aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel
diff options
context:
space:
mode:
authorIngo Molnar <mingo@elte.hu>2008-09-24 03:56:20 -0400
committerIngo Molnar <mingo@elte.hu>2008-09-24 03:56:20 -0400
commitebdd90a8cb2e3963f55499850f02ce6003558b55 (patch)
treed153f917ed41d257ddafa22f9cc2201bfddf8f9c /arch/x86/kernel
parent3c9339049df5cc3a468c11de6c4101a1ea8c3d83 (diff)
parent72d31053f62c4bc464c2783974926969614a8649 (diff)
Merge commit 'v2.6.27-rc7' into x86/pebs
Diffstat (limited to 'arch/x86/kernel')
-rw-r--r--arch/x86/kernel/acpi/boot.c16
-rw-r--r--arch/x86/kernel/acpi/cstate.c3
-rw-r--r--arch/x86/kernel/acpi/sleep.c4
-rw-r--r--arch/x86/kernel/alternative.c36
-rw-r--r--arch/x86/kernel/amd_iommu.c38
-rw-r--r--arch/x86/kernel/amd_iommu_init.c28
-rw-r--r--arch/x86/kernel/apic_32.c22
-rw-r--r--arch/x86/kernel/apic_64.c7
-rw-r--r--arch/x86/kernel/cpu/addon_cpuid_features.c17
-rw-r--r--arch/x86/kernel/cpu/amd.c9
-rw-r--r--arch/x86/kernel/cpu/bugs.c6
-rw-r--r--arch/x86/kernel/cpu/centaur.c11
-rw-r--r--arch/x86/kernel/cpu/common.c18
-rw-r--r--arch/x86/kernel/cpu/common_64.c74
-rw-r--r--arch/x86/kernel/cpu/cpufreq/Kconfig4
-rw-r--r--arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c10
-rw-r--r--arch/x86/kernel/cpu/cpufreq/elanfreq.c2
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k8.c18
-rw-r--r--arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c12
-rw-r--r--arch/x86/kernel/cpu/cpufreq/speedstep-ich.c3
-rw-r--r--arch/x86/kernel/cpu/cyrix.c50
-rw-r--r--arch/x86/kernel/cpu/feature_names.c3
-rw-r--r--arch/x86/kernel/cpu/intel_cacheinfo.c3
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_64.c5
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_amd_64.c18
-rw-r--r--arch/x86/kernel/cpu/mtrr/generic.c20
-rw-r--r--arch/x86/kernel/cpu/mtrr/main.c5
-rw-r--r--arch/x86/kernel/cpu/perfctr-watchdog.c8
-rw-r--r--arch/x86/kernel/cpuid.c15
-rw-r--r--arch/x86/kernel/e820.c2
-rw-r--r--arch/x86/kernel/efi_32.c4
-rw-r--r--arch/x86/kernel/genapic_64.c1
-rw-r--r--arch/x86/kernel/genx2apic_uv_x.c10
-rw-r--r--arch/x86/kernel/head64.c1
-rw-r--r--arch/x86/kernel/head_32.S8
-rw-r--r--arch/x86/kernel/hpet.c43
-rw-r--r--arch/x86/kernel/io_apic_32.c6
-rw-r--r--arch/x86/kernel/io_apic_64.c25
-rw-r--r--arch/x86/kernel/io_delay.c8
-rw-r--r--arch/x86/kernel/kdebugfs.c1
-rw-r--r--arch/x86/kernel/kprobes.c6
-rw-r--r--arch/x86/kernel/ldt.c6
-rw-r--r--arch/x86/kernel/machine_kexec_32.c57
-rw-r--r--arch/x86/kernel/machine_kexec_64.c2
-rw-r--r--arch/x86/kernel/mfgpt_32.c52
-rw-r--r--arch/x86/kernel/microcode.c17
-rw-r--r--arch/x86/kernel/mmconf-fam10h_64.c2
-rw-r--r--arch/x86/kernel/mpparse.c17
-rw-r--r--arch/x86/kernel/msr.c40
-rw-r--r--arch/x86/kernel/nmi.c28
-rw-r--r--arch/x86/kernel/numaq_32.c2
-rw-r--r--arch/x86/kernel/paravirt.c2
-rw-r--r--arch/x86/kernel/pci-calgary_64.c172
-rw-r--r--arch/x86/kernel/pci-dma.c157
-rw-r--r--arch/x86/kernel/pci-gart_64.c14
-rw-r--r--arch/x86/kernel/pci-nommu.c14
-rw-r--r--arch/x86/kernel/pci-swiotlb_64.c2
-rw-r--r--arch/x86/kernel/process_32.c5
-rw-r--r--arch/x86/kernel/process_64.c5
-rw-r--r--arch/x86/kernel/reboot.c11
-rw-r--r--arch/x86/kernel/relocate_kernel_32.S178
-rw-r--r--arch/x86/kernel/setup.c42
-rw-r--r--arch/x86/kernel/setup_percpu.c21
-rw-r--r--arch/x86/kernel/signal_64.c11
-rw-r--r--arch/x86/kernel/smpboot.c68
-rw-r--r--arch/x86/kernel/smpcommon.c17
-rw-r--r--arch/x86/kernel/tlb_uv.c3
-rw-r--r--arch/x86/kernel/traps_64.c9
-rw-r--r--arch/x86/kernel/tsc.c242
-rw-r--r--arch/x86/kernel/tsc_sync.c6
-rw-r--r--arch/x86/kernel/visws_quirks.c6
-rw-r--r--arch/x86/kernel/vmi_32.c3
-rw-r--r--arch/x86/kernel/vmlinux_32.lds.S8
73 files changed, 1131 insertions, 668 deletions
diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index fa88a1d71290..bfd10fd211cd 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -97,6 +97,8 @@ static u64 acpi_lapic_addr __initdata = APIC_DEFAULT_PHYS_BASE;
97#warning ACPI uses CMPXCHG, i486 and later hardware 97#warning ACPI uses CMPXCHG, i486 and later hardware
98#endif 98#endif
99 99
100static int acpi_mcfg_64bit_base_addr __initdata = FALSE;
101
100/* -------------------------------------------------------------------------- 102/* --------------------------------------------------------------------------
101 Boot-time Configuration 103 Boot-time Configuration
102 -------------------------------------------------------------------------- */ 104 -------------------------------------------------------------------------- */
@@ -158,6 +160,14 @@ char *__init __acpi_map_table(unsigned long phys, unsigned long size)
158struct acpi_mcfg_allocation *pci_mmcfg_config; 160struct acpi_mcfg_allocation *pci_mmcfg_config;
159int pci_mmcfg_config_num; 161int pci_mmcfg_config_num;
160 162
163static int __init acpi_mcfg_oem_check(struct acpi_table_mcfg *mcfg)
164{
165 if (!strcmp(mcfg->header.oem_id, "SGI"))
166 acpi_mcfg_64bit_base_addr = TRUE;
167
168 return 0;
169}
170
161int __init acpi_parse_mcfg(struct acpi_table_header *header) 171int __init acpi_parse_mcfg(struct acpi_table_header *header)
162{ 172{
163 struct acpi_table_mcfg *mcfg; 173 struct acpi_table_mcfg *mcfg;
@@ -190,8 +200,12 @@ int __init acpi_parse_mcfg(struct acpi_table_header *header)
190 } 200 }
191 201
192 memcpy(pci_mmcfg_config, &mcfg[1], config_size); 202 memcpy(pci_mmcfg_config, &mcfg[1], config_size);
203
204 acpi_mcfg_oem_check(mcfg);
205
193 for (i = 0; i < pci_mmcfg_config_num; ++i) { 206 for (i = 0; i < pci_mmcfg_config_num; ++i) {
194 if (pci_mmcfg_config[i].address > 0xFFFFFFFF) { 207 if ((pci_mmcfg_config[i].address > 0xFFFFFFFF) &&
208 !acpi_mcfg_64bit_base_addr) {
195 printk(KERN_ERR PREFIX 209 printk(KERN_ERR PREFIX
196 "MMCONFIG not in low 4GB of memory\n"); 210 "MMCONFIG not in low 4GB of memory\n");
197 kfree(pci_mmcfg_config); 211 kfree(pci_mmcfg_config);
diff --git a/arch/x86/kernel/acpi/cstate.c b/arch/x86/kernel/acpi/cstate.c
index 9220cf46aa10..c2502eb9aa83 100644
--- a/arch/x86/kernel/acpi/cstate.c
+++ b/arch/x86/kernel/acpi/cstate.c
@@ -73,7 +73,6 @@ int acpi_processor_ffh_cstate_probe(unsigned int cpu,
73 struct cpuinfo_x86 *c = &cpu_data(cpu); 73 struct cpuinfo_x86 *c = &cpu_data(cpu);
74 74
75 cpumask_t saved_mask; 75 cpumask_t saved_mask;
76 cpumask_of_cpu_ptr(new_mask, cpu);
77 int retval; 76 int retval;
78 unsigned int eax, ebx, ecx, edx; 77 unsigned int eax, ebx, ecx, edx;
79 unsigned int edx_part; 78 unsigned int edx_part;
@@ -92,7 +91,7 @@ int acpi_processor_ffh_cstate_probe(unsigned int cpu,
92 91
93 /* Make sure we are running on right CPU */ 92 /* Make sure we are running on right CPU */
94 saved_mask = current->cpus_allowed; 93 saved_mask = current->cpus_allowed;
95 retval = set_cpus_allowed_ptr(current, new_mask); 94 retval = set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
96 if (retval) 95 if (retval)
97 return -1; 96 return -1;
98 97
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index fa2161d5003b..426e5d91b63a 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -20,7 +20,7 @@ unsigned long acpi_realmode_flags;
20/* address in low memory of the wakeup routine. */ 20/* address in low memory of the wakeup routine. */
21static unsigned long acpi_realmode; 21static unsigned long acpi_realmode;
22 22
23#ifdef CONFIG_64BIT 23#if defined(CONFIG_SMP) && defined(CONFIG_64BIT)
24static char temp_stack[10240]; 24static char temp_stack[10240];
25#endif 25#endif
26 26
@@ -86,7 +86,7 @@ int acpi_save_state_mem(void)
86#endif /* !CONFIG_64BIT */ 86#endif /* !CONFIG_64BIT */
87 87
88 header->pmode_cr0 = read_cr0(); 88 header->pmode_cr0 = read_cr0();
89 header->pmode_cr4 = read_cr4(); 89 header->pmode_cr4 = read_cr4_safe();
90 header->realmode_flags = acpi_realmode_flags; 90 header->realmode_flags = acpi_realmode_flags;
91 header->real_magic = 0x12345678; 91 header->real_magic = 0x12345678;
92 92
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 2763cb37b553..65a0c1b48696 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -145,35 +145,25 @@ static const unsigned char *const p6_nops[ASM_NOP_MAX+1] = {
145extern char __vsyscall_0; 145extern char __vsyscall_0;
146const unsigned char *const *find_nop_table(void) 146const unsigned char *const *find_nop_table(void)
147{ 147{
148 return boot_cpu_data.x86_vendor != X86_VENDOR_INTEL || 148 if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
149 boot_cpu_data.x86 < 6 ? k8_nops : p6_nops; 149 boot_cpu_has(X86_FEATURE_NOPL))
150 return p6_nops;
151 else
152 return k8_nops;
150} 153}
151 154
152#else /* CONFIG_X86_64 */ 155#else /* CONFIG_X86_64 */
153 156
154static const struct nop {
155 int cpuid;
156 const unsigned char *const *noptable;
157} noptypes[] = {
158 { X86_FEATURE_K8, k8_nops },
159 { X86_FEATURE_K7, k7_nops },
160 { X86_FEATURE_P4, p6_nops },
161 { X86_FEATURE_P3, p6_nops },
162 { -1, NULL }
163};
164
165const unsigned char *const *find_nop_table(void) 157const unsigned char *const *find_nop_table(void)
166{ 158{
167 const unsigned char *const *noptable = intel_nops; 159 if (boot_cpu_has(X86_FEATURE_K8))
168 int i; 160 return k8_nops;
169 161 else if (boot_cpu_has(X86_FEATURE_K7))
170 for (i = 0; noptypes[i].cpuid >= 0; i++) { 162 return k7_nops;
171 if (boot_cpu_has(noptypes[i].cpuid)) { 163 else if (boot_cpu_has(X86_FEATURE_NOPL))
172 noptable = noptypes[i].noptable; 164 return p6_nops;
173 break; 165 else
174 } 166 return intel_nops;
175 }
176 return noptable;
177} 167}
178 168
179#endif /* CONFIG_X86_64 */ 169#endif /* CONFIG_X86_64 */
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index c25210e6ac88..69b4d060b21c 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -29,9 +29,6 @@
29 29
30#define CMD_SET_TYPE(cmd, t) ((cmd)->data[1] |= ((t) << 28)) 30#define CMD_SET_TYPE(cmd, t) ((cmd)->data[1] |= ((t) << 28))
