca_bus(int x)
1325{ 490{
@@ -1329,6 +494,12 @@ static void set_mca_bus(int x)
1329static void set_mca_bus(int x) { } 494static void set_mca_bus(int x) { }
1330#endif 495#endif
1331 496
497/* Overridden in paravirt.c if CONFIG_PARAVIRT */
498char * __attribute__((weak)) memory_setup(void)
499{
500 return machine_specific_memory_setup();
501}
502
1332/* 503/*
1333 * Determine if we were loaded by an EFI loader. If so, then we have also been 504 * Determine if we were loaded by an EFI loader. If so, then we have also been
1334 * passed the efi memmap, systab, etc., so we should use these data structures 505 * passed the efi memmap, systab, etc., so we should use these data structures
@@ -1381,7 +552,7 @@ void __init setup_arch(char **cmdline_p)
1381 efi_init(); 552 efi_init();
1382 else { 553 else {
1383 printk(KERN_INFO "BIOS-provided physical RAM map:\n"); 554 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
1384 print_memory_map(machine_specific_memory_setup()); 555 print_memory_map(memory_setup());
1385 } 556 }
1386 557
1387 copy_edd(); 558 copy_edd();
diff --git a/arch/i386/kernel/signal.c b/arch/i386/kernel/signal.c
index 43002cfb40c4..65d7620eaa09 100644
--- a/arch/i386/kernel/signal.c
+++ b/arch/i386/kernel/signal.c
@@ -128,7 +128,7 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, int *peax
128 X86_EFLAGS_TF | X86_EFLAGS_SF | X86_EFLAGS_ZF | \ 128 X86_EFLAGS_TF | X86_EFLAGS_SF | X86_EFLAGS_ZF | \
129 X86_EFLAGS_AF | X86_EFLAGS_PF | X86_EFLAGS_CF) 129 X86_EFLAGS_AF | X86_EFLAGS_PF | X86_EFLAGS_CF)
130 130
131 GET_SEG(gs); 131 COPY_SEG(gs);
132 GET_SEG(fs); 132 GET_SEG(fs);
133 COPY_SEG(es); 133 COPY_SEG(es);
134 COPY_SEG(ds); 134 COPY_SEG(ds);
@@ -244,9 +244,7 @@ setup_sigcontext(struct sigcontext __user *sc, struct _fpstate __user *fpstate,
244{ 244{
245 int tmp, err = 0; 245 int tmp, err = 0;
246 246
247 tmp = 0; 247 err |= __put_user(regs->xgs, (unsigned int __user *)&sc->gs);
248 savesegment(gs, tmp);
249 err |= __put_user(tmp, (unsigned int __user *)&sc->gs);
250 savesegment(fs, tmp); 248 savesegment(fs, tmp);
251 err |= __put_user(tmp, (unsigned int __user *)&sc->fs); 249 err |= __put_user(tmp, (unsigned int __user *)&sc->fs);
252 250
diff --git a/arch/i386/kernel/smp.c b/arch/i386/kernel/smp.c
index 31e5c6573aae..5285aff8367f 100644
--- a/arch/i386/kernel/smp.c
+++ b/arch/i386/kernel/smp.c
@@ -321,7 +321,6 @@ static inline void leave_mm (unsigned long cpu)
321 321
322fastcall void smp_invalidate_interrupt(struct pt_regs *regs) 322fastcall void smp_invalidate_interrupt(struct pt_regs *regs)
323{ 323{
324 struct pt_regs *old_regs = set_irq_regs(regs);
325 unsigned long cpu; 324 unsigned long cpu;
326 325
327 cpu = get_cpu(); 326 cpu = get_cpu();
@@ -352,7 +351,6 @@ fastcall void smp_invalidate_interrupt(struct pt_regs *regs)
352 smp_mb__after_clear_bit(); 351 smp_mb__after_clear_bit();
353out: 352out:
354 put_cpu_no_resched(); 353 put_cpu_no_resched();
355 set_irq_regs(old_regs);
356} 354}
357 355
358static void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm, 356static void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm,
@@ -607,14 +605,11 @@ void smp_send_stop(void)
607 */ 605 */
608fastcall void smp_reschedule_interrupt(struct pt_regs *regs) 606fastcall void smp_reschedule_interrupt(struct pt_regs *regs)
609{ 607{
610 struct pt_regs *old_regs = set_irq_regs(regs);
611 ack_APIC_irq(); 608 ack_APIC_irq();
612 set_irq_regs(old_regs);
613} 609}
614 610
615fastcall void smp_call_function_interrupt(struct pt_regs *regs) 611fastcall void smp_call_function_interrupt(struct pt_regs *regs)
616{ 612{
617 struct pt_regs *old_regs = set_irq_regs(regs);
618 void (*func) (void *info) = call_data->func; 613 void (*func) (void *info) = call_data->func;
619 void *info = call_data->info; 614 void *info = call_data->info;
620 int wait = call_data->wait; 615 int wait = call_data->wait;
@@ -637,7 +632,6 @@ fastcall void smp_call_function_interrupt(struct pt_regs *regs)
637 mb(); 632 mb();
638 atomic_inc(&call_data->finished); 633 atomic_inc(&call_data->finished);
639 } 634 }
640 set_irq_regs(old_regs);
641} 635}
642 636
643/* 637/*
@@ -699,6 +693,10 @@ int smp_call_function_single(int cpu, void (*func) (void *info), void *info,
699 put_cpu(); 693 put_cpu();
700 return -EBUSY; 694 return -EBUSY;
701 } 695 }
696
697 /* Can deadlock when called with interrupts disabled */
698 WARN_ON(irqs_disabled());
699
702 spin_lock_bh(&call_lock); 700 spin_lock_bh(&call_lock);
703 __smp_call_function_single(cpu, func, info, nonatomic, wait); 701 __smp_call_function_single(cpu, func, info, nonatomic, wait);
704 spin_unlock_bh(&call_lock); 702 spin_unlock_bh(&call_lock);
diff --git a/arch/i386/kernel/smpboot.c b/arch/i386/kernel/smpboot.c
index 02a9b66b6ac3..4bf0e3c83b8b 100644
--- a/arch/i386/kernel/smpboot.c
+++ b/arch/i386/kernel/smpboot.c
@@ -33,6 +33,11 @@
33 * Dave Jones : Report invalid combinations of Athlon CPUs. 33 * Dave Jones : Report invalid combinations of Athlon CPUs.
34* Rusty Russell : Hacked into shape for new "hotplug" boot process. */ 34* Rusty Russell : Hacked into shape for new "hotplug" boot process. */
35 35
36
37/* SMP boot always wants to use real time delay to allow sufficient time for
38 * the APs to come online */
39#define USE_REAL_TIME_DELAY
40
36#include <linux/module.h> 41#include <linux/module.h>
37#include <linux/init.h> 42#include <linux/init.h>
38#include <linux/kernel.h> 43#include <linux/kernel.h>
@@ -52,6 +57,8 @@
52#include <asm/desc.h> 57#include <asm/desc.h>
53#include <asm/arch_hooks.h> 58#include <asm/arch_hooks.h>
54#include <asm/nmi.h> 59#include <asm/nmi.h>
60#include <asm/pda.h>
61#include <asm/genapic.h>
55 62
56#include <mach_apic.h> 63#include <mach_apic.h>
57#include <mach_wakecpu.h> 64#include <mach_wakecpu.h>
@@ -536,11 +543,11 @@ set_cpu_sibling_map(int cpu)
536static void __devinit start_secondary(void *unused) 543static void __devinit start_secondary(void *unused)
537{ 544{
538 /* 545 /*
539 * Dont put anything before smp_callin(), SMP 546 * Don't put *anything* before secondary_cpu_init(), SMP
540 * booting is too fragile that we want to limit the 547 * booting is too fragile that we want to limit the
541 * things done here to the most necessary things. 548 * things done here to the most necessary things.
542 */ 549 */
543 cpu_init(); 550 secondary_cpu_init();
544 preempt_disable(); 551 preempt_disable();
545 smp_callin(); 552 smp_callin();
546 while (!cpu_isset(smp_processor_id(), smp_commenced_mask)) 553 while (!cpu_isset(smp_processor_id(), smp_commenced_mask))
@@ -599,13 +606,16 @@ void __devinit initialize_secondary(void)
599 "movl %0,%%esp\n\t" 606 "movl %0,%%esp\n\t"
600 "jmp *%1" 607 "jmp *%1"
601 : 608 :
602 :"r" (current->thread.esp),"r" (current->thread.eip)); 609 :"m" (current->thread.esp),"m" (current->thread.eip));
603} 610}
604 611
612/* Static state in head.S used to set up a CPU */
605extern struct { 613extern struct {
606 void * esp; 614 void * esp;
607 unsigned short ss; 615 unsigned short ss;
608} stack_start; 616} stack_start;
617extern struct i386_pda *start_pda;
618extern struct Xgt_desc_struct cpu_gdt_descr;
609 619
610#ifdef CONFIG_NUMA 620#ifdef CONFIG_NUMA
611 621
@@ -936,9 +946,6 @@ static int __devinit do_boot_cpu(int apicid, int cpu)
936 unsigned long start_eip; 946 unsigned long start_eip;
937 unsigned short nmi_high = 0, nmi_low = 0; 947 unsigned short nmi_high = 0, nmi_low = 0;
938 948
939 ++cpucount;
940 alternatives_smp_switch(1);
941
942 /* 949 /*
943 * We can't use kernel_thread since we must avoid to 950 * We can't use kernel_thread since we must avoid to
944 * reschedule the child. 951 * reschedule the child.
