aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86
diff options
context:
space:
mode:
authorIngo Molnar <mingo@elte.hu>2009-05-07 05:17:13 -0400
committerIngo Molnar <mingo@elte.hu>2009-05-07 05:17:34 -0400
commit44347d947f628060b92449702071bfe1d31dfb75 (patch)
treec6ed74610d5b3295df4296659f80f5feb94b28cc /arch/x86
parentd94fc523f3c35bd8013f04827e94756cbc0212f4 (diff)
parent413f81eba35d6ede9289b0c8a920c013a84fac71 (diff)
Merge branch 'linus' into tracing/core
Merge reason: tracing/core was on a .30-rc1 base and was missing out on on a handful of tracing fixes present in .30-rc5-almost. Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'arch/x86')
-rw-r--r--arch/x86/Kconfig3
-rw-r--r--arch/x86/Kconfig.cpu1
-rw-r--r--arch/x86/Makefile6
-rw-r--r--arch/x86/ia32/sys_ia32.c19
-rw-r--r--arch/x86/include/asm/desc.h2
-rw-r--r--arch/x86/include/asm/hardirq.h2
-rw-r--r--arch/x86/include/asm/lguest_hcall.h2
-rw-r--r--arch/x86/include/asm/mce.h1
-rw-r--r--arch/x86/include/asm/pat.h4
-rw-r--r--arch/x86/include/asm/processor.h6
-rw-r--r--arch/x86/include/asm/required-features.h2
-rw-r--r--arch/x86/include/asm/tlbflush.h2
-rw-r--r--arch/x86/include/asm/topology.h2
-rw-r--r--arch/x86/include/asm/uv/uv_mmrs.h5
-rw-r--r--arch/x86/include/asm/xen/page.h3
-rw-r--r--arch/x86/kernel/amd_iommu_init.c16
-rw-r--r--arch/x86/kernel/apic/io_apic.c2
-rw-r--r--arch/x86/kernel/apic/nmi.c5
-rw-r--r--arch/x86/kernel/apic/x2apic_uv_x.c31
-rw-r--r--arch/x86/kernel/bios_uv.c3
-rw-r--r--arch/x86/kernel/cpu/common.c2
-rw-r--r--arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c38
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_64.c33
-rw-r--r--arch/x86/kernel/cpu/proc.c2
-rw-r--r--arch/x86/kernel/entry_64.S3
-rw-r--r--arch/x86/kernel/hpet.c24
-rw-r--r--arch/x86/kernel/i8253.c2
-rw-r--r--arch/x86/kernel/kvmclock.c7
-rw-r--r--arch/x86/kernel/microcode_core.c35
-rw-r--r--arch/x86/kernel/pci-swiotlb.c2
-rw-r--r--arch/x86/kernel/quirks.c2
-rw-r--r--arch/x86/kernel/tlb_uv.c189
-rw-r--r--arch/x86/kernel/tsc.c2
-rw-r--r--arch/x86/kernel/uv_sysfs.c4
-rw-r--r--arch/x86/kernel/uv_time.c10
-rw-r--r--arch/x86/kernel/vmiclock_32.c2
-rw-r--r--arch/x86/kernel/xsave.c4
-rw-r--r--arch/x86/kvm/mmu.c2
-rw-r--r--arch/x86/kvm/x86.c8
-rw-r--r--arch/x86/lguest/boot.c18
-rw-r--r--arch/x86/mm/ioremap.c10
-rw-r--r--arch/x86/mm/kmmio.c2
-rw-r--r--arch/x86/mm/numa_32.c2
-rw-r--r--arch/x86/mm/numa_64.c3
-rw-r--r--arch/x86/mm/pageattr.c127
-rw-r--r--arch/x86/mm/pat.c189
-rw-r--r--arch/x86/mm/srat_64.c6
-rw-r--r--arch/x86/pci/amd_bus.c6
-rw-r--r--arch/x86/pci/common.c5
-rw-r--r--arch/x86/pci/i386.c4
-rw-r--r--arch/x86/pci/mmconfig-shared.c6
-rw-r--r--arch/x86/vdso/vclock_gettime.c12
-rw-r--r--arch/x86/xen/enlighten.c89
-rw-r--r--arch/x86/xen/mmu.c116
-rw-r--r--arch/x86/xen/mmu.h3
-rw-r--r--arch/x86/xen/smp.c4
-rw-r--r--arch/x86/xen/time.c7
-rw-r--r--arch/x86/xen/xen-ops.h2
58 files changed, 632 insertions, 467 deletions
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index bc25b9f5e4cd..df9e885eee14 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -277,6 +277,7 @@ config SPARSE_IRQ
277config NUMA_MIGRATE_IRQ_DESC 277config NUMA_MIGRATE_IRQ_DESC
278 bool "Move irq desc when changing irq smp_affinity" 278 bool "Move irq desc when changing irq smp_affinity"
279 depends on SPARSE_IRQ && NUMA 279 depends on SPARSE_IRQ && NUMA
280 depends on BROKEN
280 default n 281 default n
281 ---help--- 282 ---help---
282 This enables moving irq_desc to cpu/node that irq will use handled. 283 This enables moving irq_desc to cpu/node that irq will use handled.
@@ -353,6 +354,7 @@ config X86_UV
353 bool "SGI Ultraviolet" 354 bool "SGI Ultraviolet"
354 depends on X86_64 355 depends on X86_64
355 depends on X86_EXTENDED_PLATFORM 356 depends on X86_EXTENDED_PLATFORM
357 depends on NUMA
356 select X86_X2APIC 358 select X86_X2APIC
357 ---help--- 359 ---help---
358 This option is needed in order to support SGI Ultraviolet systems. 360 This option is needed in order to support SGI Ultraviolet systems.
@@ -663,6 +665,7 @@ config MAXSMP
663 665
664config NR_CPUS 666config NR_CPUS
665 int "Maximum number of CPUs" if SMP && !MAXSMP 667 int "Maximum number of CPUs" if SMP && !MAXSMP
668 range 2 8 if SMP && X86_32 && !X86_BIGSMP
666 range 2 512 if SMP && !MAXSMP 669 range 2 512 if SMP && !MAXSMP
667 default "1" if !SMP 670 default "1" if !SMP
668 default "4096" if MAXSMP 671 default "4096" if MAXSMP
diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu
index 924e156a85ab..8130334329c0 100644
--- a/arch/x86/Kconfig.cpu
+++ b/arch/x86/Kconfig.cpu
@@ -506,6 +506,7 @@ config X86_PTRACE_BTS
506 bool "Branch Trace Store" 506 bool "Branch Trace Store"
507 default y 507 default y
508 depends on X86_DEBUGCTLMSR 508 depends on X86_DEBUGCTLMSR
509 depends on BROKEN
509 ---help--- 510 ---help---
510 This adds a ptrace interface to the hardware's branch trace store. 511 This adds a ptrace interface to the hardware's branch trace store.
511 512
diff --git a/arch/x86/Makefile b/arch/x86/Makefile
index f05d8c91d9e5..8c86b72afdc2 100644
--- a/arch/x86/Makefile
+++ b/arch/x86/Makefile
@@ -153,7 +153,7 @@ endif
153 153
154boot := arch/x86/boot 154boot := arch/x86/boot
155 155
156BOOT_TARGETS = bzlilo bzdisk fdimage fdimage144 fdimage288 isoimage install 156BOOT_TARGETS = bzlilo bzdisk fdimage fdimage144 fdimage288 isoimage
157 157
158PHONY += bzImage $(BOOT_TARGETS) 158PHONY += bzImage $(BOOT_TARGETS)
159 159
@@ -171,6 +171,10 @@ bzImage: vmlinux
171$(BOOT_TARGETS): vmlinux 171$(BOOT_TARGETS): vmlinux
172 $(Q)$(MAKE) $(build)=$(boot) $@ 172 $(Q)$(MAKE) $(build)=$(boot) $@
173 173
174PHONY += install
175install:
176 $(Q)$(MAKE) $(build)=$(boot) $@
177
174PHONY += vdso_install 178PHONY += vdso_install
175vdso_install: 179vdso_install:
176 $(Q)$(MAKE) $(build)=arch/x86/vdso $@ 180 $(Q)$(MAKE) $(build)=arch/x86/vdso $@
diff --git a/arch/x86/ia32/sys_ia32.c b/arch/x86/ia32/sys_ia32.c
index efac92fd1efb..085a8c35f149 100644
--- a/arch/x86/ia32/sys_ia32.c
+++ b/arch/x86/ia32/sys_ia32.c
@@ -129,21 +129,12 @@ asmlinkage long sys32_fstatat(unsigned int dfd, char __user *filename,
129 struct stat64 __user *statbuf, int flag) 129 struct stat64 __user *statbuf, int flag)
130{ 130{
131 struct kstat stat; 131 struct kstat stat;
132 int error = -EINVAL; 132 int error;
133 133
134 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0) 134 error = vfs_fstatat(dfd, filename, &stat, flag);
135 goto out; 135 if (error)
136 136 return error;
137 if (flag & AT_SYMLINK_NOFOLLOW) 137 return cp_stat64(statbuf, &stat);
138 error = vfs_lstat_fd(dfd, filename, &stat);
139 else
140 error = vfs_stat_fd(dfd, filename, &stat);
141
142 if (!error)
143 error = cp_stat64(statbuf, &stat);
144
145out:
146 return error;
147} 138}
148 139
149/* 140/*
diff --git a/arch/x86/include/asm/desc.h b/arch/x86/include/asm/desc.h
index 5623c50d67b2..c45f415ce315 100644
--- a/arch/x86/include/asm/desc.h
+++ b/arch/x86/include/asm/desc.h
@@ -37,7 +37,7 @@ extern gate_desc idt_table[];
37struct gdt_page { 37struct gdt_page {
38 struct desc_struct gdt[GDT_ENTRIES]; 38 struct desc_struct gdt[GDT_ENTRIES];
39} __attribute__((aligned(PAGE_SIZE))); 39} __attribute__((aligned(PAGE_SIZE)));
40DECLARE_PER_CPU(struct gdt_page, gdt_page); 40DECLARE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page);
41 41
42static inline struct desc_struct *get_cpu_gdt_table(unsigned int cpu) 42static inline struct desc_struct *get_cpu_gdt_table(unsigned int cpu)
43{ 43{
diff --git a/arch/x86/include/asm/hardirq.h b/arch/x86/include/asm/hardirq.h
index 039db6aa8e02..37555e52f980 100644
--- a/arch/x86/include/asm/hardirq.h
+++ b/arch/x86/include/asm/hardirq.h
@@ -26,7 +26,7 @@ typedef struct {
26#endif 26#endif
27} ____cacheline_aligned irq_cpustat_t; 27} ____cacheline_aligned irq_cpustat_t;
28 28
29DECLARE_PER_CPU(irq_cpustat_t, irq_stat); 29DECLARE_PER_CPU_SHARED_ALIGNED(irq_cpustat_t, irq_stat);
30 30
31/* We can have at most NR_VECTORS irqs routed to a cpu at a time */ 31/* We can have at most NR_VECTORS irqs routed to a cpu at a time */
32#define MAX_HARDIRQS_PER_CPU NR_VECTORS 32#define MAX_HARDIRQS_PER_CPU NR_VECTORS
diff --git a/arch/x86/include/asm/lguest_hcall.h b/arch/x86/include/asm/lguest_hcall.h
index 0f4ee7148afe..faae1996487b 100644
--- a/arch/x86/include/asm/lguest_hcall.h
+++ b/arch/x86/include/asm/lguest_hcall.h
@@ -5,7 +5,6 @@
5#define LHCALL_FLUSH_ASYNC 0 5#define LHCALL_FLUSH_ASYNC 0
6#define LHCALL_LGUEST_INIT 1 6#define LHCALL_LGUEST_INIT 1
7#define LHCALL_SHUTDOWN 2 7#define LHCALL_SHUTDOWN 2
8#define LHCALL_LOAD_GDT 3
9#define LHCALL_NEW_PGTABLE 4 8#define LHCALL_NEW_PGTABLE 4
10#define LHCALL_FLUSH_TLB 5 9#define LHCALL_FLUSH_TLB 5
11#define LHCALL_LOAD_IDT_ENTRY 6 10#define LHCALL_LOAD_IDT_ENTRY 6
@@ -17,6 +16,7 @@
17#define LHCALL_SET_PMD 15 16#define LHCALL_SET_PMD 15
18#define LHCALL_LOAD_TLS 16 17#define LHCALL_LOAD_TLS 16
19#define LHCALL_NOTIFY 17 18#define LHCALL_NOTIFY 17
19#define LHCALL_LOAD_GDT_ENTRY 18
20 20
21#define LGUEST_TRAP_ENTRY 0x1F 21#define LGUEST_TRAP_ENTRY 0x1F
22 22
diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index 563933e06a35..4f8c199584e7 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -137,6 +137,7 @@ DECLARE_PER_CPU(mce_banks_t, mce_poll_banks);
137enum mcp_flags { 137enum mcp_flags {
138 MCP_TIMESTAMP = (1 << 0), /* log time stamp */ 138 MCP_TIMESTAMP = (1 << 0), /* log time stamp */
139 MCP_UC = (1 << 1), /* log uncorrected errors */ 139 MCP_UC = (1 << 1), /* log uncorrected errors */
140 MCP_DONTLOG = (1 << 2), /* only clear, don't log */
140}; 141};
141extern void machine_check_poll(enum mcp_flags flags, mce_banks_t *b); 142extern void machine_check_poll(enum mcp_flags flags, mce_banks_t *b);
142 143
diff --git a/arch/x86/include/asm/pat.h b/arch/x86/include/asm/pat.h
index 2cd07b9422f4..7af14e512f97 100644
--- a/arch/x86/include/asm/pat.h
+++ b/arch/x86/include/asm/pat.h
@@ -18,9 +18,5 @@ extern int free_memtype(u64 start, u64 end);
18 18
19extern int kernel_map_sync_memtype(u64 base, unsigned long size, 19extern int kernel_map_sync_memtype(u64 base, unsigned long size,
20 unsigned long flag); 20 unsigned long flag);
21extern void map_devmem(unsigned long pfn, unsigned long size,
22 struct pgprot vma_prot);
23extern void unmap_devmem(unsigned long pfn, unsigned long size,
24 struct pgprot vma_prot);
25 21
26#endif /* _ASM_X86_PAT_H */ 22#endif /* _ASM_X86_PAT_H */
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index fcf4d92e7e04..c2cceae709c8 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -138,7 +138,7 @@ extern struct tss_struct doublefault_tss;
138extern __u32 cleared_cpu_caps[NCAPINTS]; 138extern __u32 cleared_cpu_caps[NCAPINTS];
139 139
140#ifdef CONFIG_SMP 140#ifdef CONFIG_SMP
141DECLARE_PER_CPU(struct cpuinfo_x86, cpu_info); 141DECLARE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
142#define cpu_data(cpu) per_cpu(cpu_info, cpu) 142#define cpu_data(cpu) per_cpu(cpu_info, cpu)
143#define current_cpu_data __get_cpu_var(cpu_info) 143#define current_cpu_data __get_cpu_var(cpu_info)
144#else 144#else
@@ -270,7 +270,7 @@ struct tss_struct {
270 270
271} ____cacheline_aligned; 271} ____cacheline_aligned;
272 272
273DECLARE_PER_CPU(struct tss_struct, init_tss); 273DECLARE_PER_CPU_SHARED_ALIGNED(struct tss_struct, init_tss);
274 274
275/* 275/*
276 * Save the original ist values for checking stack pointers during debugging 276 * Save the original ist values for checking stack pointers during debugging
@@ -393,7 +393,7 @@ union irq_stack_union {
393 }; 393 };
394}; 394};
395 395
396DECLARE_PER_CPU(union irq_stack_union, irq_stack_union); 396DECLARE_PER_CPU_FIRST(union irq_stack_union, irq_stack_union);
397DECLARE_INIT_PER_CPU(irq_stack_union); 397DECLARE_INIT_PER_CPU(irq_stack_union);
398 398
399DECLARE_PER_CPU(char *, irq_stack_ptr); 399DECLARE_PER_CPU(char *, irq_stack_ptr);
diff --git a/arch/x86/include/asm/required-features.h b/arch/x86/include/asm/required-features.h
index d5cd6c586881..a4737dddfd58 100644
--- a/arch/x86/include/asm/required-features.h
+++ b/arch/x86/include/asm/required-features.h
@@ -50,7 +50,7 @@
50#ifdef CONFIG_X86_64 50#ifdef CONFIG_X86_64
51#define NEED_PSE 0 51#define NEED_PSE 0
52#define NEED_MSR (1<<(X86_FEATURE_MSR & 31)) 52#define NEED_MSR (1<<(X86_FEATURE_MSR & 31))
53#define NEED_PGE (1<<(X86_FEATURE_PGE & 31)) 53#define NEED_PGE 0
54#define NEED_FXSR (1<<(X86_FEATURE_FXSR & 31)) 54#define NEED_FXSR (1<<(X86_FEATURE_FXSR & 31))
55#define NEED_XMM (1<<(X86_FEATURE_XMM & 31)) 55#define NEED_XMM (1<<(X86_FEATURE_XMM & 31))
56#define NEED_XMM2 (1<<(X86_FEATURE_XMM2 & 31)) 56#define NEED_XMM2 (1<<(X86_FEATURE_XMM2 & 31))
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index e2927c5f45b1..a5ecc9c33e92 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -152,7 +152,7 @@ struct tlb_state {
152 struct mm_struct *active_mm; 152 struct mm_struct *active_mm;
153 int state; 153 int state;
154}; 154};
155DECLARE_PER_CPU(struct tlb_state, cpu_tlbstate); 155DECLARE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate);
156 156
157static inline void reset_lazy_tlbstate(void) 157static inline void reset_lazy_tlbstate(void)
158{ 158{
diff --git a/arch/x86/include/asm/topology.h b/arch/x86/include/asm/topology.h
index 892b119dba6f..f44b49abca49 100644
--- a/arch/x86/include/asm/topology.h
+++ b/arch/x86/include/asm/topology.h
@@ -200,7 +200,7 @@ static inline void arch_fix_phys_package_id(int num, u32 slot)
200} 200}
201 201
202struct pci_bus; 202struct pci_bus;
203void set_pci_bus_resources_arch_default(struct pci_bus *b); 203void x86_pci_root_bus_res_quirks(struct pci_bus *b);
204 204
205#ifdef CONFIG_SMP 205#ifdef CONFIG_SMP
206#define mc_capable() (cpumask_weight(cpu_core_mask(0)) != nr_cpu_ids) 206#define mc_capable() (cpumask_weight(cpu_core_mask(0)) != nr_cpu_ids)
diff --git a/arch/x86/include/asm/uv/uv_mmrs.h b/arch/x86/include/asm/uv/uv_mmrs.h
index db68ac8a5ac2..2cae46c7c8a2 100644
--- a/arch/x86/include/asm/uv/uv_mmrs.h
+++ b/arch/x86/include/asm/uv/uv_mmrs.h
@@ -17,6 +17,11 @@
17/* ========================================================================= */ 17/* ========================================================================= */
18/* UVH_BAU_DATA_CONFIG */ 18/* UVH_BAU_DATA_CONFIG */
19/* ========================================================================= */ 19/* ========================================================================= */
20#define UVH_LB_BAU_MISC_CONTROL 0x320170UL
21#define UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT 15
22#define UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT 16
23#define UV_INTD_SOFT_ACK_TIMEOUT_PERIOD 0x000000000bUL
24/* 1011 timebase 7 (168millisec) * 3 ticks -> 500ms */
20#define UVH_BAU_DATA_CONFIG 0x61680UL 25#define UVH_BAU_DATA_CONFIG 0x61680UL
21#define UVH_BAU_DATA_CONFIG_32 0x0438 26#define UVH_BAU_DATA_CONFIG_32 0x0438
22 27
diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
index 1a918dde46b5..018a0a400799 100644
--- a/arch/x86/include/asm/xen/page.h
+++ b/arch/x86/include/asm/xen/page.h
@@ -124,7 +124,8 @@ static inline unsigned long mfn_to_local_pfn(unsigned long mfn)
124 124
125/* VIRT <-> MACHINE conversion */ 125/* VIRT <-> MACHINE conversion */
126#define virt_to_machine(v) (phys_to_machine(XPADDR(__pa(v)))) 126#define virt_to_machine(v) (phys_to_machine(XPADDR(__pa(v))))
127#define virt_to_mfn(v) (pfn_to_mfn(PFN_DOWN(__pa(v)))) 127#define virt_to_pfn(v) (PFN_DOWN(__pa(v)))
128#define virt_to_mfn(v) (pfn_to_mfn(virt_to_pfn(v)))
128#define mfn_to_virt(m) (__va(mfn_to_pfn(m) << PAGE_SHIFT)) 129#define mfn_to_virt(m) (__va(mfn_to_pfn(m) << PAGE_SHIFT))
129 130
130static inline unsigned long pte_mfn(pte_t pte) 131static inline unsigned long pte_mfn(pte_t pte)
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 42c33cebf00f..8c0be0902dac 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -49,10 +49,10 @@
49#define IVHD_DEV_EXT_SELECT 0x46 49#define IVHD_DEV_EXT_SELECT 0x46
50#define IVHD_DEV_EXT_SELECT_RANGE 0x47 50#define IVHD_DEV_EXT_SELECT_RANGE 0x47
51 51
52#define IVHD_FLAG_HT_TUN_EN 0x00 52#define IVHD_FLAG_HT_TUN_EN_MASK 0x01
53#define IVHD_FLAG_PASSPW_EN 0x01 53#define IVHD_FLAG_PASSPW_EN_MASK 0x02
54#define IVHD_FLAG_RESPASSPW_EN 0x02 54#define IVHD_FLAG_RESPASSPW_EN_MASK 0x04
55#define IVHD_FLAG_ISOC_EN 0x03 55#define IVHD_FLAG_ISOC_EN_MASK 0x08
56 56
57#define IVMD_FLAG_EXCL_RANGE 0x08 57#define IVMD_FLAG_EXCL_RANGE 0x08
58#define IVMD_FLAG_UNITY_MAP 0x01 58#define IVMD_FLAG_UNITY_MAP 0x01
@@ -569,19 +569,19 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu,
569 * First set the recommended feature enable bits from ACPI 569 * First set the recommended feature enable bits from ACPI
570 * into the IOMMU control registers 570 * into the IOMMU control registers
571 */ 571 */
572 h->flags & IVHD_FLAG_HT_TUN_EN ? 572 h->flags & IVHD_FLAG_HT_TUN_EN_MASK ?