31 31
32#define to_pages(addr, size) \
33 (round_up(((addr) & ~PAGE_MASK) + (size), PAGE_SIZE) >> PAGE_SHIFT)
34
35#define EXIT_LOOP_COUNT 10000000 32#define EXIT_LOOP_COUNT 10000000
36 33
37static DEFINE_RWLOCK(amd_iommu_devtable_lock); 34static DEFINE_RWLOCK(amd_iommu_devtable_lock);
@@ -68,7 +65,7 @@ static int __iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd)
68 u8 *target; 65 u8 *target;
69 66
70 tail = readl(iommu->mmio_base + MMIO_CMD_TAIL_OFFSET); 67 tail = readl(iommu->mmio_base + MMIO_CMD_TAIL_OFFSET);
71 target = (iommu->cmd_buf + tail); 68 target = iommu->cmd_buf + tail;
72 memcpy_toio(target, cmd, sizeof(*cmd)); 69 memcpy_toio(target, cmd, sizeof(*cmd));
73 tail = (tail + sizeof(*cmd)) % iommu->cmd_buf_size; 70 tail = (tail + sizeof(*cmd)) % iommu->cmd_buf_size;
74 head = readl(iommu->mmio_base + MMIO_CMD_HEAD_OFFSET); 71 head = readl(iommu->mmio_base + MMIO_CMD_HEAD_OFFSET);
@@ -104,16 +101,13 @@ static int iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd)
104 */ 101 */
105static int iommu_completion_wait(struct amd_iommu *iommu) 102static int iommu_completion_wait(struct amd_iommu *iommu)
106{ 103{
107 int ret; 104 int ret, ready = 0;
105 unsigned status = 0;
108 struct iommu_cmd cmd; 106 struct iommu_cmd cmd;
109 volatile u64 ready = 0;
110 unsigned long ready_phys = virt_to_phys(&ready);
111 unsigned long i = 0; 107 unsigned long i = 0;
112 108
113 memset(&cmd, 0, sizeof(cmd)); 109 memset(&cmd, 0, sizeof(cmd));
114 cmd.data[0] = LOW_U32(ready_phys) | CMD_COMPL_WAIT_STORE_MASK; 110 cmd.data[0] = CMD_COMPL_WAIT_INT_MASK;
115 cmd.data[1] = upper_32_bits(ready_phys);
116 cmd.data[2] = 1; /* value written to 'ready' */
117 CMD_SET_TYPE(&cmd, CMD_COMPL_WAIT); 111 CMD_SET_TYPE(&cmd, CMD_COMPL_WAIT);
118 112
119 iommu->need_sync = 0; 113 iommu->need_sync = 0;
@@ -125,9 +119,15 @@ static int iommu_completion_wait(struct amd_iommu *iommu)
125 119
126 while (!ready && (i < EXIT_LOOP_COUNT)) { 120 while (!ready && (i < EXIT_LOOP_COUNT)) {
127 ++i; 121 ++i;
128 cpu_relax(); 122 /* wait for the bit to become one */
123 status = readl(iommu->mmio_base + MMIO_STATUS_OFFSET);
124 ready = status & MMIO_STATUS_COM_WAIT_INT_MASK;
129 } 125 }
130 126
127 /* set bit back to zero */
128 status &= ~MMIO_STATUS_COM_WAIT_INT_MASK;
129 writel(status, iommu->mmio_base + MMIO_STATUS_OFFSET);
130
131 if (unlikely((i == EXIT_LOOP_COUNT) && printk_ratelimit())) 131 if (unlikely((i == EXIT_LOOP_COUNT) && printk_ratelimit()))
132 printk(KERN_WARNING "AMD IOMMU: Completion wait loop failed\n"); 132 printk(KERN_WARNING "AMD IOMMU: Completion wait loop failed\n");
133 133
@@ -164,7 +164,7 @@ static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu,
164 address &= PAGE_MASK; 164 address &= PAGE_MASK;
165 CMD_SET_TYPE(&cmd, CMD_INV_IOMMU_PAGES); 165 CMD_SET_TYPE(&cmd, CMD_INV_IOMMU_PAGES);
166 cmd.data[1] |= domid; 166 cmd.data[1] |= domid;
167 cmd.data[2] = LOW_U32(address); 167 cmd.data[2] = lower_32_bits(address);
168 cmd.data[3] = upper_32_bits(address); 168 cmd.data[3] = upper_32_bits(address);
169 if (s) /* size bit - we flush more than one 4kb page */ 169 if (s) /* size bit - we flush more than one 4kb page */
170 cmd.data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK; 170 cmd.data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
@@ -185,7 +185,7 @@ static int iommu_flush_pages(struct amd_iommu *iommu, u16 domid,
185 u64 address, size_t size) 185 u64 address, size_t size)
186{ 186{
187 int s = 0; 187 int s = 0;
188 unsigned pages = to_pages(address, size); 188 unsigned pages = iommu_num_pages(address, size);
189 189
190 address &= PAGE_MASK; 190 address &= PAGE_MASK;
191 191
@@ -557,8 +557,8 @@ static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu,
557 if (iommu->exclusion_start && 557 if (iommu->exclusion_start &&
558 iommu->exclusion_start < dma_dom->aperture_size) { 558 iommu->exclusion_start < dma_dom->aperture_size) {
559 unsigned long startpage = iommu->exclusion_start >> PAGE_SHIFT; 559 unsigned long startpage = iommu->exclusion_start >> PAGE_SHIFT;
560 int pages = to_pages(iommu->exclusion_start, 560 int pages = iommu_num_pages(iommu->exclusion_start,
561 iommu->exclusion_length); 561 iommu->exclusion_length);
562 dma_ops_reserve_addresses(dma_dom, startpage, pages); 562 dma_ops_reserve_addresses(dma_dom, startpage, pages);
563 } 563 }
564 564
@@ -667,7 +667,7 @@ static int get_device_resources(struct device *dev,
667 _bdf = calc_devid(pcidev->bus->number, pcidev->devfn); 667 _bdf = calc_devid(pcidev->bus->number, pcidev->devfn);
668 668
669 /* device not translated by any IOMMU in the system? */ 669 /* device not translated by any IOMMU in the system? */
670 if (_bdf >= amd_iommu_last_bdf) { 670 if (_bdf > amd_iommu_last_bdf) {
671 *iommu = NULL; 671 *iommu = NULL;
672 *domain = NULL; 672 *domain = NULL;
673 *bdf = 0xffff; 673 *bdf = 0xffff;
@@ -767,7 +767,7 @@ static dma_addr_t __map_single(struct device *dev,
767 unsigned int pages; 767 unsigned int pages;
768 int i; 768 int i;
769 769
770 pages = to_pages(paddr, size); 770 pages = iommu_num_pages(paddr, size);
771 paddr &= PAGE_MASK; 771 paddr &= PAGE_MASK;
772 772
773 address = dma_ops_alloc_addresses(dev, dma_dom, pages); 773 address = dma_ops_alloc_addresses(dev, dma_dom, pages);
@@ -802,7 +802,7 @@ static void __unmap_single(struct amd_iommu *iommu,
802 if ((dma_addr == 0) || (dma_addr + size > dma_dom->aperture_size)) 802 if ((dma_addr == 0) || (dma_addr + size > dma_dom->aperture_size))
803 return; 803 return;
804 804
805 pages = to_pages(dma_addr, size); 805 pages = iommu_num_pages(dma_addr, size);
806 dma_addr &= PAGE_MASK; 806 dma_addr &= PAGE_MASK;
807 start = dma_addr; 807 start = dma_addr;
808 808
@@ -1085,7 +1085,7 @@ void prealloc_protection_domains(void)
1085 1085
1086 while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) { 1086 while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
1087 devid = (dev->bus->number << 8) | dev->devfn; 1087 devid = (dev->bus->number << 8) | dev->devfn;
1088 if (devid >= amd_iommu_last_bdf) 1088 if (devid > amd_iommu_last_bdf)
1089 continue; 1089 continue;
1090 devid = amd_iommu_alias_table[devid]; 1090 devid = amd_iommu_alias_table[devid];
1091 if (domain_for_device(devid)) 1091 if (domain_for_device(devid))
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index c9d8ff2eb130..a69cc0f52042 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -732,7 +732,7 @@ static int __init init_exclusion_range(struct ivmd_header *m)
732 set_device_exclusion_range(m->devid, m); 732 set_device_exclusion_range(m->devid, m);
733 break; 733 break;
734 case ACPI_IVMD_TYPE_ALL: 734 case ACPI_IVMD_TYPE_ALL:
735 for (i = 0; i < amd_iommu_last_bdf; ++i) 735 for (i = 0; i <= amd_iommu_last_bdf; ++i)
736 set_device_exclusion_range(i, m); 736 set_device_exclusion_range(i, m);
737 break; 737 break;
738 case ACPI_IVMD_TYPE_RANGE: 738 case ACPI_IVMD_TYPE_RANGE:
@@ -801,6 +801,21 @@ static int __init init_memory_definitions(struct acpi_table_header *table)
801} 801}
802 802
803/* 803/*
804 * Init the device table to not allow DMA access for devices and
805 * suppress all page faults
806 */
807static void init_device_table(void)
808{
809 u16 devid;
810
811 for (devid = 0; devid <= amd_iommu_last_bdf; ++devid) {
812 set_dev_entry_bit(devid, DEV_ENTRY_VALID);
813 set_dev_entry_bit(devid, DEV_ENTRY_TRANSLATION);
814 set_dev_entry_bit(devid, DEV_ENTRY_NO_PAGE_FAULT);
815 }
816}
817
818/*
804 * This function finally enables all IOMMUs found in the system after 819 * This function finally enables all IOMMUs found in the system after
805 * they have been initialized 820 * they have been initialized
806 */ 821 */
@@ -931,10 +946,13 @@ int __init amd_iommu_init(void)
931 if (amd_iommu_pd_alloc_bitmap == NULL) 946 if (amd_iommu_pd_alloc_bitmap == NULL)
932 goto free; 947 goto free;
933 948
949 /* init the device table */
950 init_device_table();
951
934 /* 952 /*
935 * let all alias entries point to itself 953 * let all alias entries point to itself
936 */ 954 */
937 for (i = 0; i < amd_iommu_last_bdf; ++i) 955 for (i = 0; i <= amd_iommu_last_bdf; ++i)
938 amd_iommu_alias_table[i] = i; 956 amd_iommu_alias_table[i] = i;
939 957
940 /* 958 /*
@@ -954,15 +972,15 @@ int __init amd_iommu_init(void)
954 if (acpi_table_parse("IVRS", init_memory_definitions) != 0) 972 if (acpi_table_parse("IVRS", init_memory_definitions) != 0)
955 goto free; 973 goto free;
956 974
957 ret = amd_iommu_init_dma_ops(); 975 ret = sysdev_class_register(&amd_iommu_sysdev_class);
958 if (ret) 976 if (ret)
959 goto free; 977 goto free;
960 978
961 ret = sysdev_class_register(&amd_iommu_sysdev_class); 979 ret = sysdev_register(&device_amd_iommu);
962 if (ret) 980 if (ret)
963 goto free; 981 goto free;
964 982
965 ret = sysdev_register(&device_amd_iommu); 983 ret = amd_iommu_init_dma_ops();
966 if (ret) 984 if (ret)
967 goto free; 985 goto free;
968 986
diff --git a/arch/x86/kernel/apic_32.c b/arch/x86/kernel/apic_32.c
index d6c898358371..f88bd0d982b0 100644
--- a/arch/x86/kernel/apic_32.c
+++ b/arch/x86/kernel/apic_32.c
@@ -1454,8 +1454,6 @@ void disconnect_bsp_APIC(int virt_wire_setup)
1454 } 1454 }
1455} 1455}
1456 1456
1457unsigned int __cpuinitdata maxcpus = NR_CPUS;
1458
1459void __cpuinit generic_processor_info(int apicid, int version) 1457void __cpuinit generic_processor_info(int apicid, int version)
1460{ 1458{
1461 int cpu; 1459 int cpu;
@@ -1482,12 +1480,6 @@ void __cpuinit generic_processor_info(int apicid, int version)
1482 return; 1480 return;
1483 } 1481 }
1484 1482
1485 if (num_processors >= maxcpus) {
1486 printk(KERN_WARNING "WARNING: maxcpus limit of %i reached."