@@ -946,15 +953,30 @@ static int __devinit do_boot_cpu(int apicid, int cpu)
946 idle = alloc_idle_task(cpu); 953 idle = alloc_idle_task(cpu);
947 if (IS_ERR(idle)) 954 if (IS_ERR(idle))
948 panic("failed fork for CPU %d", cpu); 955 panic("failed fork for CPU %d", cpu);
956
957 /* Pre-allocate and initialize the CPU's GDT and PDA so it
958 doesn't have to do any memory allocation during the
959 delicate CPU-bringup phase. */
960 if (!init_gdt(cpu, idle)) {
961 printk(KERN_INFO "Couldn't allocate GDT/PDA for CPU %d\n", cpu);
962 return -1; /* ? */
963 }
964
949 idle->thread.eip = (unsigned long) start_secondary; 965 idle->thread.eip = (unsigned long) start_secondary;
950 /* start_eip had better be page-aligned! */ 966 /* start_eip had better be page-aligned! */
951 start_eip = setup_trampoline(); 967 start_eip = setup_trampoline();
952 968
969 ++cpucount;
970 alternatives_smp_switch(1);
971
953 /* So we see what's up */ 972 /* So we see what's up */
954 printk("Booting processor %d/%d eip %lx\n", cpu, apicid, start_eip); 973 printk("Booting processor %d/%d eip %lx\n", cpu, apicid, start_eip);
955 /* Stack for startup_32 can be just as for start_secondary onwards */ 974 /* Stack for startup_32 can be just as for start_secondary onwards */
956 stack_start.esp = (void *) idle->thread.esp; 975 stack_start.esp = (void *) idle->thread.esp;
957 976
977 start_pda = cpu_pda(cpu);
978 cpu_gdt_descr = per_cpu(cpu_gdt_descr, cpu);
979
958 irq_ctx_init(cpu); 980 irq_ctx_init(cpu);
959 981
960 x86_cpu_to_apicid[cpu] = apicid; 982 x86_cpu_to_apicid[cpu] = apicid;
@@ -1109,34 +1131,15 @@ exit:
1109} 1131}
1110#endif 1132#endif
1111 1133
1112static void smp_tune_scheduling (void) 1134static void smp_tune_scheduling(void)
1113{ 1135{
1114 unsigned long cachesize; /* kB */ 1136 unsigned long cachesize; /* kB */
1115 unsigned long bandwidth = 350; /* MB/s */
1116 /*
1117 * Rough estimation for SMP scheduling, this is the number of
1118 * cycles it takes for a fully memory-limited process to flush
1119 * the SMP-local cache.
1120 *
1121 * (For a P5 this pretty much means we will choose another idle
1122 * CPU almost always at wakeup time (this is due to the small
1123 * L1 cache), on PIIs it's around 50-100 usecs, depending on
1124 * the cache size)
1125 */
1126 1137
1127 if (!cpu_khz) { 1138 if (cpu_khz) {
1128 /*
1129 * this basically disables processor-affinity
1130 * scheduling on SMP without a TSC.
1131 */
1132 return;
1133 } else {
1134 cachesize = boot_cpu_data.x86_cache_size; 1139 cachesize = boot_cpu_data.x86_cache_size;
1135 if (cachesize == -1) { 1140
1136 cachesize = 16; /* Pentiums, 2x8kB cache */ 1141 if (cachesize > 0)
1137 bandwidth = 100; 1142 max_cache_size = cachesize * 1024;
1138 }
1139 max_cache_size = cachesize * 1024;
1140 } 1143 }
1141} 1144}
1142 1145
@@ -1462,6 +1465,12 @@ int __devinit __cpu_up(unsigned int cpu)
1462 cpu_set(cpu, smp_commenced_mask); 1465 cpu_set(cpu, smp_commenced_mask);
1463 while (!cpu_isset(cpu, cpu_online_map)) 1466 while (!cpu_isset(cpu, cpu_online_map))
1464 cpu_relax(); 1467 cpu_relax();
1468
1469#ifdef CONFIG_X86_GENERICARCH
1470 if (num_online_cpus() > 8 && genapic == &apic_default)
1471 panic("Default flat APIC routing can't be used with > 8 cpus\n");
1472#endif
1473
1465 return 0; 1474 return 0;
1466} 1475}
1467 1476
diff --git a/arch/i386/kernel/sysenter.c b/arch/i386/kernel/sysenter.c
index 713ba39d32c6..7de9117b5a3a 100644
--- a/arch/i386/kernel/sysenter.c
+++ b/arch/i386/kernel/sysenter.c
@@ -27,7 +27,11 @@
27 * Should the kernel map a VDSO page into processes and pass its 27 * Should the kernel map a VDSO page into processes and pass its
28 * address down to glibc upon exec()? 28 * address down to glibc upon exec()?
29 */ 29 */
30#ifdef CONFIG_PARAVIRT
31unsigned int __read_mostly vdso_enabled = 0;
32#else
30unsigned int __read_mostly vdso_enabled = 1; 33unsigned int __read_mostly vdso_enabled = 1;
34#endif
31 35
32EXPORT_SYMBOL_GPL(vdso_enabled); 36EXPORT_SYMBOL_GPL(vdso_enabled);
33 37
@@ -132,7 +136,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
132 goto up_fail; 136 goto up_fail;
133 } 137 }
134 138
135 vma = kmem_cache_zalloc(vm_area_cachep, SLAB_KERNEL); 139 vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
136 if (!vma) { 140 if (!vma) {
137 ret = -ENOMEM; 141 ret = -ENOMEM;
138 goto up_fail; 142 goto up_fail;
diff --git a/arch/i386/kernel/time.c b/arch/i386/kernel/time.c
index 78af572fd17c..c505b16c0990 100644
--- a/arch/i386/kernel/time.c
+++ b/arch/i386/kernel/time.c
@@ -56,6 +56,7 @@
56#include <asm/uaccess.h> 56#include <asm/uaccess.h>
57#include <asm/processor.h> 57#include <asm/processor.h>
58#include <asm/timer.h> 58#include <asm/timer.h>
59#include <asm/time.h>
59 60
60#include "mach_time.h" 61#include "mach_time.h"
61 62
@@ -116,10 +117,7 @@ static int set_rtc_mmss(unsigned long nowtime)
116 /* gets recalled with irq locally disabled */ 117 /* gets recalled with irq locally disabled */
117 /* XXX - does irqsave resolve this? -johnstul */ 118 /* XXX - does irqsave resolve this? -johnstul */
118 spin_lock_irqsave(&rtc_lock, flags); 119 spin_lock_irqsave(&rtc_lock, flags);
119 if (efi_enabled) 120 retval = set_wallclock(nowtime);
120 retval = efi_set_rtc_mmss(nowtime);
121 else
122 retval = mach_set_rtc_mmss(nowtime);
123 spin_unlock_irqrestore(&rtc_lock, flags); 121 spin_unlock_irqrestore(&rtc_lock, flags);
124 122
125 return retval; 123 return retval;
@@ -223,10 +221,7 @@ unsigned long get_cmos_time(void)
223 221
224 spin_lock_irqsave(&rtc_lock, flags); 222 spin_lock_irqsave(&rtc_lock, flags);
225 223
226 if (efi_enabled) 224 retval = get_wallclock();
227 retval = efi_get_time();
228 else
229 retval = mach_get_cmos_time();
230 225
231 spin_unlock_irqrestore(&rtc_lock, flags); 226 spin_unlock_irqrestore(&rtc_lock, flags);
232 227
@@ -370,7 +365,7 @@ static void __init hpet_time_init(void)
370 printk("Using HPET for base-timer\n"); 365 printk("Using HPET for base-timer\n");
371 } 366 }
372 367
373 time_init_hook(); 368 do_time_init();
374} 369}
375#endif 370#endif
376 371
@@ -392,5 +387,5 @@ void __init time_init(void)
392 387
393 do_settimeofday(&ts); 388 do_settimeofday(&ts);
394 389
395 time_init_hook(); 390 do_time_init();
396} 391}
diff --git a/arch/i386/kernel/time_hpet.c b/arch/i386/kernel/time_hpet.c
index 1a2a979cf6a3..1e4702dfcd01 100644
--- a/arch/i386/kernel/time_hpet.c
+++ b/arch/i386/kernel/time_hpet.c
@@ -132,14 +132,20 @@ int __init hpet_enable(void)
132 * the single HPET timer for system time. 132 * the single HPET timer for system time.
