diff options
Diffstat (limited to 'arch/x86')
-rw-r--r-- | arch/x86/Kconfig | 15 | ||||
-rw-r--r-- | arch/x86/include/asm/boot.h | 6 | ||||
-rw-r--r-- | arch/x86/include/asm/pci.h | 2 | ||||
-rw-r--r-- | arch/x86/include/asm/percpu.h | 10 | ||||
-rw-r--r-- | arch/x86/include/asm/perf_counter.h | 3 | ||||
-rw-r--r-- | arch/x86/include/asm/proto.h | 11 | ||||
-rw-r--r-- | arch/x86/kernel/cpu/amd.c | 4 | ||||
-rw-r--r-- | arch/x86/kernel/cpu/common.c | 3 | ||||
-rw-r--r-- | arch/x86/kernel/cpu/mcheck/mce.c | 4 | ||||
-rw-r--r-- | arch/x86/kernel/cpu/perf_counter.c | 22 | ||||
-rw-r--r-- | arch/x86/kernel/dumpstack.c | 1 | ||||
-rw-r--r-- | arch/x86/kernel/e820.c | 16 | ||||
-rw-r--r-- | arch/x86/kernel/pci-dma.c | 2 | ||||
-rw-r--r-- | arch/x86/kernel/setup.c | 16 | ||||
-rw-r--r-- | arch/x86/kernel/setup_percpu.c | 219 | ||||
-rw-r--r-- | arch/x86/kernel/tlb_uv.c | 9 | ||||
-rw-r--r-- | arch/x86/kernel/traps.c | 3 | ||||
-rw-r--r-- | arch/x86/kvm/mmu.c | 6 | ||||
-rw-r--r-- | arch/x86/kvm/paging_tmpl.h | 2 | ||||
-rw-r--r-- | arch/x86/kvm/vmx.c | 15 | ||||
-rw-r--r-- | arch/x86/kvm/x86.c | 1 | ||||
-rw-r--r-- | arch/x86/kvm/x86_emulate.c | 2 | ||||
-rw-r--r-- | arch/x86/lib/delay.c | 3 | ||||
-rw-r--r-- | arch/x86/mm/init.c | 17 | ||||
-rw-r--r-- | arch/x86/mm/init_64.c | 2 | ||||
-rw-r--r-- | arch/x86/mm/pageattr.c | 65 | ||||
-rw-r--r-- | arch/x86/power/cpu.c | 2 |
27 files changed, 305 insertions, 156 deletions
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index d1430ef6b4f9..c07f72205909 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig | |||
@@ -1913,25 +1913,14 @@ config DMAR_DEFAULT_ON | |||
1913 | recommended you say N here while the DMAR code remains | 1913 | recommended you say N here while the DMAR code remains |
1914 | experimental. | 1914 | experimental. |
1915 | 1915 | ||
1916 | config DMAR_GFX_WA | ||
1917 | def_bool y | ||
1918 | prompt "Support for Graphics workaround" | ||
1919 | depends on DMAR | ||
1920 | ---help--- | ||
1921 | Current Graphics drivers tend to use physical address | ||
1922 | for DMA and avoid using DMA APIs. Setting this config | ||
1923 | option permits the IOMMU driver to set a unity map for | ||
1924 | all the OS-visible memory. Hence the driver can continue | ||
1925 | to use physical addresses for DMA. | ||
1926 | |||
1927 | config DMAR_FLOPPY_WA | 1916 | config DMAR_FLOPPY_WA |
1928 | def_bool y | 1917 | def_bool y |
1929 | depends on DMAR | 1918 | depends on DMAR |
1930 | ---help--- | 1919 | ---help--- |
1931 | Floppy disk drivers are know to bypass DMA API calls | 1920 | Floppy disk drivers are known to bypass DMA API calls |
1932 | thereby failing to work when IOMMU is enabled. This | 1921 | thereby failing to work when IOMMU is enabled. This |
1933 | workaround will setup a 1:1 mapping for the first | 1922 | workaround will setup a 1:1 mapping for the first |
1934 | 16M to make floppy (an ISA device) work. | 1923 | 16MiB to make floppy (an ISA device) work. |
1935 | 1924 | ||
1936 | config INTR_REMAP | 1925 | config INTR_REMAP |
1937 | bool "Support for Interrupt Remapping (EXPERIMENTAL)" | 1926 | bool "Support for Interrupt Remapping (EXPERIMENTAL)" |
diff --git a/arch/x86/include/asm/boot.h b/arch/x86/include/asm/boot.h index 418e632d4a80..7a1065958ba9 100644 --- a/arch/x86/include/asm/boot.h +++ b/arch/x86/include/asm/boot.h | |||
@@ -8,7 +8,7 @@ | |||
8 | 8 | ||
9 | #ifdef __KERNEL__ | 9 | #ifdef __KERNEL__ |
10 | 10 | ||
11 | #include <asm/page_types.h> | 11 | #include <asm/pgtable_types.h> |
12 | 12 | ||
13 | /* Physical address where kernel should be loaded. */ | 13 | /* Physical address where kernel should be loaded. */ |
14 | #define LOAD_PHYSICAL_ADDR ((CONFIG_PHYSICAL_START \ | 14 | #define LOAD_PHYSICAL_ADDR ((CONFIG_PHYSICAL_START \ |
@@ -16,10 +16,10 @@ | |||
16 | & ~(CONFIG_PHYSICAL_ALIGN - 1)) | 16 | & ~(CONFIG_PHYSICAL_ALIGN - 1)) |
17 | 17 | ||
18 | /* Minimum kernel alignment, as a power of two */ | 18 | /* Minimum kernel alignment, as a power of two */ |
19 | #ifdef CONFIG_x86_64 | 19 | #ifdef CONFIG_X86_64 |
20 | #define MIN_KERNEL_ALIGN_LG2 PMD_SHIFT | 20 | #define MIN_KERNEL_ALIGN_LG2 PMD_SHIFT |
21 | #else | 21 | #else |
22 | #define MIN_KERNEL_ALIGN_LG2 (PAGE_SHIFT+1) | 22 | #define MIN_KERNEL_ALIGN_LG2 (PAGE_SHIFT + THREAD_ORDER) |
23 | #endif | 23 | #endif |
24 | #define MIN_KERNEL_ALIGN (_AC(1, UL) << MIN_KERNEL_ALIGN_LG2) | 24 | #define MIN_KERNEL_ALIGN (_AC(1, UL) << MIN_KERNEL_ALIGN_LG2) |
25 | 25 | ||
diff --git a/arch/x86/include/asm/pci.h b/arch/x86/include/asm/pci.h index 927958d13c19..1ff685ca221c 100644 --- a/arch/x86/include/asm/pci.h +++ b/arch/x86/include/asm/pci.h | |||
@@ -91,7 +91,7 @@ extern void pci_iommu_alloc(void); | |||
91 | 91 | ||
92 | #define PCI_DMA_BUS_IS_PHYS (dma_ops->is_phys) | 92 | #define PCI_DMA_BUS_IS_PHYS (dma_ops->is_phys) |
93 | 93 | ||
94 | #if defined(CONFIG_X86_64) || defined(CONFIG_DMA_API_DEBUG) | 94 | #if defined(CONFIG_X86_64) || defined(CONFIG_DMAR) || defined(CONFIG_DMA_API_DEBUG) |
95 | 95 | ||
96 | #define DECLARE_PCI_UNMAP_ADDR(ADDR_NAME) \ | 96 | #define DECLARE_PCI_UNMAP_ADDR(ADDR_NAME) \ |
97 | dma_addr_t ADDR_NAME; | 97 | dma_addr_t ADDR_NAME; |
diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h index 02ecb30982a3..103f1ddb0d85 100644 --- a/arch/x86/include/asm/percpu.h +++ b/arch/x86/include/asm/percpu.h | |||
@@ -42,6 +42,7 @@ | |||
42 | 42 | ||
43 | #else /* ...!ASSEMBLY */ | 43 | #else /* ...!ASSEMBLY */ |
44 | 44 | ||
45 | #include <linux/kernel.h> | ||
45 | #include <linux/stringify.h> | 46 | #include <linux/stringify.h> |
46 | 47 | ||
47 | #ifdef CONFIG_SMP | 48 | #ifdef CONFIG_SMP |
@@ -155,6 +156,15 @@ do { \ | |||
155 | /* We can use this directly for local CPU (faster). */ | 156 | /* We can use this directly for local CPU (faster). */ |
156 | DECLARE_PER_CPU(unsigned long, this_cpu_off); | 157 | DECLARE_PER_CPU(unsigned long, this_cpu_off); |
157 | 158 | ||
159 | #ifdef CONFIG_NEED_MULTIPLE_NODES | ||
160 | void *pcpu_lpage_remapped(void *kaddr); | ||
161 | #else | ||
162 | static inline void *pcpu_lpage_remapped(void *kaddr) | ||
163 | { | ||
164 | return NULL; | ||
165 | } | ||
166 | #endif | ||
167 | |||
158 | #endif /* !__ASSEMBLY__ */ | 168 | #endif /* !__ASSEMBLY__ */ |
159 | 169 | ||
160 | #ifdef CONFIG_SMP | 170 | #ifdef CONFIG_SMP |
diff --git a/arch/x86/include/asm/perf_counter.h b/arch/x86/include/asm/perf_counter.h index 5fb33e160ea0..fa64e401589d 100644 --- a/arch/x86/include/asm/perf_counter.h +++ b/arch/x86/include/asm/perf_counter.h | |||
@@ -87,6 +87,9 @@ union cpuid10_edx { | |||