1487 " Processor ignored.\n", maxcpus);
1488 return;
1489 }
1490
1491 num_processors++; 1483 num_processors++;
1492 cpus_complement(tmp_map, cpu_present_map); 1484 cpus_complement(tmp_map, cpu_present_map);
1493 cpu = first_cpu(tmp_map); 1485 cpu = first_cpu(tmp_map);
@@ -1720,15 +1712,19 @@ static int __init parse_lapic_timer_c2_ok(char *arg)
1720} 1712}
1721early_param("lapic_timer_c2_ok", parse_lapic_timer_c2_ok); 1713early_param("lapic_timer_c2_ok", parse_lapic_timer_c2_ok);
1722 1714
1723static int __init apic_set_verbosity(char *str) 1715static int __init apic_set_verbosity(char *arg)
1724{ 1716{
1725 if (strcmp("debug", str) == 0) 1717 if (!arg)
1718 return -EINVAL;
1719
1720 if (strcmp(arg, "debug") == 0)
1726 apic_verbosity = APIC_DEBUG; 1721 apic_verbosity = APIC_DEBUG;
1727 else if (strcmp("verbose", str) == 0) 1722 else if (strcmp(arg, "verbose") == 0)
1728 apic_verbosity = APIC_VERBOSE; 1723 apic_verbosity = APIC_VERBOSE;
1729 return 1; 1724
1725 return 0;
1730} 1726}
1731__setup("apic=", apic_set_verbosity); 1727early_param("apic", apic_set_verbosity);
1732 1728
1733static int __init lapic_insert_resource(void) 1729static int __init lapic_insert_resource(void)
1734{ 1730{
diff --git a/arch/x86/kernel/apic_64.c b/arch/x86/kernel/apic_64.c
index 7f1f030da7ee..446c062e831c 100644
--- a/arch/x86/kernel/apic_64.c
+++ b/arch/x86/kernel/apic_64.c
@@ -90,7 +90,6 @@ static unsigned long apic_phys;
90 90
91unsigned long mp_lapic_addr; 91unsigned long mp_lapic_addr;
92 92
93unsigned int __cpuinitdata maxcpus = NR_CPUS;
94/* 93/*
95 * Get the LAPIC version 94 * Get the LAPIC version
96 */ 95 */
@@ -1062,12 +1061,6 @@ void __cpuinit generic_processor_info(int apicid, int version)
1062 return; 1061 return;
1063 } 1062 }
1064 1063
1065 if (num_processors >= maxcpus) {
1066 printk(KERN_WARNING "WARNING: maxcpus limit of %i reached."
1067 " Processor ignored.\n", maxcpus);
1068 return;
1069 }
1070
1071 num_processors++; 1064 num_processors++;
1072 cpus_complement(tmp_map, cpu_present_map); 1065 cpus_complement(tmp_map, cpu_present_map);
1073 cpu = first_cpu(tmp_map); 1066 cpu = first_cpu(tmp_map);
diff --git a/arch/x86/kernel/cpu/addon_cpuid_features.c b/arch/x86/kernel/cpu/addon_cpuid_features.c
index 84a8220a6072..a6ef672adbba 100644
--- a/arch/x86/kernel/cpu/addon_cpuid_features.c
+++ b/arch/x86/kernel/cpu/addon_cpuid_features.c
@@ -56,9 +56,22 @@ void __cpuinit validate_pat_support(struct cpuinfo_x86 *c)
56 56
57 switch (c->x86_vendor) { 57 switch (c->x86_vendor) {
58 case X86_VENDOR_INTEL: 58 case X86_VENDOR_INTEL:
59 if (c->x86 == 0xF || (c->x86 == 6 && c->x86_model >= 15)) 59 /*
60 * There is a known erratum on Pentium III and Core Solo
61 * and Core Duo CPUs.
62 * " Page with PAT set to WC while associated MTRR is UC
63 * may consolidate to UC "
64 * Because of this erratum, it is better to stick with
65 * setting WC in MTRR rather than using PAT on these CPUs.
66 *
67 * Enable PAT WC only on P4, Core 2 or later CPUs.
68 */
69 if (c->x86 > 0x6 || (c->x86 == 6 && c->x86_model >= 15))
60 return; 70 return;
61 break; 71
72 pat_disable("PAT WC disabled due to known CPU erratum.");
73 return;
74
62 case X86_VENDOR_AMD: 75 case X86_VENDOR_AMD:
63 case X86_VENDOR_CENTAUR: 76 case X86_VENDOR_CENTAUR:
64 case X86_VENDOR_TRANSMETA: 77 case X86_VENDOR_TRANSMETA:
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index cae9cabc3031..18514ed26104 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -31,6 +31,11 @@ static void __cpuinit early_init_amd(struct cpuinfo_x86 *c)
31 if (c->x86_power & (1<<8)) 31 if (c->x86_power & (1<<8))
32 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); 32 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
33 } 33 }
34
35 /* Set MTRR capability flag if appropriate */
36 if (c->x86_model == 13 || c->x86_model == 9 ||
37 (c->x86_model == 8 && c->x86_mask >= 8))
38 set_cpu_cap(c, X86_FEATURE_K6_MTRR);
34} 39}
35 40
36static void __cpuinit init_amd(struct cpuinfo_x86 *c) 41static void __cpuinit init_amd(struct cpuinfo_x86 *c)
@@ -166,10 +171,6 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
166 mbytes); 171 mbytes);
167 } 172 }
168 173
169 /* Set MTRR capability flag if appropriate */
170 if (c->x86_model == 13 || c->x86_model == 9 ||
171 (c->x86_model == 8 && c->x86_mask >= 8))
172 set_cpu_cap(c, X86_FEATURE_K6_MTRR);
173 break; 174 break;
174 } 175 }
175 176
diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c
index c9b58a806e85..c8e315f1aa83 100644
--- a/arch/x86/kernel/cpu/bugs.c
+++ b/arch/x86/kernel/cpu/bugs.c
@@ -50,6 +50,8 @@ static double __initdata y = 3145727.0;
50 */ 50 */
51static void __init check_fpu(void) 51static void __init check_fpu(void)
52{ 52{
53 s32 fdiv_bug;
54
53 if (!boot_cpu_data.hard_math) { 55 if (!boot_cpu_data.hard_math) {
54#ifndef CONFIG_MATH_EMULATION 56#ifndef CONFIG_MATH_EMULATION
55 printk(KERN_EMERG "No coprocessor found and no math emulation present.\n"); 57 printk(KERN_EMERG "No coprocessor found and no math emulation present.\n");
@@ -74,8 +76,10 @@ static void __init check_fpu(void)
74 "fistpl %0\n\t" 76 "fistpl %0\n\t"
75 "fwait\n\t" 77 "fwait\n\t"
76 "fninit" 78 "fninit"
77 : "=m" (*&boot_cpu_data.fdiv_bug) 79 : "=m" (*&fdiv_bug)
78 : "m" (*&x), "m" (*&y)); 80 : "m" (*&x), "m" (*&y));
81
82 boot_cpu_data.fdiv_bug = fdiv_bug;
79 if (boot_cpu_data.fdiv_bug) 83 if (boot_cpu_data.fdiv_bug)
80 printk("Hmm, FPU with FDIV bug.\n"); 84 printk("Hmm, FPU with FDIV bug.\n");
81} 85}
diff --git a/arch/x86/kernel/cpu/centaur.c b/arch/x86/kernel/cpu/centaur.c
index e0f45edd6a55..a0534c04d38a 100644
--- a/arch/x86/kernel/cpu/centaur.c
+++ b/arch/x86/kernel/cpu/centaur.c
@@ -314,6 +314,16 @@ enum {
314 EAMD3D = 1<<20, 314 EAMD3D = 1<<20,
315}; 315};
316 316
317static void __cpuinit early_init_centaur(struct cpuinfo_x86 *c)
318{
319 switch (c->x86) {
320 case 5:
321 /* Emulate MTRRs using Centaur's MCR. */
322 set_cpu_cap(c, X86_FEATURE_CENTAUR_MCR);
323 break;
324 }
325}
326
317static void __cpuinit init_centaur(struct cpuinfo_x86 *c) 327static void __cpuinit init_centaur(struct cpuinfo_x86 *c)
318{ 328{
319 329
@@ -462,6 +472,7 @@ centaur_size_cache(struct cpuinfo_x86 *c, unsigned int size)
462static struct cpu_dev centaur_cpu_dev __cpuinitdata = { 472static struct cpu_dev centaur_cpu_dev __cpuinitdata = {
463 .c_vendor = "Centaur", 473 .c_vendor = "Centaur",
464 .c_ident = { "CentaurHauls" }, 474 .c_ident = { "CentaurHauls" },
475 .c_early_init = early_init_centaur,
465 .c_init = init_centaur, 476 .c_init = init_centaur,
466 .c_size_cache = centaur_size_cache, 477 .c_size_cache = centaur_size_cache,
467}; 478};
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 80ab20d4fa39..4e456bd955bb 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -13,6 +13,7 @@
13#include <asm/mtrr.h> 13#include <asm/mtrr.h>
14#include <asm/mce.h> 14#include <asm/mce.h>
15#include <asm/pat.h> 15#include <asm/pat.h>
16#include <asm/asm.h>
16#ifdef CONFIG_X86_LOCAL_APIC 17#ifdef CONFIG_X86_LOCAL_APIC
17#include <asm/mpspec.h> 18#include <asm/mpspec.h>
18#include <asm/apic.h> 19#include <asm/apic.h>
@@ -334,11 +335,24 @@ static void __init early_cpu_detect(void)
334 335
335 get_cpu_vendor(c, 1); 336 get_cpu_vendor(c, 1);
336 337
338 early_get_cap(c);
339
337 if (c->x86_vendor != X86_VENDOR_UNKNOWN && 340 if (c->x86_vendor != X86_VENDOR_UNKNOWN &&
338 cpu_devs[c->x86_vendor]->c_early_init) 341 cpu_devs[c->x86_vendor]->c_early_init)
339 cpu_devs[c->x86_vendor]->c_early_init(c); 342 cpu_devs[c->x86_vendor]->c_early_init(c);
343}
340 344
341 early_get_cap(c); 345/*
346 * The NOPL instruction is supposed to exist on all CPUs with
347 * family >= 6; unfortunately, that's not true in practice because
348 * of early VIA chips and (more importantly) broken virtualizers that
349 * are not easy to detect. In the latter case it doesn't even *fail*
350 * reliably, so probing for it doesn't even work. Disable it completely
351 * unless we can find a reliable way to detect all the broken cases.