573 iommu_feature_enable(iommu, CONTROL_HT_TUN_EN) : 573 iommu_feature_enable(iommu, CONTROL_HT_TUN_EN) :
574 iommu_feature_disable(iommu, CONTROL_HT_TUN_EN); 574 iommu_feature_disable(iommu, CONTROL_HT_TUN_EN);
575 575
576 h->flags & IVHD_FLAG_PASSPW_EN ? 576 h->flags & IVHD_FLAG_PASSPW_EN_MASK ?
577 iommu_feature_enable(iommu, CONTROL_PASSPW_EN) : 577 iommu_feature_enable(iommu, CONTROL_PASSPW_EN) :
578 iommu_feature_disable(iommu, CONTROL_PASSPW_EN); 578 iommu_feature_disable(iommu, CONTROL_PASSPW_EN);
579 579
580 h->flags & IVHD_FLAG_RESPASSPW_EN ? 580 h->flags & IVHD_FLAG_RESPASSPW_EN_MASK ?
581 iommu_feature_enable(iommu, CONTROL_RESPASSPW_EN) : 581 iommu_feature_enable(iommu, CONTROL_RESPASSPW_EN) :
582 iommu_feature_disable(iommu, CONTROL_RESPASSPW_EN); 582 iommu_feature_disable(iommu, CONTROL_RESPASSPW_EN);
583 583
584 h->flags & IVHD_FLAG_ISOC_EN ? 584 h->flags & IVHD_FLAG_ISOC_EN_MASK ?
585 iommu_feature_enable(iommu, CONTROL_ISOC_EN) : 585 iommu_feature_enable(iommu, CONTROL_ISOC_EN) :
586 iommu_feature_disable(iommu, CONTROL_ISOC_EN); 586 iommu_feature_disable(iommu, CONTROL_ISOC_EN);
587 587
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index a2789e42e162..30da617d18e4 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -3670,12 +3670,14 @@ int arch_setup_hpet_msi(unsigned int irq)
3670{ 3670{
3671 int ret; 3671 int ret;
3672 struct msi_msg msg; 3672 struct msi_msg msg;
3673 struct irq_desc *desc = irq_to_desc(irq);
3673 3674
3674 ret = msi_compose_msg(NULL, irq, &msg); 3675 ret = msi_compose_msg(NULL, irq, &msg);
3675 if (ret < 0) 3676 if (ret < 0)
3676 return ret; 3677 return ret;
3677 3678
3678 hpet_msi_write(irq, &msg); 3679 hpet_msi_write(irq, &msg);
3680 desc->status |= IRQ_MOVE_PCNTXT;
3679 set_irq_chip_and_handler_name(irq, &hpet_msi_type, handle_edge_irq, 3681 set_irq_chip_and_handler_name(irq, &hpet_msi_type, handle_edge_irq,
3680 "edge"); 3682 "edge");
3681 3683
diff --git a/arch/x86/kernel/apic/nmi.c b/arch/x86/kernel/apic/nmi.c
index d6bd62407152..ce4fbfa315a1 100644
--- a/arch/x86/kernel/apic/nmi.c
+++ b/arch/x86/kernel/apic/nmi.c
@@ -138,7 +138,7 @@ int __init check_nmi_watchdog(void)
138 if (!prev_nmi_count) 138 if (!prev_nmi_count)
139 goto error; 139 goto error;
140 140
141 alloc_cpumask_var(&backtrace_mask, GFP_KERNEL); 141 alloc_cpumask_var(&backtrace_mask, GFP_KERNEL|__GFP_ZERO);
142 printk(KERN_INFO "Testing NMI watchdog ... "); 142 printk(KERN_INFO "Testing NMI watchdog ... ");
143 143
144#ifdef CONFIG_SMP 144#ifdef CONFIG_SMP
@@ -414,7 +414,8 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason)
414 touched = 1; 414 touched = 1;
415 } 415 }
416 416
417 if (cpumask_test_cpu(cpu, backtrace_mask)) { 417 /* We can be called before check_nmi_watchdog, hence NULL check. */
418 if (backtrace_mask != NULL && cpumask_test_cpu(cpu, backtrace_mask)) {
418 static DEFINE_SPINLOCK(lock); /* Serialise the printks */ 419 static DEFINE_SPINLOCK(lock); /* Serialise the printks */
419 420
420 spin_lock(&lock); 421 spin_lock(&lock);
diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c
index 1248318436e8..2bda69352976 100644
--- a/arch/x86/kernel/apic/x2apic_uv_x.c
+++ b/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -19,6 +19,7 @@
19#include <linux/timer.h> 19#include <linux/timer.h>
20#include <linux/cpu.h> 20#include <linux/cpu.h>
21#include <linux/init.h> 21#include <linux/init.h>
22#include <linux/io.h>
22 23
23#include <asm/uv/uv_mmrs.h> 24#include <asm/uv/uv_mmrs.h>
24#include <asm/uv/uv_hub.h> 25#include <asm/uv/uv_hub.h>
@@ -34,6 +35,17 @@ DEFINE_PER_CPU(int, x2apic_extra_bits);
34 35
35static enum uv_system_type uv_system_type; 36static enum uv_system_type uv_system_type;
36 37
38static int early_get_nodeid(void)
39{
40 union uvh_node_id_u node_id;
41 unsigned long *mmr;
42
43 mmr = early_ioremap(UV_LOCAL_MMR_BASE | UVH_NODE_ID, sizeof(*mmr));
44 node_id.v = *mmr;
45 early_iounmap(mmr, sizeof(*mmr));
46 return node_id.s.node_id;
47}
48
37static int uv_acpi_madt_oem_check(char *oem_id, char *oem_table_id) 49static int uv_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
38{ 50{
39 if (!strcmp(oem_id, "SGI")) { 51 if (!strcmp(oem_id, "SGI")) {
@@ -42,6 +54,8 @@ static int uv_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
42 else if (!strcmp(oem_table_id, "UVX")) 54 else if (!strcmp(oem_table_id, "UVX"))
43 uv_system_type = UV_X2APIC; 55 uv_system_type = UV_X2APIC;
44 else if (!strcmp(oem_table_id, "UVH")) { 56 else if (!strcmp(oem_table_id, "UVH")) {
57 __get_cpu_var(x2apic_extra_bits) =
58 early_get_nodeid() << (UV_APIC_PNODE_SHIFT - 1);
45 uv_system_type = UV_NON_UNIQUE_APIC; 59 uv_system_type = UV_NON_UNIQUE_APIC;
46 return 1; 60 return 1;
47 } 61 }
@@ -549,7 +563,8 @@ void __init uv_system_init(void)
549 unsigned long gnode_upper, lowmem_redir_base, lowmem_redir_size; 563 unsigned long gnode_upper, lowmem_redir_base, lowmem_redir_size;
550 int bytes, nid, cpu, lcpu, pnode, blade, i, j, m_val, n_val; 564 int bytes, nid, cpu, lcpu, pnode, blade, i, j, m_val, n_val;
551 int max_pnode = 0; 565 int max_pnode = 0;
552 unsigned long mmr_base, present; 566 unsigned long mmr_base, present, paddr;
567 unsigned short pnode_mask;
553 568
554 map_low_mmrs(); 569 map_low_mmrs();
555 570
@@ -592,6 +607,7 @@ void __init uv_system_init(void)
592 } 607 }
593 } 608 }
594 609
610 pnode_mask = (1 << n_val) - 1;
595 node_id.v = uv_read_local_mmr(UVH_NODE_ID); 611 node_id.v = uv_read_local_mmr(UVH_NODE_ID);
596 gnode_upper = (((unsigned long)node_id.s.node_id) & 612 gnode_upper = (((unsigned long)node_id.s.node_id) &
597 ~((1 << n_val) - 1)) << m_val; 613 ~((1 << n_val) - 1)) << m_val;
@@ -615,7 +631,7 @@ void __init uv_system_init(void)
615 uv_cpu_hub_info(cpu)->numa_blade_id = blade; 631 uv_cpu_hub_info(cpu)->numa_blade_id = blade;
616 uv_cpu_hub_info(cpu)->blade_processor_id = lcpu; 632 uv_cpu_hub_info(cpu)->blade_processor_id = lcpu;
617 uv_cpu_hub_info(cpu)->pnode = pnode; 633 uv_cpu_hub_info(cpu)->pnode = pnode;
618 uv_cpu_hub_info(cpu)->pnode_mask = (1 << n_val) - 1; 634 uv_cpu_hub_info(cpu)->pnode_mask = pnode_mask;
619 uv_cpu_hub_info(cpu)->gpa_mask = (1 << (m_val + n_val)) - 1; 635 uv_cpu_hub_info(cpu)->gpa_mask = (1 << (m_val + n_val)) - 1;
620 uv_cpu_hub_info(cpu)->gnode_upper = gnode_upper; 636 uv_cpu_hub_info(cpu)->gnode_upper = gnode_upper;
621 uv_cpu_hub_info(cpu)->global_mmr_base = mmr_base; 637 uv_cpu_hub_info(cpu)->global_mmr_base = mmr_base;
@@ -631,6 +647,17 @@ void __init uv_system_init(void)
631 lcpu, blade); 647 lcpu, blade);
632 } 648 }
633 649
650 /* Add blade/pnode info for nodes without cpus */
651 for_each_online_node(nid) {
652 if (uv_node_to_blade[nid] >= 0)
653 continue;
654 paddr = node_start_pfn(nid) << PAGE_SHIFT;
655 paddr = uv_soc_phys_ram_to_gpa(paddr);
656 pnode = (paddr >> m_val) & pnode_mask;
657 blade = boot_pnode_to_blade(pnode);
658 uv_node_to_blade[nid] = blade;
659 }
660
634 map_gru_high(max_pnode); 661 map_gru_high(max_pnode);
635 map_mmr_high(max_pnode); 662 map_mmr_high(max_pnode);
636 map_config_high(max_pnode); 663 map_config_high(max_pnode);
diff --git a/arch/x86/kernel/bios_uv.c b/arch/x86/kernel/bios_uv.c
index f63882728d91..63a88e1f987d 100644
--- a/arch/x86/kernel/bios_uv.c
+++ b/arch/x86/kernel/bios_uv.c
@@ -182,7 +182,8 @@ void uv_bios_init(void)
182 memcpy(&uv_systab, tab, sizeof(struct uv_systab)); 182 memcpy(&uv_systab, tab, sizeof(struct uv_systab));
183 iounmap(tab); 183 iounmap(tab);
184 184
185 printk(KERN_INFO "EFI UV System Table Revision %d\n", tab->revision); 185 printk(KERN_INFO "EFI UV System Table Revision %d\n",
186 uv_systab.revision);
186} 187}
187#else /* !CONFIG_EFI */ 188#else /* !CONFIG_EFI */
188 189
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index c4f667896c28..c1caefc82e62 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -1203,6 +1203,8 @@ void __cpuinit cpu_init(void)
1203 load_TR_desc(); 1203 load_TR_desc();
1204 load_LDT(&init_mm.context); 1204 load_LDT(&init_mm.context);
1205 1205
1206 t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap);
1207
1206#ifdef CONFIG_DOUBLEFAULT 1208#ifdef CONFIG_DOUBLEFAULT
1207 /* Set up doublefault TSS pointer in the GDT */ 1209 /* Set up doublefault TSS pointer in the GDT */
1208 __set_tss_desc(cpu, GDT_ENTRY_DOUBLEFAULT_TSS, &doublefault_tss); 1210 __set_tss_desc(cpu, GDT_ENTRY_DOUBLEFAULT_TSS, &doublefault_tss);
diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
index 3e3cd3db7a0c..208ecf6643df 100644
--- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
+++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
@@ -65,14 +65,18 @@ enum {
65struct acpi_cpufreq_data { 65struct acpi_cpufreq_data {
66 struct acpi_processor_performance *acpi_data; 66 struct acpi_processor_performance *acpi_data;
67 struct cpufreq_frequency_table *freq_table; 67 struct cpufreq_frequency_table *freq_table;
68 unsigned int max_freq;
69 unsigned int resume; 68 unsigned int resume;
70 unsigned int cpu_feature; 69 unsigned int cpu_feature;
71 u64 saved_aperf, saved_mperf;
72}; 70};
73 71
74static DEFINE_PER_CPU(struct acpi_cpufreq_data *, drv_data); 72static DEFINE_PER_CPU(struct acpi_cpufreq_data *, drv_data);
75 73
74struct acpi_msr_data {
75 u64 saved_aperf, saved_mperf;
76};
77
78static DEFINE_PER_CPU(struct acpi_msr_data, msr_data);
79
76DEFINE_TRACE(power_mark); 80DEFINE_TRACE(power_mark);
77 81
78/* acpi_perf_data is a pointer to percpu data. */ 82/* acpi_perf_data is a pointer to percpu data. */
@@ -204,7 +208,13 @@ static void drv_read(struct drv_cmd *cmd)
204 208
205static void drv_write(struct drv_cmd *cmd) 209static void drv_write(struct drv_cmd *cmd)
206{ 210{
211 int this_cpu;
212
213 this_cpu = get_cpu();
214 if (cpumask_test_cpu(this_cpu, cmd->mask))
215 do_drv_write(cmd);
207 smp_call_function_many(cmd->mask, do_drv_write, cmd, 1); 216 smp_call_function_many(cmd->mask, do_drv_write, cmd, 1);
217 put_cpu();
208} 218}
209 219
210static u32 get_cur_val(const struct cpumask *mask) 220static u32 get_cur_val(const struct cpumask *mask)
@@ -277,15 +287,15 @@ static unsigned int get_measured_perf(struct cpufreq_policy *policy,
277 unsigned int perf_percent; 287 unsigned int perf_percent;
278 unsigned int retval; 288 unsigned int retval;
279 289
280 if (smp_call_function_single(cpu, read_measured_perf_ctrs, &cur, 1)) 290 if (smp_call_function_single(cpu, read_measured_perf_ctrs, &readin, 1))
281 return 0; 291 return 0;
282 292
283 cur.aperf.whole = readin.aperf.whole - 293 cur.aperf.whole = readin.aperf.whole -
284 per_cpu(drv_data, cpu)->saved_aperf; 294 per_cpu(msr_data, cpu).saved_aperf;
285 cur.mperf.whole = readin.mperf.whole - 295 cur.mperf.whole = readin.mperf.whole -
286 per_cpu(drv_data, cpu)->saved_mperf; 296 per_cpu(msr_data, cpu).saved_mperf;
287 per_cpu(drv_data, cpu)->saved_aperf = readin.aperf.whole; 297 per_cpu(msr_data, cpu).saved_aperf = readin.aperf.whole;
288 per_cpu(drv_data, cpu)->saved_mperf = readin.mperf.whole; 298 per_cpu(msr_data, cpu).saved_mperf = readin.mperf.whole;
289 299
290#ifdef __i386__ 300#ifdef __i386__
291 /* 301 /*
@@ -329,7 +339,7 @@ static unsigned int get_measured_perf(struct cpufreq_policy *policy,
329 339
330#endif 340#endif
331 341
332 retval = per_cpu(drv_data, policy->cpu)->max_freq * perf_percent / 100; 342 retval = (policy->cpuinfo.max_freq * perf_percent) / 100;
333 343
334 return retval; 344 return retval;
335} 345}
@@ -682,16 +692,11 @@ static int acpi_cpufreq_cpu_init(struct cpufreq_policy *policy)
682 /* Check for high latency (>20uS) from buggy BIOSes, like on T42 */ 692 /* Check for high latency (>20uS) from buggy BIOSes, like on T42 */
683 if (perf->control_register.space_id == ACPI_ADR_SPACE_FIXED_HARDWARE && 693 if (perf->control_register.space_id == ACPI_ADR_SPACE_FIXED_HARDWARE &&
684 policy->cpuinfo.transition_latency > 20 * 1000) { 694 policy->cpuinfo.transition_latency > 20 * 1000) {
685 static int print_once;
686 policy->cpuinfo.transition_latency = 20 * 1000; 695 policy->cpuinfo.transition_latency = 20 * 1000;
687 if (!print_once) { 696 printk_once(KERN_INFO "Capping off P-state tranision"
688 print_once = 1; 697 " latency at 20 uS\n");
689 printk(KERN_INFO "Capping off P-state tranision latency"
690 " at 20 uS\n");
691 }
692 } 698 }
693 699
694 data->max_freq = perf->states[0].core_frequency * 1000;
695 /* table init */ 700 /* table init */
696 for (i = 0; i < perf->state_count; i++) { 701 for (i = 0; i < perf->state_count; i++) {
697 if (i > 0 && perf->states[i].core_frequency >= 702 if (i > 0 && perf->states[i].core_frequency >=
@@ -710,6 +715,9 @@ static int acpi_cpufreq_cpu_init(struct cpufreq_policy *policy)
710 if (result) 715 if (result)
711 goto err_freqfree; 716 goto err_freqfree;
712 717
718 if (perf->states[0].core_frequency * 1000 != policy->cpuinfo.max_freq)
719 printk(KERN_WARNING FW_WARN "P-state 0 is not max freq\n");
720
713 switch (perf->control_register.space_id) { 721 switch (perf->control_register.space_id) {
714 case ACPI_ADR_SPACE_SYSTEM_IO: 722 case ACPI_ADR_SPACE_SYSTEM_IO:
715 /* Current speed is unknown and not detectable by IO port */ 723 /* Current speed is unknown and not detectable by IO port */
diff --git a/arch/x86/kernel/cpu/mcheck/mce_64.c b/arch/x86/kernel/cpu/mcheck/mce_64.c
index 863f89568b1a..6fb0b359d2a5 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_64.c
@@ -239,9 +239,10 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b)
239 * Don't get the IP here because it's unlikely to 239 * Don't get the IP here because it's unlikely to
240 * have anything to do with the actual error location. 240 * have anything to do with the actual error location.