87 | #ifdef CONFIG_PERF_COUNTERS | 87 | #ifdef CONFIG_PERF_COUNTERS |
88 | extern void init_hw_perf_counters(void); | 88 | extern void init_hw_perf_counters(void); |
89 | extern void perf_counters_lapic_init(void); | 89 | extern void perf_counters_lapic_init(void); |
90 | |||
91 | #define PERF_COUNTER_INDEX_OFFSET 0 | ||
92 | |||
90 | #else | 93 | #else |
91 | static inline void init_hw_perf_counters(void) { } | 94 | static inline void init_hw_perf_counters(void) { } |
92 | static inline void perf_counters_lapic_init(void) { } | 95 | static inline void perf_counters_lapic_init(void) { } |
diff --git a/arch/x86/include/asm/proto.h b/arch/x86/include/asm/proto.h index 49fb3ecf3bb3..621f56d73121 100644 --- a/arch/x86/include/asm/proto.h +++ b/arch/x86/include/asm/proto.h | |||
@@ -22,7 +22,14 @@ extern int reboot_force; | |||
22 | 22 | ||
23 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr); | 23 | long do_arch_prctl(struct task_struct *task, int code, unsigned long addr); |
24 | 24 | ||
25 | #define round_up(x, y) (((x) + (y) - 1) & ~((y) - 1)) | 25 | /* |
26 | #define round_down(x, y) ((x) & ~((y) - 1)) | 26 | * This looks more complex than it should be. But we need to |
27 | * get the type for the ~ right in round_down (it needs to be | ||
28 | * as wide as the result!), and we want to evaluate the macro | ||
29 | * arguments just once each. | ||
30 | */ | ||
31 | #define __round_mask(x,y) ((__typeof__(x))((y)-1)) | ||
32 | #define round_up(x,y) ((((x)-1) | __round_mask(x,y))+1) | ||
33 | #define round_down(x,y) ((x) & ~__round_mask(x,y)) | ||
27 | 34 | ||
28 | #endif /* _ASM_X86_PROTO_H */ | 35 | #endif /* _ASM_X86_PROTO_H */ |
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index e5b27d8f1b47..28e5f5956042 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c | |||
@@ -258,13 +258,15 @@ static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c) | |||
258 | { | 258 | { |
259 | #ifdef CONFIG_X86_HT | 259 | #ifdef CONFIG_X86_HT |
260 | unsigned bits; | 260 | unsigned bits; |
261 | int cpu = smp_processor_id(); | ||
261 | 262 | ||
262 | bits = c->x86_coreid_bits; | 263 | bits = c->x86_coreid_bits; |
263 | |||
264 | /* Low order bits define the core id (index of core in socket) */ | 264 | /* Low order bits define the core id (index of core in socket) */ |
265 | c->cpu_core_id = c->initial_apicid & ((1 << bits)-1); | 265 | c->cpu_core_id = c->initial_apicid & ((1 << bits)-1); |
266 | /* Convert the initial APIC ID into the socket ID */ | 266 | /* Convert the initial APIC ID into the socket ID */ |
267 | c->phys_proc_id = c->initial_apicid >> bits; | 267 | c->phys_proc_id = c->initial_apicid >> bits; |
268 | /* use socket ID also for last level cache */ | ||
269 | per_cpu(cpu_llc_id, cpu) = c->phys_proc_id; | ||
268 | #endif | 270 | #endif |
269 | } | 271 | } |
270 | 272 | ||
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 6b26d4deada0..f1961c07af9a 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c | |||
@@ -848,9 +848,6 @@ static void __cpuinit identify_cpu(struct cpuinfo_x86 *c) | |||
848 | #if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) | 848 | #if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) |
849 | numa_add_cpu(smp_processor_id()); | 849 | numa_add_cpu(smp_processor_id()); |
850 | #endif | 850 | #endif |
851 | |||
852 | /* Cap the iomem address space to what is addressable on all CPUs */ | ||
853 | iomem_resource.end &= (1ULL << c->x86_phys_bits) - 1; | ||
854 | } | 851 | } |
855 | 852 | ||
856 | #ifdef CONFIG_X86_64 | 853 | #ifdef CONFIG_X86_64 |
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c index 284d1de968bc..af425b83202b 100644 --- a/arch/x86/kernel/cpu/mcheck/mce.c +++ b/arch/x86/kernel/cpu/mcheck/mce.c | |||
@@ -1117,7 +1117,7 @@ static void mcheck_timer(unsigned long data) | |||
1117 | *n = min(*n*2, (int)round_jiffies_relative(check_interval*HZ)); | 1117 | *n = min(*n*2, (int)round_jiffies_relative(check_interval*HZ)); |
1118 | 1118 | ||
1119 | t->expires = jiffies + *n; | 1119 | t->expires = jiffies + *n; |
1120 | add_timer(t); | 1120 | add_timer_on(t, smp_processor_id()); |
1121 | } | 1121 | } |
1122 | 1122 | ||
1123 | static void mce_do_trigger(struct work_struct *work) | 1123 | static void mce_do_trigger(struct work_struct *work) |
@@ -1321,7 +1321,7 @@ static void mce_init_timer(void) | |||
1321 | return; | 1321 | return; |
1322 | setup_timer(t, mcheck_timer, smp_processor_id()); | 1322 | setup_timer(t, mcheck_timer, smp_processor_id()); |
1323 | t->expires = round_jiffies(jiffies + *n); | 1323 | t->expires = round_jiffies(jiffies + *n); |
1324 | add_timer(t); | 1324 | add_timer_on(t, smp_processor_id()); |
1325 | } | 1325 | } |
1326 | 1326 | ||
1327 | /* | 1327 | /* |
diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 76dfef23f789..d4cf4ce19aac 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c | |||
@@ -401,7 +401,7 @@ static const u64 amd_hw_cache_event_ids | |||
401 | [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */ | 401 | [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */ |
402 | }, | 402 | }, |
403 | [ C(OP_WRITE) ] = { | 403 | [ C(OP_WRITE) ] = { |
404 | [ C(RESULT_ACCESS) ] = 0x0042, /* Data Cache Refills from L2 */ | 404 | [ C(RESULT_ACCESS) ] = 0x0142, /* Data Cache Refills :system */ |
405 | [ C(RESULT_MISS) ] = 0, | 405 | [ C(RESULT_MISS) ] = 0, |
406 | }, | 406 | }, |
407 | [ C(OP_PREFETCH) ] = { | 407 | [ C(OP_PREFETCH) ] = { |
@@ -912,6 +912,8 @@ x86_perf_counter_set_period(struct perf_counter *counter, | |||
912 | err = checking_wrmsrl(hwc->counter_base + idx, | 912 | err = checking_wrmsrl(hwc->counter_base + idx, |
913 | (u64)(-left) & x86_pmu.counter_mask); | 913 | (u64)(-left) & x86_pmu.counter_mask); |
914 | 914 | ||
915 | perf_counter_update_userpage(counter); | ||
916 | |||
915 | return ret; | 917 | return ret; |
916 | } | 918 | } |
917 | 919 | ||
@@ -969,13 +971,6 @@ fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc) | |||
969 | if (!x86_pmu.num_counters_fixed) | 971 | if (!x86_pmu.num_counters_fixed) |
970 | return -1; | 972 | return -1; |
971 | 973 | ||
972 | /* | ||
973 | * Quirk, IA32_FIXED_CTRs do not work on current Atom processors: | ||
974 | */ | ||
975 | if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL && | ||
976 | boot_cpu_data.x86_model == 28) | ||
977 | return -1; | ||
978 | |||
979 | event = hwc->config & ARCH_PERFMON_EVENT_MASK; | 974 | event = hwc->config & ARCH_PERFMON_EVENT_MASK; |
980 | 975 | ||
981 | if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS))) | 976 | if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS))) |
@@ -1041,6 +1036,8 @@ try_generic: | |||