352 */
353static void __cpuinit detect_nopl(struct cpuinfo_x86 *c)
354{
355 clear_cpu_cap(c, X86_FEATURE_NOPL);
342} 356}
343 357
344static void __cpuinit generic_identify(struct cpuinfo_x86 *c) 358static void __cpuinit generic_identify(struct cpuinfo_x86 *c)
@@ -395,8 +409,8 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 *c)
395 } 409 }
396 410
397 init_scattered_cpuid_features(c); 411 init_scattered_cpuid_features(c);
412 detect_nopl(c);
398 } 413 }
399
400} 414}
401 415
402static void __cpuinit squash_the_stupid_serial_number(struct cpuinfo_x86 *c) 416static void __cpuinit squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c
index dd6e3f15017e..a11f5d4477cd 100644
--- a/arch/x86/kernel/cpu/common_64.c
+++ b/arch/x86/kernel/cpu/common_64.c
@@ -18,6 +18,7 @@
18#include <asm/mtrr.h> 18#include <asm/mtrr.h>
19#include <asm/mce.h> 19#include <asm/mce.h>
20#include <asm/pat.h> 20#include <asm/pat.h>
21#include <asm/asm.h>
21#include <asm/numa.h> 22#include <asm/numa.h>
22#ifdef CONFIG_X86_LOCAL_APIC 23#ifdef CONFIG_X86_LOCAL_APIC
23#include <asm/mpspec.h> 24#include <asm/mpspec.h>
@@ -215,6 +216,39 @@ static void __init early_cpu_support_print(void)
215 } 216 }
216} 217}
217 218
219/*
220 * The NOPL instruction is supposed to exist on all CPUs with
221 * family >= 6, unfortunately, that's not true in practice because
222 * of early VIA chips and (more importantly) broken virtualizers that
223 * are not easy to detect. Hence, probe for it based on first
224 * principles.
225 *
226 * Note: no 64-bit chip is known to lack these, but put the code here
227 * for consistency with 32 bits, and to make it utterly trivial to
228 * diagnose the problem should it ever surface.
229 */
230static void __cpuinit detect_nopl(struct cpuinfo_x86 *c)
231{
232 const u32 nopl_signature = 0x888c53b1; /* Random number */
233 u32 has_nopl = nopl_signature;
234
235 clear_cpu_cap(c, X86_FEATURE_NOPL);
236 if (c->x86 >= 6) {
237 asm volatile("\n"
238 "1: .byte 0x0f,0x1f,0xc0\n" /* nopl %eax */
239 "2:\n"
240 " .section .fixup,\"ax\"\n"
241 "3: xor %0,%0\n"
242 " jmp 2b\n"
243 " .previous\n"
244 _ASM_EXTABLE(1b,3b)
245 : "+a" (has_nopl));
246
247 if (has_nopl == nopl_signature)
248 set_cpu_cap(c, X86_FEATURE_NOPL);
249 }
250}
251
218static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c); 252static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c);
219 253
220void __init early_cpu_init(void) 254void __init early_cpu_init(void)
@@ -313,6 +347,8 @@ static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c)
313 c->x86_phys_bits = eax & 0xff; 347 c->x86_phys_bits = eax & 0xff;
314 } 348 }
315 349
350 detect_nopl(c);
351
316 if (c->x86_vendor != X86_VENDOR_UNKNOWN && 352 if (c->x86_vendor != X86_VENDOR_UNKNOWN &&
317 cpu_devs[c->x86_vendor]->c_early_init) 353 cpu_devs[c->x86_vendor]->c_early_init)
318 cpu_devs[c->x86_vendor]->c_early_init(c); 354 cpu_devs[c->x86_vendor]->c_early_init(c);
@@ -493,17 +529,20 @@ void pda_init(int cpu)
493 /* others are initialized in smpboot.c */ 529 /* others are initialized in smpboot.c */
494 pda->pcurrent = &init_task; 530 pda->pcurrent = &init_task;
495 pda->irqstackptr = boot_cpu_stack; 531 pda->irqstackptr = boot_cpu_stack;
532 pda->irqstackptr += IRQSTACKSIZE - 64;
496 } else { 533 } else {
497 pda->irqstackptr = (char *) 534 if (!pda->irqstackptr) {
498 __get_free_pages(GFP_ATOMIC, IRQSTACK_ORDER); 535 pda->irqstackptr = (char *)
499 if (!pda->irqstackptr) 536 __get_free_pages(GFP_ATOMIC, IRQSTACK_ORDER);
500 panic("cannot allocate irqstack for cpu %d", cpu); 537 if (!pda->irqstackptr)
538 panic("cannot allocate irqstack for cpu %d",
539 cpu);
540 pda->irqstackptr += IRQSTACKSIZE - 64;
541 }
501 542
502 if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE) 543 if (pda->nodenumber == 0 && cpu_to_node(cpu) != NUMA_NO_NODE)
503 pda->nodenumber = cpu_to_node(cpu); 544 pda->nodenumber = cpu_to_node(cpu);
504 } 545 }
505
506 pda->irqstackptr += IRQSTACKSIZE-64;
507} 546}
508 547
509char boot_exception_stacks[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + 548char boot_exception_stacks[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ +
@@ -601,19 +640,22 @@ void __cpuinit cpu_init(void)
601 /* 640 /*
602 * set up and load the per-CPU TSS 641 * set up and load the per-CPU TSS
603 */ 642 */
604 for (v = 0; v < N_EXCEPTION_STACKS; v++) { 643 if (!orig_ist->ist[0]) {
605 static const unsigned int order[N_EXCEPTION_STACKS] = { 644 static const unsigned int order[N_EXCEPTION_STACKS] = {
606 [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER, 645 [0 ... N_EXCEPTION_STACKS - 1] = EXCEPTION_STACK_ORDER,
607 [DEBUG_STACK - 1] = DEBUG_STACK_ORDER 646 [DEBUG_STACK - 1] = DEBUG_STACK_ORDER
608 }; 647 };
609 if (cpu) { 648 for (v = 0; v < N_EXCEPTION_STACKS; v++) {
610 estacks = (char *)__get_free_pages(GFP_ATOMIC, order[v]); 649 if (cpu) {
611 if (!estacks) 650 estacks = (char *)__get_free_pages(GFP_ATOMIC, order[v]);
612 panic("Cannot allocate exception stack %ld %d\n", 651 if (!estacks)
613 v, cpu); 652 panic("Cannot allocate exception "
653 "stack %ld %d\n", v, cpu);
654 }
655 estacks += PAGE_SIZE << order[v];
656 orig_ist->ist[v] = t->x86_tss.ist[v] =
657 (unsigned long)estacks;
614 } 658 }
615 estacks += PAGE_SIZE << order[v];
616 orig_ist->ist[v] = t->x86_tss.ist[v] = (unsigned long)estacks;
617 } 659 }
618 660
619 t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap); 661 t->x86_tss.io_bitmap_base = offsetof(struct tss_struct, io_bitmap);
diff --git a/arch/x86/kernel/cpu/cpufreq/Kconfig b/arch/x86/kernel/cpu/cpufreq/Kconfig
index cb7a5715596d..efae3b22a0ff 100644
--- a/arch/x86/kernel/cpu/cpufreq/Kconfig
+++ b/arch/x86/kernel/cpu/cpufreq/Kconfig
@@ -235,9 +235,9 @@ config X86_LONGHAUL
235 If in doubt, say N. 235 If in doubt, say N.
236 236
237config X86_E_POWERSAVER 237config X86_E_POWERSAVER
238 tristate "VIA C7 Enhanced PowerSaver (EXPERIMENTAL)" 238 tristate "VIA C7 Enhanced PowerSaver"
239 select CPU_FREQ_TABLE 239 select CPU_FREQ_TABLE
240 depends on X86_32 && EXPERIMENTAL 240 depends on X86_32
241 help 241 help
242 This adds the CPUFreq driver for VIA C7 processors. 242 This adds the CPUFreq driver for VIA C7 processors.
243 243
diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
index ff2fff56f0a8..dd097b835839 100644
--- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
+++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
@@ -200,12 +200,10 @@ static void drv_read(struct drv_cmd *cmd)
200static void drv_write(struct drv_cmd *cmd) 200static void drv_write(struct drv_cmd *cmd)
201{ 201{
202 cpumask_t saved_mask = current->cpus_allowed; 202 cpumask_t saved_mask = current->cpus_allowed;
203 cpumask_of_cpu_ptr_declare(cpu_mask);
204 unsigned int i; 203 unsigned int i;
205 204
206 for_each_cpu_mask_nr(i, cmd->mask) { 205 for_each_cpu_mask_nr(i, cmd->mask) {
207 cpumask_of_cpu_ptr_next(cpu_mask, i); 206 set_cpus_allowed_ptr(current, &cpumask_of_cpu(i));
208 set_cpus_allowed_ptr(current, cpu_mask);
209 do_drv_write(cmd); 207 do_drv_write(cmd);
210 } 208 }
211 209
@@ -269,12 +267,11 @@ static unsigned int get_measured_perf(unsigned int cpu)
269 } aperf_cur, mperf_cur; 267 } aperf_cur, mperf_cur;
270 268
271 cpumask_t saved_mask; 269 cpumask_t saved_mask;
272 cpumask_of_cpu_ptr(cpu_mask, cpu);
273 unsigned int perf_percent; 270 unsigned int perf_percent;
274 unsigned int retval; 271 unsigned int retval;
275 272
276 saved_mask = current->cpus_allowed; 273 saved_mask = current->cpus_allowed;
277 set_cpus_allowed_ptr(current, cpu_mask); 274 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
278 if (get_cpu() != cpu) { 275 if (get_cpu() != cpu) {
279 /* We were not able to run on requested processor */ 276 /* We were not able to run on requested processor */
280 put_cpu(); 277 put_cpu();
@@ -340,7 +337,6 @@ static unsigned int get_measured_perf(unsigned int cpu)
340 337
341static unsigned int get_cur_freq_on_cpu(unsigned int cpu) 338static unsigned int get_cur_freq_on_cpu(unsigned int cpu)
342{ 339{
343 cpumask_of_cpu_ptr(cpu_mask, cpu);
344 struct acpi_cpufreq_data *data = per_cpu(drv_data, cpu); 340 struct acpi_cpufreq_data *data = per_cpu(drv_data, cpu);
345 unsigned int freq; 341 unsigned int freq;
346 unsigned int cached_freq; 342 unsigned int cached_freq;
@@ -353,7 +349,7 @@ static unsigned int get_cur_freq_on_cpu(unsigned int cpu)
353 } 349 }
354 350
355 cached_freq = data->freq_table[data->acpi_data->state].frequency; 351 cached_freq = data->freq_table[data->acpi_data->state].frequency;
356 freq = extract_freq(get_cur_val(cpu_mask), data); 352 freq = extract_freq(get_cur_val(&cpumask_of_cpu(cpu)), data);
357 if (freq != cached_freq) { 353 if (freq != cached_freq) {
358 /* 354 /*
359 * The dreaded BIOS frequency change behind our back. 355 * The dreaded BIOS frequency change behind our back.
diff --git a/arch/x86/kernel/cpu/cpufreq/elanfreq.c b/arch/x86/kernel/cpu/cpufreq/elanfreq.c
index 94619c22f563..e4a4bf870e94 100644
--- a/arch/x86/kernel/cpu/cpufreq/elanfreq.c
+++ b/arch/x86/kernel/cpu/cpufreq/elanfreq.c
@@ -44,7 +44,7 @@ struct s_elan_multiplier {
44 * It is important that the frequencies 44 * It is important that the frequencies
45 * are listed in ascending order here! 45 * are listed in ascending order here!