133 */ 133 */
134#ifdef CONFIG_HPET_EMULATE_RTC 134#ifdef CONFIG_HPET_EMULATE_RTC
135 if (!(id & HPET_ID_NUMBER)) 135 if (!(id & HPET_ID_NUMBER)) {
136 iounmap(hpet_virt_address);
137 hpet_virt_address = NULL;
136 return -1; 138 return -1;
139 }
137#endif 140#endif
138 141
139 142
140 hpet_period = hpet_readl(HPET_PERIOD); 143 hpet_period = hpet_readl(HPET_PERIOD);
141 if ((hpet_period < HPET_MIN_PERIOD) || (hpet_period > HPET_MAX_PERIOD)) 144 if ((hpet_period < HPET_MIN_PERIOD) || (hpet_period > HPET_MAX_PERIOD)) {
145 iounmap(hpet_virt_address);
146 hpet_virt_address = NULL;
142 return -1; 147 return -1;
148 }
143 149
144 /* 150 /*
145 * 64 bit math 151 * 64 bit math
@@ -156,8 +162,11 @@ int __init hpet_enable(void)
156 162
157 hpet_use_timer = id & HPET_ID_LEGSUP; 163 hpet_use_timer = id & HPET_ID_LEGSUP;
158 164
159 if (hpet_timer_stop_set_go(hpet_tick)) 165 if (hpet_timer_stop_set_go(hpet_tick)) {
166 iounmap(hpet_virt_address);
167 hpet_virt_address = NULL;
160 return -1; 168 return -1;
169 }
161 170
162 use_hpet = 1; 171 use_hpet = 1;
163 172
diff --git a/arch/i386/kernel/topology.c b/arch/i386/kernel/topology.c
index 07d6da36a825..79cf608e14ca 100644
--- a/arch/i386/kernel/topology.c
+++ b/arch/i386/kernel/topology.c
@@ -40,14 +40,18 @@ int arch_register_cpu(int num)
40 * restrictions and assumptions in kernel. This basically 40 * restrictions and assumptions in kernel. This basically
41 * doesnt add a control file, one cannot attempt to offline 41 * doesnt add a control file, one cannot attempt to offline
42 * BSP. 42 * BSP.
43 *
44 * Also certain PCI quirks require not to enable hotplug control
45 * for all CPU's.
43 */ 46 */
44 if (!num) 47 if (num && enable_cpu_hotplug)
45 cpu_devices[num].cpu.no_control = 1; 48 cpu_devices[num].cpu.hotpluggable = 1;
46 49
47 return register_cpu(&cpu_devices[num].cpu, num); 50 return register_cpu(&cpu_devices[num].cpu, num);
48} 51}
49 52
50#ifdef CONFIG_HOTPLUG_CPU 53#ifdef CONFIG_HOTPLUG_CPU
54int enable_cpu_hotplug = 1;
51 55
52void arch_unregister_cpu(int num) { 56void arch_unregister_cpu(int num) {
53 return unregister_cpu(&cpu_devices[num].cpu); 57 return unregister_cpu(&cpu_devices[num].cpu);
diff --git a/arch/i386/kernel/traps.c b/arch/i386/kernel/traps.c
index fe9c5e8e7e6f..68de48e498ca 100644
--- a/arch/i386/kernel/traps.c
+++ b/arch/i386/kernel/traps.c
@@ -29,6 +29,7 @@
29#include <linux/kexec.h> 29#include <linux/kexec.h>
30#include <linux/unwind.h> 30#include <linux/unwind.h>
31#include <linux/uaccess.h> 31#include <linux/uaccess.h>
32#include <linux/nmi.h>
32 33
33#ifdef CONFIG_EISA 34#ifdef CONFIG_EISA
34#include <linux/ioport.h> 35#include <linux/ioport.h>
@@ -61,9 +62,6 @@ int panic_on_unrecovered_nmi;
61 62
62asmlinkage int system_call(void); 63asmlinkage int system_call(void);
63 64
64struct desc_struct default_ldt[] = { { 0, 0 }, { 0, 0 }, { 0, 0 },
65 { 0, 0 }, { 0, 0 } };
66
67/* Do we ignore FPU interrupts ? */ 65/* Do we ignore FPU interrupts ? */
68char ignore_fpu_irq = 0; 66char ignore_fpu_irq = 0;
69 67
@@ -94,7 +92,7 @@ asmlinkage void alignment_check(void);
94asmlinkage void spurious_interrupt_bug(void); 92asmlinkage void spurious_interrupt_bug(void);
95asmlinkage void machine_check(void); 93asmlinkage void machine_check(void);
96 94
97static int kstack_depth_to_print = 24; 95int kstack_depth_to_print = 24;
98#ifdef CONFIG_STACK_UNWIND 96#ifdef CONFIG_STACK_UNWIND
99static int call_trace = 1; 97static int call_trace = 1;
100#else 98#else
@@ -163,16 +161,25 @@ dump_trace_unwind(struct unwind_frame_info *info, void *data)
163{ 161{
164 struct ops_and_data *oad = (struct ops_and_data *)data; 162 struct ops_and_data *oad = (struct ops_and_data *)data;
165 int n = 0; 163 int n = 0;
164 unsigned long sp = UNW_SP(info);
166 165
166 if (arch_unw_user_mode(info))
167 return -1;
167 while (unwind(info) == 0 && UNW_PC(info)) { 168 while (unwind(info) == 0 && UNW_PC(info)) {
168 n++; 169 n++;
169 oad->ops->address(oad->data, UNW_PC(info)); 170 oad->ops->address(oad->data, UNW_PC(info));
170 if (arch_unw_user_mode(info)) 171 if (arch_unw_user_mode(info))
171 break; 172 break;
173 if ((sp & ~(PAGE_SIZE - 1)) == (UNW_SP(info) & ~(PAGE_SIZE - 1))
174 && sp > UNW_SP(info))
175 break;
176 sp = UNW_SP(info);
172 } 177 }
173 return n; 178 return n;
174} 179}
175 180
181#define MSG(msg) ops->warning(data, msg)
182
176void dump_trace(struct task_struct *task, struct pt_regs *regs, 183void dump_trace(struct task_struct *task, struct pt_regs *regs,
177 unsigned long *stack, 184 unsigned long *stack,
178 struct stacktrace_ops *ops, void *data) 185 struct stacktrace_ops *ops, void *data)
@@ -191,29 +198,31 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
191 if (unwind_init_frame_info(&info, task, regs) == 0) 198 if (unwind_init_frame_info(&info, task, regs) == 0)
192 unw_ret = dump_trace_unwind(&info, &oad); 199 unw_ret = dump_trace_unwind(&info, &oad);
193 } else if (task == current) 200 } else if (task == current)
194 unw_ret = unwind_init_running(&info, dump_trace_unwind, &oad); 201 unw_ret = unwind_init_running(&info, dump_trace_unwind,
202 &oad);
195 else { 203 else {
196 if (unwind_init_blocked(&info, task) == 0) 204 if (unwind_init_blocked(&info, task) == 0)
197 unw_ret = dump_trace_unwind(&info, &oad); 205 unw_ret = dump_trace_unwind(&info, &oad);
198 } 206 }
199 if (unw_ret > 0) { 207 if (unw_ret > 0) {
200 if (call_trace == 1 && !arch_unw_user_mode(&info)) { 208 if (call_trace == 1 && !arch_unw_user_mode(&info)) {
201 ops->warning_symbol(data, "DWARF2 unwinder stuck at %s\n", 209 ops->warning_symbol(data,
210 "DWARF2 unwinder stuck at %s",
202 UNW_PC(&info)); 211 UNW_PC(&info));
203 if (UNW_SP(&info) >= PAGE_OFFSET) { 212 if (UNW_SP(&info) >= PAGE_OFFSET) {
204 ops->warning(data, "Leftover inexact backtrace:\n"); 213 MSG("Leftover inexact backtrace:");
205 stack = (void *)UNW_SP(&info); 214 stack = (void *)UNW_SP(&info);
206 if (!stack) 215 if (!stack)
207 return; 216 return;
208 ebp = UNW_FP(&info); 217 ebp = UNW_FP(&info);
209 } else 218 } else
210 ops->warning(data, "Full inexact backtrace again:\n"); 219 MSG("Full inexact backtrace again:");
211 } else if (call_trace >= 1) 220 } else if (call_trace >= 1)
212 return; 221 return;
213 else 222 else
214 ops->warning(data, "Full inexact backtrace again:\n"); 223 MSG("Full inexact backtrace again:");
215 } else 224 } else
216 ops->warning(data, "Inexact backtrace:\n"); 225 MSG("Inexact backtrace:");
217 } 226 }
218 if (!stack) { 227 if (!stack) {
219 unsigned long dummy; 228 unsigned long dummy;
@@ -247,6 +256,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
247 stack = (unsigned long*)context->previous_esp; 256 stack = (unsigned long*)context->previous_esp;
248 if (!stack) 257 if (!stack)
249 break; 258 break;
259 touch_nmi_watchdog();
250 } 260 }
251} 261}
252EXPORT_SYMBOL(dump_trace); 262EXPORT_SYMBOL(dump_trace);
@@ -379,7 +389,7 @@ void show_registers(struct pt_regs *regs)
379 * time of the fault.. 389 * time of the fault..