1041 | x86_perf_counter_set_period(counter, hwc, idx); | 1036 | x86_perf_counter_set_period(counter, hwc, idx); |
1042 | x86_pmu.enable(hwc, idx); | 1037 | x86_pmu.enable(hwc, idx); |
1043 | 1038 | ||
1039 | perf_counter_update_userpage(counter); | ||
1040 | |||
1044 | return 0; | 1041 | return 0; |
1045 | } | 1042 | } |
1046 | 1043 | ||
@@ -1133,6 +1130,8 @@ static void x86_pmu_disable(struct perf_counter *counter) | |||
1133 | x86_perf_counter_update(counter, hwc, idx); | 1130 | x86_perf_counter_update(counter, hwc, idx); |
1134 | cpuc->counters[idx] = NULL; | 1131 | cpuc->counters[idx] = NULL; |
1135 | clear_bit(idx, cpuc->used_mask); | 1132 | clear_bit(idx, cpuc->used_mask); |
1133 | |||
1134 | perf_counter_update_userpage(counter); | ||
1136 | } | 1135 | } |
1137 | 1136 | ||
1138 | /* | 1137 | /* |
@@ -1428,8 +1427,6 @@ static int intel_pmu_init(void) | |||
1428 | */ | 1427 | */ |
1429 | x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3); | 1428 | x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3); |
1430 | 1429 | ||
1431 | rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl); | ||
1432 | |||
1433 | /* | 1430 | /* |
1434 | * Install the hw-cache-events table: | 1431 | * Install the hw-cache-events table: |
1435 | */ | 1432 | */ |
@@ -1499,21 +1496,22 @@ void __init init_hw_perf_counters(void) | |||
1499 | pr_cont("%s PMU driver.\n", x86_pmu.name); | 1496 | pr_cont("%s PMU driver.\n", x86_pmu.name); |
1500 | 1497 | ||
1501 | if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { | 1498 | if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { |
1502 | x86_pmu.num_counters = X86_PMC_MAX_GENERIC; | ||
1503 | WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!", | 1499 | WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!", |
1504 | x86_pmu.num_counters, X86_PMC_MAX_GENERIC); | 1500 | x86_pmu.num_counters, X86_PMC_MAX_GENERIC); |
1501 | x86_pmu.num_counters = X86_PMC_MAX_GENERIC; | ||
1505 | } | 1502 | } |
1506 | perf_counter_mask = (1 << x86_pmu.num_counters) - 1; | 1503 | perf_counter_mask = (1 << x86_pmu.num_counters) - 1; |
1507 | perf_max_counters = x86_pmu.num_counters; | 1504 | perf_max_counters = x86_pmu.num_counters; |
1508 | 1505 | ||
1509 | if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) { | 1506 | if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) { |
1510 | x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED; | ||
1511 | WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!", | 1507 | WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!", |
1512 | x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED); | 1508 | x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED); |
1509 | x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED; | ||
1513 | } | 1510 | } |
1514 | 1511 | ||
1515 | perf_counter_mask |= | 1512 | perf_counter_mask |= |
1516 | ((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED; | 1513 | ((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED; |
1514 | x86_pmu.intel_ctrl = perf_counter_mask; | ||
1517 | 1515 | ||
1518 | perf_counters_lapic_init(); | 1516 | perf_counters_lapic_init(); |
1519 | register_die_notifier(&perf_counter_nmi_notifier); | 1517 | register_die_notifier(&perf_counter_nmi_notifier); |
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c index 95ea5fa7d444..c8405718a4c3 100644 --- a/arch/x86/kernel/dumpstack.c +++ b/arch/x86/kernel/dumpstack.c | |||
@@ -22,6 +22,7 @@ | |||
22 | #include "dumpstack.h" | 22 | #include "dumpstack.h" |
23 | 23 | ||
24 | int panic_on_unrecovered_nmi; | 24 | int panic_on_unrecovered_nmi; |
25 | int panic_on_io_nmi; | ||
25 | unsigned int code_bytes = 64; | 26 | unsigned int code_bytes = 64; |
26 | int kstack_depth_to_print = 3 * STACKSLOTS_PER_LINE; | 27 | int kstack_depth_to_print = 3 * STACKSLOTS_PER_LINE; |
27 | static int die_counter; | 28 | static int die_counter; |
diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c index 7271fa33d791..c4ca89d9aaf4 100644 --- a/arch/x86/kernel/e820.c +++ b/arch/x86/kernel/e820.c | |||
@@ -1383,6 +1383,8 @@ static unsigned long ram_alignment(resource_size_t pos) | |||
1383 | return 32*1024*1024; | 1383 | return 32*1024*1024; |
1384 | } | 1384 | } |
1385 | 1385 | ||
1386 | #define MAX_RESOURCE_SIZE ((resource_size_t)-1) | ||
1387 | |||
1386 | void __init e820_reserve_resources_late(void) | 1388 | void __init e820_reserve_resources_late(void) |
1387 | { | 1389 | { |
1388 | int i; | 1390 | int i; |
@@ -1400,17 +1402,19 @@ void __init e820_reserve_resources_late(void) | |||
1400 | * avoid stolen RAM: | 1402 | * avoid stolen RAM: |
1401 | */ | 1403 | */ |
1402 | for (i = 0; i < e820.nr_map; i++) { | 1404 | for (i = 0; i < e820.nr_map; i++) { |
1403 | struct e820entry *entry = &e820_saved.map[i]; | 1405 | struct e820entry *entry = &e820.map[i]; |
1404 | resource_size_t start, end; | 1406 | u64 start, end; |
1405 | 1407 | ||
1406 | if (entry->type != E820_RAM) | 1408 | if (entry->type != E820_RAM) |
1407 | continue; | 1409 | continue; |
1408 | start = entry->addr + entry->size; | 1410 | start = entry->addr + entry->size; |
1409 | end = round_up(start, ram_alignment(start)); | 1411 | end = round_up(start, ram_alignment(start)) - 1; |
1410 | if (start == end) | 1412 | if (end > MAX_RESOURCE_SIZE) |
1413 | end = MAX_RESOURCE_SIZE; | ||
1414 | if (start >= end) | ||
1411 | continue; | 1415 | continue; |
1412 | reserve_region_with_split(&iomem_resource, start, | 1416 | reserve_region_with_split(&iomem_resource, start, end, |
1413 | end - 1, "RAM buffer"); | 1417 | "RAM buffer"); |
1414 | } | 1418 | } |
1415 | } | 1419 | } |
1416 | 1420 | ||
diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index 47630479b067..1a041bcf506b 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c | |||
@@ -211,11 +211,11 @@ static __init int iommu_setup(char *p) | |||
211 | #ifdef CONFIG_SWIOTLB | 211 | #ifdef CONFIG_SWIOTLB |
212 | if (!strncmp(p, "soft", 4)) | 212 | if (!strncmp(p, "soft", 4)) |
213 | swiotlb = 1; | 213 | swiotlb = 1; |
214 | #endif | ||
214 | if (!strncmp(p, "pt", 2)) { | 215 | if (!strncmp(p, "pt", 2)) { |
215 | iommu_pass_through = 1; | 216 | iommu_pass_through = 1; |
216 | return 1; | 217 | return 1; |
217 | } | 218 | } |
218 | #endif | ||
219 | 219 | ||
220 | gart_parse_options(p); | 220 | gart_parse_options(p); |
221 | 221 | ||
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index be5ae80f897f..de2cab132844 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c | |||
@@ -289,6 +289,20 @@ void * __init extend_brk(size_t size, size_t align) | |||
289 | return ret; | 289 | return ret; |
290 | } | 290 | } |