46 */ 46 */
47struct s_elan_multiplier elan_multiplier[] = { 47static struct s_elan_multiplier elan_multiplier[] = {
48 {1000, 0x02, 0x18}, 48 {1000, 0x02, 0x18},
49 {2000, 0x02, 0x10}, 49 {2000, 0x02, 0x10},
50 {4000, 0x02, 0x08}, 50 {4000, 0x02, 0x08},
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
index 53c7b6936973..84bb395038d8 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
@@ -66,7 +66,6 @@ static u32 find_freq_from_fid(u32 fid)
66 return 800 + (fid * 100); 66 return 800 + (fid * 100);
67} 67}
68 68
69
70/* Return a frequency in KHz, given an input fid */ 69/* Return a frequency in KHz, given an input fid */
71static u32 find_khz_freq_from_fid(u32 fid) 70static u32 find_khz_freq_from_fid(u32 fid)
72{ 71{
@@ -78,7 +77,6 @@ static u32 find_khz_freq_from_pstate(struct cpufreq_frequency_table *data, u32 p
78 return data[pstate].frequency; 77 return data[pstate].frequency;
79} 78}
80 79
81
82/* Return the vco fid for an input fid 80/* Return the vco fid for an input fid
83 * 81 *
84 * Each "low" fid has corresponding "high" fid, and you can get to "low" fids 82 * Each "low" fid has corresponding "high" fid, and you can get to "low" fids
@@ -166,7 +164,6 @@ static void fidvid_msr_init(void)
166 wrmsr(MSR_FIDVID_CTL, lo, hi); 164 wrmsr(MSR_FIDVID_CTL, lo, hi);
167} 165}
168 166
169
170/* write the new fid value along with the other control fields to the msr */ 167/* write the new fid value along with the other control fields to the msr */
171static int write_new_fid(struct powernow_k8_data *data, u32 fid) 168static int write_new_fid(struct powernow_k8_data *data, u32 fid)
172{ 169{
@@ -479,12 +476,11 @@ static int core_voltage_post_transition(struct powernow_k8_data *data, u32 reqvi
479static int check_supported_cpu(unsigned int cpu) 476static int check_supported_cpu(unsigned int cpu)
480{ 477{
481 cpumask_t oldmask; 478 cpumask_t oldmask;
482 cpumask_of_cpu_ptr(cpu_mask, cpu);
483 u32 eax, ebx, ecx, edx; 479 u32 eax, ebx, ecx, edx;
484 unsigned int rc = 0; 480 unsigned int rc = 0;
485 481
486 oldmask = current->cpus_allowed; 482 oldmask = current->cpus_allowed;
487 set_cpus_allowed_ptr(current, cpu_mask); 483 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
488 484
489 if (smp_processor_id() != cpu) { 485 if (smp_processor_id() != cpu) {
490 printk(KERN_ERR PFX "limiting to cpu %u failed\n", cpu); 486 printk(KERN_ERR PFX "limiting to cpu %u failed\n", cpu);
@@ -1017,7 +1013,6 @@ static int transition_frequency_pstate(struct powernow_k8_data *data, unsigned i
1017static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsigned relation) 1013static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsigned relation)
1018{ 1014{
1019 cpumask_t oldmask; 1015 cpumask_t oldmask;
1020 cpumask_of_cpu_ptr(cpu_mask, pol->cpu);
1021 struct powernow_k8_data *data = per_cpu(powernow_data, pol->cpu); 1016 struct powernow_k8_data *data = per_cpu(powernow_data, pol->cpu);
1022 u32 checkfid; 1017 u32 checkfid;
1023 u32 checkvid; 1018 u32 checkvid;
@@ -1032,7 +1027,7 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi
1032 1027
1033 /* only run on specific CPU from here on */ 1028 /* only run on specific CPU from here on */
1034 oldmask = current->cpus_allowed; 1029 oldmask = current->cpus_allowed;
1035 set_cpus_allowed_ptr(current, cpu_mask); 1030 set_cpus_allowed_ptr(current, &cpumask_of_cpu(pol->cpu));
1036 1031
1037 if (smp_processor_id() != pol->cpu) { 1032 if (smp_processor_id() != pol->cpu) {
1038 printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu); 1033 printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu);
@@ -1107,7 +1102,6 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1107{ 1102{
1108 struct powernow_k8_data *data; 1103 struct powernow_k8_data *data;
1109 cpumask_t oldmask; 1104 cpumask_t oldmask;
1110 cpumask_of_cpu_ptr_declare(newmask);
1111 int rc; 1105 int rc;
1112 1106
1113 if (!cpu_online(pol->cpu)) 1107 if (!cpu_online(pol->cpu))
@@ -1159,8 +1153,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1159 1153
1160 /* only run on specific CPU from here on */ 1154 /* only run on specific CPU from here on */
1161 oldmask = current->cpus_allowed; 1155 oldmask = current->cpus_allowed;
1162 cpumask_of_cpu_ptr_next(newmask, pol->cpu); 1156 set_cpus_allowed_ptr(current, &cpumask_of_cpu(pol->cpu));
1163 set_cpus_allowed_ptr(current, newmask);
1164 1157
1165 if (smp_processor_id() != pol->cpu) { 1158 if (smp_processor_id() != pol->cpu) {
1166 printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu); 1159 printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu);
@@ -1182,7 +1175,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1182 set_cpus_allowed_ptr(current, &oldmask); 1175 set_cpus_allowed_ptr(current, &oldmask);
1183 1176
1184 if (cpu_family == CPU_HW_PSTATE) 1177 if (cpu_family == CPU_HW_PSTATE)
1185 pol->cpus = *newmask; 1178 pol->cpus = cpumask_of_cpu(pol->cpu);
1186 else 1179 else
1187 pol->cpus = per_cpu(cpu_core_map, pol->cpu); 1180 pol->cpus = per_cpu(cpu_core_map, pol->cpu);
1188 data->available_cores = &(pol->cpus); 1181 data->available_cores = &(pol->cpus);
@@ -1248,7 +1241,6 @@ static unsigned int powernowk8_get (unsigned int cpu)
1248{ 1241{
1249 struct powernow_k8_data *data; 1242 struct powernow_k8_data *data;
1250 cpumask_t oldmask = current->cpus_allowed; 1243 cpumask_t oldmask = current->cpus_allowed;
1251 cpumask_of_cpu_ptr(newmask, cpu);
1252 unsigned int khz = 0; 1244 unsigned int khz = 0;
1253 unsigned int first; 1245 unsigned int first;
1254 1246
@@ -1258,7 +1250,7 @@ static unsigned int powernowk8_get (unsigned int cpu)
1258 if (!data) 1250 if (!data)
1259 return -EINVAL; 1251 return -EINVAL;
1260 1252
1261 set_cpus_allowed_ptr(current, newmask); 1253 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
1262 if (smp_processor_id() != cpu) { 1254 if (smp_processor_id() != cpu) {
1263 printk(KERN_ERR PFX 1255 printk(KERN_ERR PFX
1264 "limiting to CPU %d failed in powernowk8_get\n", cpu); 1256 "limiting to CPU %d failed in powernowk8_get\n", cpu);
diff --git a/arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c b/arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c
index ca2ac13b7af2..15e13c01cc36 100644
--- a/arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c
+++ b/arch/x86/kernel/cpu/cpufreq/speedstep-centrino.c
@@ -324,10 +324,9 @@ static unsigned int get_cur_freq(unsigned int cpu)
324 unsigned l, h; 324 unsigned l, h;
325 unsigned clock_freq; 325 unsigned clock_freq;
326 cpumask_t saved_mask; 326 cpumask_t saved_mask;
327 cpumask_of_cpu_ptr(new_mask, cpu);
328 327
329 saved_mask = current->cpus_allowed; 328 saved_mask = current->cpus_allowed;
330 set_cpus_allowed_ptr(current, new_mask); 329 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
331 if (smp_processor_id() != cpu) 330 if (smp_processor_id() != cpu)
332 return 0; 331 return 0;
333 332
@@ -585,15 +584,12 @@ static int centrino_target (struct cpufreq_policy *policy,
585 * Best effort undo.. 584 * Best effort undo..
586 */ 585 */
587 586
588 if (!cpus_empty(*covered_cpus)) { 587 if (!cpus_empty(*covered_cpus))
589 cpumask_of_cpu_ptr_declare(new_mask);
590
591 for_each_cpu_mask_nr(j, *covered_cpus) { 588 for_each_cpu_mask_nr(j, *covered_cpus) {
592 cpumask_of_cpu_ptr_next(new_mask, j); 589 set_cpus_allowed_ptr(current,
593 set_cpus_allowed_ptr(current, new_mask); 590 &cpumask_of_cpu(j));
594 wrmsr(MSR_IA32_PERF_CTL, oldmsr, h); 591 wrmsr(MSR_IA32_PERF_CTL, oldmsr, h);
595 } 592 }
596 }
597 593
598 tmp = freqs.new; 594 tmp = freqs.new;
599 freqs.new = freqs.old; 595 freqs.new = freqs.old;
diff --git a/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c b/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c
index 2f3728dc24f6..191f7263c61d 100644
--- a/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c
+++ b/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c
@@ -244,8 +244,7 @@ static unsigned int _speedstep_get(const cpumask_t *cpus)
244 244
245static unsigned int speedstep_get(unsigned int cpu) 245static unsigned int speedstep_get(unsigned int cpu)
246{ 246{
247 cpumask_of_cpu_ptr(newmask, cpu); 247 return _speedstep_get(&cpumask_of_cpu(cpu));
248 return _speedstep_get(newmask);
249} 248}
250 249
251/** 250/**
diff --git a/arch/x86/kernel/cpu/cyrix.c b/arch/x86/kernel/cpu/cyrix.c
index 3fd7a67bb06a..898a5a2002ed 100644
--- a/arch/x86/kernel/cpu/cyrix.c
+++ b/arch/x86/kernel/cpu/cyrix.c
@@ -15,13 +15,11 @@
15/* 15/*
16 * Read NSC/Cyrix DEVID registers (DIR) to get more detailed info. about the CPU 16 * Read NSC/Cyrix DEVID registers (DIR) to get more detailed info. about the CPU
17 */ 17 */
18static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1) 18static void __cpuinit __do_cyrix_devid(unsigned char *dir0, unsigned char *dir1)
19{ 19{
20 unsigned char ccr2, ccr3; 20 unsigned char ccr2, ccr3;
21 unsigned long flags;
22 21
23 /* we test for DEVID by checking whether CCR3 is writable */ 22 /* we test for DEVID by checking whether CCR3 is writable */
24 local_irq_save(flags);
25 ccr3 = getCx86(CX86_CCR3); 23 ccr3 = getCx86(CX86_CCR3);
26 setCx86(CX86_CCR3, ccr3 ^ 0x80); 24 setCx86(CX86_CCR3, ccr3 ^ 0x80);
27 getCx86(0xc0); /* dummy to change bus */ 25 getCx86(0xc0); /* dummy to change bus */
@@ -44,9 +42,16 @@ static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1)
44 *dir0 = getCx86(CX86_DIR0); 42 *dir0 = getCx86(CX86_DIR0);
45 *dir1 = getCx86(CX86_DIR1); 43 *dir1 = getCx86(CX86_DIR1);
46 } 44 }
47 local_irq_restore(flags);
48} 45}
49 46
47static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1)
48{
49 unsigned long flags;
50
51 local_irq_save(flags);
52 __do_cyrix_devid(dir0, dir1);
53 local_irq_restore(flags);
54}
50/* 55/*
51 * Cx86_dir0_msb is a HACK needed by check_cx686_cpuid/slop in bugs.h in 56 * Cx86_dir0_msb is a HACK needed by check_cx686_cpuid/slop in bugs.h in
52 * order to identify the Cyrix CPU model after we're out of setup.c 57 * order to identify the Cyrix CPU model after we're out of setup.c
@@ -134,23 +139,6 @@ static void __cpuinit set_cx86_memwb(void)
134 setCx86(CX86_CCR2, getCx86(CX86_CCR2) | 0x14); 139 setCx86(CX86_CCR2, getCx86(CX86_CCR2) | 0x14);
135} 140}
136 141
137static void __cpuinit set_cx86_inc(void)
138{
139 unsigned char ccr3;
140
141 printk(KERN_INFO "Enable Incrementor on Cyrix/NSC processor.\n");
142
143 ccr3 = getCx86(CX86_CCR3);
144 setCx86(CX86_CCR3, (ccr3 & 0x0f) | 0x10); /* enable MAPEN */
145 /* PCR1 -- Performance Control */
146 /* Incrementor on, whatever that is */
147 setCx86(CX86_PCR1, getCx86(CX86_PCR1) | 0x02);
148 /* PCR0 -- Performance Control */
149 /* Incrementor Margin 10 */
150 setCx86(CX86_PCR0, getCx86(CX86_PCR0) | 0x04);
151 setCx86(CX86_CCR3, ccr3); /* disable MAPEN */
152}
153
154/* 142/*
155 * Configure later MediaGX and/or Geode processor. 143 * Configure later MediaGX and/or Geode processor.