241 */ 241 */
242 242 if (!(flags & MCP_DONTLOG)) {
243 mce_log(&m); 243 mce_log(&m);
244 add_taint(TAINT_MACHINE_CHECK); 244 add_taint(TAINT_MACHINE_CHECK);
245 }
245 246
246 /* 247 /*
247 * Clear state for this bank. 248 * Clear state for this bank.
@@ -452,13 +453,14 @@ void mce_log_therm_throt_event(__u64 status)
452 */ 453 */
453 454
454static int check_interval = 5 * 60; /* 5 minutes */ 455static int check_interval = 5 * 60; /* 5 minutes */
455static int next_interval; /* in jiffies */ 456static DEFINE_PER_CPU(int, next_interval); /* in jiffies */
456static void mcheck_timer(unsigned long); 457static void mcheck_timer(unsigned long);
457static DEFINE_PER_CPU(struct timer_list, mce_timer); 458static DEFINE_PER_CPU(struct timer_list, mce_timer);
458 459
459static void mcheck_timer(unsigned long data) 460static void mcheck_timer(unsigned long data)
460{ 461{
461 struct timer_list *t = &per_cpu(mce_timer, data); 462 struct timer_list *t = &per_cpu(mce_timer, data);
463 int *n;
462 464
463 WARN_ON(smp_processor_id() != data); 465 WARN_ON(smp_processor_id() != data);
464 466
@@ -470,14 +472,14 @@ static void mcheck_timer(unsigned long data)
470 * Alert userspace if needed. If we logged an MCE, reduce the 472 * Alert userspace if needed. If we logged an MCE, reduce the
471 * polling interval, otherwise increase the polling interval. 473 * polling interval, otherwise increase the polling interval.
472 */ 474 */
475 n = &__get_cpu_var(next_interval);
473 if (mce_notify_user()) { 476 if (mce_notify_user()) {
474 next_interval = max(next_interval/2, HZ/100); 477 *n = max(*n/2, HZ/100);
475 } else { 478 } else {
476 next_interval = min(next_interval * 2, 479 *n = min(*n*2, (int)round_jiffies_relative(check_interval*HZ));
477 (int)round_jiffies_relative(check_interval*HZ));
478 } 480 }
479 481
480 t->expires = jiffies + next_interval; 482 t->expires = jiffies + *n;
481 add_timer(t); 483 add_timer(t);
482} 484}
483 485
@@ -584,7 +586,7 @@ static void mce_init(void *dummy)
584 * Log the machine checks left over from the previous reset. 586 * Log the machine checks left over from the previous reset.
585 */ 587 */
586 bitmap_fill(all_banks, MAX_NR_BANKS); 588 bitmap_fill(all_banks, MAX_NR_BANKS);
587 machine_check_poll(MCP_UC, &all_banks); 589 machine_check_poll(MCP_UC|(!mce_bootlog ? MCP_DONTLOG : 0), &all_banks);
588 590
589 set_in_cr4(X86_CR4_MCE); 591 set_in_cr4(X86_CR4_MCE);
590 592
@@ -632,14 +634,13 @@ static void mce_cpu_features(struct cpuinfo_x86 *c)
632static void mce_init_timer(void) 634static void mce_init_timer(void)
633{ 635{
634 struct timer_list *t = &__get_cpu_var(mce_timer); 636 struct timer_list *t = &__get_cpu_var(mce_timer);
637 int *n = &__get_cpu_var(next_interval);
635 638
636 /* data race harmless because everyone sets to the same value */ 639 *n = check_interval * HZ;
637 if (!next_interval) 640 if (!*n)
638 next_interval = check_interval * HZ;
639 if (!next_interval)
640 return; 641 return;
641 setup_timer(t, mcheck_timer, smp_processor_id()); 642 setup_timer(t, mcheck_timer, smp_processor_id());
642 t->expires = round_jiffies(jiffies + next_interval); 643 t->expires = round_jiffies(jiffies + *n);
643 add_timer(t); 644 add_timer(t);
644} 645}
645 646
@@ -907,7 +908,6 @@ static void mce_cpu_restart(void *data)
907/* Reinit MCEs after user configuration changes */ 908/* Reinit MCEs after user configuration changes */
908static void mce_restart(void) 909static void mce_restart(void)
909{ 910{
910 next_interval = check_interval * HZ;
911 on_each_cpu(mce_cpu_restart, NULL, 1); 911 on_each_cpu(mce_cpu_restart, NULL, 1);
912} 912}
913 913
@@ -1110,7 +1110,8 @@ static int __cpuinit mce_cpu_callback(struct notifier_block *nfb,
1110 break; 1110 break;
1111 case CPU_DOWN_FAILED: 1111 case CPU_DOWN_FAILED:
1112 case CPU_DOWN_FAILED_FROZEN: 1112 case CPU_DOWN_FAILED_FROZEN:
1113 t->expires = round_jiffies(jiffies + next_interval); 1113 t->expires = round_jiffies(jiffies +
1114 __get_cpu_var(next_interval));
1114 add_timer_on(t, cpu); 1115 add_timer_on(t, cpu);
1115 smp_call_function_single(cpu, mce_reenable_cpu, &action, 1); 1116 smp_call_function_single(cpu, mce_reenable_cpu, &action, 1);
1116 break; 1117 break;
diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c
index f93047fed791..d5e30397246b 100644
--- a/arch/x86/kernel/cpu/proc.c
+++ b/arch/x86/kernel/cpu/proc.c
@@ -14,7 +14,7 @@ static void show_cpuinfo_core(struct seq_file *m, struct cpuinfo_x86 *c,
14 if (c->x86_max_cores * smp_num_siblings > 1) { 14 if (c->x86_max_cores * smp_num_siblings > 1) {
15 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id); 15 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id);
16 seq_printf(m, "siblings\t: %d\n", 16 seq_printf(m, "siblings\t: %d\n",
17 cpumask_weight(cpu_sibling_mask(cpu))); 17 cpumask_weight(cpu_core_mask(cpu)));
18 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id); 18 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id);
19 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores); 19 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
20 seq_printf(m, "apicid\t\t: %d\n", c->apicid); 20 seq_printf(m, "apicid\t\t: %d\n", c->apicid);
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index 1ac99865591c..987f91f0f755 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -1397,7 +1397,10 @@ ENTRY(paranoid_exit)
1397paranoid_swapgs: 1397paranoid_swapgs:
1398 TRACE_IRQS_IRETQ 0 1398 TRACE_IRQS_IRETQ 0
1399 SWAPGS_UNSAFE_STACK 1399 SWAPGS_UNSAFE_STACK
1400 RESTORE_ALL 8
1401 jmp irq_return
1400paranoid_restore: 1402paranoid_restore:
1403 TRACE_IRQS_IRETQ 0
1401 RESTORE_ALL 8 1404 RESTORE_ALL 8
1402 jmp irq_return 1405 jmp irq_return
1403paranoid_userspace: 1406paranoid_userspace:
diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c
index 648b3a2a3a44..81408b93f887 100644
--- a/arch/x86/kernel/hpet.c
+++ b/arch/x86/kernel/hpet.c
@@ -236,6 +236,10 @@ static void hpet_stop_counter(void)
236 unsigned long cfg = hpet_readl(HPET_CFG); 236 unsigned long cfg = hpet_readl(HPET_CFG);
237 cfg &= ~HPET_CFG_ENABLE; 237 cfg &= ~HPET_CFG_ENABLE;
238 hpet_writel(cfg, HPET_CFG); 238 hpet_writel(cfg, HPET_CFG);
239}
240
241static void hpet_reset_counter(void)
242{
239 hpet_writel(0, HPET_COUNTER); 243 hpet_writel(0, HPET_COUNTER);
240 hpet_writel(0, HPET_COUNTER + 4); 244 hpet_writel(0, HPET_COUNTER + 4);
241} 245}
@@ -250,6 +254,7 @@ static void hpet_start_counter(void)
250static void hpet_restart_counter(void) 254static void hpet_restart_counter(void)
251{ 255{
252 hpet_stop_counter(); 256 hpet_stop_counter();
257 hpet_reset_counter();
253 hpet_start_counter(); 258 hpet_start_counter();
254} 259}
255 260
@@ -309,7 +314,7 @@ static int hpet_setup_msi_irq(unsigned int irq);
309static void hpet_set_mode(enum clock_event_mode mode, 314static void hpet_set_mode(enum clock_event_mode mode,
310 struct clock_event_device *evt, int timer) 315 struct clock_event_device *evt, int timer)
311{ 316{
312 unsigned long cfg; 317 unsigned long cfg, cmp, now;
313 uint64_t delta; 318 uint64_t delta;
314 319
315 switch (mode) { 320 switch (mode) {
@@ -317,12 +322,23 @@ static void hpet_set_mode(enum clock_event_mode mode,
317 hpet_stop_counter(); 322 hpet_stop_counter();
318 delta = ((uint64_t)(NSEC_PER_SEC/HZ)) * evt->mult; 323 delta = ((uint64_t)(NSEC_PER_SEC/HZ)) * evt->mult;
319 delta >>= evt->shift; 324 delta >>= evt->shift;
325 now = hpet_readl(HPET_COUNTER);
326 cmp = now + (unsigned long) delta;
320 cfg = hpet_readl(HPET_Tn_CFG(timer)); 327 cfg = hpet_readl(HPET_Tn_CFG(timer));
321 /* Make sure we use edge triggered interrupts */ 328 /* Make sure we use edge triggered interrupts */
322 cfg &= ~HPET_TN_LEVEL; 329 cfg &= ~HPET_TN_LEVEL;
323 cfg |= HPET_TN_ENABLE | HPET_TN_PERIODIC | 330 cfg |= HPET_TN_ENABLE | HPET_TN_PERIODIC |
324 HPET_TN_SETVAL | HPET_TN_32BIT; 331 HPET_TN_SETVAL | HPET_TN_32BIT;
325 hpet_writel(cfg, HPET_Tn_CFG(timer)); 332 hpet_writel(cfg, HPET_Tn_CFG(timer));
333 hpet_writel(cmp, HPET_Tn_CMP(timer));
334 udelay(1);
335 /*
336 * HPET on AMD 81xx needs a second write (with HPET_TN_SETVAL
337 * cleared) to T0_CMP to set the period. The HPET_TN_SETVAL
338 * bit is automatically cleared after the first write.
339 * (See AMD-8111 HyperTransport I/O Hub Data Sheet,
340 * Publication # 24674)
341 */
326 hpet_writel((unsigned long) delta, HPET_Tn_CMP(timer)); 342 hpet_writel((unsigned long) delta, HPET_Tn_CMP(timer));
327 hpet_start_counter(); 343 hpet_start_counter();
328 hpet_print_config(); 344 hpet_print_config();
@@ -722,7 +738,7 @@ static int hpet_cpuhp_notify(struct notifier_block *n,
722/* 738/*
723 * Clock source related code 739 * Clock source related code
724 */ 740 */
725static cycle_t read_hpet(void) 741static cycle_t read_hpet(struct clocksource *cs)
726{ 742{
727 return (cycle_t)hpet_readl(HPET_COUNTER); 743 return (cycle_t)hpet_readl(HPET_COUNTER);
728} 744}
@@ -756,7 +772,7 @@ static int hpet_clocksource_register(void)
756 hpet_restart_counter(); 772 hpet_restart_counter();
757 773
758 /* Verify whether hpet counter works */ 774 /* Verify whether hpet counter works */
759 t1 = read_hpet(); 775 t1 = hpet_readl(HPET_COUNTER);
760 rdtscll(start); 776 rdtscll(start);
761 777
762 /* 778 /*
@@ -770,7 +786,7 @@ static int hpet_clocksource_register(void)
770 rdtscll(now); 786 rdtscll(now);
771 } while ((now - start) < 200000UL); 787 } while ((now - start) < 200000UL);
772 788
773 if (t1 == read_hpet()) { 789 if (t1 == hpet_readl(HPET_COUNTER)) {
774 printk(KERN_WARNING 790 printk(KERN_WARNING
775 "HPET counter not counting. HPET disabled\n"); 791 "HPET counter not counting. HPET disabled\n");
776 return -ENODEV; 792 return -ENODEV;
diff --git a/arch/x86/kernel/i8253.c b/arch/x86/kernel/i8253.c
index 3475440baa54..c2e0bb0890d4 100644
--- a/arch/x86/kernel/i8253.c
+++ b/arch/x86/kernel/i8253.c
@@ -129,7 +129,7 @@ void __init setup_pit_timer(void)
129 * to just read by itself. So use jiffies to emulate a free 129 * to just read by itself. So use jiffies to emulate a free
130 * running counter: 130 * running counter:
131 */ 131 */
132static cycle_t pit_read(void) 132static cycle_t pit_read(struct clocksource *cs)
133{ 133{
134 static int old_count; 134 static int old_count;
135 static u32 old_jifs; 135 static u32 old_jifs;
diff --git a/arch/x86/kernel/kvmclock.c b/arch/x86/kernel/kvmclock.c
index 137f2e8132df..223af43f1526 100644
--- a/arch/x86/kernel/kvmclock.c
+++ b/arch/x86/kernel/kvmclock.c
@@ -77,6 +77,11 @@ static cycle_t kvm_clock_read(void)
77 return ret; 77 return ret;
78} 78}
79 79
80static cycle_t kvm_clock_get_cycles(struct clocksource *cs)
81{
82 return kvm_clock_read();
83}
84
80/* 85/*
81 * If we don't do that, there is the possibility that the guest 86 * If we don't do that, there is the possibility that the guest
82 * will calibrate under heavy load - thus, getting a lower lpj - 87 * will calibrate under heavy load - thus, getting a lower lpj -
@@ -107,7 +112,7 @@ static void kvm_get_preset_lpj(void)
107 112
108static struct clocksource kvm_clock = { 113static struct clocksource kvm_clock = {
109 .name = "kvm-clock", 114 .name = "kvm-clock",
110 .read = kvm_clock_read, 115 .read = kvm_clock_get_cycles,
111 .rating = 400, 116 .rating = 400,
112 .mask = CLOCKSOURCE_MASK(64), 117 .mask = CLOCKSOURCE_MASK(64),
113 .mult = 1 << KVM_SCALE, 118 .mult = 1 << KVM_SCALE,
diff --git a/arch/x86/kernel/microcode_core.c b/arch/x86/kernel/microcode_core.c
index a0f3851ef310..98c470c069d1 100644
--- a/arch/x86/kernel/microcode_core.c
+++ b/arch/x86/kernel/microcode_core.c
@@ -108,40 +108,29 @@ struct ucode_cpu_info ucode_cpu_info[NR_CPUS];
108EXPORT_SYMBOL_GPL(ucode_cpu_info); 108EXPORT_SYMBOL_GPL(ucode_cpu_info);
109 109
110#ifdef CONFIG_MICROCODE_OLD_INTERFACE 110#ifdef CONFIG_MICROCODE_OLD_INTERFACE
111struct update_for_cpu {
112 const void __user *buf;
113 size_t size;
114};
115
116static long update_for_cpu(void *_ufc)
117{
118 struct update_for_cpu *ufc = _ufc;
119 int error;
120
121 error = microcode_ops->request_microcode_user(smp_processor_id(),
122 ufc->buf, ufc->size);
123 if (error < 0)
124 return error;
125 if (!error)
126 microcode_ops->apply_microcode(smp_processor_id());
127 return error;
128}
129
130static int do_microcode_update(const void __user *buf, size_t size) 111static int do_microcode_update(const void __user *buf, size_t size)
131{ 112{
113 cpumask_t old;
132 int error = 0; 114 int error = 0;
133 int cpu; 115 int cpu;
134 struct update_for_cpu ufc = { .buf = buf, .size = size }; 116
117 old = current->cpus_allowed;
135 118
136 for_each_online_cpu(cpu) { 119 for_each_online_cpu(cpu) {
137 struct ucode_cpu_info *uci = ucode_cpu_info + cpu; 120 struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
138 121
139 if (!uci->valid) 122 if (!uci->valid)
140 continue; 123 continue;
141 error = work_on_cpu(cpu, update_for_cpu, &ufc); 124
125 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
126 error = microcode_ops->request_microcode_user(cpu, buf, size);
142 if (error < 0) 127 if (error < 0)
143 break; 128 goto out;
129 if (!error)
130 microcode_ops->apply_microcode(cpu);
144 } 131 }
132out:
133 set_cpus_allowed_ptr(current, &old);
145 return error; 134 return error;
146} 135}
147 136
@@ -391,8 +380,6 @@ static int mc_sysdev_add(struct sys_device *sys_dev)
391 return err; 380 return err;
392 381
393 err = microcode_init_cpu(cpu); 382 err = microcode_init_cpu(cpu);
394 if (err)
395 sysfs_remove_group(&sys_dev->kobj, &mc_attr_group);
396 383
397 return err; 384 return err;
398} 385}
diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
index 34f12e9996ed..221a3853e268 100644
--- a/arch/x86/kernel/pci-swiotlb.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -50,7 +50,7 @@ static void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
50 return swiotlb_alloc_coherent(hwdev, size, dma_handle, flags); 50 return swiotlb_alloc_coherent(hwdev, size, dma_handle, flags);
51} 51}
52 52
53struct dma_map_ops swiotlb_dma_ops = { 53static struct dma_map_ops swiotlb_dma_ops = {
54 .mapping_error = swiotlb_dma_mapping_error, 54 .mapping_error = swiotlb_dma_mapping_error,
55 .alloc_coherent = x86_swiotlb_alloc_coherent, 55 .alloc_coherent = x86_swiotlb_alloc_coherent,
56 .free_coherent = swiotlb_free_coherent, 56 .free_coherent = swiotlb_free_coherent,
diff --git a/arch/x86/kernel/quirks.c b/arch/x86/kernel/quirks.c
index e95022e4f5d5..7563b31b4f03 100644
--- a/arch/x86/kernel/quirks.c
+++ b/arch/x86/kernel/quirks.c
@@ -261,8 +261,6 @@ static void old_ich_force_enable_hpet_user(struct pci_dev *dev)
261{ 261{
262 if (hpet_force_user) 262 if (hpet_force_user)
263 old_ich_force_enable_hpet(dev); 263 old_ich_force_enable_hpet(dev);
264 else
265 hpet_print_force_info();
266} 264}
267 265
268DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ESB_1, 266DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ESB_1,
diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c
index deb5ebb32c3b..ed0c33761e6d 100644
--- a/arch/x86/kernel/tlb_uv.c
+++ b/arch/x86/kernel/tlb_uv.c
@@ -25,6 +25,8 @@ static int uv_bau_retry_limit __read_mostly;
25 25
26/* position of pnode (which is nasid>>1): */ 26/* position of pnode (which is nasid>>1): */
27static int uv_nshift __read_mostly; 27static int uv_nshift __read_mostly;
28/* base pnode in this partition */
29static int uv_partition_base_pnode __read_mostly;
28 30
29static unsigned long uv_mmask __read_mostly; 31static unsigned long uv_mmask __read_mostly;
30 32
@@ -32,6 +34,34 @@ static DEFINE_PER_CPU(struct ptc_stats, ptcstats);
32static DEFINE_PER_CPU(struct bau_control, bau_control); 34static DEFINE_PER_CPU(struct bau_control, bau_control);
33 35
34/* 36/*
37 * Determine the first node on a blade.