291 | 291 | ||
292 | #ifdef CONFIG_X86_64 | ||
293 | static void __init init_gbpages(void) | ||
294 | { | ||
295 | if (direct_gbpages && cpu_has_gbpages) | ||
296 | printk(KERN_INFO "Using GB pages for direct mapping\n"); | ||
297 | else | ||
298 | direct_gbpages = 0; | ||
299 | } | ||
300 | #else | ||
301 | static inline void init_gbpages(void) | ||
302 | { | ||
303 | } | ||
304 | #endif | ||
305 | |||
292 | static void __init reserve_brk(void) | 306 | static void __init reserve_brk(void) |
293 | { | 307 | { |
294 | if (_brk_end > _brk_start) | 308 | if (_brk_end > _brk_start) |
@@ -871,6 +885,8 @@ void __init setup_arch(char **cmdline_p) | |||
871 | 885 | ||
872 | reserve_brk(); | 886 | reserve_brk(); |
873 | 887 | ||
888 | init_gbpages(); | ||
889 | |||
874 | /* max_pfn_mapped is updated here */ | 890 | /* max_pfn_mapped is updated here */ |
875 | max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT); | 891 | max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT); |
876 | max_pfn_mapped = max_low_pfn_mapped; | 892 | max_pfn_mapped = max_low_pfn_mapped; |
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c index 9c3f0823e6aa..29a3eef7cf4a 100644 --- a/arch/x86/kernel/setup_percpu.c +++ b/arch/x86/kernel/setup_percpu.c | |||
@@ -124,7 +124,7 @@ static void * __init pcpu_alloc_bootmem(unsigned int cpu, unsigned long size, | |||
124 | } | 124 | } |
125 | 125 | ||
126 | /* | 126 | /* |
127 | * Remap allocator | 127 | * Large page remap allocator |
128 | * | 128 | * |
129 | * This allocator uses PMD page as unit. A PMD page is allocated for | 129 | * This allocator uses PMD page as unit. A PMD page is allocated for |
130 | * each cpu and each is remapped into vmalloc area using PMD mapping. | 130 | * each cpu and each is remapped into vmalloc area using PMD mapping. |
@@ -137,105 +137,185 @@ static void * __init pcpu_alloc_bootmem(unsigned int cpu, unsigned long size, | |||
137 | * better than only using 4k mappings while still being NUMA friendly. | 137 | * better than only using 4k mappings while still being NUMA friendly. |
138 | */ | 138 | */ |
139 | #ifdef CONFIG_NEED_MULTIPLE_NODES | 139 | #ifdef CONFIG_NEED_MULTIPLE_NODES |
140 | static size_t pcpur_size __initdata; | 140 | struct pcpul_ent { |
141 | static void **pcpur_ptrs __initdata; | 141 | unsigned int cpu; |
142 | void *ptr; | ||
143 | }; | ||
144 | |||
145 | static size_t pcpul_size; | ||
146 | static struct pcpul_ent *pcpul_map; | ||
147 | static struct vm_struct pcpul_vm; | ||
142 | 148 | ||
143 | static struct page * __init pcpur_get_page(unsigned int cpu, int pageno) | 149 | static struct page * __init pcpul_get_page(unsigned int cpu, int pageno) |
144 | { | 150 | { |
145 | size_t off = (size_t)pageno << PAGE_SHIFT; | 151 | size_t off = (size_t)pageno << PAGE_SHIFT; |
146 | 152 | ||
147 | if (off >= pcpur_size) | 153 | if (off >= pcpul_size) |
148 | return NULL; | 154 | return NULL; |
149 | 155 | ||
150 | return virt_to_page(pcpur_ptrs[cpu] + off); | 156 | return virt_to_page(pcpul_map[cpu].ptr + off); |
151 | } | 157 | } |
152 | 158 | ||
153 | static ssize_t __init setup_pcpu_remap(size_t static_size) | 159 | static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen) |
154 | { | 160 | { |
155 | static struct vm_struct vm; | 161 | size_t map_size, dyn_size; |
156 | size_t ptrs_size, dyn_size; | ||
157 | unsigned int cpu; | 162 | unsigned int cpu; |
163 | int i, j; | ||
158 | ssize_t ret; | 164 | ssize_t ret; |
159 | 165 | ||
160 | /* | 166 | if (!chosen) { |
161 | * If large page isn't supported, there's no benefit in doing | 167 | size_t vm_size = VMALLOC_END - VMALLOC_START; |
162 | * this. Also, on non-NUMA, embedding is better. | 168 | size_t tot_size = num_possible_cpus() * PMD_SIZE; |
163 | * | 169 | |
164 | * NOTE: disabled for now. | 170 | /* on non-NUMA, embedding is better */ |
165 | */ | 171 | if (!pcpu_need_numa()) |
166 | if (true || !cpu_has_pse || !pcpu_need_numa()) | 172 | return -EINVAL; |
173 | |||
174 | /* don't consume more than 20% of vmalloc area */ | ||
175 | if (tot_size > vm_size / 5) { | ||
176 | pr_info("PERCPU: too large chunk size %zuMB for " | ||
177 | "large page remap\n", tot_size >> 20); | ||
178 | return -EINVAL; | ||
179 | } | ||
180 | } | ||
181 | |||
182 | /* need PSE */ | ||
183 | if (!cpu_has_pse) { | ||
184 | pr_warning("PERCPU: lpage allocator requires PSE\n"); | ||
167 | return -EINVAL; | 185 | return -EINVAL; |
186 | } | ||
168 | 187 | ||
169 | /* | 188 | /* |
170 | * Currently supports only single page. Supporting multiple | 189 | * Currently supports only single page. Supporting multiple |
171 | * pages won't be too difficult if it ever becomes necessary. | 190 | * pages won't be too difficult if it ever becomes necessary. |
172 | */ | 191 | */ |
173 | pcpur_size = PFN_ALIGN(static_size + PERCPU_MODULE_RESERVE + | 192 | pcpul_size = PFN_ALIGN(static_size + PERCPU_MODULE_RESERVE + |
174 | PERCPU_DYNAMIC_RESERVE); | 193 | PERCPU_DYNAMIC_RESERVE); |
175 | if (pcpur_size > PMD_SIZE) { | 194 | if (pcpul_size > PMD_SIZE) { |
176 | pr_warning("PERCPU: static data is larger than large page, " | 195 | pr_warning("PERCPU: static data is larger than large page, " |
177 | "can't use large page\n"); | 196 | "can't use large page\n"); |
178 | return -EINVAL; | 197 | return -EINVAL; |
179 | } | 198 | } |
180 | dyn_size = pcpur_size - static_size - PERCPU_FIRST_CHUNK_RESERVE; | 199 | dyn_size = pcpul_size - static_size - PERCPU_FIRST_CHUNK_RESERVE; |
181 | 200 | ||
182 | /* allocate pointer array and alloc large pages */ | 201 | /* allocate pointer array and alloc large pages */ |
183 | ptrs_size = PFN_ALIGN(num_possible_cpus() * sizeof(pcpur_ptrs[0])); | 202 | map_size = PFN_ALIGN(num_possible_cpus() * sizeof(pcpul_map[0])); |
184 | pcpur_ptrs = alloc_bootmem(ptrs_size); | 203 | pcpul_map = alloc_bootmem(map_size); |
185 | 204 | ||
186 | for_each_possible_cpu(cpu) { | 205 | for_each_possible_cpu(cpu) { |
187 | pcpur_ptrs[cpu] = pcpu_alloc_bootmem(cpu, PMD_SIZE, PMD_SIZE); | 206 | pcpul_map[cpu].cpu = cpu; |
188 | if (!pcpur_ptrs[cpu]) | 207 | pcpul_map[cpu].ptr = pcpu_alloc_bootmem(cpu, PMD_SIZE, |
208 | PMD_SIZE); | ||
209 | if (!pcpul_map[cpu].ptr) { | ||
210 | pr_warning("PERCPU: failed to allocate large page " | ||
211 | "for cpu%u\n", cpu); | ||
189 | goto enomem; | 212 | goto enomem; |
213 | } | ||
190 | 214 | ||
191 | /* | 215 | /* |
192 | * Only use pcpur_size bytes and give back the rest. | 216 | * Only use pcpul_size bytes and give back the rest. |
193 | * | 217 | * |
194 | * Ingo: The 2MB up-rounding bootmem is needed to make | 218 | * Ingo: The 2MB up-rounding bootmem is needed to make |