156 */ 144 */
@@ -174,11 +162,28 @@ static void __cpuinit geode_configure(void)
174 162
175 set_cx86_memwb(); 163 set_cx86_memwb();
176 set_cx86_reorder(); 164 set_cx86_reorder();
177 set_cx86_inc();
178 165
179 local_irq_restore(flags); 166 local_irq_restore(flags);
180} 167}
181 168
169static void __cpuinit early_init_cyrix(struct cpuinfo_x86 *c)
170{
171 unsigned char dir0, dir0_msn, dir1 = 0;
172
173 __do_cyrix_devid(&dir0, &dir1);
174 dir0_msn = dir0 >> 4; /* identifies CPU "family" */
175
176 switch (dir0_msn) {
177 case 3: /* 6x86/6x86L */
178 /* Emulate MTRRs using Cyrix's ARRs. */
179 set_cpu_cap(c, X86_FEATURE_CYRIX_ARR);
180 break;
181 case 5: /* 6x86MX/M II */
182 /* Emulate MTRRs using Cyrix's ARRs. */
183 set_cpu_cap(c, X86_FEATURE_CYRIX_ARR);
184 break;
185 }
186}
182 187
183static void __cpuinit init_cyrix(struct cpuinfo_x86 *c) 188static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
184{ 189{
@@ -434,6 +439,7 @@ static void __cpuinit cyrix_identify(struct cpuinfo_x86 *c)
434static struct cpu_dev cyrix_cpu_dev __cpuinitdata = { 439static struct cpu_dev cyrix_cpu_dev __cpuinitdata = {
435 .c_vendor = "Cyrix", 440 .c_vendor = "Cyrix",
436 .c_ident = { "CyrixInstead" }, 441 .c_ident = { "CyrixInstead" },
442 .c_early_init = early_init_cyrix,
437 .c_init = init_cyrix, 443 .c_init = init_cyrix,
438 .c_identify = cyrix_identify, 444 .c_identify = cyrix_identify,
439}; 445};
diff --git a/arch/x86/kernel/cpu/feature_names.c b/arch/x86/kernel/cpu/feature_names.c
index e43ad4ad4cba..c9017799497c 100644
--- a/arch/x86/kernel/cpu/feature_names.c
+++ b/arch/x86/kernel/cpu/feature_names.c
@@ -39,7 +39,8 @@ const char * const x86_cap_flags[NCAPINTS*32] = {
39 NULL, NULL, NULL, NULL, 39 NULL, NULL, NULL, NULL,
40 "constant_tsc", "up", NULL, "arch_perfmon", 40 "constant_tsc", "up", NULL, "arch_perfmon",
41 "pebs", "bts", NULL, NULL, 41 "pebs", "bts", NULL, NULL,
42 "rep_good", NULL, NULL, NULL, NULL, NULL, NULL, NULL, 42 "rep_good", NULL, NULL, NULL,
43 "nopl", NULL, NULL, NULL,
43 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, 44 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
44 45
45 /* Intel-defined (#2) */ 46 /* Intel-defined (#2) */
diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c
index 650d40f7912b..6b0a10b002f1 100644
--- a/arch/x86/kernel/cpu/intel_cacheinfo.c
+++ b/arch/x86/kernel/cpu/intel_cacheinfo.c
@@ -516,7 +516,6 @@ static int __cpuinit detect_cache_attributes(unsigned int cpu)
516 unsigned long j; 516 unsigned long j;
517 int retval; 517 int retval;
518 cpumask_t oldmask; 518 cpumask_t oldmask;
519 cpumask_of_cpu_ptr(newmask, cpu);
520 519
521 if (num_cache_leaves == 0) 520 if (num_cache_leaves == 0)
522 return -ENOENT; 521 return -ENOENT;
@@ -527,7 +526,7 @@ static int __cpuinit detect_cache_attributes(unsigned int cpu)
527 return -ENOMEM; 526 return -ENOMEM;
528 527
529 oldmask = current->cpus_allowed; 528 oldmask = current->cpus_allowed;
530 retval = set_cpus_allowed_ptr(current, newmask); 529 retval = set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
531 if (retval) 530 if (retval)
532 goto out; 531 goto out;
533 532
diff --git a/arch/x86/kernel/cpu/mcheck/mce_64.c b/arch/x86/kernel/cpu/mcheck/mce_64.c
index 65a339678ece..726a5fcdf341 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_64.c
@@ -759,6 +759,7 @@ static struct sysdev_class mce_sysclass = {
759}; 759};
760 760
761DEFINE_PER_CPU(struct sys_device, device_mce); 761DEFINE_PER_CPU(struct sys_device, device_mce);
762void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu) __cpuinitdata;
762 763
763/* Why are there no generic functions for this? */ 764/* Why are there no generic functions for this? */
764#define ACCESSOR(name, var, start) \ 765#define ACCESSOR(name, var, start) \
@@ -883,9 +884,13 @@ static int __cpuinit mce_cpu_callback(struct notifier_block *nfb,
883 case CPU_ONLINE: 884 case CPU_ONLINE:
884 case CPU_ONLINE_FROZEN: 885 case CPU_ONLINE_FROZEN:
885 mce_create_device(cpu); 886 mce_create_device(cpu);
887 if (threshold_cpu_callback)
888 threshold_cpu_callback(action, cpu);
886 break; 889 break;
887 case CPU_DEAD: 890 case CPU_DEAD:
888 case CPU_DEAD_FROZEN: 891 case CPU_DEAD_FROZEN:
892 if (threshold_cpu_callback)
893 threshold_cpu_callback(action, cpu);
889 mce_remove_device(cpu); 894 mce_remove_device(cpu);
890 break; 895 break;
891 } 896 }
diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
index 88736cadbaa6..5eb390a4b2e9 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
@@ -628,6 +628,7 @@ static void threshold_remove_bank(unsigned int cpu, int bank)
628 deallocate_threshold_block(cpu, bank); 628 deallocate_threshold_block(cpu, bank);
629 629
630free_out: 630free_out:
631 kobject_del(b->kobj);
631 kobject_put(b->kobj); 632 kobject_put(b->kobj);
632 kfree(b); 633 kfree(b);
633 per_cpu(threshold_banks, cpu)[bank] = NULL; 634 per_cpu(threshold_banks, cpu)[bank] = NULL;
@@ -645,14 +646,11 @@ static void threshold_remove_device(unsigned int cpu)
645} 646}
646 647
647/* get notified when a cpu comes on/off */ 648/* get notified when a cpu comes on/off */
648static int __cpuinit threshold_cpu_callback(struct notifier_block *nfb, 649static void __cpuinit amd_64_threshold_cpu_callback(unsigned long action,
649 unsigned long action, void *hcpu) 650 unsigned int cpu)
650{ 651{
651 /* cpu was unsigned int to begin with */
652 unsigned int cpu = (unsigned long)hcpu;
653
654 if (cpu >= NR_CPUS) 652 if (cpu >= NR_CPUS)
655 goto out; 653 return;
656 654
657 switch (action) { 655 switch (action) {
658 case CPU_ONLINE: 656 case CPU_ONLINE:
@@ -666,14 +664,8 @@ static int __cpuinit threshold_cpu_callback(struct notifier_block *nfb,
666 default: 664 default:
667 break; 665 break;
668 } 666 }
669 out:
670 return NOTIFY_OK;
671} 667}
672 668
673static struct notifier_block threshold_cpu_notifier __cpuinitdata = {
674 .notifier_call = threshold_cpu_callback,
675};
676
677static __init int threshold_init_device(void) 669static __init int threshold_init_device(void)
678{ 670{
679 unsigned lcpu = 0; 671 unsigned lcpu = 0;
@@ -684,7 +676,7 @@ static __init int threshold_init_device(void)
684 if (err) 676 if (err)
685 return err; 677 return err;
686 } 678 }
687 register_hotcpu_notifier(&threshold_cpu_notifier); 679 threshold_cpu_callback = amd_64_threshold_cpu_callback;
688 return 0; 680 return 0;
689} 681}
690 682
diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index 509bd3d9eacd..cb7d3b6a80eb 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -379,6 +379,7 @@ static void generic_get_mtrr(unsigned int reg, unsigned long *base,
379 unsigned long *size, mtrr_type *type) 379 unsigned long *size, mtrr_type *type)
380{ 380{
381 unsigned int mask_lo, mask_hi, base_lo, base_hi; 381 unsigned int mask_lo, mask_hi, base_lo, base_hi;
382 unsigned int tmp, hi;
382 383
383 rdmsr(MTRRphysMask_MSR(reg), mask_lo, mask_hi); 384 rdmsr(MTRRphysMask_MSR(reg), mask_lo, mask_hi);
384 if ((mask_lo & 0x800) == 0) { 385 if ((mask_lo & 0x800) == 0) {
@@ -392,8 +393,23 @@ static void generic_get_mtrr(unsigned int reg, unsigned long *base,
392 rdmsr(MTRRphysBase_MSR(reg), base_lo, base_hi); 393 rdmsr(MTRRphysBase_MSR(reg), base_lo, base_hi);
393 394
394 /* Work out the shifted address mask. */ 395 /* Work out the shifted address mask. */
395 mask_lo = size_or_mask | mask_hi << (32 - PAGE_SHIFT) 396 tmp = mask_hi << (32 - PAGE_SHIFT) | mask_lo >> PAGE_SHIFT;
396 | mask_lo >> PAGE_SHIFT; 397 mask_lo = size_or_mask | tmp;
398 /* Expand tmp with high bits to all 1s*/
399 hi = fls(tmp);
400 if (hi > 0) {
401 tmp |= ~((1<<(hi - 1)) - 1);
402
403 if (tmp != mask_lo) {
404 static int once = 1;
405
406 if (once) {
407 printk(KERN_INFO "mtrr: your BIOS has set up an incorrect mask, fixing it up.\n");
408 once = 0;
409 }
410 mask_lo = tmp;
411 }
412 }
397 413
398 /* This works correctly if size is a power of two, i.e. a 414 /* This works correctly if size is a power of two, i.e. a