380 */ 390 */
381 if (in_kernel) { 391 if (in_kernel) {
382 u8 __user *eip; 392 u8 *eip;
383 int code_bytes = 64; 393 int code_bytes = 64;
384 unsigned char c; 394 unsigned char c;
385 395
@@ -388,18 +398,20 @@ void show_registers(struct pt_regs *regs)
388 398
389 printk(KERN_EMERG "Code: "); 399 printk(KERN_EMERG "Code: ");
390 400
391 eip = (u8 __user *)regs->eip - 43; 401 eip = (u8 *)regs->eip - 43;
392 if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) { 402 if (eip < (u8 *)PAGE_OFFSET ||
403 probe_kernel_address(eip, c)) {
393 /* try starting at EIP */ 404 /* try starting at EIP */
394 eip = (u8 __user *)regs->eip; 405 eip = (u8 *)regs->eip;
395 code_bytes = 32; 406 code_bytes = 32;
396 } 407 }
397 for (i = 0; i < code_bytes; i++, eip++) { 408 for (i = 0; i < code_bytes; i++, eip++) {
398 if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) { 409 if (eip < (u8 *)PAGE_OFFSET ||
410 probe_kernel_address(eip, c)) {
399 printk(" Bad EIP value."); 411 printk(" Bad EIP value.");
400 break; 412 break;
401 } 413 }
402 if (eip == (u8 __user *)regs->eip) 414 if (eip == (u8 *)regs->eip)
403 printk("<%02x> ", c); 415 printk("<%02x> ", c);
404 else 416 else
405 printk("%02x ", c); 417 printk("%02x ", c);
@@ -415,7 +427,7 @@ static void handle_BUG(struct pt_regs *regs)
415 427
416 if (eip < PAGE_OFFSET) 428 if (eip < PAGE_OFFSET)
417 return; 429 return;
418 if (probe_kernel_address((unsigned short __user *)eip, ud2)) 430 if (probe_kernel_address((unsigned short *)eip, ud2))
419 return; 431 return;
420 if (ud2 != 0x0b0f) 432 if (ud2 != 0x0b0f)
421 return; 433 return;
@@ -428,11 +440,11 @@ static void handle_BUG(struct pt_regs *regs)
428 char *file; 440 char *file;
429 char c; 441 char c;
430 442
431 if (probe_kernel_address((unsigned short __user *)(eip + 2), 443 if (probe_kernel_address((unsigned short *)(eip + 2), line))
432 line))
433 break; 444 break;
434 if (__get_user(file, (char * __user *)(eip + 4)) || 445 if (probe_kernel_address((char **)(eip + 4), file) ||
435 (unsigned long)file < PAGE_OFFSET || __get_user(c, file)) 446 (unsigned long)file < PAGE_OFFSET ||
447 probe_kernel_address(file, c))
436 file = "<bad filename>"; 448 file = "<bad filename>";
437 449
438 printk(KERN_EMERG "kernel BUG at %s:%d!\n", file, line); 450 printk(KERN_EMERG "kernel BUG at %s:%d!\n", file, line);
@@ -452,7 +464,7 @@ void die(const char * str, struct pt_regs * regs, long err)
452 u32 lock_owner; 464 u32 lock_owner;
453 int lock_owner_depth; 465 int lock_owner_depth;
454 } die = { 466 } die = {
455 .lock = SPIN_LOCK_UNLOCKED, 467 .lock = __SPIN_LOCK_UNLOCKED(die.lock),
456 .lock_owner = -1, 468 .lock_owner = -1,
457 .lock_owner_depth = 0 469 .lock_owner_depth = 0
458 }; 470 };
@@ -707,8 +719,7 @@ mem_parity_error(unsigned char reason, struct pt_regs * regs)
707{ 719{
708 printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on " 720 printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on "
709 "CPU %d.\n", reason, smp_processor_id()); 721 "CPU %d.\n", reason, smp_processor_id());
710 printk(KERN_EMERG "You probably have a hardware problem with your RAM " 722 printk(KERN_EMERG "You have some hardware problem, likely on the PCI bus.\n");
711 "chips\n");
712 if (panic_on_unrecovered_nmi) 723 if (panic_on_unrecovered_nmi)
713 panic("NMI: Not continuing"); 724 panic("NMI: Not continuing");
714 725
@@ -773,7 +784,6 @@ void __kprobes die_nmi(struct pt_regs *regs, const char *msg)
773 printk(" on CPU%d, eip %08lx, registers:\n", 784 printk(" on CPU%d, eip %08lx, registers:\n",
774 smp_processor_id(), regs->eip); 785 smp_processor_id(), regs->eip);
775 show_registers(regs); 786 show_registers(regs);
776 printk(KERN_EMERG "console shuts up ...\n");
777 console_silent(); 787 console_silent();
778 spin_unlock(&nmi_print_lock); 788 spin_unlock(&nmi_print_lock);
779 bust_spinlocks(0); 789 bust_spinlocks(0);
@@ -1088,49 +1098,24 @@ fastcall void do_spurious_interrupt_bug(struct pt_regs * regs,
1088#endif 1098#endif
1089} 1099}
1090 1100
1091fastcall void setup_x86_bogus_stack(unsigned char * stk) 1101fastcall unsigned long patch_espfix_desc(unsigned long uesp,
1092{ 1102 unsigned long kesp)
1093 unsigned long *switch16_ptr, *switch32_ptr;
1094 struct pt_regs *regs;
1095 unsigned long stack_top, stack_bot;
1096 unsigned short iret_frame16_off;
1097 int cpu = smp_processor_id();
1098 /* reserve the space on 32bit stack for the magic switch16 pointer */
1099 memmove(stk, stk + 8, sizeof(struct pt_regs));
1100 switch16_ptr = (unsigned long *)(stk + sizeof(struct pt_regs));
1101 regs = (struct pt_regs *)stk;
1102 /* now the switch32 on 16bit stack */
1103 stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu);
1104 stack_top = stack_bot + CPU_16BIT_STACK_SIZE;
1105 switch32_ptr = (unsigned long *)(stack_top - 8);
1106 iret_frame16_off = CPU_16BIT_STACK_SIZE - 8 - 20;
1107 /* copy iret frame on 16bit stack */
1108 memcpy((void *)(stack_bot + iret_frame16_off), &regs->eip, 20);
1109 /* fill in the switch pointers */
1110 switch16_ptr[0] = (regs->esp & 0xffff0000) | iret_frame16_off;
1111 switch16_ptr[1] = __ESPFIX_SS;
1112 switch32_ptr[0] = (unsigned long)stk + sizeof(struct pt_regs) +
1113 8 - CPU_16BIT_STACK_SIZE;
1114 switch32_ptr[1] = __KERNEL_DS;
1115}
1116
1117fastcall unsigned char * fixup_x86_bogus_stack(unsigned short sp)
1118{ 1103{
1119 unsigned long *switch32_ptr;
1120 unsigned char *stack16, *stack32;
1121 unsigned long stack_top, stack_bot;
1122 int len;
1123 int cpu = smp_processor_id(); 1104 int cpu = smp_processor_id();
1124 stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu); 1105 struct Xgt_desc_struct *cpu_gdt_descr = &per_cpu(cpu_gdt_descr, cpu);
1125 stack_top = stack_bot + CPU_16BIT_STACK_SIZE; 1106 struct desc_struct *gdt = (struct desc_struct *)cpu_gdt_descr->address;
1126 switch32_ptr = (unsigned long *)(stack_top - 8); 1107 unsigned long base = (kesp - uesp) & -THREAD_SIZE;
1127 /* copy the data from 16bit stack to 32bit stack */ 1108 unsigned long new_kesp = kesp - base;
1128 len = CPU_16BIT_STACK_SIZE - 8 - sp; 1109 unsigned long lim_pages = (new_kesp | (THREAD_SIZE - 1)) >> PAGE_SHIFT;
1129 stack16 = (unsigned char *)(stack_bot + sp); 1110 __u64 desc = *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS];
1130 stack32 = (unsigned char *) 1111 /* Set up base for espfix segment */
1131 (switch32_ptr[0] + CPU_16BIT_STACK_SIZE - 8 - len); 1112 desc &= 0x00f0ff0000000000ULL;
1132 memcpy(stack32, stack16, len); 1113 desc |= ((((__u64)base) << 16) & 0x000000ffffff0000ULL) |
1133 return stack32; 1114 ((((__u64)base) << 32) & 0xff00000000000000ULL) |
1115 ((((__u64)lim_pages) << 32) & 0x000f000000000000ULL) |
1116 (lim_pages & 0xffff);
1117 *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS] = desc;
1118 return new_kesp;
1134} 1119}
1135 1120
1136/* 1121/*
@@ -1143,7 +1128,7 @@ fastcall unsigned char * fixup_x86_bogus_stack(unsigned short sp)
1143 * Must be called with kernel preemption disabled (in this case, 1128 * Must be called with kernel preemption disabled (in this case,
1144 * local interrupts are disabled at the call-site in entry.S). 1129 * local interrupts are disabled at the call-site in entry.S).