195 | * sure the partial 2MB page is still fully RAM - it's | 219 | * sure the partial 2MB page is still fully RAM - it's |
196 | * not well-specified to have a PAT-incompatible area | 220 | * not well-specified to have a PAT-incompatible area |
197 | * (unmapped RAM, device memory, etc.) in that hole. | 221 | * (unmapped RAM, device memory, etc.) in that hole. |
198 | */ | 222 | */ |
199 | free_bootmem(__pa(pcpur_ptrs[cpu] + pcpur_size), | 223 | free_bootmem(__pa(pcpul_map[cpu].ptr + pcpul_size), |
200 | PMD_SIZE - pcpur_size); | 224 | PMD_SIZE - pcpul_size); |
201 | 225 | ||
202 | memcpy(pcpur_ptrs[cpu], __per_cpu_load, static_size); | 226 | memcpy(pcpul_map[cpu].ptr, __per_cpu_load, static_size); |
203 | } | 227 | } |
204 | 228 | ||
205 | /* allocate address and map */ | 229 | /* allocate address and map */ |
206 | vm.flags = VM_ALLOC; | 230 | pcpul_vm.flags = VM_ALLOC; |
207 | vm.size = num_possible_cpus() * PMD_SIZE; | 231 | pcpul_vm.size = num_possible_cpus() * PMD_SIZE; |
208 | vm_area_register_early(&vm, PMD_SIZE); | 232 | vm_area_register_early(&pcpul_vm, PMD_SIZE); |
209 | 233 | ||
210 | for_each_possible_cpu(cpu) { | 234 | for_each_possible_cpu(cpu) { |
211 | pmd_t *pmd; | 235 | pmd_t *pmd, pmd_v; |
212 | 236 | ||
213 | pmd = populate_extra_pmd((unsigned long)vm.addr | 237 | pmd = populate_extra_pmd((unsigned long)pcpul_vm.addr + |
214 | + cpu * PMD_SIZE); | 238 | cpu * PMD_SIZE); |
215 | set_pmd(pmd, pfn_pmd(page_to_pfn(virt_to_page(pcpur_ptrs[cpu])), | 239 | pmd_v = pfn_pmd(page_to_pfn(virt_to_page(pcpul_map[cpu].ptr)), |
216 | PAGE_KERNEL_LARGE)); | 240 | PAGE_KERNEL_LARGE); |
241 | set_pmd(pmd, pmd_v); | ||
217 | } | 242 | } |
218 | 243 | ||
219 | /* we're ready, commit */ | 244 | /* we're ready, commit */ |
220 | pr_info("PERCPU: Remapped at %p with large pages, static data " | 245 | pr_info("PERCPU: Remapped at %p with large pages, static data " |
221 | "%zu bytes\n", vm.addr, static_size); | 246 | "%zu bytes\n", pcpul_vm.addr, static_size); |
222 | 247 | ||
223 | ret = pcpu_setup_first_chunk(pcpur_get_page, static_size, | 248 | ret = pcpu_setup_first_chunk(pcpul_get_page, static_size, |
224 | PERCPU_FIRST_CHUNK_RESERVE, dyn_size, | 249 | PERCPU_FIRST_CHUNK_RESERVE, dyn_size, |
225 | PMD_SIZE, vm.addr, NULL); | 250 | PMD_SIZE, pcpul_vm.addr, NULL); |
226 | goto out_free_ar; | 251 | |
252 | /* sort pcpul_map array for pcpu_lpage_remapped() */ | ||
253 | for (i = 0; i < num_possible_cpus() - 1; i++) | ||
254 | for (j = i + 1; j < num_possible_cpus(); j++) | ||
255 | if (pcpul_map[i].ptr > pcpul_map[j].ptr) { | ||
256 | struct pcpul_ent tmp = pcpul_map[i]; | ||
257 | pcpul_map[i] = pcpul_map[j]; | ||
258 | pcpul_map[j] = tmp; | ||
259 | } | ||
260 | |||
261 | return ret; | ||
227 | 262 | ||
228 | enomem: | 263 | enomem: |
229 | for_each_possible_cpu(cpu) | 264 | for_each_possible_cpu(cpu) |
230 | if (pcpur_ptrs[cpu]) | 265 | if (pcpul_map[cpu].ptr) |
231 | free_bootmem(__pa(pcpur_ptrs[cpu]), PMD_SIZE); | 266 | free_bootmem(__pa(pcpul_map[cpu].ptr), pcpul_size); |
232 | ret = -ENOMEM; | 267 | free_bootmem(__pa(pcpul_map), map_size); |
233 | out_free_ar: | 268 | return -ENOMEM; |
234 | free_bootmem(__pa(pcpur_ptrs), ptrs_size); | 269 | } |
235 | return ret; | 270 | |
271 | /** | ||
272 | * pcpu_lpage_remapped - determine whether a kaddr is in pcpul recycled area | ||
273 | * @kaddr: the kernel address in question | ||
274 | * | ||
275 | * Determine whether @kaddr falls in the pcpul recycled area. This is | ||
276 | * used by pageattr to detect VM aliases and break up the pcpu PMD | ||
277 | * mapping such that the same physical page is not mapped under | ||
278 | * different attributes. | ||
279 | * | ||
280 | * The recycled area is always at the tail of a partially used PMD | ||
281 | * page. | ||
282 | * | ||
283 | * RETURNS: | ||
284 | * Address of corresponding remapped pcpu address if match is found; | ||
285 | * otherwise, NULL. | ||
286 | */ | ||
287 | void *pcpu_lpage_remapped(void *kaddr) | ||
288 | { | ||
289 | void *pmd_addr = (void *)((unsigned long)kaddr & PMD_MASK); | ||
290 | unsigned long offset = (unsigned long)kaddr & ~PMD_MASK; | ||
291 | int left = 0, right = num_possible_cpus() - 1; | ||
292 | int pos; | ||
293 | |||
294 | /* pcpul in use at all? */ | ||
295 | if (!pcpul_map) | ||
296 | return NULL; | ||
297 | |||
298 | /* okay, perform binary search */ | ||
299 | while (left <= right) { | ||
300 | pos = (left + right) / 2; | ||
301 | |||
302 | if (pcpul_map[pos].ptr < pmd_addr) | ||
303 | left = pos + 1; | ||
304 | else if (pcpul_map[pos].ptr > pmd_addr) | ||
305 | right = pos - 1; | ||
306 | else { | ||
307 | /* it shouldn't be in the area for the first chunk */ | ||
308 | WARN_ON(offset < pcpul_size); | ||
309 | |||
310 | return pcpul_vm.addr + | ||
311 | pcpul_map[pos].cpu * PMD_SIZE + offset; | ||
312 | } | ||
313 | } | ||
314 | |||
315 | return NULL; | ||
236 | } | 316 | } |
237 | #else | 317 | #else |
238 | static ssize_t __init setup_pcpu_remap(size_t static_size) | 318 | static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen) |
239 | { | 319 | { |
240 | return -EINVAL; | 320 | return -EINVAL; |
241 | } | 321 | } |
@@ -249,7 +329,7 @@ static ssize_t __init setup_pcpu_remap(size_t static_size) | |||
249 | * mapping so that it can use PMD mapping without additional TLB | 329 | * mapping so that it can use PMD mapping without additional TLB |
250 | * pressure. | 330 | * pressure. |
251 | */ | 331 | */ |
252 | static ssize_t __init setup_pcpu_embed(size_t static_size) | 332 | static ssize_t __init setup_pcpu_embed(size_t static_size, bool chosen) |
253 | { | 333 | { |
254 | size_t reserve = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE; | 334 | size_t reserve = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE; |
255 | 335 | ||
@@ -258,7 +338,7 @@ static ssize_t __init setup_pcpu_embed(size_t static_size) | |||
258 | * this. Also, embedding allocation doesn't play well with | 338 | * this. Also, embedding allocation doesn't play well with |
259 | * NUMA. | 339 | * NUMA. |
260 | */ | 340 | */ |
261 | if (!cpu_has_pse || pcpu_need_numa()) | 341 | if (!chosen && (!cpu_has_pse || pcpu_need_numa())) |
262 | return -EINVAL; | 342 | return -EINVAL; |
263 | 343 | ||
264 | return pcpu_embed_first_chunk(static_size, PERCPU_FIRST_CHUNK_RESERVE, | 344 | return pcpu_embed_first_chunk(static_size, PERCPU_FIRST_CHUNK_RESERVE, |
@@ -308,8 +388,11 @@ static ssize_t __init setup_pcpu_4k(size_t static_size) | |||
308 | void *ptr; | 388 | void *ptr; |
309 | 389 | ||