38 */
39static int __init blade_to_first_node(int blade)
40{
41 int node, b;
42
43 for_each_online_node(node) {
44 b = uv_node_to_blade_id(node);
45 if (blade == b)
46 return node;
47 }
48 return -1; /* shouldn't happen */
49}
50
51/*
52 * Determine the apicid of the first cpu on a blade.
53 */
54static int __init blade_to_first_apicid(int blade)
55{
56 int cpu;
57
58 for_each_present_cpu(cpu)
59 if (blade == uv_cpu_to_blade_id(cpu))
60 return per_cpu(x86_cpu_to_apicid, cpu);
61 return -1;
62}
63
64/*
35 * Free a software acknowledge hardware resource by clearing its Pending 65 * Free a software acknowledge hardware resource by clearing its Pending
36 * bit. This will return a reply to the sender. 66 * bit. This will return a reply to the sender.
37 * If the message has timed out, a reply has already been sent by the 67 * If the message has timed out, a reply has already been sent by the
@@ -67,7 +97,7 @@ static void uv_bau_process_message(struct bau_payload_queue_entry *msg,
67 msp = __get_cpu_var(bau_control).msg_statuses + msg_slot; 97 msp = __get_cpu_var(bau_control).msg_statuses + msg_slot;
68 cpu = uv_blade_processor_id(); 98 cpu = uv_blade_processor_id();
69 msg->number_of_cpus = 99 msg->number_of_cpus =
70 uv_blade_nr_online_cpus(uv_node_to_blade_id(numa_node_id())); 100 uv_blade_nr_online_cpus(uv_node_to_blade_id(numa_node_id()));
71 this_cpu_mask = 1UL << cpu; 101 this_cpu_mask = 1UL << cpu;
72 if (msp->seen_by.bits & this_cpu_mask) 102 if (msp->seen_by.bits & this_cpu_mask)
73 return; 103 return;
@@ -215,14 +245,14 @@ static int uv_wait_completion(struct bau_desc *bau_desc,
215 * Returns @flush_mask if some remote flushing remains to be done. The 245 * Returns @flush_mask if some remote flushing remains to be done. The
216 * mask will have some bits still set. 246 * mask will have some bits still set.
217 */ 247 */
218const struct cpumask *uv_flush_send_and_wait(int cpu, int this_blade, 248const struct cpumask *uv_flush_send_and_wait(int cpu, int this_pnode,
219 struct bau_desc *bau_desc, 249 struct bau_desc *bau_desc,
220 struct cpumask *flush_mask) 250 struct cpumask *flush_mask)
221{ 251{
222 int completion_status = 0; 252 int completion_status = 0;
223 int right_shift; 253 int right_shift;
224 int tries = 0; 254 int tries = 0;
225 int blade; 255 int pnode;
226 int bit; 256 int bit;
227 unsigned long mmr_offset; 257 unsigned long mmr_offset;
228 unsigned long index; 258 unsigned long index;
@@ -265,8 +295,8 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_blade,
265 * use the IPI method of shootdown on them. 295 * use the IPI method of shootdown on them.
266 */ 296 */
267 for_each_cpu(bit, flush_mask) { 297 for_each_cpu(bit, flush_mask) {
268 blade = uv_cpu_to_blade_id(bit); 298 pnode = uv_cpu_to_pnode(bit);
269 if (blade == this_blade) 299 if (pnode == this_pnode)
270 continue; 300 continue;
271 cpumask_clear_cpu(bit, flush_mask); 301 cpumask_clear_cpu(bit, flush_mask);
272 } 302 }
@@ -309,16 +339,16 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
309 struct cpumask *flush_mask = __get_cpu_var(uv_flush_tlb_mask); 339 struct cpumask *flush_mask = __get_cpu_var(uv_flush_tlb_mask);
310 int i; 340 int i;
311 int bit; 341 int bit;
312 int blade; 342 int pnode;
313 int uv_cpu; 343 int uv_cpu;
314 int this_blade; 344 int this_pnode;
315 int locals = 0; 345 int locals = 0;
316 struct bau_desc *bau_desc; 346 struct bau_desc *bau_desc;
317 347
318 cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu)); 348 cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu));
319 349
320 uv_cpu = uv_blade_processor_id(); 350 uv_cpu = uv_blade_processor_id();
321 this_blade = uv_numa_blade_id(); 351 this_pnode = uv_hub_info->pnode;
322 bau_desc = __get_cpu_var(bau_control).descriptor_base; 352 bau_desc = __get_cpu_var(bau_control).descriptor_base;
323 bau_desc += UV_ITEMS_PER_DESCRIPTOR * uv_cpu; 353 bau_desc += UV_ITEMS_PER_DESCRIPTOR * uv_cpu;
324 354
@@ -326,13 +356,14 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
326 356
327 i = 0; 357 i = 0;
328 for_each_cpu(bit, flush_mask) { 358 for_each_cpu(bit, flush_mask) {
329 blade = uv_cpu_to_blade_id(bit); 359 pnode = uv_cpu_to_pnode(bit);
330 BUG_ON(blade > (UV_DISTRIBUTION_SIZE - 1)); 360 BUG_ON(pnode > (UV_DISTRIBUTION_SIZE - 1));
331 if (blade == this_blade) { 361 if (pnode == this_pnode) {
332 locals++; 362 locals++;
333 continue; 363 continue;
334 } 364 }
335 bau_node_set(blade, &bau_desc->distribution); 365 bau_node_set(pnode - uv_partition_base_pnode,
366 &bau_desc->distribution);
336 i++; 367 i++;
337 } 368 }
338 if (i == 0) { 369 if (i == 0) {
@@ -350,7 +381,7 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
350 bau_desc->payload.address = va; 381 bau_desc->payload.address = va;
351 bau_desc->payload.sending_cpu = cpu; 382 bau_desc->payload.sending_cpu = cpu;
352 383
353 return uv_flush_send_and_wait(uv_cpu, this_blade, bau_desc, flush_mask); 384 return uv_flush_send_and_wait(uv_cpu, this_pnode, bau_desc, flush_mask);
354} 385}
355 386
356/* 387/*
@@ -418,24 +449,58 @@ void uv_bau_message_interrupt(struct pt_regs *regs)
418 set_irq_regs(old_regs); 449 set_irq_regs(old_regs);
419} 450}
420 451
452/*
453 * uv_enable_timeouts
454 *
455 * Each target blade (i.e. blades that have cpu's) needs to have
456 * shootdown message timeouts enabled. The timeout does not cause
457 * an interrupt, but causes an error message to be returned to
458 * the sender.
459 */
421static void uv_enable_timeouts(void) 460static void uv_enable_timeouts(void)
422{ 461{
423 int i;
424 int blade; 462 int blade;
425 int last_blade; 463 int nblades;
426 int pnode; 464 int pnode;
427 int cur_cpu = 0; 465 unsigned long mmr_image;
428 unsigned long apicid;
429 466
430 last_blade = -1; 467 nblades = uv_num_possible_blades();
431 for_each_online_node(i) { 468
432 blade = uv_node_to_blade_id(i); 469 for (blade = 0; blade < nblades; blade++) {
433 if (blade == last_blade) 470 if (!uv_blade_nr_possible_cpus(blade))
434 continue; 471 continue;
435 last_blade = blade; 472
436 apicid = per_cpu(x86_cpu_to_apicid, cur_cpu);
437 pnode = uv_blade_to_pnode(blade); 473 pnode = uv_blade_to_pnode(blade);
438 cur_cpu += uv_blade_nr_possible_cpus(i); 474 mmr_image =
475 uv_read_global_mmr64(pnode, UVH_LB_BAU_MISC_CONTROL);
476 /*
477 * Set the timeout period and then lock it in, in three
478 * steps; captures and locks in the period.
479 *
480 * To program the period, the SOFT_ACK_MODE must be off.
481 */
482 mmr_image &= ~((unsigned long)1 <<
483 UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT);
484 uv_write_global_mmr64
485 (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
486 /*
487 * Set the 4-bit period.
488 */
489 mmr_image &= ~((unsigned long)0xf <<
490 UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT);
491 mmr_image |= (UV_INTD_SOFT_ACK_TIMEOUT_PERIOD <<
492 UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT);
493 uv_write_global_mmr64
494 (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
495 /*
496 * Subsequent reversals of the timebase bit (3) cause an
497 * immediate timeout of one or all INTD resources as
498 * indicated in bits 2:0 (7 causes all of them to timeout).
499 */
500 mmr_image |= ((unsigned long)1 <<
501 UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT);
502 uv_write_global_mmr64
503 (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
439 } 504 }
440} 505}
441 506
@@ -482,8 +547,7 @@ static int uv_ptc_seq_show(struct seq_file *file, void *data)
482 stat->requestee, stat->onetlb, stat->alltlb, 547 stat->requestee, stat->onetlb, stat->alltlb,
483 stat->s_retry, stat->d_retry, stat->ptc_i); 548 stat->s_retry, stat->d_retry, stat->ptc_i);
484 seq_printf(file, "%lx %ld %ld %ld %ld %ld %ld\n", 549 seq_printf(file, "%lx %ld %ld %ld %ld %ld %ld\n",
485 uv_read_global_mmr64(uv_blade_to_pnode 550 uv_read_global_mmr64(uv_cpu_to_pnode(cpu),
486 (uv_cpu_to_blade_id(cpu)),
487 UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE), 551 UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE),
488 stat->sflush, stat->dflush, 552 stat->sflush, stat->dflush,
489 stat->retriesok, stat->nomsg, 553 stat->retriesok, stat->nomsg,
@@ -617,16 +681,18 @@ static struct bau_control * __init uv_table_bases_init(int blade, int node)
617 * finish the initialization of the per-blade control structures 681 * finish the initialization of the per-blade control structures
618 */ 682 */
619static void __init 683static void __init
620uv_table_bases_finish(int blade, int node, int cur_cpu, 684uv_table_bases_finish(int blade,
621 struct bau_control *bau_tablesp, 685 struct bau_control *bau_tablesp,
622 struct bau_desc *adp) 686 struct bau_desc *adp)
623{ 687{
624 struct bau_control *bcp; 688 struct bau_control *bcp;
625 int i; 689 int cpu;
626 690
627 for (i = cur_cpu; i < cur_cpu + uv_blade_nr_possible_cpus(blade); i++) { 691 for_each_present_cpu(cpu) {
628 bcp = (struct bau_control *)&per_cpu(bau_control, i); 692 if (blade != uv_cpu_to_blade_id(cpu))
693 continue;
629 694
695 bcp = (struct bau_control *)&per_cpu(bau_control, cpu);
630 bcp->bau_msg_head = bau_tablesp->va_queue_first; 696 bcp->bau_msg_head = bau_tablesp->va_queue_first;
631 bcp->va_queue_first = bau_tablesp->va_queue_first; 697 bcp->va_queue_first = bau_tablesp->va_queue_first;
632 bcp->va_queue_last = bau_tablesp->va_queue_last; 698 bcp->va_queue_last = bau_tablesp->va_queue_last;
@@ -649,11 +715,10 @@ uv_activation_descriptor_init(int node, int pnode)
649 struct bau_desc *adp; 715 struct bau_desc *adp;
650 struct bau_desc *ad2; 716 struct bau_desc *ad2;
651 717
652 adp = (struct bau_desc *) 718 adp = (struct bau_desc *)kmalloc_node(16384, GFP_KERNEL, node);
653 kmalloc_node(16384, GFP_KERNEL, node);
654 BUG_ON(!adp); 719 BUG_ON(!adp);
655 720
656 pa = __pa((unsigned long)adp); 721 pa = uv_gpa(adp); /* need the real nasid*/
657 n = pa >> uv_nshift; 722 n = pa >> uv_nshift;
658 m = pa & uv_mmask; 723 m = pa & uv_mmask;
659 724
@@ -667,8 +732,12 @@ uv_activation_descriptor_init(int node, int pnode)
667 for (i = 0, ad2 = adp; i < UV_ACTIVATION_DESCRIPTOR_SIZE; i++, ad2++) { 732 for (i = 0, ad2 = adp; i < UV_ACTIVATION_DESCRIPTOR_SIZE; i++, ad2++) {
668 memset(ad2, 0, sizeof(struct bau_desc)); 733 memset(ad2, 0, sizeof(struct bau_desc));
669 ad2->header.sw_ack_flag = 1; 734 ad2->header.sw_ack_flag = 1;
670 ad2->header.base_dest_nodeid = 735 /*
671 uv_blade_to_pnode(uv_cpu_to_blade_id(0)); 736 * base_dest_nodeid is the first node in the partition, so
737 * the bit map will indicate partition-relative node numbers.
738 * note that base_dest_nodeid is actually a nasid.
739 */
740 ad2->header.base_dest_nodeid = uv_partition_base_pnode << 1;
672 ad2->header.command = UV_NET_ENDPOINT_INTD; 741 ad2->header.command = UV_NET_ENDPOINT_INTD;
673 ad2->header.int_both = 1; 742 ad2->header.int_both = 1;
674 /* 743 /*
@@ -686,6 +755,8 @@ static struct bau_payload_queue_entry * __init
686uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp) 755uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp)
687{ 756{
688 struct bau_payload_queue_entry *pqp; 757 struct bau_payload_queue_entry *pqp;
758 unsigned long pa;
759 int pn;
689 char *cp; 760 char *cp;
690 761
691 pqp = (struct bau_payload_queue_entry *) kmalloc_node( 762 pqp = (struct bau_payload_queue_entry *) kmalloc_node(
@@ -696,10 +767,14 @@ uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp)
696 cp = (char *)pqp + 31; 767 cp = (char *)pqp + 31;
697 pqp = (struct bau_payload_queue_entry *)(((unsigned long)cp >> 5) << 5); 768 pqp = (struct bau_payload_queue_entry *)(((unsigned long)cp >> 5) << 5);
698 bau_tablesp->va_queue_first = pqp; 769 bau_tablesp->va_queue_first = pqp;
770 /*
771 * need the pnode of where the memory was really allocated
772 */
773 pa = uv_gpa(pqp);
774 pn = pa >> uv_nshift;
699 uv_write_global_mmr64(pnode, 775 uv_write_global_mmr64(pnode,
700 UVH_LB_BAU_INTD_PAYLOAD_QUEUE_FIRST, 776 UVH_LB_BAU_INTD_PAYLOAD_QUEUE_FIRST,
701 ((unsigned long)pnode << 777 ((unsigned long)pn << UV_PAYLOADQ_PNODE_SHIFT) |
702 UV_PAYLOADQ_PNODE_SHIFT) |
703 uv_physnodeaddr(pqp)); 778 uv_physnodeaddr(pqp));
704 uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_TAIL, 779 uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_TAIL,
705 uv_physnodeaddr(pqp)); 780 uv_physnodeaddr(pqp));
@@ -715,8 +790,9 @@ uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp)
715/* 790/*
716 * Initialization of each UV blade's structures 791 * Initialization of each UV blade's structures
717 */ 792 */
718static int __init uv_init_blade(int blade, int node, int cur_cpu) 793static int __init uv_init_blade(int blade)
719{ 794{
795 int node;
720 int pnode; 796 int pnode;
721 unsigned long pa; 797 unsigned long pa;
722 unsigned long apicid; 798 unsigned long apicid;
@@ -724,16 +800,17 @@ static int __init uv_init_blade(int blade, int node, int cur_cpu)
724 struct bau_payload_queue_entry *pqp; 800 struct bau_payload_queue_entry *pqp;
725 struct bau_control *bau_tablesp; 801 struct bau_control *bau_tablesp;
726 802
803 node = blade_to_first_node(blade);
727 bau_tablesp = uv_table_bases_init(blade, node); 804 bau_tablesp = uv_table_bases_init(blade, node);
728 pnode = uv_blade_to_pnode(blade); 805 pnode = uv_blade_to_pnode(blade);
729 adp = uv_activation_descriptor_init(node, pnode); 806 adp = uv_activation_descriptor_init(node, pnode);
730 pqp = uv_payload_queue_init(node, pnode, bau_tablesp); 807 pqp = uv_payload_queue_init(node, pnode, bau_tablesp);
731 uv_table_bases_finish(blade, node, cur_cpu, bau_tablesp, adp); 808 uv_table_bases_finish(blade, bau_tablesp, adp);
732 /* 809 /*
733 * the below initialization can't be in firmware because the 810 * the below initialization can't be in firmware because the
734 * messaging IRQ will be determined by the OS 811 * messaging IRQ will be determined by the OS
735 */ 812 */
736 apicid = per_cpu(x86_cpu_to_apicid, cur_cpu); 813 apicid = blade_to_first_apicid(blade);
737 pa = uv_read_global_mmr64(pnode, UVH_BAU_DATA_CONFIG); 814 pa = uv_read_global_mmr64(pnode, UVH_BAU_DATA_CONFIG);
738 if ((pa & 0xff) != UV_BAU_MESSAGE) { 815 if ((pa & 0xff) != UV_BAU_MESSAGE) {
739 uv_write_global_mmr64(pnode, UVH_BAU_DATA_CONFIG, 816 uv_write_global_mmr64(pnode, UVH_BAU_DATA_CONFIG,
@@ -748,9 +825,7 @@ static int __init uv_init_blade(int blade, int node, int cur_cpu)
748static int __init uv_bau_init(void) 825static int __init uv_bau_init(void)
749{ 826{
750 int blade; 827 int blade;
751 int node;
752 int nblades; 828 int nblades;
753 int last_blade;
754 int cur_cpu; 829 int cur_cpu;
755 830
756 if (!is_uv_system()) 831 if (!is_uv_system())
@@ -763,29 +838,21 @@ static int __init uv_bau_init(void)
763 uv_bau_retry_limit = 1; 838 uv_bau_retry_limit = 1;
764 uv_nshift = uv_hub_info->n_val; 839 uv_nshift = uv_hub_info->n_val;
765 uv_mmask = (1UL << uv_hub_info->n_val) - 1; 840 uv_mmask = (1UL << uv_hub_info->n_val) - 1;
766 nblades = 0; 841 nblades = uv_num_possible_blades();
767 last_blade = -1; 842
768 cur_cpu = 0;
769 for_each_online_node(node) {
770 blade = uv_node_to_blade_id(node);
771 if (blade == last_blade)
772 continue;
773 last_blade = blade;
774 nblades++;
775 }
776 uv_bau_table_bases = (struct bau_control **) 843 uv_bau_table_bases = (struct bau_control **)
777 kmalloc(nblades * sizeof(struct bau_control *), GFP_KERNEL); 844 kmalloc(nblades * sizeof(struct bau_control *), GFP_KERNEL);
778 BUG_ON(!uv_bau_table_bases); 845 BUG_ON(!uv_bau_table_bases);
779 846
780 last_blade = -1; 847 uv_partition_base_pnode = 0x7fffffff;
781 for_each_online_node(node) { 848 for (blade = 0; blade < nblades; blade++)
782 blade = uv_node_to_blade_id(node); 849 if (uv_blade_nr_possible_cpus(blade) &&
783 if (blade == last_blade) 850 (uv_blade_to_pnode(blade) < uv_partition_base_pnode))
784 continue; 851 uv_partition_base_pnode = uv_blade_to_pnode(blade);
785 last_blade = blade; 852 for (blade = 0; blade < nblades; blade++)
786 uv_init_blade(blade, node, cur_cpu); 853 if (uv_blade_nr_possible_cpus(blade))
787 cur_cpu += uv_blade_nr_possible_cpus(blade); 854 uv_init_blade(blade);
788 } 855
789 alloc_intr_gate(UV_BAU_MESSAGE, uv_bau_message_intr1); 856 alloc_intr_gate(UV_BAU_MESSAGE, uv_bau_message_intr1);
790 uv_enable_timeouts(); 857 uv_enable_timeouts();
791 858
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 7a567ebe6361..d57de05dc430 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -699,7 +699,7 @@ static struct clocksource clocksource_tsc;
699 * code, which is necessary to support wrapping clocksources like pm 699 * code, which is necessary to support wrapping clocksources like pm
700 * timer. 700 * timer.