1145 */ 1130 */
1146asmlinkage void math_state_restore(struct pt_regs regs) 1131asmlinkage void math_state_restore(void)
1147{ 1132{
1148 struct thread_info *thread = current_thread_info(); 1133 struct thread_info *thread = current_thread_info();
1149 struct task_struct *tsk = thread->task; 1134 struct task_struct *tsk = thread->task;
@@ -1153,6 +1138,7 @@ asmlinkage void math_state_restore(struct pt_regs regs)
1153 init_fpu(tsk); 1138 init_fpu(tsk);
1154 restore_fpu(tsk); 1139 restore_fpu(tsk);
1155 thread->status |= TS_USEDFPU; /* So we fnsave on switch_to() */ 1140 thread->status |= TS_USEDFPU; /* So we fnsave on switch_to() */
1141 tsk->fpu_counter++;
1156} 1142}
1157 1143
1158#ifndef CONFIG_MATH_EMULATION 1144#ifndef CONFIG_MATH_EMULATION
diff --git a/arch/i386/kernel/tsc.c b/arch/i386/kernel/tsc.c
index 9810c8c90750..1bbe45dca7a0 100644
--- a/arch/i386/kernel/tsc.c
+++ b/arch/i386/kernel/tsc.c
@@ -13,7 +13,6 @@
13 13
14#include <asm/delay.h> 14#include <asm/delay.h>
15#include <asm/tsc.h> 15#include <asm/tsc.h>
16#include <asm/delay.h>
17#include <asm/io.h> 16#include <asm/io.h>
18 17
19#include "mach_timer.h" 18#include "mach_timer.h"
diff --git a/arch/i386/kernel/vm86.c b/arch/i386/kernel/vm86.c
index cbcd61d6120b..be2f96e67f78 100644
--- a/arch/i386/kernel/vm86.c
+++ b/arch/i386/kernel/vm86.c
@@ -43,6 +43,7 @@
43#include <linux/highmem.h> 43#include <linux/highmem.h>
44#include <linux/ptrace.h> 44#include <linux/ptrace.h>
45#include <linux/audit.h> 45#include <linux/audit.h>
46#include <linux/stddef.h>
46 47
47#include <asm/uaccess.h> 48#include <asm/uaccess.h>
48#include <asm/io.h> 49#include <asm/io.h>
@@ -72,10 +73,10 @@
72/* 73/*
73 * 8- and 16-bit register defines.. 74 * 8- and 16-bit register defines..
74 */ 75 */
75#define AL(regs) (((unsigned char *)&((regs)->eax))[0]) 76#define AL(regs) (((unsigned char *)&((regs)->pt.eax))[0])
76#define AH(regs) (((unsigned char *)&((regs)->eax))[1]) 77#define AH(regs) (((unsigned char *)&((regs)->pt.eax))[1])
77#define IP(regs) (*(unsigned short *)&((regs)->eip)) 78#define IP(regs) (*(unsigned short *)&((regs)->pt.eip))
78#define SP(regs) (*(unsigned short *)&((regs)->esp)) 79#define SP(regs) (*(unsigned short *)&((regs)->pt.esp))
79 80
80/* 81/*
81 * virtual flags (16 and 32-bit versions) 82 * virtual flags (16 and 32-bit versions)
@@ -89,10 +90,37 @@
89#define SAFE_MASK (0xDD5) 90#define SAFE_MASK (0xDD5)
90#define RETURN_MASK (0xDFF) 91#define RETURN_MASK (0xDFF)
91 92
92#define VM86_REGS_PART2 orig_eax 93/* convert kernel_vm86_regs to vm86_regs */
93#define VM86_REGS_SIZE1 \ 94static int copy_vm86_regs_to_user(struct vm86_regs __user *user,
94 ( (unsigned)( & (((struct kernel_vm86_regs *)0)->VM86_REGS_PART2) ) ) 95 const struct kernel_vm86_regs *regs)
95#define VM86_REGS_SIZE2 (sizeof(struct kernel_vm86_regs) - VM86_REGS_SIZE1) 96{
97 int ret = 0;
98
99 /* kernel_vm86_regs is missing xfs, so copy everything up to
100 (but not including) xgs, and then rest after xgs. */
101 ret += copy_to_user(user, regs, offsetof(struct kernel_vm86_regs, pt.xgs));
102 ret += copy_to_user(&user->__null_gs, &regs->pt.xgs,
103 sizeof(struct kernel_vm86_regs) -
104 offsetof(struct kernel_vm86_regs, pt.xgs));
105
106 return ret;
107}
108
109/* convert vm86_regs to kernel_vm86_regs */
110static int copy_vm86_regs_from_user(struct kernel_vm86_regs *regs,
111 const struct vm86_regs __user *user,
112 unsigned extra)
113{
114 int ret = 0;
115
116 ret += copy_from_user(regs, user, offsetof(struct kernel_vm86_regs, pt.xgs));
117 ret += copy_from_user(&regs->pt.xgs, &user->__null_gs,
118 sizeof(struct kernel_vm86_regs) -
119 offsetof(struct kernel_vm86_regs, pt.xgs) +
120 extra);
121
122 return ret;
123}
96 124
97struct pt_regs * FASTCALL(save_v86_state(struct kernel_vm86_regs * regs)); 125struct pt_regs * FASTCALL(save_v86_state(struct kernel_vm86_regs * regs));
98struct pt_regs * fastcall save_v86_state(struct kernel_vm86_regs * regs) 126struct pt_regs * fastcall save_v86_state(struct kernel_vm86_regs * regs)
@@ -112,10 +140,8 @@ struct pt_regs * fastcall save_v86_state(struct kernel_vm86_regs * regs)
112 printk("no vm86_info: BAD\n"); 140 printk("no vm86_info: BAD\n");
113 do_exit(SIGSEGV); 141 do_exit(SIGSEGV);
114 } 142 }
115 set_flags(regs->eflags, VEFLAGS, VIF_MASK | current->thread.v86mask); 143 set_flags(regs->pt.eflags, VEFLAGS, VIF_MASK | current->thread.v86mask);
116 tmp = copy_to_user(&current->thread.vm86_info->regs,regs, VM86_REGS_SIZE1); 144 tmp = copy_vm86_regs_to_user(&current->thread.vm86_info->regs,regs);
117 tmp += copy_to_user(&current->thread.vm86_info->regs.VM86_REGS_PART2,
118 &regs->VM86_REGS_PART2, VM86_REGS_SIZE2);
119 tmp += put_user(current->thread.screen_bitmap,&current->thread.vm86_info->screen_bitmap); 145 tmp += put_user(current->thread.screen_bitmap,&current->thread.vm86_info->screen_bitmap);
120 if (tmp) { 146 if (tmp) {
121 printk("vm86: could not access userspace vm86_info\n"); 147 printk("vm86: could not access userspace vm86_info\n");
@@ -129,9 +155,11 @@ struct pt_regs * fastcall save_v86_state(struct kernel_vm86_regs * regs)
129 current->thread.saved_esp0 = 0; 155 current->thread.saved_esp0 = 0;
130 put_cpu(); 156 put_cpu();
131 157
132 loadsegment(fs, current->thread.saved_fs);
133 loadsegment(gs, current->thread.saved_gs);
134 ret = KVM86->regs32; 158 ret = KVM86->regs32;
159
160 loadsegment(fs, current->thread.saved_fs);
161 ret->xgs = current->thread.saved_gs;
162
135 return ret; 163 return ret;
136} 164}
137 165
@@ -183,9 +211,9 @@ asmlinkage int sys_vm86old(struct pt_regs regs)
183 tsk = current; 211 tsk = current;
184 if (tsk->thread.saved_esp0) 212 if (tsk->thread.saved_esp0)
185 goto out; 213 goto out;
186 tmp = copy_from_user(&info, v86, VM86_REGS_SIZE1); 214 tmp = copy_vm86_regs_from_user(&info.regs, &v86->regs,
187 tmp += copy_from_user(&info.regs.VM86_REGS_PART2, &v86->regs.VM86_REGS_PART2, 215 offsetof(struct kernel_vm86_struct, vm86plus) -
188 (long)&info.vm86plus - (long)&info.regs.VM86_REGS_PART2); 216 sizeof(info.regs));
189 ret = -EFAULT; 217 ret = -EFAULT;
190 if (tmp) 218 if (tmp)
191 goto out; 219 goto out;
@@ -233,9 +261,9 @@ asmlinkage int sys_vm86(struct pt_regs regs)
233 if (tsk->thread.saved_esp0) 261 if (tsk->thread.saved_esp0)
234 goto out; 262 goto out;
235 v86 = (struct vm86plus_struct __user *)regs.ecx; 263 v86 = (struct vm86plus_struct __user *)regs.ecx;
236 tmp = copy_from_user(&info, v86, VM86_REGS_SIZE1); 264 tmp = copy_vm86_regs_from_user(&info.regs, &v86->regs,
237 tmp += copy_from_user(&info.regs.VM86_REGS_PART2, &v86->regs.VM86_REGS_PART2, 265 offsetof(struct kernel_vm86_struct, regs32) -
238 (long)&info.regs32 - (long)&info.regs.VM86_REGS_PART2); 266 sizeof(info.regs));
239 ret = -EFAULT; 267 ret = -EFAULT;
240 if (tmp) 268 if (tmp)
241 goto out; 269 goto out;
@@ -252,15 +280,15 @@ out:
252static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk) 280static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk)
253{ 281{
254 struct tss_struct *tss; 282 struct tss_struct *tss;
255 long eax;
256/* 283/*
257 * make sure the vm86() system call doesn't try to do anything silly 284 * make sure the vm86() system call doesn't try to do anything silly
258 */ 285 */
259 info->regs.__null_ds = 0; 286 info->regs.pt.xds = 0;
260 info->regs.__null_es = 0; 287 info->regs.pt.xes = 0;
288 info->regs.pt.xgs = 0;
261 289
262/* we are clearing fs,gs later just before "jmp resume_userspace", 290/* we are clearing fs later just before "jmp resume_userspace",
263 * because starting with Linux 2.1.x they aren't no longer saved/restored 291 * because it is not saved/restored.