399 contiguous range. */ 415 contiguous range. */
diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c
index 6f23969c8faf..b117d7f8a564 100644
--- a/arch/x86/kernel/cpu/mtrr/main.c
+++ b/arch/x86/kernel/cpu/mtrr/main.c
@@ -1496,11 +1496,8 @@ int __init mtrr_trim_uncached_memory(unsigned long end_pfn)
1496 1496
1497 /* kvm/qemu doesn't have mtrr set right, don't trim them all */ 1497 /* kvm/qemu doesn't have mtrr set right, don't trim them all */
1498 if (!highest_pfn) { 1498 if (!highest_pfn) {
1499 if (!kvm_para_available()) { 1499 WARN(!kvm_para_available(), KERN_WARNING
1500 printk(KERN_WARNING
1501 "WARNING: strange, CPU MTRRs all blank?\n"); 1500 "WARNING: strange, CPU MTRRs all blank?\n");
1502 WARN_ON(1);
1503 }
1504 return 0; 1501 return 0;
1505 } 1502 }
1506 1503
diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c
index de7439f82b92..05cc22dbd4ff 100644
--- a/arch/x86/kernel/cpu/perfctr-watchdog.c
+++ b/arch/x86/kernel/cpu/perfctr-watchdog.c
@@ -478,7 +478,13 @@ static int setup_p4_watchdog(unsigned nmi_hz)
478 perfctr_msr = MSR_P4_IQ_PERFCTR1; 478 perfctr_msr = MSR_P4_IQ_PERFCTR1;
479 evntsel_msr = MSR_P4_CRU_ESCR0; 479 evntsel_msr = MSR_P4_CRU_ESCR0;
480 cccr_msr = MSR_P4_IQ_CCCR1; 480 cccr_msr = MSR_P4_IQ_CCCR1;
481 cccr_val = P4_CCCR_OVF_PMI1 | P4_CCCR_ESCR_SELECT(4); 481
482 /* Pentium 4 D processors don't support P4_CCCR_OVF_PMI1 */
483 if (boot_cpu_data.x86_model == 4 && boot_cpu_data.x86_mask == 4)
484 cccr_val = P4_CCCR_OVF_PMI0;
485 else
486 cccr_val = P4_CCCR_OVF_PMI1;
487 cccr_val |= P4_CCCR_ESCR_SELECT(4);
482 } 488 }
483 489
484 evntsel = P4_ESCR_EVENT_SELECT(0x3F) 490 evntsel = P4_ESCR_EVENT_SELECT(0x3F)
diff --git a/arch/x86/kernel/cpuid.c b/arch/x86/kernel/cpuid.c
index 14b11b3be31c..8e9cd6a8ec12 100644
--- a/arch/x86/kernel/cpuid.c
+++ b/arch/x86/kernel/cpuid.c
@@ -89,6 +89,8 @@ static ssize_t cpuid_read(struct file *file, char __user *buf,
89 struct cpuid_regs cmd; 89 struct cpuid_regs cmd;
90 int cpu = iminor(file->f_path.dentry->d_inode); 90 int cpu = iminor(file->f_path.dentry->d_inode);
91 u64 pos = *ppos; 91 u64 pos = *ppos;
92 ssize_t bytes = 0;
93 int err = 0;
92 94
93 if (count % 16) 95 if (count % 16)
94 return -EINVAL; /* Invalid chunk size */ 96 return -EINVAL; /* Invalid chunk size */
@@ -96,14 +98,19 @@ static ssize_t cpuid_read(struct file *file, char __user *buf,
96 for (; count; count -= 16) { 98 for (; count; count -= 16) {
97 cmd.eax = pos; 99 cmd.eax = pos;
98 cmd.ecx = pos >> 32; 100 cmd.ecx = pos >> 32;
99 smp_call_function_single(cpu, cpuid_smp_cpuid, &cmd, 1); 101 err = smp_call_function_single(cpu, cpuid_smp_cpuid, &cmd, 1);
100 if (copy_to_user(tmp, &cmd, 16)) 102 if (err)
101 return -EFAULT; 103 break;
104 if (copy_to_user(tmp, &cmd, 16)) {
105 err = -EFAULT;
106 break;
107 }
102 tmp += 16; 108 tmp += 16;
109 bytes += 16;
103 *ppos = ++pos; 110 *ppos = ++pos;
104 } 111 }
105 112
106 return tmp - buf; 113 return bytes ? bytes : err;
107} 114}
108 115
109static int cpuid_open(struct inode *inode, struct file *file) 116static int cpuid_open(struct inode *inode, struct file *file)
diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c
index 9af89078f7bb..66e48aa2dd1b 100644
--- a/arch/x86/kernel/e820.c
+++ b/arch/x86/kernel/e820.c
@@ -1203,7 +1203,7 @@ static int __init parse_memmap_opt(char *p)
1203 if (!p) 1203 if (!p)
1204 return -EINVAL; 1204 return -EINVAL;
1205 1205
1206 if (!strcmp(p, "exactmap")) { 1206 if (!strncmp(p, "exactmap", 8)) {
1207#ifdef CONFIG_CRASH_DUMP 1207#ifdef CONFIG_CRASH_DUMP
1208 /* 1208 /*
1209 * If we are doing a crash dump, we still need to know 1209 * If we are doing a crash dump, we still need to know
diff --git a/arch/x86/kernel/efi_32.c b/arch/x86/kernel/efi_32.c
index 4b63c8e1f13b..5cab48ee61a4 100644
--- a/arch/x86/kernel/efi_32.c
+++ b/arch/x86/kernel/efi_32.c
@@ -53,7 +53,7 @@ void efi_call_phys_prelog(void)
53 * directory. If I have PAE, I just need to duplicate one entry in 53 * directory. If I have PAE, I just need to duplicate one entry in
54 * page directory. 54 * page directory.
55 */ 55 */
56 cr4 = read_cr4(); 56 cr4 = read_cr4_safe();
57 57
58 if (cr4 & X86_CR4_PAE) { 58 if (cr4 & X86_CR4_PAE) {
59 efi_bak_pg_dir_pointer[0].pgd = 59 efi_bak_pg_dir_pointer[0].pgd =
@@ -91,7 +91,7 @@ void efi_call_phys_epilog(void)
91 gdt_descr.size = GDT_SIZE - 1; 91 gdt_descr.size = GDT_SIZE - 1;
92 load_gdt(&gdt_descr); 92 load_gdt(&gdt_descr);
93 93
94 cr4 = read_cr4(); 94 cr4 = read_cr4_safe();
95 95
96 if (cr4 & X86_CR4_PAE) { 96 if (cr4 & X86_CR4_PAE) {
97 swapper_pg_dir[pgd_index(0)].pgd = 97 swapper_pg_dir[pgd_index(0)].pgd =
diff --git a/arch/x86/kernel/genapic_64.c b/arch/x86/kernel/genapic_64.c
index 1fa8be5bd217..eaff0bbb1444 100644
--- a/arch/x86/kernel/genapic_64.c
+++ b/arch/x86/kernel/genapic_64.c
@@ -99,3 +99,4 @@ int is_uv_system(void)
99{ 99{
100 return uv_system_type != UV_NONE; 100 return uv_system_type != UV_NONE;
101} 101}
102EXPORT_SYMBOL_GPL(is_uv_system);
diff --git a/arch/x86/kernel/genx2apic_uv_x.c b/arch/x86/kernel/genx2apic_uv_x.c
index 2cfcbded888a..bfa837cb16be 100644
--- a/arch/x86/kernel/genx2apic_uv_x.c
+++ b/arch/x86/kernel/genx2apic_uv_x.c
@@ -222,7 +222,7 @@ static __init void map_low_mmrs(void)
222 222
223enum map_type {map_wb, map_uc}; 223enum map_type {map_wb, map_uc};
224 224
225static void map_high(char *id, unsigned long base, int shift, enum map_type map_type) 225static __init void map_high(char *id, unsigned long base, int shift, enum map_type map_type)
226{ 226{
227 unsigned long bytes, paddr; 227 unsigned long bytes, paddr;
228 228
@@ -293,7 +293,9 @@ static __init void uv_rtc_init(void)
293 sn_rtc_cycles_per_second = ticks_per_sec; 293 sn_rtc_cycles_per_second = ticks_per_sec;
294} 294}
295 295
296static __init void uv_system_init(void) 296static bool uv_system_inited;
297
298void __init uv_system_init(void)
297{ 299{
298 union uvh_si_addr_map_config_u m_n_config; 300 union uvh_si_addr_map_config_u m_n_config;
299 union uvh_node_id_u node_id; 301 union uvh_node_id_u node_id;
@@ -383,6 +385,7 @@ static __init void uv_system_init(void)
383 map_mmr_high(max_pnode); 385 map_mmr_high(max_pnode);
384 map_config_high(max_pnode); 386 map_config_high(max_pnode);
385 map_mmioh_high(max_pnode); 387 map_mmioh_high(max_pnode);
388 uv_system_inited = true;
386} 389}
387 390
388/* 391/*
@@ -391,8 +394,7 @@ static __init void uv_system_init(void)
391 */ 394 */
392void __cpuinit uv_cpu_init(void) 395void __cpuinit uv_cpu_init(void)
393{ 396{
394 if (!uv_node_to_blade) 397 BUG_ON(!uv_system_inited);
395 uv_system_init();
396 398
397 uv_blade_info[uv_numa_blade_id()].nr_online_cpus++; 399 uv_blade_info[uv_numa_blade_id()].nr_online_cpus++;
398 400
diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index 1b318e903bf6..9bfc4d72fb2e 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -88,6 +88,7 @@ void __init x86_64_start_kernel(char * real_mode_data)
88 BUILD_BUG_ON(!(MODULES_VADDR > __START_KERNEL)); 88 BUILD_BUG_ON(!(MODULES_VADDR > __START_KERNEL));
89 BUILD_BUG_ON(!(((MODULES_END - 1) & PGDIR_MASK) == 89 BUILD_BUG_ON(!(((MODULES_END - 1) & PGDIR_MASK) ==
90 (__START_KERNEL & PGDIR_MASK))); 90 (__START_KERNEL & PGDIR_MASK)));
91 BUILD_BUG_ON(__fix_to_virt(__end_of_fixed_addresses) <= MODULES_END);
91 92
92 /* clear bss before set_intr_gate with early_idt_handler */ 93 /* clear bss before set_intr_gate with early_idt_handler */
93 clear_bss(); 94 clear_bss();
diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S
index f67e93441caf..a7010c3a377a 100644
--- a/arch/x86/kernel/head_32.S
+++ b/arch/x86/kernel/head_32.S
@@ -456,9 +456,6 @@ is386: movl $2,%ecx # set MP
4561: 4561:
457#endif /* CONFIG_SMP */ 457#endif /* CONFIG_SMP */
458 jmp *(initial_code) 458 jmp *(initial_code)
459.align 4
460ENTRY(initial_code)
461 .long i386_start_kernel
462 459
463/* 460/*
464 * We depend on ET to be correct. This checks for 287/387. 461 * We depend on ET to be correct. This checks for 287/387.