701 */ 701 */
702static cycle_t read_tsc(void) 702static cycle_t read_tsc(struct clocksource *cs)
703{ 703{
704 cycle_t ret = (cycle_t)get_cycles(); 704 cycle_t ret = (cycle_t)get_cycles();
705 705
diff --git a/arch/x86/kernel/uv_sysfs.c b/arch/x86/kernel/uv_sysfs.c
index 67f9b9dbf800..36afb98675a4 100644
--- a/arch/x86/kernel/uv_sysfs.c
+++ b/arch/x86/kernel/uv_sysfs.c
@@ -21,6 +21,7 @@
21 21
22#include <linux/sysdev.h> 22#include <linux/sysdev.h>
23#include <asm/uv/bios.h> 23#include <asm/uv/bios.h>
24#include <asm/uv/uv.h>
24 25
25struct kobject *sgi_uv_kobj; 26struct kobject *sgi_uv_kobj;
26 27
@@ -47,6 +48,9 @@ static int __init sgi_uv_sysfs_init(void)
47{ 48{
48 unsigned long ret; 49 unsigned long ret;
49 50
51 if (!is_uv_system())
52 return -ENODEV;
53
50 if (!sgi_uv_kobj) 54 if (!sgi_uv_kobj)
51 sgi_uv_kobj = kobject_create_and_add("sgi_uv", firmware_kobj); 55 sgi_uv_kobj = kobject_create_and_add("sgi_uv", firmware_kobj);
52 if (!sgi_uv_kobj) { 56 if (!sgi_uv_kobj) {
diff --git a/arch/x86/kernel/uv_time.c b/arch/x86/kernel/uv_time.c
index 2ffb6c53326e..583f11d5c480 100644
--- a/arch/x86/kernel/uv_time.c
+++ b/arch/x86/kernel/uv_time.c
@@ -29,7 +29,7 @@
29 29
30#define RTC_NAME "sgi_rtc" 30#define RTC_NAME "sgi_rtc"
31 31
32static cycle_t uv_read_rtc(void); 32static cycle_t uv_read_rtc(struct clocksource *cs);
33static int uv_rtc_next_event(unsigned long, struct clock_event_device *); 33static int uv_rtc_next_event(unsigned long, struct clock_event_device *);
34static void uv_rtc_timer_setup(enum clock_event_mode, 34static void uv_rtc_timer_setup(enum clock_event_mode,
35 struct clock_event_device *); 35 struct clock_event_device *);
@@ -123,7 +123,7 @@ static int uv_setup_intr(int cpu, u64 expires)
123 /* Initialize comparator value */ 123 /* Initialize comparator value */
124 uv_write_global_mmr64(pnode, UVH_INT_CMPB, expires); 124 uv_write_global_mmr64(pnode, UVH_INT_CMPB, expires);
125 125
126 return (expires < uv_read_rtc() && !uv_intr_pending(pnode)); 126 return (expires < uv_read_rtc(NULL) && !uv_intr_pending(pnode));
127} 127}
128 128
129/* 129/*
@@ -256,7 +256,7 @@ static int uv_rtc_unset_timer(int cpu)
256 256
257 spin_lock_irqsave(&head->lock, flags); 257 spin_lock_irqsave(&head->lock, flags);
258 258
259 if (head->next_cpu == bcpu && uv_read_rtc() >= *t) 259 if (head->next_cpu == bcpu && uv_read_rtc(NULL) >= *t)
260 rc = 1; 260 rc = 1;
261 261
262 *t = ULLONG_MAX; 262 *t = ULLONG_MAX;
@@ -278,7 +278,7 @@ static int uv_rtc_unset_timer(int cpu)
278/* 278/*
279 * Read the RTC. 279 * Read the RTC.
280 */ 280 */
281static cycle_t uv_read_rtc(void) 281static cycle_t uv_read_rtc(struct clocksource *cs)
282{ 282{
283 return (cycle_t)uv_read_local_mmr(UVH_RTC); 283 return (cycle_t)uv_read_local_mmr(UVH_RTC);
284} 284}
@@ -291,7 +291,7 @@ static int uv_rtc_next_event(unsigned long delta,
291{ 291{
292 int ced_cpu = cpumask_first(ced->cpumask); 292 int ced_cpu = cpumask_first(ced->cpumask);
293 293
294 return uv_rtc_set_timer(ced_cpu, delta + uv_read_rtc()); 294 return uv_rtc_set_timer(ced_cpu, delta + uv_read_rtc(NULL));
295} 295}
296 296
297/* 297/*
diff --git a/arch/x86/kernel/vmiclock_32.c b/arch/x86/kernel/vmiclock_32.c
index d303369a7bad..2b3eb82efeeb 100644
--- a/arch/x86/kernel/vmiclock_32.c
+++ b/arch/x86/kernel/vmiclock_32.c
@@ -283,7 +283,7 @@ void __devinit vmi_time_ap_init(void)
283/** vmi clocksource */ 283/** vmi clocksource */
284static struct clocksource clocksource_vmi; 284static struct clocksource clocksource_vmi;
285 285
286static cycle_t read_real_cycles(void) 286static cycle_t read_real_cycles(struct clocksource *cs)
287{ 287{
288 cycle_t ret = (cycle_t)vmi_timer_ops.get_cycle_counter(VMI_CYCLES_REAL); 288 cycle_t ret = (cycle_t)vmi_timer_ops.get_cycle_counter(VMI_CYCLES_REAL);
289 return max(ret, clocksource_vmi.cycle_last); 289 return max(ret, clocksource_vmi.cycle_last);
diff --git a/arch/x86/kernel/xsave.c b/arch/x86/kernel/xsave.c
index 0a5b04aa98f1..c5ee17e8c6d9 100644
--- a/arch/x86/kernel/xsave.c
+++ b/arch/x86/kernel/xsave.c
@@ -89,7 +89,7 @@ int save_i387_xstate(void __user *buf)
89 89
90 if (!used_math()) 90 if (!used_math())
91 return 0; 91 return 0;
92 clear_used_math(); /* trigger finit */ 92
93 if (task_thread_info(tsk)->status & TS_USEDFPU) { 93 if (task_thread_info(tsk)->status & TS_USEDFPU) {
94 /* 94 /*
95 * Start with clearing the user buffer. This will present a 95 * Start with clearing the user buffer. This will present a
@@ -114,6 +114,8 @@ int save_i387_xstate(void __user *buf)
114 return -1; 114 return -1;
115 } 115 }
116 116
117 clear_used_math(); /* trigger finit */
118
117 if (task_thread_info(tsk)->status & TS_XSAVE) { 119 if (task_thread_info(tsk)->status & TS_XSAVE) {
118 struct _fpstate __user *fx = buf; 120 struct _fpstate __user *fx = buf;
119 struct _xstate __user *x = buf; 121 struct _xstate __user *x = buf;
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 2a36f7f7c4c7..b6caf1329b1b 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -1248,7 +1248,7 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
1248 pgprintk("%s: adding gfn %lx role %x\n", __func__, gfn, role.word); 1248 pgprintk("%s: adding gfn %lx role %x\n", __func__, gfn, role.word);
1249 sp->gfn = gfn; 1249 sp->gfn = gfn;
1250 sp->role = role; 1250 sp->role = role;
1251 sp->global = role.cr4_pge; 1251 sp->global = 0;
1252 hlist_add_head(&sp->hash_link, bucket); 1252 hlist_add_head(&sp->hash_link, bucket);
1253 if (!direct) { 1253 if (!direct) {
1254 if (rmap_write_protect(vcpu->kvm, gfn)) 1254 if (rmap_write_protect(vcpu->kvm, gfn))
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 8ca100a9ecac..7c1ce5ac6131 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2775,6 +2775,9 @@ out:
2775 2775
2776void kvm_arch_exit(void) 2776void kvm_arch_exit(void)
2777{ 2777{
2778 if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
2779 cpufreq_unregister_notifier(&kvmclock_cpufreq_notifier_block,
2780 CPUFREQ_TRANSITION_NOTIFIER);
2778 kvm_x86_ops = NULL; 2781 kvm_x86_ops = NULL;
2779 kvm_mmu_module_exit(); 2782 kvm_mmu_module_exit();
2780} 2783}
@@ -4159,6 +4162,11 @@ EXPORT_SYMBOL_GPL(kvm_put_guest_fpu);
4159 4162
4160void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu) 4163void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
4161{ 4164{
4165 if (vcpu->arch.time_page) {
4166 kvm_release_page_dirty(vcpu->arch.time_page);
4167 vcpu->arch.time_page = NULL;
4168 }
4169
4162 kvm_x86_ops->vcpu_free(vcpu); 4170 kvm_x86_ops->vcpu_free(vcpu);
4163} 4171}
4164 4172
diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c
index e94a11e42f98..ca7ec44bafc3 100644
--- a/arch/x86/lguest/boot.c
+++ b/arch/x86/lguest/boot.c
@@ -273,15 +273,15 @@ static void lguest_load_idt(const struct desc_ptr *desc)
273 * controls the entire thing and the Guest asks it to make changes using the 273 * controls the entire thing and the Guest asks it to make changes using the
274 * LOAD_GDT hypercall. 274 * LOAD_GDT hypercall.
275 * 275 *
276 * This is the opposite of the IDT code where we have a LOAD_IDT_ENTRY 276 * This is the exactly like the IDT code.
277 * hypercall and use that repeatedly to load a new IDT. I don't think it
278 * really matters, but wouldn't it be nice if they were the same? Wouldn't
279 * it be even better if you were the one to send the patch to fix it?
280 */ 277 */
281static void lguest_load_gdt(const struct desc_ptr *desc) 278static void lguest_load_gdt(const struct desc_ptr *desc)
282{ 279{
283 BUG_ON((desc->size + 1) / 8 != GDT_ENTRIES); 280 unsigned int i;
284 kvm_hypercall2(LHCALL_LOAD_GDT, __pa(desc->address), GDT_ENTRIES); 281 struct desc_struct *gdt = (void *)desc->address;
282
283 for (i = 0; i < (desc->size+1)/8; i++)
284 kvm_hypercall3(LHCALL_LOAD_GDT_ENTRY, i, gdt[i].a, gdt[i].b);
285} 285}
286 286
287/* For a single GDT entry which changes, we do the lazy thing: alter our GDT, 287/* For a single GDT entry which changes, we do the lazy thing: alter our GDT,
@@ -291,7 +291,9 @@ static void lguest_write_gdt_entry(struct desc_struct *dt, int entrynum,
291 const void *desc, int type) 291 const void *desc, int type)
292{ 292{
293 native_write_gdt_entry(dt, entrynum, desc, type); 293 native_write_gdt_entry(dt, entrynum, desc, type);
294 kvm_hypercall2(LHCALL_LOAD_GDT, __pa(dt), GDT_ENTRIES); 294 /* Tell Host about this new entry. */
295 kvm_hypercall3(LHCALL_LOAD_GDT_ENTRY, entrynum,
296 dt[entrynum].a, dt[entrynum].b);
295} 297}
296 298
297/* OK, I lied. There are three "thread local storage" GDT entries which change 299/* OK, I lied. There are three "thread local storage" GDT entries which change
@@ -661,7 +663,7 @@ static unsigned long lguest_tsc_khz(void)
661 663
662/* If we can't use the TSC, the kernel falls back to our lower-priority 664/* If we can't use the TSC, the kernel falls back to our lower-priority
663 * "lguest_clock", where we read the time value given to us by the Host. */ 665 * "lguest_clock", where we read the time value given to us by the Host. */
664static cycle_t lguest_clock_read(void) 666static cycle_t lguest_clock_read(struct clocksource *cs)
665{ 667{
666 unsigned long sec, nsec; 668 unsigned long sec, nsec;
667 669
diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
index 09daebfdb11c..8a450930834f 100644
--- a/arch/x86/mm/ioremap.c
+++ b/arch/x86/mm/ioremap.c
@@ -280,15 +280,16 @@ static void __iomem *__ioremap_caller(resource_size_t phys_addr,
280 return NULL; 280 return NULL;
281 area->phys_addr = phys_addr; 281 area->phys_addr = phys_addr;
282 vaddr = (unsigned long) area->addr; 282 vaddr = (unsigned long) area->addr;
283 if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot)) { 283
284 if (kernel_map_sync_memtype(phys_addr, size, prot_val)) {
284 free_memtype(phys_addr, phys_addr + size); 285 free_memtype(phys_addr, phys_addr + size);
285 free_vm_area(area); 286 free_vm_area(area);
286 return NULL; 287 return NULL;
287 } 288 }
288 289
289 if (ioremap_change_attr(vaddr, size, prot_val) < 0) { 290 if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot)) {
290 free_memtype(phys_addr, phys_addr + size); 291 free_memtype(phys_addr, phys_addr + size);
291 vunmap(area->addr); 292 free_vm_area(area);
292 return NULL; 293 return NULL;
293 } 294 }
294 295
@@ -374,7 +375,8 @@ static void __iomem *ioremap_default(resource_size_t phys_addr,
374 * - UC_MINUS for non-WB-able memory with no other conflicting mappings 375 * - UC_MINUS for non-WB-able memory with no other conflicting mappings
375 * - Inherit from confliting mappings otherwise 376 * - Inherit from confliting mappings otherwise
376 */ 377 */
377 err = reserve_memtype(phys_addr, phys_addr + size, -1, &flags); 378 err = reserve_memtype(phys_addr, phys_addr + size,
379 _PAGE_CACHE_WB, &flags);
378 if (err < 0) 380 if (err < 0)
379 return NULL; 381 return NULL;
380 382
diff --git a/arch/x86/mm/kmmio.c b/arch/x86/mm/kmmio.c
index 256ce643b0ba..16ccbd77917f 100644
--- a/arch/x86/mm/kmmio.c
+++ b/arch/x86/mm/kmmio.c
@@ -87,7 +87,7 @@ static struct kmmio_probe *get_kmmio_probe(unsigned long addr)
87{ 87{
88 struct kmmio_probe *p; 88 struct kmmio_probe *p;
89 list_for_each_entry_rcu(p, &kmmio_probes, list) { 89 list_for_each_entry_rcu(p, &kmmio_probes, list) {
90 if (addr >= p->addr && addr <= (p->addr + p->len)) 90 if (addr >= p->addr && addr < (p->addr + p->len))
91 return p; 91 return p;
92 } 92 }
93 return NULL; 93 return NULL;
diff --git a/arch/x86/mm/numa_32.c b/arch/x86/mm/numa_32.c
index 3daefa04ace5..d2530062fe00 100644
--- a/arch/x86/mm/numa_32.c
+++ b/arch/x86/mm/numa_32.c
@@ -257,7 +257,7 @@ void resume_map_numa_kva(pgd_t *pgd_base)
257} 257}
258#endif 258#endif
259 259
260static unsigned long calculate_numa_remap_pages(void) 260static __init unsigned long calculate_numa_remap_pages(void)
261{ 261{
262 int nid; 262 int nid;
263 unsigned long size, reserve_pages = 0; 263 unsigned long size, reserve_pages = 0;
diff --git a/arch/x86/mm/numa_64.c b/arch/x86/mm/numa_64.c
index d73aaa892371..2d05a12029dc 100644
--- a/arch/x86/mm/numa_64.c
+++ b/arch/x86/mm/numa_64.c
@@ -188,6 +188,9 @@ void __init setup_node_bootmem(int nodeid, unsigned long start,
188 const int pgdat_size = roundup(sizeof(pg_data_t), PAGE_SIZE); 188 const int pgdat_size = roundup(sizeof(pg_data_t), PAGE_SIZE);
189 int nid; 189 int nid;
190 190
191 if (!end)
192 return;
193
191 start = roundup(start, ZONE_ALIGN); 194 start = roundup(start, ZONE_ALIGN);
192 195
193 printk(KERN_INFO "Bootmem setup node %d %016lx-%016lx\n", nodeid, 196 printk(KERN_INFO "Bootmem setup node %d %016lx-%016lx\n", nodeid,
diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index d71e1b636ce6..797f9f107cb6 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -945,71 +945,94 @@ int _set_memory_uc(unsigned long addr, int numpages)
945 945
946int set_memory_uc(unsigned long addr, int numpages) 946int set_memory_uc(unsigned long addr, int numpages)
947{ 947{
948 int ret;
949
948 /* 950 /*
949 * for now UC MINUS. see comments in ioremap_nocache() 951 * for now UC MINUS. see comments in ioremap_nocache()
950 */ 952 */
951 if (reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE, 953 ret = reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE,
952 _PAGE_CACHE_UC_MINUS, NULL)) 954 _PAGE_CACHE_UC_MINUS, NULL);
953 return -EINVAL; 955 if (ret)
956 goto out_err;
957
958 ret = _set_memory_uc(addr, numpages);
959 if (ret)
960 goto out_free;
954 961
955 return _set_memory_uc(addr, numpages); 962 return 0;
963
964out_free:
965 free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
966out_err:
967 return ret;
956} 968}
957EXPORT_SYMBOL(set_memory_uc); 969EXPORT_SYMBOL(set_memory_uc);
958 970
959int set_memory_array_uc(unsigned long *addr, int addrinarray) 971int set_memory_array_uc(unsigned long *addr, int addrinarray)
960{ 972{
961 unsigned long start; 973 int i, j;
962 unsigned long end; 974 int ret;
963 int i; 975
964 /* 976 /*
965 * for now UC MINUS. see comments in ioremap_nocache() 977 * for now UC MINUS. see comments in ioremap_nocache()
966 */ 978 */
967 for (i = 0; i < addrinarray; i++) { 979 for (i = 0; i < addrinarray; i++) {