264 */ 292 */
265 293
266/* 294/*
@@ -268,10 +296,10 @@ static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk
268 * has set it up safely, so this makes sure interrupt etc flags are 296 * has set it up safely, so this makes sure interrupt etc flags are
269 * inherited from protected mode. 297 * inherited from protected mode.
270 */ 298 */
271 VEFLAGS = info->regs.eflags; 299 VEFLAGS = info->regs.pt.eflags;
272 info->regs.eflags &= SAFE_MASK; 300 info->regs.pt.eflags &= SAFE_MASK;
273 info->regs.eflags |= info->regs32->eflags & ~SAFE_MASK; 301 info->regs.pt.eflags |= info->regs32->eflags & ~SAFE_MASK;
274 info->regs.eflags |= VM_MASK; 302 info->regs.pt.eflags |= VM_MASK;
275 303
276 switch (info->cpu_type) { 304 switch (info->cpu_type) {
277 case CPU_286: 305 case CPU_286:
@@ -294,7 +322,7 @@ static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk
294 info->regs32->eax = 0; 322 info->regs32->eax = 0;
295 tsk->thread.saved_esp0 = tsk->thread.esp0; 323 tsk->thread.saved_esp0 = tsk->thread.esp0;
296 savesegment(fs, tsk->thread.saved_fs); 324 savesegment(fs, tsk->thread.saved_fs);
297 savesegment(gs, tsk->thread.saved_gs); 325 tsk->thread.saved_gs = info->regs32->xgs;
298 326
299 tss = &per_cpu(init_tss, get_cpu()); 327 tss = &per_cpu(init_tss, get_cpu());
300 tsk->thread.esp0 = (unsigned long) &info->VM86_TSS_ESP0; 328 tsk->thread.esp0 = (unsigned long) &info->VM86_TSS_ESP0;
@@ -306,19 +334,18 @@ static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk
306 tsk->thread.screen_bitmap = info->screen_bitmap; 334 tsk->thread.screen_bitmap = info->screen_bitmap;
307 if (info->flags & VM86_SCREEN_BITMAP) 335 if (info->flags & VM86_SCREEN_BITMAP)
308 mark_screen_rdonly(tsk->mm); 336 mark_screen_rdonly(tsk->mm);
309 __asm__ __volatile__("xorl %eax,%eax; movl %eax,%fs; movl %eax,%gs\n\t");
310 __asm__ __volatile__("movl %%eax, %0\n" :"=r"(eax));
311 337
312 /*call audit_syscall_exit since we do not exit via the normal paths */ 338 /*call audit_syscall_exit since we do not exit via the normal paths */
313 if (unlikely(current->audit_context)) 339 if (unlikely(current->audit_context))
314 audit_syscall_exit(AUDITSC_RESULT(eax), eax); 340 audit_syscall_exit(AUDITSC_RESULT(0), 0);
315 341
316 __asm__ __volatile__( 342 __asm__ __volatile__(
317 "movl %0,%%esp\n\t" 343 "movl %0,%%esp\n\t"
318 "movl %1,%%ebp\n\t" 344 "movl %1,%%ebp\n\t"
345 "mov %2, %%fs\n\t"
319 "jmp resume_userspace" 346 "jmp resume_userspace"
320 : /* no outputs */ 347 : /* no outputs */
321 :"r" (&info->regs), "r" (task_thread_info(tsk))); 348 :"r" (&info->regs), "r" (task_thread_info(tsk)), "r" (0));
322 /* we never return here */ 349 /* we never return here */
323} 350}
324 351
@@ -348,12 +375,12 @@ static inline void clear_IF(struct kernel_vm86_regs * regs)
348 375
349static inline void clear_TF(struct kernel_vm86_regs * regs) 376static inline void clear_TF(struct kernel_vm86_regs * regs)
350{ 377{
351 regs->eflags &= ~TF_MASK; 378 regs->pt.eflags &= ~TF_MASK;
352} 379}
353 380
354static inline void clear_AC(struct kernel_vm86_regs * regs) 381static inline void clear_AC(struct kernel_vm86_regs * regs)
355{ 382{
356 regs->eflags &= ~AC_MASK; 383 regs->pt.eflags &= ~AC_MASK;
357} 384}
358 385
359/* It is correct to call set_IF(regs) from the set_vflags_* 386/* It is correct to call set_IF(regs) from the set_vflags_*
@@ -370,7 +397,7 @@ static inline void clear_AC(struct kernel_vm86_regs * regs)
370static inline void set_vflags_long(unsigned long eflags, struct kernel_vm86_regs * regs) 397static inline void set_vflags_long(unsigned long eflags, struct kernel_vm86_regs * regs)
371{ 398{
372 set_flags(VEFLAGS, eflags, current->thread.v86mask); 399 set_flags(VEFLAGS, eflags, current->thread.v86mask);
373 set_flags(regs->eflags, eflags, SAFE_MASK); 400 set_flags(regs->pt.eflags, eflags, SAFE_MASK);
374 if (eflags & IF_MASK) 401 if (eflags & IF_MASK)
375 set_IF(regs); 402 set_IF(regs);
376 else 403 else
@@ -380,7 +407,7 @@ static inline void set_vflags_long(unsigned long eflags, struct kernel_vm86_regs
380static inline void set_vflags_short(unsigned short flags, struct kernel_vm86_regs * regs) 407static inline void set_vflags_short(unsigned short flags, struct kernel_vm86_regs * regs)
381{ 408{
382 set_flags(VFLAGS, flags, current->thread.v86mask); 409 set_flags(VFLAGS, flags, current->thread.v86mask);
383 set_flags(regs->eflags, flags, SAFE_MASK); 410 set_flags(regs->pt.eflags, flags, SAFE_MASK);
384 if (flags & IF_MASK) 411 if (flags & IF_MASK)
385 set_IF(regs); 412 set_IF(regs);
386 else 413 else
@@ -389,7 +416,7 @@ static inline void set_vflags_short(unsigned short flags, struct kernel_vm86_reg
389 416
390static inline unsigned long get_vflags(struct kernel_vm86_regs * regs) 417static inline unsigned long get_vflags(struct kernel_vm86_regs * regs)
391{ 418{
392 unsigned long flags = regs->eflags & RETURN_MASK; 419 unsigned long flags = regs->pt.eflags & RETURN_MASK;
393 420
394 if (VEFLAGS & VIF_MASK) 421 if (VEFLAGS & VIF_MASK)
395 flags |= IF_MASK; 422 flags |= IF_MASK;
@@ -493,7 +520,7 @@ static void do_int(struct kernel_vm86_regs *regs, int i,
493 unsigned long __user *intr_ptr; 520 unsigned long __user *intr_ptr;
494 unsigned long segoffs; 521 unsigned long segoffs;
495 522
496 if (regs->cs == BIOSSEG) 523 if (regs->pt.xcs == BIOSSEG)
497 goto cannot_handle; 524 goto cannot_handle;
498 if (is_revectored(i, &KVM86->int_revectored)) 525 if (is_revectored(i, &KVM86->int_revectored))
499 goto cannot_handle; 526 goto cannot_handle;
@@ -505,9 +532,9 @@ static void do_int(struct kernel_vm86_regs *regs, int i,
505 if ((segoffs >> 16) == BIOSSEG) 532 if ((segoffs >> 16) == BIOSSEG)
506 goto cannot_handle; 533 goto cannot_handle;
507 pushw(ssp, sp, get_vflags(regs), cannot_handle); 534 pushw(ssp, sp, get_vflags(regs), cannot_handle);
508 pushw(ssp, sp, regs->cs, cannot_handle); 535 pushw(ssp, sp, regs->pt.xcs, cannot_handle);
509 pushw(ssp, sp, IP(regs), cannot_handle); 536 pushw(ssp, sp, IP(regs), cannot_handle);
510 regs->cs = segoffs >> 16; 537 regs->pt.xcs = segoffs >> 16;
511 SP(regs) -= 6; 538 SP(regs) -= 6;