310 | ptr = pcpu_alloc_bootmem(cpu, PAGE_SIZE, PAGE_SIZE); | 390 | ptr = pcpu_alloc_bootmem(cpu, PAGE_SIZE, PAGE_SIZE); |
311 | if (!ptr) | 391 | if (!ptr) { |
392 | pr_warning("PERCPU: failed to allocate " | ||
393 | "4k page for cpu%u\n", cpu); | ||
312 | goto enomem; | 394 | goto enomem; |
395 | } | ||
313 | 396 | ||
314 | memcpy(ptr, __per_cpu_load + i * PAGE_SIZE, PAGE_SIZE); | 397 | memcpy(ptr, __per_cpu_load + i * PAGE_SIZE, PAGE_SIZE); |
315 | pcpu4k_pages[j++] = virt_to_page(ptr); | 398 | pcpu4k_pages[j++] = virt_to_page(ptr); |
@@ -333,6 +416,16 @@ out_free_ar: | |||
333 | return ret; | 416 | return ret; |
334 | } | 417 | } |
335 | 418 | ||
419 | /* for explicit first chunk allocator selection */ | ||
420 | static char pcpu_chosen_alloc[16] __initdata; | ||
421 | |||
422 | static int __init percpu_alloc_setup(char *str) | ||
423 | { | ||
424 | strncpy(pcpu_chosen_alloc, str, sizeof(pcpu_chosen_alloc) - 1); | ||
425 | return 0; | ||
426 | } | ||
427 | early_param("percpu_alloc", percpu_alloc_setup); | ||
428 | |||
336 | static inline void setup_percpu_segment(int cpu) | 429 | static inline void setup_percpu_segment(int cpu) |
337 | { | 430 | { |
338 | #ifdef CONFIG_X86_32 | 431 | #ifdef CONFIG_X86_32 |
@@ -346,11 +439,6 @@ static inline void setup_percpu_segment(int cpu) | |||
346 | #endif | 439 | #endif |
347 | } | 440 | } |
348 | 441 | ||
349 | /* | ||
350 | * Great future plan: | ||
351 | * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data. | ||
352 | * Always point %gs to its beginning | ||
353 | */ | ||
354 | void __init setup_per_cpu_areas(void) | 442 | void __init setup_per_cpu_areas(void) |
355 | { | 443 | { |
356 | size_t static_size = __per_cpu_end - __per_cpu_start; | 444 | size_t static_size = __per_cpu_end - __per_cpu_start; |
@@ -367,9 +455,26 @@ void __init setup_per_cpu_areas(void) | |||
367 | * of large page mappings. Please read comments on top of | 455 | * of large page mappings. Please read comments on top of |
368 | * each allocator for details. | 456 | * each allocator for details. |
369 | */ | 457 | */ |
370 | ret = setup_pcpu_remap(static_size); | 458 | ret = -EINVAL; |
371 | if (ret < 0) | 459 | if (strlen(pcpu_chosen_alloc)) { |
372 | ret = setup_pcpu_embed(static_size); | 460 | if (strcmp(pcpu_chosen_alloc, "4k")) { |
461 | if (!strcmp(pcpu_chosen_alloc, "lpage")) | ||
462 | ret = setup_pcpu_lpage(static_size, true); | ||
463 | else if (!strcmp(pcpu_chosen_alloc, "embed")) | ||
464 | ret = setup_pcpu_embed(static_size, true); | ||
465 | else | ||
466 | pr_warning("PERCPU: unknown allocator %s " | ||
467 | "specified\n", pcpu_chosen_alloc); | ||
468 | if (ret < 0) | ||
469 | pr_warning("PERCPU: %s allocator failed (%zd), " | ||
470 | "falling back to 4k\n", | ||
471 | pcpu_chosen_alloc, ret); | ||
472 | } | ||
473 | } else { | ||
474 | ret = setup_pcpu_lpage(static_size, false); | ||
475 | if (ret < 0) | ||
476 | ret = setup_pcpu_embed(static_size, false); | ||
477 | } | ||
373 | if (ret < 0) | 478 | if (ret < 0) |
374 | ret = setup_pcpu_4k(static_size); | 479 | ret = setup_pcpu_4k(static_size); |
375 | if (ret < 0) | 480 | if (ret < 0) |
diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c index 124d40c575df..8ccabb8a2f6a 100644 --- a/arch/x86/kernel/tlb_uv.c +++ b/arch/x86/kernel/tlb_uv.c | |||
@@ -711,7 +711,6 @@ uv_activation_descriptor_init(int node, int pnode) | |||
711 | unsigned long pa; | 711 | unsigned long pa; |
712 | unsigned long m; | 712 | unsigned long m; |
713 | unsigned long n; | 713 | unsigned long n; |
714 | unsigned long mmr_image; | ||
715 | struct bau_desc *adp; | 714 | struct bau_desc *adp; |
716 | struct bau_desc *ad2; | 715 | struct bau_desc *ad2; |
717 | 716 | ||
@@ -727,12 +726,8 @@ uv_activation_descriptor_init(int node, int pnode) | |||
727 | n = pa >> uv_nshift; | 726 | n = pa >> uv_nshift; |
728 | m = pa & uv_mmask; | 727 | m = pa & uv_mmask; |
729 | 728 | ||
730 | mmr_image = uv_read_global_mmr64(pnode, UVH_LB_BAU_SB_DESCRIPTOR_BASE); | 729 | uv_write_global_mmr64(pnode, UVH_LB_BAU_SB_DESCRIPTOR_BASE, |
731 | if (mmr_image) { | 730 | (n << UV_DESC_BASE_PNODE_SHIFT | m)); |
732 | uv_write_global_mmr64(pnode, (unsigned long) | ||
733 | UVH_LB_BAU_SB_DESCRIPTOR_BASE, | ||
734 | (n << UV_DESC_BASE_PNODE_SHIFT | m)); | ||
735 | } | ||
736 | 731 | ||
737 | /* | 732 | /* |
738 | * initializing all 8 (UV_ITEMS_PER_DESCRIPTOR) descriptors for each | 733 | * initializing all 8 (UV_ITEMS_PER_DESCRIPTOR) descriptors for each |
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index a0f48f5671c0..5204332f475d 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c | |||
@@ -346,6 +346,9 @@ io_check_error(unsigned char reason, struct pt_regs *regs) | |||
346 | printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n"); | 346 | printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n"); |
347 | show_registers(regs); | 347 | show_registers(regs); |
348 | 348 | ||
349 | if (panic_on_io_nmi) | ||
350 | panic("NMI IOCK error: Not continuing"); | ||
351 | |||
349 | /* Re-enable the IOCK line, wait for a few seconds */ | 352 | /* Re-enable the IOCK line, wait for a few seconds */ |
350 | reason = (reason & 0xf) | 8; | 353 | reason = (reason & 0xf) | 8; |
351 | outb(reason, 0x61); | 354 | outb(reason, 0x61); |
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c index 5c3d6e81a7dc..7030b5f911bf 100644 --- a/arch/x86/kvm/mmu.c +++ b/arch/x86/kvm/mmu.c | |||
@@ -2157,7 +2157,7 @@ static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu, int level) | |||
2157 | else | 2157 | else |
2158 | /* 32 bits PSE 4MB page */ | 2158 | /* 32 bits PSE 4MB page */ |
2159 | context->rsvd_bits_mask[1][1] = rsvd_bits(13, 21); | 2159 | context->rsvd_bits_mask[1][1] = rsvd_bits(13, 21); |
2160 | context->rsvd_bits_mask[1][0] = ~0ull; | 2160 | context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[1][0]; |
2161 | break; | 2161 | break; |
2162 | case PT32E_ROOT_LEVEL: | 2162 | case PT32E_ROOT_LEVEL: |
2163 | context->rsvd_bits_mask[0][2] = | 2163 | context->rsvd_bits_mask[0][2] = |
@@ -2170,7 +2170,7 @@ static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu, int level) | |||
2170 | context->rsvd_bits_mask[1][1] = exb_bit_rsvd | | 2170 | context->rsvd_bits_mask[1][1] = exb_bit_rsvd | |
2171 | rsvd_bits(maxphyaddr, 62) | | 2171 | rsvd_bits(maxphyaddr, 62) | |
2172 | rsvd_bits(13, 20); /* large page */ | 2172 | rsvd_bits(13, 20); /* large page */ |
2173 | context->rsvd_bits_mask[1][0] = ~0ull; | 2173 | context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[1][0]; |
2174 | break; | 2174 | break; |
2175 | case PT64_ROOT_LEVEL: | 2175 | case PT64_ROOT_LEVEL: |