968 start = __pa(addr[i]); 980 ret = reserve_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE,
969 for (end = start + PAGE_SIZE; i < addrinarray - 1; end += PAGE_SIZE) { 981 _PAGE_CACHE_UC_MINUS, NULL);
970 if (end != __pa(addr[i + 1])) 982 if (ret)
971 break; 983 goto out_free;
972 i++;
973 }
974 if (reserve_memtype(start, end, _PAGE_CACHE_UC_MINUS, NULL))
975 goto out;
976 } 984 }
977 985
978 return change_page_attr_set(addr, addrinarray, 986 ret = change_page_attr_set(addr, addrinarray,
979 __pgprot(_PAGE_CACHE_UC_MINUS), 1); 987 __pgprot(_PAGE_CACHE_UC_MINUS), 1);
980out: 988 if (ret)
981 for (i = 0; i < addrinarray; i++) { 989 goto out_free;
982 unsigned long tmp = __pa(addr[i]); 990
983 991 return 0;
984 if (tmp == start) 992
985 break; 993out_free:
986 for (end = tmp + PAGE_SIZE; i < addrinarray - 1; end += PAGE_SIZE) { 994 for (j = 0; j < i; j++)
987 if (end != __pa(addr[i + 1])) 995 free_memtype(__pa(addr[j]), __pa(addr[j]) + PAGE_SIZE);
988 break; 996
989 i++; 997 return ret;
990 }
991 free_memtype(tmp, end);
992 }
993 return -EINVAL;
994} 998}
995EXPORT_SYMBOL(set_memory_array_uc); 999EXPORT_SYMBOL(set_memory_array_uc);
996 1000
997int _set_memory_wc(unsigned long addr, int numpages) 1001int _set_memory_wc(unsigned long addr, int numpages)
998{ 1002{
999 return change_page_attr_set(&addr, numpages, 1003 int ret;
1004 ret = change_page_attr_set(&addr, numpages,
1005 __pgprot(_PAGE_CACHE_UC_MINUS), 0);
1006
1007 if (!ret) {
1008 ret = change_page_attr_set(&addr, numpages,
1000 __pgprot(_PAGE_CACHE_WC), 0); 1009 __pgprot(_PAGE_CACHE_WC), 0);
1010 }
1011 return ret;
1001} 1012}
1002 1013
1003int set_memory_wc(unsigned long addr, int numpages) 1014int set_memory_wc(unsigned long addr, int numpages)
1004{ 1015{
1016 int ret;
1017
1005 if (!pat_enabled) 1018 if (!pat_enabled)
1006 return set_memory_uc(addr, numpages); 1019 return set_memory_uc(addr, numpages);
1007 1020
1008 if (reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE, 1021 ret = reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE,
1009 _PAGE_CACHE_WC, NULL)) 1022 _PAGE_CACHE_WC, NULL);
1010 return -EINVAL; 1023 if (ret)
1024 goto out_err;
1025
1026 ret = _set_memory_wc(addr, numpages);
1027 if (ret)
1028 goto out_free;
1029
1030 return 0;
1011 1031
1012 return _set_memory_wc(addr, numpages); 1032out_free:
1033 free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
1034out_err:
1035 return ret;
1013} 1036}
1014EXPORT_SYMBOL(set_memory_wc); 1037EXPORT_SYMBOL(set_memory_wc);
1015 1038
@@ -1021,29 +1044,31 @@ int _set_memory_wb(unsigned long addr, int numpages)
1021 1044
1022int set_memory_wb(unsigned long addr, int numpages) 1045int set_memory_wb(unsigned long addr, int numpages)
1023{ 1046{
1024 free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE); 1047 int ret;
1048
1049 ret = _set_memory_wb(addr, numpages);
1050 if (ret)
1051 return ret;
1025 1052
1026 return _set_memory_wb(addr, numpages); 1053 free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
1054 return 0;
1027} 1055}
1028EXPORT_SYMBOL(set_memory_wb); 1056EXPORT_SYMBOL(set_memory_wb);
1029 1057
1030int set_memory_array_wb(unsigned long *addr, int addrinarray) 1058int set_memory_array_wb(unsigned long *addr, int addrinarray)
1031{ 1059{
1032 int i; 1060 int i;
1061 int ret;
1033 1062
1034 for (i = 0; i < addrinarray; i++) { 1063 ret = change_page_attr_clear(addr, addrinarray,
1035 unsigned long start = __pa(addr[i]);
1036 unsigned long end;
1037
1038 for (end = start + PAGE_SIZE; i < addrinarray - 1; end += PAGE_SIZE) {
1039 if (end != __pa(addr[i + 1]))
1040 break;
1041 i++;
1042 }
1043 free_memtype(start, end);
1044 }
1045 return change_page_attr_clear(addr, addrinarray,
1046 __pgprot(_PAGE_CACHE_MASK), 1); 1064 __pgprot(_PAGE_CACHE_MASK), 1);
1065 if (ret)
1066 return ret;
1067
1068 for (i = 0; i < addrinarray; i++)
1069 free_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE);
1070
1071 return 0;
1047} 1072}
1048EXPORT_SYMBOL(set_memory_array_wb); 1073EXPORT_SYMBOL(set_memory_array_wb);
1049 1074
@@ -1136,6 +1161,8 @@ int set_pages_array_wb(struct page **pages, int addrinarray)
1136 1161
1137 retval = cpa_clear_pages_array(pages, addrinarray, 1162 retval = cpa_clear_pages_array(pages, addrinarray,
1138 __pgprot(_PAGE_CACHE_MASK)); 1163 __pgprot(_PAGE_CACHE_MASK));
1164 if (retval)
1165 return retval;
1139 1166
1140 for (i = 0; i < addrinarray; i++) { 1167 for (i = 0; i < addrinarray; i++) {
1141 start = (unsigned long)page_address(pages[i]); 1168 start = (unsigned long)page_address(pages[i]);
@@ -1143,7 +1170,7 @@ int set_pages_array_wb(struct page **pages, int addrinarray)
1143 free_memtype(start, end); 1170 free_memtype(start, end);
1144 } 1171 }
1145 1172
1146 return retval; 1173 return 0;
1147} 1174}
1148EXPORT_SYMBOL(set_pages_array_wb); 1175EXPORT_SYMBOL(set_pages_array_wb);
1149 1176
diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c
index c009a241d562..e6718bb28065 100644
--- a/arch/x86/mm/pat.c
+++ b/arch/x86/mm/pat.c
@@ -182,10 +182,10 @@ static unsigned long pat_x_mtrr_type(u64 start, u64 end, unsigned long req_type)
182 u8 mtrr_type; 182 u8 mtrr_type;
183 183
184 mtrr_type = mtrr_type_lookup(start, end); 184 mtrr_type = mtrr_type_lookup(start, end);
185 if (mtrr_type == MTRR_TYPE_UNCACHABLE) 185 if (mtrr_type != MTRR_TYPE_WRBACK)
186 return _PAGE_CACHE_UC; 186 return _PAGE_CACHE_UC_MINUS;
187 if (mtrr_type == MTRR_TYPE_WRCOMB) 187
188 return _PAGE_CACHE_WC; 188 return _PAGE_CACHE_WB;
189 } 189 }
190 190
191 return req_type; 191 return req_type;
@@ -352,23 +352,13 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type,
352 return 0; 352 return 0;
353 } 353 }
354 354
355 if (req_type == -1) { 355 /*
356 /* 356 * Call mtrr_lookup to get the type hint. This is an
357 * Call mtrr_lookup to get the type hint. This is an 357 * optimization for /dev/mem mmap'ers into WB memory (BIOS
358 * optimization for /dev/mem mmap'ers into WB memory (BIOS 358 * tools and ACPI tools). Use WB request for WB memory and use
359 * tools and ACPI tools). Use WB request for WB memory and use 359 * UC_MINUS otherwise.
360 * UC_MINUS otherwise. 360 */
361 */ 361 actual_type = pat_x_mtrr_type(start, end, req_type & _PAGE_CACHE_MASK);
362 u8 mtrr_type = mtrr_type_lookup(start, end);
363
364 if (mtrr_type == MTRR_TYPE_WRBACK)
365 actual_type = _PAGE_CACHE_WB;
366 else
367 actual_type = _PAGE_CACHE_UC_MINUS;
368 } else {
369 actual_type = pat_x_mtrr_type(start, end,
370 req_type & _PAGE_CACHE_MASK);
371 }
372 362
373 if (new_type) 363 if (new_type)
374 *new_type = actual_type; 364 *new_type = actual_type;
@@ -546,9 +536,7 @@ static inline int range_is_allowed(unsigned long pfn, unsigned long size)
546int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn, 536int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
547 unsigned long size, pgprot_t *vma_prot) 537 unsigned long size, pgprot_t *vma_prot)
548{ 538{
549 u64 offset = ((u64) pfn) << PAGE_SHIFT; 539 unsigned long flags = _PAGE_CACHE_WB;
550 unsigned long flags = -1;
551 int retval;
552 540
553 if (!range_is_allowed(pfn, size)) 541 if (!range_is_allowed(pfn, size))
554 return 0; 542 return 0;
@@ -576,64 +564,11 @@ int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
576 } 564 }
577#endif 565#endif
578 566
579 /*
580 * With O_SYNC, we can only take UC_MINUS mapping. Fail if we cannot.
581 *
582 * Without O_SYNC, we want to get
583 * - WB for WB-able memory and no other conflicting mappings
584 * - UC_MINUS for non-WB-able memory with no other conflicting mappings
585 * - Inherit from confliting mappings otherwise
586 */
587 if (flags != -1) {
588 retval = reserve_memtype(offset, offset + size, flags, NULL);
589 } else {
590 retval = reserve_memtype(offset, offset + size, -1, &flags);
591 }
592
593 if (retval < 0)
594 return 0;
595
596 if (((pfn < max_low_pfn_mapped) ||
597 (pfn >= (1UL<<(32 - PAGE_SHIFT)) && pfn < max_pfn_mapped)) &&
598 ioremap_change_attr((unsigned long)__va(offset), size, flags) < 0) {
599 free_memtype(offset, offset + size);
600 printk(KERN_INFO
601 "%s:%d /dev/mem ioremap_change_attr failed %s for %Lx-%Lx\n",
602 current->comm, current->pid,
603 cattr_name(flags),
604 offset, (unsigned long long)(offset + size));
605 return 0;
606 }
607
608 *vma_prot = __pgprot((pgprot_val(*vma_prot) & ~_PAGE_CACHE_MASK) | 567 *vma_prot = __pgprot((pgprot_val(*vma_prot) & ~_PAGE_CACHE_MASK) |
609 flags); 568 flags);
610 return 1; 569 return 1;
611} 570}
612 571
613void map_devmem(unsigned long pfn, unsigned long size, pgprot_t vma_prot)
614{
615 unsigned long want_flags = (pgprot_val(vma_prot) & _PAGE_CACHE_MASK);
616 u64 addr = (u64)pfn << PAGE_SHIFT;
617 unsigned long flags;
618
619 reserve_memtype(addr, addr + size, want_flags, &flags);
620 if (flags != want_flags) {
621 printk(KERN_INFO
622 "%s:%d /dev/mem expected mapping type %s for %Lx-%Lx, got %s\n",
623 current->comm, current->pid,
624 cattr_name(want_flags),
625 addr, (unsigned long long)(addr + size),
626 cattr_name(flags));
627 }
628}
629
630void unmap_devmem(unsigned long pfn, unsigned long size, pgprot_t vma_prot)
631{
632 u64 addr = (u64)pfn << PAGE_SHIFT;
633
634 free_memtype(addr, addr + size);
635}
636
637/* 572/*
638 * Change the memory type for the physial address range in kernel identity 573 * Change the memory type for the physial address range in kernel identity
639 * mapping space if that range is a part of identity map. 574 * mapping space if that range is a part of identity map.
@@ -671,8 +606,8 @@ static int reserve_pfn_range(u64 paddr, unsigned long size, pgprot_t *vma_prot,
671{ 606{
672 int is_ram = 0; 607 int is_ram = 0;
673 int ret; 608 int ret;
674 unsigned long flags;
675 unsigned long want_flags = (pgprot_val(*vma_prot) & _PAGE_CACHE_MASK); 609 unsigned long want_flags = (pgprot_val(*vma_prot) & _PAGE_CACHE_MASK);
610 unsigned long flags = want_flags;
676 611
677 is_ram = pat_pagerange_is_ram(paddr, paddr + size); 612 is_ram = pat_pagerange_is_ram(paddr, paddr + size);
678 613
@@ -734,29 +669,28 @@ static void free_pfn_range(u64 paddr, unsigned long size)
734 * 669 *
735 * If the vma has a linear pfn mapping for the entire range, we get the prot 670 * If the vma has a linear pfn mapping for the entire range, we get the prot
736 * from pte and reserve the entire vma range with single reserve_pfn_range call. 671 * from pte and reserve the entire vma range with single reserve_pfn_range call.
737 * Otherwise, we reserve the entire vma range, my ging through the PTEs page
738 * by page to get physical address and protection.
739 */ 672 */
740int track_pfn_vma_copy(struct vm_area_struct *vma) 673int track_pfn_vma_copy(struct vm_area_struct *vma)
741{ 674{
742 int retval = 0;
743 unsigned long i, j;
744 resource_size_t paddr; 675 resource_size_t paddr;
745 unsigned long prot; 676 unsigned long prot;
746 unsigned long vma_start = vma->vm_start; 677 unsigned long vma_size = vma->vm_end - vma->vm_start;
747 unsigned long vma_end = vma->vm_end;
748 unsigned long vma_size = vma_end - vma_start;
749 pgprot_t pgprot; 678 pgprot_t pgprot;
750 679
751 if (!pat_enabled) 680 if (!pat_enabled)
752 return 0; 681 return 0;
753 682
683 /*
684 * For now, only handle remap_pfn_range() vmas where
685 * is_linear_pfn_mapping() == TRUE. Handling of
686 * vm_insert_pfn() is TBD.
687 */
754 if (is_linear_pfn_mapping(vma)) { 688 if (is_linear_pfn_mapping(vma)) {
755 /* 689 /*
756 * reserve the whole chunk covered by vma. We need the 690 * reserve the whole chunk covered by vma. We need the
757 * starting address and protection from pte. 691 * starting address and protection from pte.
758 */ 692 */
759 if (follow_phys(vma, vma_start, 0, &prot, &paddr)) { 693 if (follow_phys(vma, vma->vm_start, 0, &prot, &paddr)) {
760 WARN_ON_ONCE(1); 694 WARN_ON_ONCE(1);
761 return -EINVAL; 695 return -EINVAL;
762 } 696 }
@@ -764,28 +698,7 @@ int track_pfn_vma_copy(struct vm_area_struct *vma)
764 return reserve_pfn_range(paddr, vma_size, &pgprot, 1); 698 return reserve_pfn_range(paddr, vma_size, &pgprot, 1);
765 } 699 }
766 700
767 /* reserve entire vma page by page, using pfn and prot from pte */
768 for (i = 0; i < vma_size; i += PAGE_SIZE) {
769 if (follow_phys(vma, vma_start + i, 0, &prot, &paddr))
770 continue;
771
772 pgprot = __pgprot(prot);
773 retval = reserve_pfn_range(paddr, PAGE_SIZE, &pgprot, 1);
774 if (retval)
775 goto cleanup_ret;
776 }
777 return 0; 701 return 0;
778
779cleanup_ret:
780 /* Reserve error: Cleanup partial reservation and return error */
781 for (j = 0; j < i; j += PAGE_SIZE) {
782 if (follow_phys(vma, vma_start + j, 0, &prot, &paddr))
783 continue;
784
785 free_pfn_range(paddr, PAGE_SIZE);
786 }
787
788 return retval;
789} 702}
790 703
791/* 704/*
@@ -795,50 +708,28 @@ cleanup_ret:
795 * prot is passed in as a parameter for the new mapping. If the vma has a 708 * prot is passed in as a parameter for the new mapping. If the vma has a
796 * linear pfn mapping for the entire range reserve the entire vma range with 709 * linear pfn mapping for the entire range reserve the entire vma range with
797 * single reserve_pfn_range call. 710 * single reserve_pfn_range call.
798 * Otherwise, we look t the pfn and size and reserve only the specified range
799 * page by page.
800 *
801 * Note that this function can be called with caller trying to map only a
802 * subrange/page inside the vma.
803 */ 711 */
804int track_pfn_vma_new(struct vm_area_struct *vma, pgprot_t *prot, 712int track_pfn_vma_new(struct vm_area_struct *vma, pgprot_t *prot,
805 unsigned long pfn, unsigned long size) 713 unsigned long pfn, unsigned long size)
806{ 714{
807 int retval = 0;
808 unsigned long i, j;
809 resource_size_t base_paddr;
810 resource_size_t paddr; 715 resource_size_t paddr;
811 unsigned long vma_start = vma->vm_start; 716 unsigned long vma_size = vma->vm_end - vma->vm_start;
812 unsigned long vma_end = vma->vm_end;
813 unsigned long vma_size = vma_end - vma_start;
814 717
815 if (!pat_enabled) 718 if (!pat_enabled)
816 return 0; 719 return 0;
817 720
721 /*
722 * For now, only handle remap_pfn_range() vmas where
723 * is_linear_pfn_mapping() == TRUE. Handling of
724 * vm_insert_pfn() is TBD.