512 IP(regs) = segoffs & 0xffff; 539 IP(regs) = segoffs & 0xffff;
513 clear_TF(regs); 540 clear_TF(regs);
@@ -524,7 +551,7 @@ int handle_vm86_trap(struct kernel_vm86_regs * regs, long error_code, int trapno
524 if (VMPI.is_vm86pus) { 551 if (VMPI.is_vm86pus) {
525 if ( (trapno==3) || (trapno==1) ) 552 if ( (trapno==3) || (trapno==1) )
526 return_to_32bit(regs, VM86_TRAP + (trapno << 8)); 553 return_to_32bit(regs, VM86_TRAP + (trapno << 8));
527 do_int(regs, trapno, (unsigned char __user *) (regs->ss << 4), SP(regs)); 554 do_int(regs, trapno, (unsigned char __user *) (regs->pt.xss << 4), SP(regs));
528 return 0; 555 return 0;
529 } 556 }
530 if (trapno !=1) 557 if (trapno !=1)
@@ -560,10 +587,10 @@ void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code)
560 handle_vm86_trap(regs, 0, 1); \ 587 handle_vm86_trap(regs, 0, 1); \
561 return; } while (0) 588 return; } while (0)
562 589
563 orig_flags = *(unsigned short *)&regs->eflags; 590 orig_flags = *(unsigned short *)&regs->pt.eflags;
564 591
565 csp = (unsigned char __user *) (regs->cs << 4); 592 csp = (unsigned char __user *) (regs->pt.xcs << 4);
566 ssp = (unsigned char __user *) (regs->ss << 4); 593 ssp = (unsigned char __user *) (regs->pt.xss << 4);
567 sp = SP(regs); 594 sp = SP(regs);
568 ip = IP(regs); 595 ip = IP(regs);
569 596
@@ -650,7 +677,7 @@ void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code)
650 SP(regs) += 6; 677 SP(regs) += 6;
651 } 678 }
652 IP(regs) = newip; 679 IP(regs) = newip;
653 regs->cs = newcs; 680 regs->pt.xcs = newcs;
654 CHECK_IF_IN_TRAP; 681 CHECK_IF_IN_TRAP;
655 if (data32) { 682 if (data32) {
656 set_vflags_long(newflags, regs); 683 set_vflags_long(newflags, regs);
diff --git a/arch/i386/kernel/vmlinux.lds.S b/arch/i386/kernel/vmlinux.lds.S
index c6f84a0322ba..56e6ad5cb045 100644
--- a/arch/i386/kernel/vmlinux.lds.S
+++ b/arch/i386/kernel/vmlinux.lds.S
@@ -1,13 +1,26 @@
1/* ld script to make i386 Linux kernel 1/* ld script to make i386 Linux kernel
2 * Written by Martin Mares <mj@atrey.karlin.mff.cuni.cz>; 2 * Written by Martin Mares <mj@atrey.karlin.mff.cuni.cz>;
3 *
4 * Don't define absolute symbols until and unless you know that symbol
5 * value is should remain constant even if kernel image is relocated
6 * at run time. Absolute symbols are not relocated. If symbol value should
7 * change if kernel is relocated, make the symbol section relative and
8 * put it inside the section definition.
3 */ 9 */
4 10
11/* Don't define absolute symbols until and unless you know that symbol
12 * value is should remain constant even if kernel image is relocated
13 * at run time. Absolute symbols are not relocated. If symbol value should
14 * change if kernel is relocated, make the symbol section relative and
15 * put it inside the section definition.
16 */
5#define LOAD_OFFSET __PAGE_OFFSET 17#define LOAD_OFFSET __PAGE_OFFSET
6 18
7#include <asm-generic/vmlinux.lds.h> 19#include <asm-generic/vmlinux.lds.h>
8#include <asm/thread_info.h> 20#include <asm/thread_info.h>
9#include <asm/page.h> 21#include <asm/page.h>
10#include <asm/cache.h> 22#include <asm/cache.h>
23#include <asm/boot.h>
11 24
12OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386") 25OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
13OUTPUT_ARCH(i386) 26OUTPUT_ARCH(i386)
@@ -21,34 +34,35 @@ PHDRS {
21} 34}
22SECTIONS 35SECTIONS
23{ 36{
24 . = __KERNEL_START; 37 . = LOAD_OFFSET + LOAD_PHYSICAL_ADDR;
25 phys_startup_32 = startup_32 - LOAD_OFFSET; 38 phys_startup_32 = startup_32 - LOAD_OFFSET;
26 /* read-only */ 39 /* read-only */
27 _text = .; /* Text and read-only data */
28 .text : AT(ADDR(.text) - LOAD_OFFSET) { 40 .text : AT(ADDR(.text) - LOAD_OFFSET) {
41 _text = .; /* Text and read-only data */
29 *(.text) 42 *(.text)
30 SCHED_TEXT 43 SCHED_TEXT
31 LOCK_TEXT 44 LOCK_TEXT
32 KPROBES_TEXT 45 KPROBES_TEXT
33 *(.fixup) 46 *(.fixup)
34 *(.gnu.warning) 47 *(.gnu.warning)
35 } :text = 0x9090 48 _etext = .; /* End of text section */
36 49 } :text = 0x9090
37 _etext = .; /* End of text section */
38 50
39 . = ALIGN(16); /* Exception table */ 51 . = ALIGN(16); /* Exception table */
40 __start___ex_table = .; 52 __ex_table : AT(ADDR(__ex_table) - LOAD_OFFSET) {
41 __ex_table : AT(ADDR(__ex_table) - LOAD_OFFSET) { *(__ex_table) } 53 __start___ex_table = .;
42 __stop___ex_table = .; 54 *(__ex_table)
55 __stop___ex_table = .;
56 }
43 57
44 RODATA 58 RODATA
45 59
46 . = ALIGN(4); 60 . = ALIGN(4);
47 __tracedata_start = .;
48 .tracedata : AT(ADDR(.tracedata) - LOAD_OFFSET) { 61 .tracedata : AT(ADDR(.tracedata) - LOAD_OFFSET) {
62 __tracedata_start = .;
49 *(.tracedata) 63 *(.tracedata)
64 __tracedata_end = .;
50 } 65 }
51 __tracedata_end = .;
52 66
53 /* writeable */ 67 /* writeable */
54 . = ALIGN(4096); 68 . = ALIGN(4096);
@@ -57,11 +71,19 @@ SECTIONS
57 CONSTRUCTORS 71 CONSTRUCTORS
58 } :data 72 } :data
59 73
74 .paravirtprobe : AT(ADDR(.paravirtprobe) - LOAD_OFFSET) {
75 __start_paravirtprobe = .;
76 *(.paravirtprobe)
77 __stop_paravirtprobe = .;
78 }
79
60 . = ALIGN(4096); 80 . = ALIGN(4096);
61 __nosave_begin = .; 81 .data_nosave : AT(ADDR(.data_nosave) - LOAD_OFFSET) {
62 .data_nosave : AT(ADDR(.data_nosave) - LOAD_OFFSET) { *(.data.nosave) } 82 __nosave_begin = .;
63 . = ALIGN(4096); 83 *(.data.nosave)
64 __nosave_end = .; 84 . = ALIGN(4096);
85 __nosave_end = .;
86 }
65 87
66 . = ALIGN(4096); 88 . = ALIGN(4096);
67 .data.page_aligned : AT(ADDR(.data.page_aligned) - LOAD_OFFSET) { 89 .data.page_aligned : AT(ADDR(.data.page_aligned) - LOAD_OFFSET) {
@@ -75,17 +97,10 @@ SECTIONS
75 97
76 /* rarely changed data like cpu maps */ 98 /* rarely changed data like cpu maps */
77 . = ALIGN(32); 99 . = ALIGN(32);
78 .data.read_mostly : AT(ADDR(.data.read_mostly) - LOAD_OFFSET) { *(.data.read_mostly) } 100 .data.read_mostly : AT(ADDR(.data.read_mostly) - LOAD_OFFSET) {
79 _edata = .; /* End of data section */ 101 *(.data.read_mostly)
80 102 _edata = .; /* End of data section */
81#ifdef CONFIG_STACK_UNWIND
82 . = ALIGN(4);
83 .eh_frame : AT(ADDR(.eh_frame) - LOAD_OFFSET) {
84 __start_unwind = .;
85 *(.eh_frame)
86 __end_unwind = .;
87 } 103 }
88#endif