2176 | context->rsvd_bits_mask[0][3] = exb_bit_rsvd | | 2176 | context->rsvd_bits_mask[0][3] = exb_bit_rsvd | |
@@ -2186,7 +2186,7 @@ static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu, int level) | |||
2186 | context->rsvd_bits_mask[1][1] = exb_bit_rsvd | | 2186 | context->rsvd_bits_mask[1][1] = exb_bit_rsvd | |
2187 | rsvd_bits(maxphyaddr, 51) | | 2187 | rsvd_bits(maxphyaddr, 51) | |
2188 | rsvd_bits(13, 20); /* large page */ | 2188 | rsvd_bits(13, 20); /* large page */ |
2189 | context->rsvd_bits_mask[1][0] = ~0ull; | 2189 | context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[1][0]; |
2190 | break; | 2190 | break; |
2191 | } | 2191 | } |
2192 | } | 2192 | } |
diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h index 258e4591e1ca..67785f635399 100644 --- a/arch/x86/kvm/paging_tmpl.h +++ b/arch/x86/kvm/paging_tmpl.h | |||
@@ -281,7 +281,7 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr, | |||
281 | { | 281 | { |
282 | unsigned access = gw->pt_access; | 282 | unsigned access = gw->pt_access; |
283 | struct kvm_mmu_page *shadow_page; | 283 | struct kvm_mmu_page *shadow_page; |
284 | u64 spte, *sptep; | 284 | u64 spte, *sptep = NULL; |
285 | int direct; | 285 | int direct; |
286 | gfn_t table_gfn; | 286 | gfn_t table_gfn; |
287 | int r; | 287 | int r; |
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index e770bf349ec4..356a0ce85c68 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c | |||
@@ -3012,6 +3012,12 @@ static int handle_vmcall(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | |||
3012 | return 1; | 3012 | return 1; |
3013 | } | 3013 | } |
3014 | 3014 | ||
3015 | static int handle_vmx_insn(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | ||
3016 | { | ||
3017 | kvm_queue_exception(vcpu, UD_VECTOR); | ||
3018 | return 1; | ||
3019 | } | ||
3020 | |||
3015 | static int handle_invlpg(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | 3021 | static int handle_invlpg(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) |
3016 | { | 3022 | { |
3017 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | 3023 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
@@ -3198,6 +3204,15 @@ static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu, | |||
3198 | [EXIT_REASON_HLT] = handle_halt, | 3204 | [EXIT_REASON_HLT] = handle_halt, |
3199 | [EXIT_REASON_INVLPG] = handle_invlpg, | 3205 | [EXIT_REASON_INVLPG] = handle_invlpg, |
3200 | [EXIT_REASON_VMCALL] = handle_vmcall, | 3206 | [EXIT_REASON_VMCALL] = handle_vmcall, |
3207 | [EXIT_REASON_VMCLEAR] = handle_vmx_insn, | ||
3208 | [EXIT_REASON_VMLAUNCH] = handle_vmx_insn, | ||
3209 | [EXIT_REASON_VMPTRLD] = handle_vmx_insn, | ||
3210 | [EXIT_REASON_VMPTRST] = handle_vmx_insn, | ||
3211 | [EXIT_REASON_VMREAD] = handle_vmx_insn, | ||
3212 | [EXIT_REASON_VMRESUME] = handle_vmx_insn, | ||
3213 | [EXIT_REASON_VMWRITE] = handle_vmx_insn, | ||
3214 | [EXIT_REASON_VMOFF] = handle_vmx_insn, | ||
3215 | [EXIT_REASON_VMON] = handle_vmx_insn, | ||
3201 | [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold, | 3216 | [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold, |
3202 | [EXIT_REASON_APIC_ACCESS] = handle_apic_access, | 3217 | [EXIT_REASON_APIC_ACCESS] = handle_apic_access, |
3203 | [EXIT_REASON_WBINVD] = handle_wbinvd, | 3218 | [EXIT_REASON_WBINVD] = handle_wbinvd, |
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 249540f98513..fe5474aec41a 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c | |||
@@ -898,6 +898,7 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) | |||
898 | case MSR_VM_HSAVE_PA: | 898 | case MSR_VM_HSAVE_PA: |
899 | case MSR_P6_EVNTSEL0: | 899 | case MSR_P6_EVNTSEL0: |
900 | case MSR_P6_EVNTSEL1: | 900 | case MSR_P6_EVNTSEL1: |
901 | case MSR_K7_EVNTSEL0: | ||
901 | data = 0; | 902 | data = 0; |
902 | break; | 903 | break; |
903 | case MSR_MTRRcap: | 904 | case MSR_MTRRcap: |
diff --git a/arch/x86/kvm/x86_emulate.c b/arch/x86/kvm/x86_emulate.c index c1b6c232e02b..616de4628d60 100644 --- a/arch/x86/kvm/x86_emulate.c +++ b/arch/x86/kvm/x86_emulate.c | |||
@@ -1361,7 +1361,7 @@ static inline int writeback(struct x86_emulate_ctxt *ctxt, | |||
1361 | return 0; | 1361 | return 0; |
1362 | } | 1362 | } |
1363 | 1363 | ||
1364 | void toggle_interruptibility(struct x86_emulate_ctxt *ctxt, u32 mask) | 1364 | static void toggle_interruptibility(struct x86_emulate_ctxt *ctxt, u32 mask) |
1365 | { | 1365 | { |
1366 | u32 int_shadow = kvm_x86_ops->get_interrupt_shadow(ctxt->vcpu, mask); | 1366 | u32 int_shadow = kvm_x86_ops->get_interrupt_shadow(ctxt->vcpu, mask); |
1367 | /* | 1367 | /* |
diff --git a/arch/x86/lib/delay.c b/arch/x86/lib/delay.c index f4568605d7d5..ff485d361182 100644 --- a/arch/x86/lib/delay.c +++ b/arch/x86/lib/delay.c | |||
@@ -55,8 +55,10 @@ static void delay_tsc(unsigned long loops) | |||
55 | 55 | ||
56 | preempt_disable(); | 56 | preempt_disable(); |
57 | cpu = smp_processor_id(); | 57 | cpu = smp_processor_id(); |
58 | rdtsc_barrier(); | ||
58 | rdtscl(bclock); | 59 | rdtscl(bclock); |
59 | for (;;) { | 60 | for (;;) { |
61 | rdtsc_barrier(); | ||
60 | rdtscl(now); | 62 | rdtscl(now); |
61 | if ((now - bclock) >= loops) | 63 | if ((now - bclock) >= loops) |
62 | break; | 64 | break; |
@@ -78,6 +80,7 @@ static void delay_tsc(unsigned long loops) | |||
78 | if (unlikely(cpu != smp_processor_id())) { | 80 | if (unlikely(cpu != smp_processor_id())) { |
79 | loops -= (now - bclock); | 81 | loops -= (now - bclock); |
80 | cpu = smp_processor_id(); | 82 | cpu = smp_processor_id(); |
83 | rdtsc_barrier(); | ||
81 | rdtscl(bclock); | 84 | rdtscl(bclock); |
82 | } | 85 | } |
83 | } | 86 | } |
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c index f53b57e4086f..47ce9a2ce5e7 100644 --- a/arch/x86/mm/init.c +++ b/arch/x86/mm/init.c | |||
@@ -177,20 +177,6 @@ static int __meminit save_mr(struct map_range *mr, int nr_range, | |||
177 | return nr_range; | 177 | return nr_range; |
178 | } | 178 | } |
179 | 179 | ||
180 | #ifdef CONFIG_X86_64 | ||
181 | static void __init init_gbpages(void) | ||
182 | { | ||
183 | if (direct_gbpages && cpu_has_gbpages) | ||
184 | printk(KERN_INFO "Using GB pages for direct mapping\n"); | ||
185 | else | ||
186 | direct_gbpages = 0; | ||
187 | } | ||
188 | #else | ||
189 | static inline void init_gbpages(void) | ||
190 | { | ||
191 | } | ||
192 | #endif | ||
193 | |||
194 | /* | 180 | /* |
195 | * Setup the direct mapping of the physical memory at PAGE_OFFSET. | 181 | * Setup the direct mapping of the physical memory at PAGE_OFFSET. |
196 | * This runs before bootmem is initialized and gets pages directly from | 182 | * This runs before bootmem is initialized and gets pages directly from |