725 */
818 if (is_linear_pfn_mapping(vma)) { 726 if (is_linear_pfn_mapping(vma)) {
819 /* reserve the whole chunk starting from vm_pgoff */ 727 /* reserve the whole chunk starting from vm_pgoff */
820 paddr = (resource_size_t)vma->vm_pgoff << PAGE_SHIFT; 728 paddr = (resource_size_t)vma->vm_pgoff << PAGE_SHIFT;
821 return reserve_pfn_range(paddr, vma_size, prot, 0); 729 return reserve_pfn_range(paddr, vma_size, prot, 0);
822 } 730 }
823 731
824 /* reserve page by page using pfn and size */
825 base_paddr = (resource_size_t)pfn << PAGE_SHIFT;
826 for (i = 0; i < size; i += PAGE_SIZE) {
827 paddr = base_paddr + i;
828 retval = reserve_pfn_range(paddr, PAGE_SIZE, prot, 0);
829 if (retval)
830 goto cleanup_ret;
831 }
832 return 0; 732 return 0;
833
834cleanup_ret:
835 /* Reserve error: Cleanup partial reservation and return error */
836 for (j = 0; j < i; j += PAGE_SIZE) {
837 paddr = base_paddr + j;
838 free_pfn_range(paddr, PAGE_SIZE);
839 }
840
841 return retval;
842} 733}
843 734
844/* 735/*
@@ -849,39 +740,23 @@ cleanup_ret:
849void untrack_pfn_vma(struct vm_area_struct *vma, unsigned long pfn, 740void untrack_pfn_vma(struct vm_area_struct *vma, unsigned long pfn,
850 unsigned long size) 741 unsigned long size)
851{ 742{
852 unsigned long i;
853 resource_size_t paddr; 743 resource_size_t paddr;
854 unsigned long prot; 744 unsigned long vma_size = vma->vm_end - vma->vm_start;
855 unsigned long vma_start = vma->vm_start;
856 unsigned long vma_end = vma->vm_end;
857 unsigned long vma_size = vma_end - vma_start;
858 745
859 if (!pat_enabled) 746 if (!pat_enabled)
860 return; 747 return;
861 748
749 /*
750 * For now, only handle remap_pfn_range() vmas where
751 * is_linear_pfn_mapping() == TRUE. Handling of
752 * vm_insert_pfn() is TBD.
753 */
862 if (is_linear_pfn_mapping(vma)) { 754 if (is_linear_pfn_mapping(vma)) {
863 /* free the whole chunk starting from vm_pgoff */ 755 /* free the whole chunk starting from vm_pgoff */
864 paddr = (resource_size_t)vma->vm_pgoff << PAGE_SHIFT; 756 paddr = (resource_size_t)vma->vm_pgoff << PAGE_SHIFT;
865 free_pfn_range(paddr, vma_size); 757 free_pfn_range(paddr, vma_size);
866 return; 758 return;
867 } 759 }
868
869 if (size != 0 && size != vma_size) {
870 /* free page by page, using pfn and size */
871 paddr = (resource_size_t)pfn << PAGE_SHIFT;
872 for (i = 0; i < size; i += PAGE_SIZE) {
873 paddr = paddr + i;
874 free_pfn_range(paddr, PAGE_SIZE);
875 }
876 } else {
877 /* free entire vma, page by page, using the pfn from pte */
878 for (i = 0; i < vma_size; i += PAGE_SIZE) {
879 if (follow_phys(vma, vma_start + i, 0, &prot, &paddr))
880 continue;
881
882 free_pfn_range(paddr, PAGE_SIZE);
883 }
884 }
885} 760}
886 761
887pgprot_t pgprot_writecombine(pgprot_t prot) 762pgprot_t pgprot_writecombine(pgprot_t prot)
diff --git a/arch/x86/mm/srat_64.c b/arch/x86/mm/srat_64.c
index c7d272b8574c..33c5fa57e43d 100644
--- a/arch/x86/mm/srat_64.c
+++ b/arch/x86/mm/srat_64.c
@@ -28,6 +28,7 @@ int acpi_numa __initdata;
28static struct acpi_table_slit *acpi_slit; 28static struct acpi_table_slit *acpi_slit;
29 29
30static nodemask_t nodes_parsed __initdata; 30static nodemask_t nodes_parsed __initdata;
31static nodemask_t cpu_nodes_parsed __initdata;
31static struct bootnode nodes[MAX_NUMNODES] __initdata; 32static struct bootnode nodes[MAX_NUMNODES] __initdata;
32static struct bootnode nodes_add[MAX_NUMNODES]; 33static struct bootnode nodes_add[MAX_NUMNODES];
33static int found_add_area __initdata; 34static int found_add_area __initdata;
@@ -141,6 +142,7 @@ acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa)
141 142
142 apic_id = pa->apic_id; 143 apic_id = pa->apic_id;
143 apicid_to_node[apic_id] = node; 144 apicid_to_node[apic_id] = node;
145 node_set(node, cpu_nodes_parsed);
144 acpi_numa = 1; 146 acpi_numa = 1;
145 printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n", 147 printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n",
146 pxm, apic_id, node); 148 pxm, apic_id, node);
@@ -174,6 +176,7 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
174 else 176 else
175 apic_id = pa->apic_id; 177 apic_id = pa->apic_id;
176 apicid_to_node[apic_id] = node; 178 apicid_to_node[apic_id] = node;
179 node_set(node, cpu_nodes_parsed);
177 acpi_numa = 1; 180 acpi_numa = 1;
178 printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n", 181 printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n",
179 pxm, apic_id, node); 182 pxm, apic_id, node);
@@ -402,7 +405,8 @@ int __init acpi_scan_nodes(unsigned long start, unsigned long end)
402 return -1; 405 return -1;
403 } 406 }
404 407
405 node_possible_map = nodes_parsed; 408 /* Account for nodes with cpus and no memory */
409 nodes_or(node_possible_map, nodes_parsed, cpu_nodes_parsed);
406 410
407 /* Finally register nodes */ 411 /* Finally register nodes */
408 for_each_node_mask(i, node_possible_map) 412 for_each_node_mask(i, node_possible_map)
diff --git a/arch/x86/pci/amd_bus.c b/arch/x86/pci/amd_bus.c
index 9bb09823b362..f893d6a6e803 100644
--- a/arch/x86/pci/amd_bus.c
+++ b/arch/x86/pci/amd_bus.c
@@ -94,12 +94,16 @@ struct pci_root_info {
94static int pci_root_num; 94static int pci_root_num;
95static struct pci_root_info pci_root_info[PCI_ROOT_NR]; 95static struct pci_root_info pci_root_info[PCI_ROOT_NR];
96 96
97void set_pci_bus_resources_arch_default(struct pci_bus *b) 97void x86_pci_root_bus_res_quirks(struct pci_bus *b)
98{ 98{
99 int i; 99 int i;
100 int j; 100 int j;
101 struct pci_root_info *info; 101 struct pci_root_info *info;
102 102
103 /* don't go for it if _CRS is used */
104 if (pci_probe & PCI_USE__CRS)
105 return;
106
103 /* if only one root bus, don't need to anything */ 107 /* if only one root bus, don't need to anything */
104 if (pci_root_num < 2) 108 if (pci_root_num < 2)
105 return; 109 return;
diff --git a/arch/x86/pci/common.c b/arch/x86/pci/common.c
index 8c362b96b644..2202b6257b82 100644
--- a/arch/x86/pci/common.c
+++ b/arch/x86/pci/common.c
@@ -147,10 +147,13 @@ static void __devinit pcibios_fixup_device_resources(struct pci_dev *dev)
147 * are examined. 147 * are examined.
148 */ 148 */
149 149
150void __devinit pcibios_fixup_bus(struct pci_bus *b) 150void __devinit pcibios_fixup_bus(struct pci_bus *b)
151{ 151{
152 struct pci_dev *dev; 152 struct pci_dev *dev;
153 153
154 /* root bus? */
155 if (!b->parent)
156 x86_pci_root_bus_res_quirks(b);
154 pci_read_bridge_bases(b); 157 pci_read_bridge_bases(b);
155 list_for_each_entry(dev, &b->devices, bus_list) 158 list_for_each_entry(dev, &b->devices, bus_list)
156 pcibios_fixup_device_resources(dev); 159 pcibios_fixup_device_resources(dev);
diff --git a/arch/x86/pci/i386.c b/arch/x86/pci/i386.c
index f1817f71e009..a85bef20a3b9 100644
--- a/arch/x86/pci/i386.c
+++ b/arch/x86/pci/i386.c
@@ -238,6 +238,10 @@ void __init pcibios_resource_survey(void)
238 */ 238 */
239fs_initcall(pcibios_assign_resources); 239fs_initcall(pcibios_assign_resources);
240 240
241void __weak x86_pci_root_bus_res_quirks(struct pci_bus *b)
242{
243}
244
241/* 245/*
242 * If we set up a device for bus mastering, we need to check the latency 246 * If we set up a device for bus mastering, we need to check the latency
243 * timer as certain crappy BIOSes forget to set it properly. 247 * timer as certain crappy BIOSes forget to set it properly.
diff --git a/arch/x86/pci/mmconfig-shared.c b/arch/x86/pci/mmconfig-shared.c
index 905bb526b133..5fa10bb9604f 100644
--- a/arch/x86/pci/mmconfig-shared.c
+++ b/arch/x86/pci/mmconfig-shared.c
@@ -375,7 +375,7 @@ static acpi_status __init check_mcfg_resource(struct acpi_resource *res,
375 if (!fixmem32) 375 if (!fixmem32)
376 return AE_OK; 376 return AE_OK;
377 if ((mcfg_res->start >= fixmem32->address) && 377 if ((mcfg_res->start >= fixmem32->address) &&
378 (mcfg_res->end < (fixmem32->address + 378 (mcfg_res->end <= (fixmem32->address +
379 fixmem32->address_length))) { 379 fixmem32->address_length))) {
380 mcfg_res->flags = 1; 380 mcfg_res->flags = 1;
381 return AE_CTRL_TERMINATE; 381 return AE_CTRL_TERMINATE;
@@ -392,7 +392,7 @@ static acpi_status __init check_mcfg_resource(struct acpi_resource *res,
392 return AE_OK; 392 return AE_OK;
393 393
394 if ((mcfg_res->start >= address.minimum) && 394 if ((mcfg_res->start >= address.minimum) &&
395 (mcfg_res->end < (address.minimum + address.address_length))) { 395 (mcfg_res->end <= (address.minimum + address.address_length))) {
396 mcfg_res->flags = 1; 396 mcfg_res->flags = 1;
397 return AE_CTRL_TERMINATE; 397 return AE_CTRL_TERMINATE;
398 } 398 }
@@ -439,7 +439,7 @@ static int __init is_mmconf_reserved(check_reserved_t is_reserved,
439 u64 old_size = size; 439 u64 old_size = size;
440 int valid = 0; 440 int valid = 0;
441 441
442 while (!is_reserved(addr, addr + size - 1, E820_RESERVED)) { 442 while (!is_reserved(addr, addr + size, E820_RESERVED)) {
443 size >>= 1; 443 size >>= 1;
444 if (size < (16UL<<20)) 444 if (size < (16UL<<20))
445 break; 445 break;
diff --git a/arch/x86/vdso/vclock_gettime.c b/arch/x86/vdso/vclock_gettime.c
index d9d35824c56f..6a40b78b46aa 100644
--- a/arch/x86/vdso/vclock_gettime.c
+++ b/arch/x86/vdso/vclock_gettime.c
@@ -104,11 +104,13 @@ notrace int __vdso_gettimeofday(struct timeval *tv, struct timezone *tz)
104{ 104{
105 long ret; 105 long ret;
106 if (likely(gtod->sysctl_enabled && gtod->clock.vread)) { 106 if (likely(gtod->sysctl_enabled && gtod->clock.vread)) {
107 BUILD_BUG_ON(offsetof(struct timeval, tv_usec) != 107 if (likely(tv != NULL)) {
108 offsetof(struct timespec, tv_nsec) || 108 BUILD_BUG_ON(offsetof(struct timeval, tv_usec) !=
109 sizeof(*tv) != sizeof(struct timespec)); 109 offsetof(struct timespec, tv_nsec) ||
110 do_realtime((struct timespec *)tv); 110 sizeof(*tv) != sizeof(struct timespec));
111 tv->tv_usec /= 1000; 111 do_realtime((struct timespec *)tv);
112 tv->tv_usec /= 1000;
113 }
112 if (unlikely(tz != NULL)) { 114 if (unlikely(tz != NULL)) {
113 /* Avoid memcpy. Some old compilers fail to inline it */ 115 /* Avoid memcpy. Some old compilers fail to inline it */
114 tz->tz_minuteswest = gtod->sys_tz.tz_minuteswest; 116 tz->tz_minuteswest = gtod->sys_tz.tz_minuteswest;
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 82cd39a6cbd3..f09e8c36ee80 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -42,6 +42,7 @@
42#include <asm/xen/hypervisor.h> 42#include <asm/xen/hypervisor.h>
43#include <asm/fixmap.h> 43#include <asm/fixmap.h>
44#include <asm/processor.h> 44#include <asm/processor.h>
45#include <asm/proto.h>
45#include <asm/msr-index.h> 46#include <asm/msr-index.h>
46#include <asm/setup.h> 47#include <asm/setup.h>
47#include <asm/desc.h> 48#include <asm/desc.h>
@@ -168,21 +169,23 @@ static void __init xen_banner(void)
168 xen_feature(XENFEAT_mmu_pt_update_preserve_ad) ? " (preserve-AD)" : ""); 169 xen_feature(XENFEAT_mmu_pt_update_preserve_ad) ? " (preserve-AD)" : "");
169} 170}
170 171
172static __read_mostly unsigned int cpuid_leaf1_edx_mask = ~0;
173static __read_mostly unsigned int cpuid_leaf1_ecx_mask = ~0;
174
171static void xen_cpuid(unsigned int *ax, unsigned int *bx, 175static void xen_cpuid(unsigned int *ax, unsigned int *bx,
172 unsigned int *cx, unsigned int *dx) 176 unsigned int *cx, unsigned int *dx)
173{ 177{
178 unsigned maskecx = ~0;
174 unsigned maskedx = ~0; 179 unsigned maskedx = ~0;
175 180
176 /* 181 /*
177 * Mask out inconvenient features, to try and disable as many 182 * Mask out inconvenient features, to try and disable as many
178 * unsupported kernel subsystems as possible. 183 * unsupported kernel subsystems as possible.
179 */ 184 */
180 if (*ax == 1) 185 if (*ax == 1) {
181 maskedx = ~((1 << X86_FEATURE_APIC) | /* disable APIC */ 186 maskecx = cpuid_leaf1_ecx_mask;
182 (1 << X86_FEATURE_ACPI) | /* disable ACPI */ 187 maskedx = cpuid_leaf1_edx_mask;
183 (1 << X86_FEATURE_MCE) | /* disable MCE */ 188 }
184 (1 << X86_FEATURE_MCA) | /* disable MCA */
185 (1 << X86_FEATURE_ACC)); /* thermal monitoring */
186 189
187 asm(XEN_EMULATE_PREFIX "cpuid" 190 asm(XEN_EMULATE_PREFIX "cpuid"
188 : "=a" (*ax), 191 : "=a" (*ax),
@@ -190,9 +193,43 @@ static void xen_cpuid(unsigned int *ax, unsigned int *bx,
190 "=c" (*cx), 193 "=c" (*cx),
191 "=d" (*dx) 194 "=d" (*dx)
192 : "0" (*ax), "2" (*cx)); 195 : "0" (*ax), "2" (*cx));
196
197 *cx &= maskecx;
193 *dx &= maskedx; 198 *dx &= maskedx;
194} 199}
195 200
201static __init void xen_init_cpuid_mask(void)
202{
203 unsigned int ax, bx, cx, dx;
204
205 cpuid_leaf1_edx_mask =
206 ~((1 << X86_FEATURE_MCE) | /* disable MCE */
207 (1 << X86_FEATURE_MCA) | /* disable MCA */
208 (1 << X86_FEATURE_ACC)); /* thermal monitoring */
209
210 if (!xen_initial_domain())
211 cpuid_leaf1_edx_mask &=
212 ~((1 << X86_FEATURE_APIC) | /* disable local APIC */
213 (1 << X86_FEATURE_ACPI)); /* disable ACPI */
214
215 ax = 1;
216 xen_cpuid(&ax, &bx, &cx, &dx);
217
218 /* cpuid claims we support xsave; try enabling it to see what happens */
219 if (cx & (1 << (X86_FEATURE_XSAVE % 32))) {
220 unsigned long cr4;
221
222 set_in_cr4(X86_CR4_OSXSAVE);
223
224 cr4 = read_cr4();
225
226 if ((cr4 & X86_CR4_OSXSAVE) == 0)
227 cpuid_leaf1_ecx_mask &= ~(1 << (X86_FEATURE_XSAVE % 32));
228
229 clear_in_cr4(X86_CR4_OSXSAVE);
230 }
231}
232
196static void xen_set_debugreg(int reg, unsigned long val) 233static void xen_set_debugreg(int reg, unsigned long val)
197{ 234{
198 HYPERVISOR_set_debugreg(reg, val); 235 HYPERVISOR_set_debugreg(reg, val);
@@ -284,12 +321,11 @@ static void xen_set_ldt(const void *addr, unsigned entries)
284 321
285static void xen_load_gdt(const struct desc_ptr *dtr) 322static void xen_load_gdt(const struct desc_ptr *dtr)
286{ 323{
287 unsigned long *frames;
288 unsigned long va = dtr->address; 324 unsigned long va = dtr->address;
289 unsigned int size = dtr->size + 1; 325 unsigned int size = dtr->size + 1;
290 unsigned pages = (size + PAGE_SIZE - 1) / PAGE_SIZE; 326 unsigned pages = (size + PAGE_SIZE - 1) / PAGE_SIZE;
327 unsigned long frames[pages];
291 int f; 328 int f;
292 struct multicall_space mcs;
293 329
294 /* A GDT can be up to 64k in size, which corresponds to 8192 330 /* A GDT can be up to 64k in size, which corresponds to 8192
295 8-byte entries, or 16 4k pages.. */ 331 8-byte entries, or 16 4k pages.. */
@@ -297,19 +333,26 @@ static void xen_load_gdt(const struct desc_ptr *dtr)
297 BUG_ON(size > 65536); 333 BUG_ON(size > 65536);
298 BUG_ON(va & ~PAGE_MASK); 334 BUG_ON(va & ~PAGE_MASK);
299 335
300 mcs = xen_mc_entry(sizeof(*frames) * pages);
301 frames = mcs.args;
302
303 for (f = 0; va < dtr->address + size; va += PAGE_SIZE, f++) { 336 for (f = 0; va < dtr->address + size; va += PAGE_SIZE, f++) {
304 frames[f] = arbitrary_virt_to_mfn((void *)va); 337 int level;
338 pte_t *ptep = lookup_address(va, &level);
339 unsigned long pfn, mfn;
340 void *virt;
341
342 BUG_ON(ptep == NULL);
343
344 pfn = pte_pfn(*ptep);
345 mfn = pfn_to_mfn(pfn);
346 virt = __va(PFN_PHYS(pfn));
347
348 frames[f] = mfn;
305 349
306 make_lowmem_page_readonly((void *)va); 350 make_lowmem_page_readonly((void *)va);
307 make_lowmem_page_readonly(mfn_to_virt(frames[f])); 351 make_lowmem_page_readonly(virt);
308 } 352 }
309 353
310 MULTI_set_gdt(mcs.mc, frames, size / sizeof(struct desc_struct)); 354 if (HYPERVISOR_set_gdt(frames, size / sizeof(struct desc_struct)))
311 355 BUG();
312 xen_mc_issue(PARAVIRT_LAZY_CPU);
313} 356}
314 357
315static void load_TLS_descriptor(struct thread_struct *t, 358static void load_TLS_descriptor(struct thread_struct *t,
@@ -385,7 +428,7 @@ static void xen_write_ldt_entry(struct desc_struct *dt, int entrynum,
385static int cvt_gate_to_trap(int vector, const gate_desc *val, 428static int cvt_gate_to_trap(int vector, const gate_desc *val,
386 struct trap_info *info) 429 struct trap_info *info)
387{ 430{
388 if (val->type != 0xf && val->type != 0xe) 431 if (val->type != GATE_TRAP && val->type != GATE_INTERRUPT)
389 return 0; 432 return 0;
390 433
391 info->vector = vector; 434 info->vector = vector;
@@ -393,8 +436,8 @@ static int cvt_gate_to_trap(int vector, const gate_desc *val,
393 info->cs = gate_segment(*val); 436 info->cs = gate_segment(*val);
394 info->flags = val->dpl; 437 info->flags = val->dpl;
395 /* interrupt gates clear IF */ 438 /* interrupt gates clear IF */
396 if (val->type == 0xe) 439 if (val->type == GATE_INTERRUPT)
397 info->flags |= 4; 440 info->flags |= 1 << 2;
398 441
399 return 1; 442 return 1;
400} 443}
@@ -872,7 +915,6 @@ static const struct machine_ops __initdata xen_machine_ops = {
872 .emergency_restart = xen_emergency_restart, 915 .emergency_restart = xen_emergency_restart,
873}; 916};
874 917
875
876/* First C function to be called on Xen boot */ 918/* First C function to be called on Xen boot */
877asmlinkage void __init xen_start_kernel(void) 919asmlinkage void __init xen_start_kernel(void)
878{ 920{
@@ -897,6 +939,8 @@ asmlinkage void __init xen_start_kernel(void)
897 939
898 xen_init_irq_ops(); 940 xen_init_irq_ops();
899 941
942 xen_init_cpuid_mask();
943
900#ifdef CONFIG_X86_LOCAL_APIC 944#ifdef CONFIG_X86_LOCAL_APIC
901 /* 945 /*
902 * set up the basic apic ops. 946 * set up the basic apic ops.