89 104
90 . = ALIGN(THREAD_SIZE); /* init_task */ 105 . = ALIGN(THREAD_SIZE); /* init_task */
91 .data.init_task : AT(ADDR(.data.init_task) - LOAD_OFFSET) { 106 .data.init_task : AT(ADDR(.data.init_task) - LOAD_OFFSET) {
@@ -94,88 +109,102 @@ SECTIONS
94 109
95 /* might get freed after init */ 110 /* might get freed after init */
96 . = ALIGN(4096); 111 . = ALIGN(4096);
97 __smp_alt_begin = .;
98 __smp_alt_instructions = .;
99 .smp_altinstructions : AT(ADDR(.smp_altinstructions) - LOAD_OFFSET) { 112 .smp_altinstructions : AT(ADDR(.smp_altinstructions) - LOAD_OFFSET) {
113 __smp_alt_begin = .;
114 __smp_alt_instructions = .;
100 *(.smp_altinstructions) 115 *(.smp_altinstructions)
116 __smp_alt_instructions_end = .;
101 } 117 }
102 __smp_alt_instructions_end = .;
103 . = ALIGN(4); 118 . = ALIGN(4);
104 __smp_locks = .;
105 .smp_locks : AT(ADDR(.smp_locks) - LOAD_OFFSET) { 119 .smp_locks : AT(ADDR(.smp_locks) - LOAD_OFFSET) {
120 __smp_locks = .;
106 *(.smp_locks) 121 *(.smp_locks)
122 __smp_locks_end = .;
107 } 123 }
108 __smp_locks_end = .;
109 .smp_altinstr_replacement : AT(ADDR(.smp_altinstr_replacement) - LOAD_OFFSET) { 124 .smp_altinstr_replacement : AT(ADDR(.smp_altinstr_replacement) - LOAD_OFFSET) {
110 *(.smp_altinstr_replacement) 125 *(.smp_altinstr_replacement)
126 __smp_alt_end = .;
111 } 127 }
128 /* will be freed after init
129 * Following ALIGN() is required to make sure no other data falls on the
130 * same page where __smp_alt_end is pointing as that page might be freed
131 * after boot. Always make sure that ALIGN() directive is present after
132 * the section which contains __smp_alt_end.
133 */
112 . = ALIGN(4096); 134 . = ALIGN(4096);
113 __smp_alt_end = .;
114 135
115 /* will be freed after init */ 136 /* will be freed after init */
116 . = ALIGN(4096); /* Init code and data */ 137 . = ALIGN(4096); /* Init code and data */
117 __init_begin = .;
118 .init.text : AT(ADDR(.init.text) - LOAD_OFFSET) { 138 .init.text : AT(ADDR(.init.text) - LOAD_OFFSET) {
139 __init_begin = .;
119 _sinittext = .; 140 _sinittext = .;
120 *(.init.text) 141 *(.init.text)
121 _einittext = .; 142 _einittext = .;
122 } 143 }
123 .init.data : AT(ADDR(.init.data) - LOAD_OFFSET) { *(.init.data) } 144 .init.data : AT(ADDR(.init.data) - LOAD_OFFSET) { *(.init.data) }
124 . = ALIGN(16); 145 . = ALIGN(16);
125 __setup_start = .; 146 .init.setup : AT(ADDR(.init.setup) - LOAD_OFFSET) {
126 .init.setup : AT(ADDR(.init.setup) - LOAD_OFFSET) { *(.init.setup) } 147 __setup_start = .;
127 __setup_end = .; 148 *(.init.setup)
128 __initcall_start = .; 149 __setup_end = .;
150 }
129 .initcall.init : AT(ADDR(.initcall.init) - LOAD_OFFSET) { 151 .initcall.init : AT(ADDR(.initcall.init) - LOAD_OFFSET) {
152 __initcall_start = .;
130 INITCALLS 153 INITCALLS
154 __initcall_end = .;
131 } 155 }
132 __initcall_end = .;
133 __con_initcall_start = .;
134 .con_initcall.init : AT(ADDR(.con_initcall.init) - LOAD_OFFSET) { 156 .con_initcall.init : AT(ADDR(.con_initcall.init) - LOAD_OFFSET) {
157 __con_initcall_start = .;
135 *(.con_initcall.init) 158 *(.con_initcall.init)
159 __con_initcall_end = .;
136 } 160 }
137 __con_initcall_end = .;
138 SECURITY_INIT 161 SECURITY_INIT
139 . = ALIGN(4); 162 . = ALIGN(4);
140 __alt_instructions = .;
141 .altinstructions : AT(ADDR(.altinstructions) - LOAD_OFFSET) { 163 .altinstructions : AT(ADDR(.altinstructions) - LOAD_OFFSET) {
164 __alt_instructions = .;
142 *(.altinstructions) 165 *(.altinstructions)
166 __alt_instructions_end = .;
143 } 167 }
144 __alt_instructions_end = .;
145 .altinstr_replacement : AT(ADDR(.altinstr_replacement) - LOAD_OFFSET) { 168 .altinstr_replacement : AT(ADDR(.altinstr_replacement) - LOAD_OFFSET) {
146 *(.altinstr_replacement) 169 *(.altinstr_replacement)
147 } 170 }
171 . = ALIGN(4);
172 .parainstructions : AT(ADDR(.parainstructions) - LOAD_OFFSET) {
173 __start_parainstructions = .;
174 *(.parainstructions)
175 __stop_parainstructions = .;
176 }
148 /* .exit.text is discard at runtime, not link time, to deal with references 177 /* .exit.text is discard at runtime, not link time, to deal with references
149 from .altinstructions and .eh_frame */ 178 from .altinstructions and .eh_frame */
150 .exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) { *(.exit.text) } 179 .exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) { *(.exit.text) }
151 .exit.data : AT(ADDR(.exit.data) - LOAD_OFFSET) { *(.exit.data) } 180 .exit.data : AT(ADDR(.exit.data) - LOAD_OFFSET) { *(.exit.data) }
152 . = ALIGN(4096); 181 . = ALIGN(4096);
153 __initramfs_start = .; 182 .init.ramfs : AT(ADDR(.init.ramfs) - LOAD_OFFSET) {
154 .init.ramfs : AT(ADDR(.init.ramfs) - LOAD_OFFSET) { *(.init.ramfs) } 183 __initramfs_start = .;
155 __initramfs_end = .; 184 *(.init.ramfs)
185 __initramfs_end = .;
186 }
156 . = ALIGN(L1_CACHE_BYTES); 187 . = ALIGN(L1_CACHE_BYTES);
157 __per_cpu_start = .; 188 .data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) {
158 .data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) { *(.data.percpu) } 189 __per_cpu_start = .;
159 __per_cpu_end = .; 190 *(.data.percpu)
191 __per_cpu_end = .;
192 }
160 . = ALIGN(4096); 193 . = ALIGN(4096);
161 __init_end = .;
162 /* freed after init ends here */ 194 /* freed after init ends here */
163 195
164 __bss_start = .; /* BSS */
165 .bss.page_aligned : AT(ADDR(.bss.page_aligned) - LOAD_OFFSET) {
166 *(.bss.page_aligned)
167 }
168 .bss : AT(ADDR(.bss) - LOAD_OFFSET) { 196 .bss : AT(ADDR(.bss) - LOAD_OFFSET) {
197 __init_end = .;
198 __bss_start = .; /* BSS */
199 *(.bss.page_aligned)
169 *(.bss) 200 *(.bss)
201 . = ALIGN(4);
202 __bss_stop = .;
203 _end = . ;
204 /* This is where the kernel creates the early boot page tables */
205 . = ALIGN(4096);
206 pg0 = . ;
170 } 207 }
171 . = ALIGN(4);
172 __bss_stop = .;
173
174 _end = . ;
175
176 /* This is where the kernel creates the early boot page tables */
177 . = ALIGN(4096);
178 pg0 = .;
179 208
180 /* Sections to be discarded */ 209 /* Sections to be discarded */
181 /DISCARD/ : { 210 /DISCARD/ : {
generated by cgit v1.2.2 (git 2.25.0) at 2025-01-07 10:17:15 -0500