@@ -210,9 +196,6 @@ unsigned long __init_refok init_memory_mapping(unsigned long start, | |||
210 | 196 | ||
211 | printk(KERN_INFO "init_memory_mapping: %016lx-%016lx\n", start, end); | 197 | printk(KERN_INFO "init_memory_mapping: %016lx-%016lx\n", start, end); |
212 | 198 | ||
213 | if (!after_bootmem) | ||
214 | init_gbpages(); | ||
215 | |||
216 | #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_KMEMCHECK) | 199 | #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_KMEMCHECK) |
217 | /* | 200 | /* |
218 | * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages. | 201 | * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages. |
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c index c4378f4fd4a5..b177652251a4 100644 --- a/arch/x86/mm/init_64.c +++ b/arch/x86/mm/init_64.c | |||
@@ -598,6 +598,8 @@ void __init paging_init(void) | |||
598 | 598 | ||
599 | sparse_memory_present_with_active_regions(MAX_NUMNODES); | 599 | sparse_memory_present_with_active_regions(MAX_NUMNODES); |
600 | sparse_init(); | 600 | sparse_init(); |
601 | /* clear the default setting with node 0 */ | ||
602 | nodes_clear(node_states[N_NORMAL_MEMORY]); | ||
601 | free_area_init_nodes(max_zone_pfns); | 603 | free_area_init_nodes(max_zone_pfns); |
602 | } | 604 | } |
603 | 605 | ||
diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c index 3cfe9ced8a4c..1b734d7a8966 100644 --- a/arch/x86/mm/pageattr.c +++ b/arch/x86/mm/pageattr.c | |||
@@ -11,6 +11,7 @@ | |||
11 | #include <linux/interrupt.h> | 11 | #include <linux/interrupt.h> |
12 | #include <linux/seq_file.h> | 12 | #include <linux/seq_file.h> |
13 | #include <linux/debugfs.h> | 13 | #include <linux/debugfs.h> |
14 | #include <linux/pfn.h> | ||
14 | 15 | ||
15 | #include <asm/e820.h> | 16 | #include <asm/e820.h> |
16 | #include <asm/processor.h> | 17 | #include <asm/processor.h> |
@@ -681,8 +682,9 @@ static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias); | |||
681 | static int cpa_process_alias(struct cpa_data *cpa) | 682 | static int cpa_process_alias(struct cpa_data *cpa) |
682 | { | 683 | { |
683 | struct cpa_data alias_cpa; | 684 | struct cpa_data alias_cpa; |
684 | int ret = 0; | 685 | unsigned long laddr = (unsigned long)__va(cpa->pfn << PAGE_SHIFT); |
685 | unsigned long temp_cpa_vaddr, vaddr; | 686 | unsigned long vaddr, remapped; |
687 | int ret; | ||
686 | 688 | ||
687 | if (cpa->pfn >= max_pfn_mapped) | 689 | if (cpa->pfn >= max_pfn_mapped) |
688 | return 0; | 690 | return 0; |
@@ -706,42 +708,55 @@ static int cpa_process_alias(struct cpa_data *cpa) | |||
706 | PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT)))) { | 708 | PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT)))) { |
707 | 709 | ||
708 | alias_cpa = *cpa; | 710 | alias_cpa = *cpa; |
709 | temp_cpa_vaddr = (unsigned long) __va(cpa->pfn << PAGE_SHIFT); | 711 | alias_cpa.vaddr = &laddr; |
710 | alias_cpa.vaddr = &temp_cpa_vaddr; | ||
711 | alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY); | 712 | alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY); |
712 | 713 | ||
713 | |||
714 | ret = __change_page_attr_set_clr(&alias_cpa, 0); | 714 | ret = __change_page_attr_set_clr(&alias_cpa, 0); |
715 | if (ret) | ||
716 | return ret; | ||
715 | } | 717 | } |
716 | 718 | ||
717 | #ifdef CONFIG_X86_64 | 719 | #ifdef CONFIG_X86_64 |
718 | if (ret) | ||
719 | return ret; | ||
720 | /* | 720 | /* |
721 | * No need to redo, when the primary call touched the high | 721 | * If the primary call didn't touch the high mapping already |
722 | * mapping already: | 722 | * and the physical address is inside the kernel map, we need |
723 | */ | ||
724 | if (within(vaddr, (unsigned long) _text, _brk_end)) | ||
725 | return 0; | ||
726 | |||
727 | /* | ||
728 | * If the physical address is inside the kernel map, we need | ||
729 | * to touch the high mapped kernel as well: | 723 | * to touch the high mapped kernel as well: |
730 | */ | 724 | */ |
731 | if (!within(cpa->pfn, highmap_start_pfn(), highmap_end_pfn())) | 725 | if (!within(vaddr, (unsigned long)_text, _brk_end) && |
732 | return 0; | 726 | within(cpa->pfn, highmap_start_pfn(), highmap_end_pfn())) { |
727 | unsigned long temp_cpa_vaddr = (cpa->pfn << PAGE_SHIFT) + | ||
728 | __START_KERNEL_map - phys_base; | ||
729 | alias_cpa = *cpa; | ||
730 | alias_cpa.vaddr = &temp_cpa_vaddr; | ||
731 | alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY); | ||
733 | 732 | ||
734 | alias_cpa = *cpa; | 733 | /* |
735 | temp_cpa_vaddr = (cpa->pfn << PAGE_SHIFT) + __START_KERNEL_map - phys_base; | 734 | * The high mapping range is imprecise, so ignore the |
736 | alias_cpa.vaddr = &temp_cpa_vaddr; | 735 | * return value. |
737 | alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY); | 736 | */ |
737 | __change_page_attr_set_clr(&alias_cpa, 0); | ||
738 | } | ||
739 | #endif | ||
738 | 740 | ||
739 | /* | 741 | /* |
740 | * The high mapping range is imprecise, so ignore the return value. | 742 | * If the PMD page was partially used for per-cpu remapping, |
743 | * the recycled area needs to be split and modified. Because | ||
744 | * the area is always proper subset of a PMD page | ||
745 | * cpa->numpages is guaranteed to be 1 for these areas, so | ||
746 | * there's no need to loop over and check for further remaps. | ||
741 | */ | 747 | */ |
742 | __change_page_attr_set_clr(&alias_cpa, 0); | 748 | remapped = (unsigned long)pcpu_lpage_remapped((void *)laddr); |
743 | #endif | 749 | if (remapped) { |
744 | return ret; | 750 | WARN_ON(cpa->numpages > 1); |
751 | alias_cpa = *cpa; | ||
752 | alias_cpa.vaddr = &remapped; | ||
753 | alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY); | ||
754 | ret = __change_page_attr_set_clr(&alias_cpa, 0); | ||
755 | if (ret) | ||
756 | return ret; | ||
757 | } | ||
758 | |||
759 | return 0; | ||
745 | } | 760 | } |
746 | 761 | ||
747 | static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias) | 762 | static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias) |
diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c index d277ef1eea51..b3d20b9cac63 100644 --- a/arch/x86/power/cpu.c +++ b/arch/x86/power/cpu.c | |||
@@ -244,7 +244,7 @@ static void __restore_processor_state(struct saved_context *ctxt) | |||
244 | do_fpu_end(); | 244 | do_fpu_end(); |
245 | mtrr_ap_init(); | 245 | mtrr_ap_init(); |
246 | 246 | ||
247 | #ifdef CONFIG_X86_32 | 247 | #ifdef CONFIG_X86_OLD_MCE |
248 | mcheck_init(&boot_cpu_data); | 248 | mcheck_init(&boot_cpu_data); |
249 | #endif | 249 | #endif |
250 | } | 250 | } |