@@ -938,6 +982,11 @@ asmlinkage void __init xen_start_kernel(void)
938 if (!xen_initial_domain()) 982 if (!xen_initial_domain())
939 __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD); 983 __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD);
940 984
985#ifdef CONFIG_X86_64
986 /* Work out if we support NX */
987 check_efer();
988#endif
989
941 /* Don't do the full vcpu_info placement stuff until we have a 990 /* Don't do the full vcpu_info placement stuff until we have a
942 possible map and a non-dummy shared_info. */ 991 possible map and a non-dummy shared_info. */
943 per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0]; 992 per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0];
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
index 2a81838a9ab7..9842b1212407 100644
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -184,7 +184,7 @@ static inline unsigned p2m_index(unsigned long pfn)
184} 184}
185 185
186/* Build the parallel p2m_top_mfn structures */ 186/* Build the parallel p2m_top_mfn structures */
187void xen_setup_mfn_list_list(void) 187static void __init xen_build_mfn_list_list(void)
188{ 188{
189 unsigned pfn, idx; 189 unsigned pfn, idx;
190 190
@@ -198,7 +198,10 @@ void xen_setup_mfn_list_list(void)
198 unsigned topidx = idx * P2M_ENTRIES_PER_PAGE; 198 unsigned topidx = idx * P2M_ENTRIES_PER_PAGE;
199 p2m_top_mfn_list[idx] = virt_to_mfn(&p2m_top_mfn[topidx]); 199 p2m_top_mfn_list[idx] = virt_to_mfn(&p2m_top_mfn[topidx]);
200 } 200 }
201}
201 202
203void xen_setup_mfn_list_list(void)
204{
202 BUG_ON(HYPERVISOR_shared_info == &xen_dummy_shared_info); 205 BUG_ON(HYPERVISOR_shared_info == &xen_dummy_shared_info);
203 206
204 HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list = 207 HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list =
@@ -218,6 +221,8 @@ void __init xen_build_dynamic_phys_to_machine(void)
218 221
219 p2m_top[topidx] = &mfn_list[pfn]; 222 p2m_top[topidx] = &mfn_list[pfn];
220 } 223 }
224
225 xen_build_mfn_list_list();
221} 226}
222 227
223unsigned long get_phys_to_machine(unsigned long pfn) 228unsigned long get_phys_to_machine(unsigned long pfn)
@@ -233,47 +238,74 @@ unsigned long get_phys_to_machine(unsigned long pfn)
233} 238}
234EXPORT_SYMBOL_GPL(get_phys_to_machine); 239EXPORT_SYMBOL_GPL(get_phys_to_machine);
235 240
236static void alloc_p2m(unsigned long **pp, unsigned long *mfnp) 241/* install a new p2m_top page */
242bool install_p2mtop_page(unsigned long pfn, unsigned long *p)
237{ 243{
238 unsigned long *p; 244 unsigned topidx = p2m_top_index(pfn);
245 unsigned long **pfnp, *mfnp;
239 unsigned i; 246 unsigned i;
240 247
241 p = (void *)__get_free_page(GFP_KERNEL | __GFP_NOFAIL); 248 pfnp = &p2m_top[topidx];
242 BUG_ON(p == NULL); 249 mfnp = &p2m_top_mfn[topidx];
243 250
244 for (i = 0; i < P2M_ENTRIES_PER_PAGE; i++) 251 for (i = 0; i < P2M_ENTRIES_PER_PAGE; i++)
245 p[i] = INVALID_P2M_ENTRY; 252 p[i] = INVALID_P2M_ENTRY;
246 253
247 if (cmpxchg(pp, p2m_missing, p) != p2m_missing) 254 if (cmpxchg(pfnp, p2m_missing, p) == p2m_missing) {
248 free_page((unsigned long)p);
249 else
250 *mfnp = virt_to_mfn(p); 255 *mfnp = virt_to_mfn(p);
256 return true;
257 }
258
259 return false;
251} 260}
252 261
253void set_phys_to_machine(unsigned long pfn, unsigned long mfn) 262static void alloc_p2m(unsigned long pfn)
254{ 263{
255 unsigned topidx, idx; 264 unsigned long *p;
256 265
257 if (unlikely(xen_feature(XENFEAT_auto_translated_physmap))) { 266 p = (void *)__get_free_page(GFP_KERNEL | __GFP_NOFAIL);
258 BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY); 267 BUG_ON(p == NULL);
259 return; 268
260 } 269 if (!install_p2mtop_page(pfn, p))
270 free_page((unsigned long)p);
271}
272
273/* Try to install p2m mapping; fail if intermediate bits missing */
274bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn)
275{
276 unsigned topidx, idx;
261 277
262 if (unlikely(pfn >= MAX_DOMAIN_PAGES)) { 278 if (unlikely(pfn >= MAX_DOMAIN_PAGES)) {
263 BUG_ON(mfn != INVALID_P2M_ENTRY); 279 BUG_ON(mfn != INVALID_P2M_ENTRY);
264 return; 280 return true;
265 } 281 }
266 282
267 topidx = p2m_top_index(pfn); 283 topidx = p2m_top_index(pfn);
268 if (p2m_top[topidx] == p2m_missing) { 284 if (p2m_top[topidx] == p2m_missing) {
269 /* no need to allocate a page to store an invalid entry */
270 if (mfn == INVALID_P2M_ENTRY) 285 if (mfn == INVALID_P2M_ENTRY)
271 return; 286 return true;
272 alloc_p2m(&p2m_top[topidx], &p2m_top_mfn[topidx]); 287 return false;
273 } 288 }
274 289
275 idx = p2m_index(pfn); 290 idx = p2m_index(pfn);
276 p2m_top[topidx][idx] = mfn; 291 p2m_top[topidx][idx] = mfn;
292
293 return true;
294}
295
296void set_phys_to_machine(unsigned long pfn, unsigned long mfn)
297{
298 if (unlikely(xen_feature(XENFEAT_auto_translated_physmap))) {
299 BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY);
300 return;
301 }
302
303 if (unlikely(!__set_phys_to_machine(pfn, mfn))) {
304 alloc_p2m(pfn);
305
306 if (!__set_phys_to_machine(pfn, mfn))
307 BUG();
308 }
277} 309}
278 310
279unsigned long arbitrary_virt_to_mfn(void *vaddr) 311unsigned long arbitrary_virt_to_mfn(void *vaddr)
@@ -987,7 +1019,7 @@ static __init int xen_mark_pinned(struct mm_struct *mm, struct page *page,
987 return 0; 1019 return 0;
988} 1020}
989 1021
990void __init xen_mark_init_mm_pinned(void) 1022static void __init xen_mark_init_mm_pinned(void)
991{ 1023{
992 xen_pgd_walk(&init_mm, xen_mark_pinned, FIXADDR_TOP); 1024 xen_pgd_walk(&init_mm, xen_mark_pinned, FIXADDR_TOP);
993} 1025}
@@ -1270,8 +1302,8 @@ static void xen_flush_tlb_others(const struct cpumask *cpus,
1270 } *args; 1302 } *args;
1271 struct multicall_space mcs; 1303 struct multicall_space mcs;
1272 1304
1273 BUG_ON(cpumask_empty(cpus)); 1305 if (cpumask_empty(cpus))
1274 BUG_ON(!mm); 1306 return; /* nothing to do */
1275 1307
1276 mcs = xen_mc_entry(sizeof(*args)); 1308 mcs = xen_mc_entry(sizeof(*args));
1277 args = mcs.args; 1309 args = mcs.args;
@@ -1438,6 +1470,15 @@ static __init void xen_set_pte_init(pte_t *ptep, pte_t pte)
1438} 1470}
1439#endif 1471#endif
1440 1472
1473static void pin_pagetable_pfn(unsigned cmd, unsigned long pfn)
1474{
1475 struct mmuext_op op;
1476 op.cmd = cmd;
1477 op.arg1.mfn = pfn_to_mfn(pfn);
1478 if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF))
1479 BUG();
1480}
1481
1441/* Early in boot, while setting up the initial pagetable, assume 1482/* Early in boot, while setting up the initial pagetable, assume
1442 everything is pinned. */ 1483 everything is pinned. */
1443static __init void xen_alloc_pte_init(struct mm_struct *mm, unsigned long pfn) 1484static __init void xen_alloc_pte_init(struct mm_struct *mm, unsigned long pfn)
@@ -1446,22 +1487,29 @@ static __init void xen_alloc_pte_init(struct mm_struct *mm, unsigned long pfn)
1446 BUG_ON(mem_map); /* should only be used early */ 1487 BUG_ON(mem_map); /* should only be used early */
1447#endif 1488#endif
1448 make_lowmem_page_readonly(__va(PFN_PHYS(pfn))); 1489 make_lowmem_page_readonly(__va(PFN_PHYS(pfn)));
1490 pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn);
1491}
1492
1493/* Used for pmd and pud */
1494static __init void xen_alloc_pmd_init(struct mm_struct *mm, unsigned long pfn)
1495{
1496#ifdef CONFIG_FLATMEM
1497 BUG_ON(mem_map); /* should only be used early */
1498#endif
1499 make_lowmem_page_readonly(__va(PFN_PHYS(pfn)));
1449} 1500}
1450 1501
1451/* Early release_pte assumes that all pts are pinned, since there's 1502/* Early release_pte assumes that all pts are pinned, since there's
1452 only init_mm and anything attached to that is pinned. */ 1503 only init_mm and anything attached to that is pinned. */
1453static void xen_release_pte_init(unsigned long pfn) 1504static __init void xen_release_pte_init(unsigned long pfn)
1454{ 1505{
1506 pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn);
1455 make_lowmem_page_readwrite(__va(PFN_PHYS(pfn))); 1507 make_lowmem_page_readwrite(__va(PFN_PHYS(pfn)));
1456} 1508}
1457 1509
1458static void pin_pagetable_pfn(unsigned cmd, unsigned long pfn) 1510static __init void xen_release_pmd_init(unsigned long pfn)
1459{ 1511{
1460 struct mmuext_op op; 1512 make_lowmem_page_readwrite(__va(PFN_PHYS(pfn)));
1461 op.cmd = cmd;
1462 op.arg1.mfn = pfn_to_mfn(pfn);
1463 if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF))
1464 BUG();
1465} 1513}
1466 1514
1467/* This needs to make sure the new pte page is pinned iff its being 1515/* This needs to make sure the new pte page is pinned iff its being
@@ -1773,6 +1821,9 @@ static void xen_set_fixmap(unsigned idx, phys_addr_t phys, pgprot_t prot)
1773#ifdef CONFIG_X86_LOCAL_APIC 1821#ifdef CONFIG_X86_LOCAL_APIC
1774 case FIX_APIC_BASE: /* maps dummy local APIC */ 1822 case FIX_APIC_BASE: /* maps dummy local APIC */
1775#endif 1823#endif
1824 case FIX_TEXT_POKE0:
1825 case FIX_TEXT_POKE1:
1826 /* All local page mappings */
1776 pte = pfn_pte(phys, prot); 1827 pte = pfn_pte(phys, prot);
1777 break; 1828 break;
1778 1829
@@ -1819,7 +1870,6 @@ __init void xen_post_allocator_init(void)
1819 xen_mark_init_mm_pinned(); 1870 xen_mark_init_mm_pinned();
1820} 1871}
1821 1872
1822
1823const struct pv_mmu_ops xen_mmu_ops __initdata = { 1873const struct pv_mmu_ops xen_mmu_ops __initdata = {
1824 .pagetable_setup_start = xen_pagetable_setup_start, 1874 .pagetable_setup_start = xen_pagetable_setup_start,
1825 .pagetable_setup_done = xen_pagetable_setup_done, 1875 .pagetable_setup_done = xen_pagetable_setup_done,
@@ -1843,9 +1893,9 @@ const struct pv_mmu_ops xen_mmu_ops __initdata = {
1843 1893
1844 .alloc_pte = xen_alloc_pte_init, 1894 .alloc_pte = xen_alloc_pte_init,
1845 .release_pte = xen_release_pte_init, 1895 .release_pte = xen_release_pte_init,
1846 .alloc_pmd = xen_alloc_pte_init, 1896 .alloc_pmd = xen_alloc_pmd_init,
1847 .alloc_pmd_clone = paravirt_nop, 1897 .alloc_pmd_clone = paravirt_nop,
1848 .release_pmd = xen_release_pte_init, 1898 .release_pmd = xen_release_pmd_init,
1849 1899
1850#ifdef CONFIG_HIGHPTE 1900#ifdef CONFIG_HIGHPTE
1851 .kmap_atomic_pte = xen_kmap_atomic_pte, 1901 .kmap_atomic_pte = xen_kmap_atomic_pte,
@@ -1883,8 +1933,8 @@ const struct pv_mmu_ops xen_mmu_ops __initdata = {
1883 .make_pud = PV_CALLEE_SAVE(xen_make_pud), 1933 .make_pud = PV_CALLEE_SAVE(xen_make_pud),
1884 .set_pgd = xen_set_pgd_hyper, 1934 .set_pgd = xen_set_pgd_hyper,
1885 1935
1886 .alloc_pud = xen_alloc_pte_init, 1936 .alloc_pud = xen_alloc_pmd_init,
1887 .release_pud = xen_release_pte_init, 1937 .release_pud = xen_release_pmd_init,
1888#endif /* PAGETABLE_LEVELS == 4 */ 1938#endif /* PAGETABLE_LEVELS == 4 */
1889 1939
1890 .activate_mm = xen_activate_mm, 1940 .activate_mm = xen_activate_mm,
diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h
index 24d1b44a337d..da7302624897 100644
--- a/arch/x86/xen/mmu.h
+++ b/arch/x86/xen/mmu.h
@@ -11,6 +11,9 @@ enum pt_level {
11}; 11};
12 12
13 13
14bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn);
15bool install_p2mtop_page(unsigned long pfn, unsigned long *p);
16
14void set_pte_mfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags); 17void set_pte_mfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags);
15 18
16 19
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index 585a6e330837..429834ec1687 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -317,7 +317,7 @@ static int __cpuinit xen_cpu_up(unsigned int cpu)
317 BUG_ON(rc); 317 BUG_ON(rc);
318 318
319 while(per_cpu(cpu_state, cpu) != CPU_ONLINE) { 319 while(per_cpu(cpu_state, cpu) != CPU_ONLINE) {
320 HYPERVISOR_sched_op(SCHEDOP_yield, 0); 320 HYPERVISOR_sched_op(SCHEDOP_yield, NULL);
321 barrier(); 321 barrier();
322 } 322 }
323 323
@@ -422,7 +422,7 @@ static void xen_smp_send_call_function_ipi(const struct cpumask *mask)
422 /* Make sure other vcpus get a chance to run if they need to. */ 422 /* Make sure other vcpus get a chance to run if they need to. */
423 for_each_cpu(cpu, mask) { 423 for_each_cpu(cpu, mask) {
424 if (xen_vcpu_stolen(cpu)) { 424 if (xen_vcpu_stolen(cpu)) {
425 HYPERVISOR_sched_op(SCHEDOP_yield, 0); 425 HYPERVISOR_sched_op(SCHEDOP_yield, NULL);
426 break; 426 break;
427 } 427 }
428 } 428 }
diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
index 14f240623497..0a5aa44299a5 100644
--- a/arch/x86/xen/time.c
+++ b/arch/x86/xen/time.c
@@ -213,6 +213,11 @@ cycle_t xen_clocksource_read(void)
213 return ret; 213 return ret;
214} 214}
215 215
216static cycle_t xen_clocksource_get_cycles(struct clocksource *cs)
217{
218 return xen_clocksource_read();
219}
220
216static void xen_read_wallclock(struct timespec *ts) 221static void xen_read_wallclock(struct timespec *ts)
217{ 222{
218 struct shared_info *s = HYPERVISOR_shared_info; 223 struct shared_info *s = HYPERVISOR_shared_info;
@@ -241,7 +246,7 @@ int xen_set_wallclock(unsigned long now)
241static struct clocksource xen_clocksource __read_mostly = { 246static struct clocksource xen_clocksource __read_mostly = {
242 .name = "xen", 247 .name = "xen",
243 .rating = 400, 248 .rating = 400,
244 .read = xen_clocksource_read, 249 .read = xen_clocksource_get_cycles,
245 .mask = ~0, 250 .mask = ~0,
246 .mult = 1<<XEN_SHIFT, /* time directly in nanoseconds */ 251 .mult = 1<<XEN_SHIFT, /* time directly in nanoseconds */
247 .shift = XEN_SHIFT, 252 .shift = XEN_SHIFT,
diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
index 2f5ef2632ea2..20139464943c 100644
--- a/arch/x86/xen/xen-ops.h
+++ b/arch/x86/xen/xen-ops.h
@@ -57,8 +57,6 @@ irqreturn_t xen_debug_interrupt(int irq, void *dev_id);
57 57
58bool xen_vcpu_stolen(int vcpu); 58bool xen_vcpu_stolen(int vcpu);
59 59
60void xen_mark_init_mm_pinned(void);
61
62void xen_setup_vcpu_info_placement(void); 60void xen_setup_vcpu_info_placement(void);
63 61
64#ifdef CONFIG_SMP 62#ifdef CONFIG_SMP
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-22 05:11:37 -0500