@@ -601,6 +598,11 @@ ignore_int:
601#endif 598#endif
602 iret 599 iret
603 600
601.section .cpuinit.data,"wa"
602.align 4
603ENTRY(initial_code)
604 .long i386_start_kernel
605
604.section .text 606.section .text
605/* 607/*
606 * Real beginning of normal "text" segment 608 * Real beginning of normal "text" segment
diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c
index ad2b15a1334d..73deaffadd03 100644
--- a/arch/x86/kernel/hpet.c
+++ b/arch/x86/kernel/hpet.c
@@ -210,8 +210,8 @@ static void hpet_legacy_clockevent_register(void)
210 /* Calculate the min / max delta */ 210 /* Calculate the min / max delta */
211 hpet_clockevent.max_delta_ns = clockevent_delta2ns(0x7FFFFFFF, 211 hpet_clockevent.max_delta_ns = clockevent_delta2ns(0x7FFFFFFF,
212 &hpet_clockevent); 212 &hpet_clockevent);
213 hpet_clockevent.min_delta_ns = clockevent_delta2ns(0x30, 213 /* 5 usec minimum reprogramming delta. */
214 &hpet_clockevent); 214 hpet_clockevent.min_delta_ns = 5000;
215 215
216 /* 216 /*
217 * Start hpet with the boot cpu mask and make it 217 * Start hpet with the boot cpu mask and make it
@@ -270,15 +270,22 @@ static void hpet_legacy_set_mode(enum clock_event_mode mode,
270} 270}
271 271
272static int hpet_legacy_next_event(unsigned long delta, 272static int hpet_legacy_next_event(unsigned long delta,
273 struct clock_event_device *evt) 273 struct clock_event_device *evt)
274{ 274{
275 unsigned long cnt; 275 u32 cnt;
276 276
277 cnt = hpet_readl(HPET_COUNTER); 277 cnt = hpet_readl(HPET_COUNTER);
278 cnt += delta; 278 cnt += (u32) delta;
279 hpet_writel(cnt, HPET_T0_CMP); 279 hpet_writel(cnt, HPET_T0_CMP);
280 280
281 return ((long)(hpet_readl(HPET_COUNTER) - cnt ) > 0) ? -ETIME : 0; 281 /*
282 * We need to read back the CMP register to make sure that
283 * what we wrote hit the chip before we compare it to the
284 * counter.
285 */
286 WARN_ON((u32)hpet_readl(HPET_T0_CMP) != cnt);
287
288 return (s32)((u32)hpet_readl(HPET_COUNTER) - cnt) >= 0 ? -ETIME : 0;
282} 289}
283 290
284/* 291/*
@@ -359,6 +366,7 @@ static int hpet_clocksource_register(void)
359int __init hpet_enable(void) 366int __init hpet_enable(void)
360{ 367{
361 unsigned long id; 368 unsigned long id;
369 int i;
362 370
363 if (!is_hpet_capable()) 371 if (!is_hpet_capable())
364 return 0; 372 return 0;
@@ -369,6 +377,29 @@ int __init hpet_enable(void)
369 * Read the period and check for a sane value: 377 * Read the period and check for a sane value:
370 */ 378 */
371 hpet_period = hpet_readl(HPET_PERIOD); 379 hpet_period = hpet_readl(HPET_PERIOD);
380
381 /*
382 * AMD SB700 based systems with spread spectrum enabled use a
383 * SMM based HPET emulation to provide proper frequency
384 * setting. The SMM code is initialized with the first HPET
385 * register access and takes some time to complete. During
386 * this time the config register reads 0xffffffff. We check
387 * for max. 1000 loops whether the config register reads a non
388 * 0xffffffff value to make sure that HPET is up and running
389 * before we go further. A counting loop is safe, as the HPET
390 * access takes thousands of CPU cycles. On non SB700 based
391 * machines this check is only done once and has no side
392 * effects.
393 */
394 for (i = 0; hpet_readl(HPET_CFG) == 0xFFFFFFFF; i++) {
395 if (i == 1000) {
396 printk(KERN_WARNING
397 "HPET config register value = 0xFFFFFFFF. "
398 "Disabling HPET\n");
399 goto out_nohpet;
400 }
401 }
402
372 if (hpet_period < HPET_MIN_PERIOD || hpet_period > HPET_MAX_PERIOD) 403 if (hpet_period < HPET_MIN_PERIOD || hpet_period > HPET_MAX_PERIOD)
373 goto out_nohpet; 404 goto out_nohpet;
374 405
diff --git a/arch/x86/kernel/io_apic_32.c b/arch/x86/kernel/io_apic_32.c
index de9aa0e3a9c5..09cddb57bec4 100644
--- a/arch/x86/kernel/io_apic_32.c
+++ b/arch/x86/kernel/io_apic_32.c
@@ -57,7 +57,7 @@ atomic_t irq_mis_count;
57static struct { int pin, apic; } ioapic_i8259 = { -1, -1 }; 57static struct { int pin, apic; } ioapic_i8259 = { -1, -1 };
58 58
59static DEFINE_SPINLOCK(ioapic_lock); 59static DEFINE_SPINLOCK(ioapic_lock);
60static DEFINE_SPINLOCK(vector_lock); 60DEFINE_SPINLOCK(vector_lock);
61 61
62int timer_through_8259 __initdata; 62int timer_through_8259 __initdata;
63 63
@@ -1209,10 +1209,6 @@ static int assign_irq_vector(int irq)
1209 return vector; 1209 return vector;
1210} 1210}
1211 1211
1212void setup_vector_irq(int cpu)
1213{
1214}
1215
1216static struct irq_chip ioapic_chip; 1212static struct irq_chip ioapic_chip;
1217 1213
1218#define IOAPIC_AUTO -1 1214#define IOAPIC_AUTO -1
diff --git a/arch/x86/kernel/io_apic_64.c b/arch/x86/kernel/io_apic_64.c
index 8269434d1707..61a83b70c18f 100644
--- a/arch/x86/kernel/io_apic_64.c
+++ b/arch/x86/kernel/io_apic_64.c
@@ -101,7 +101,7 @@ int timer_through_8259 __initdata;
101static struct { int pin, apic; } ioapic_i8259 = { -1, -1 }; 101static struct { int pin, apic; } ioapic_i8259 = { -1, -1 };
102 102
103static DEFINE_SPINLOCK(ioapic_lock); 103static DEFINE_SPINLOCK(ioapic_lock);
104DEFINE_SPINLOCK(vector_lock); 104static DEFINE_SPINLOCK(vector_lock);
105 105
106/* 106/*
107 * # of IRQ routing registers 107 * # of IRQ routing registers
@@ -697,6 +697,19 @@ static int pin_2_irq(int idx, int apic, int pin)
697 return irq; 697 return irq;
698} 698}
699 699
700void lock_vector_lock(void)
701{
702 /* Used to the online set of cpus does not change
703 * during assign_irq_vector.
704 */
705 spin_lock(&vector_lock);
706}
707
708void unlock_vector_lock(void)
709{
710 spin_unlock(&vector_lock);
711}
712
700static int __assign_irq_vector(int irq, cpumask_t mask) 713static int __assign_irq_vector(int irq, cpumask_t mask)
701{ 714{
702 /* 715 /*
@@ -802,7 +815,7 @@ static void __clear_irq_vector(int irq)
802 cpus_clear(cfg->domain); 815 cpus_clear(cfg->domain);
803} 816}
804 817
805static void __setup_vector_irq(int cpu) 818void __setup_vector_irq(int cpu)
806{ 819{
807 /* Initialize vector_irq on a new cpu */ 820 /* Initialize vector_irq on a new cpu */
808 /* This function must be called with vector_lock held */ 821 /* This function must be called with vector_lock held */
@@ -825,14 +838,6 @@ static void __setup_vector_irq(int cpu)
825 } 838 }
826} 839}
827 840
828void setup_vector_irq(int cpu)
829{
830 spin_lock(&vector_lock);
831 __setup_vector_irq(smp_processor_id());
832 spin_unlock(&vector_lock);
833}
834
835
836static struct irq_chip ioapic_chip; 841static struct irq_chip ioapic_chip;
837 842
838static void ioapic_register_intr(int irq, unsigned long trigger) 843static void ioapic_register_intr(int irq, unsigned long trigger)
diff --git a/arch/x86/kernel/io_delay.c b/arch/x86/kernel/io_delay.c
index 1c3a66a67f83..720d2607aacb 100644
--- a/arch/x86/kernel/io_delay.c
+++ b/arch/x86/kernel/io_delay.c
@@ -92,6 +92,14 @@ static struct dmi_system_id __initdata io_delay_0xed_port_dmi_table[] = {
92 DMI_MATCH(DMI_BOARD_NAME, "30BF") 92 DMI_MATCH(DMI_BOARD_NAME, "30BF")
93 } 93 }
94 }, 94 },
95 {
96 .callback = dmi_io_delay_0xed_port,
97 .ident = "Presario F700",
98 .matches = {
99 DMI_MATCH(DMI_BOARD_VENDOR, "Quanta"),
100 DMI_MATCH(DMI_BOARD_NAME, "30D3")
101 }
102 },
95 { } 103 { }
96}; 104};
97 105
diff --git a/arch/x86/kernel/kdebugfs.c b/arch/x86/kernel/kdebugfs.c
index f2d43bc75514..ff7d3b0124f1 100644
--- a/arch/x86/kernel/kdebugfs.c
+++ b/arch/x86/kernel/kdebugfs.c
@@ -139,6 +139,7 @@ static int __init create_setup_data_nodes(struct dentry *parent)
139 if (PageHighMem(pg)) { 139 if (PageHighMem(pg)) {
140 data = ioremap_cache(pa_data, sizeof(*data)); 140 data = ioremap_cache(pa_data, sizeof(*data));
141 if (!data) { 141 if (!data) {
142 kfree(node);
142 error = -ENXIO; 143 error = -ENXIO;
143 goto err_dir; 144 goto err_dir;
144 } 145 }
diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c
index 43c019f85f0d..6c27679ec6aa 100644
--- a/arch/x86/kernel/kprobes.c
+++ b/arch/x86/kernel/kprobes.c
@@ -431,7 +431,6 @@ static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
431 regs->ip = (unsigned long)p->ainsn.insn; 431 regs->ip = (unsigned long)p->ainsn.insn;
432} 432}
433 433
434/* Called with kretprobe_lock held */
435void __kprobes arch_prepare_kretprobe(struct kretprobe_instance *ri, 434void __kprobes arch_prepare_kretprobe(struct kretprobe_instance *ri,
436 struct pt_regs *regs) 435 struct pt_regs *regs)
437{ 436{
@@ -682,8 +681,7 @@ static __used __kprobes void *trampoline_handler(struct pt_regs *regs)
682 unsigned long trampoline_address = (unsigned long)&kretprobe_trampoline; 681 unsigned long trampoline_address = (unsigned long)&kretprobe_trampoline;
683 682
684 INIT_HLIST_HEAD(&empty_rp); 683 INIT_HLIST_HEAD(&empty_rp);
685 spin_lock_irqsave(&kretprobe_lock, flags); 684 kretprobe_hash_lock(current, &head, &flags);
686 head = kretprobe_inst_table_head(current);
687 /* fixup registers */ 685 /* fixup registers */
688#ifdef CONFIG_X86_64 686#ifdef CONFIG_X86_64
689 regs->cs = __KERNEL_CS; 687 regs->cs = __KERNEL_CS;
@@ -732,7 +730,7 @@ static __used __kprobes void *trampoline_handler(struct pt_regs *regs)
732 730
733 kretprobe_assert(ri, orig_ret_address, trampoline_address); 731 kretprobe_assert(ri, orig_ret_address, trampoline_address);
734 732
735 spin_unlock_irqrestore(&kretprobe_lock, flags); 733 kretprobe_hash_unlock(current, &flags);
736 734
737 hlist_for_each_entry_safe(ri, node, tmp, &empty_rp, hlist) { 735 hlist_for_each_entry_safe(ri, node, tmp, &empty_rp, hlist) {
738 hlist_del(&ri->hlist); 736 hlist_del(&ri->hlist);
diff --git a/arch/x86/kernel/ldt.c b/arch/x86/kernel/ldt.c
index 3fee2aa50f3f..b68e21f06f4f 100644
--- a/arch/x86/kernel/ldt.c
+++ b/arch/x86/kernel/ldt.c
@@ -62,12 +62,10 @@ static int alloc_ldt(mm_context_t *pc, int mincount, int reload)
62 62
63 if (reload) { 63 if (reload) {
64#ifdef CONFIG_SMP 64#ifdef CONFIG_SMP
65 cpumask_of_cpu_ptr_declare(mask);
66
67 preempt_disable(); 65 preempt_disable();
68 load_LDT(pc); 66 load_LDT(pc);
69 cpumask_of_cpu_ptr_next(mask, smp_processor_id()); 67 if (!cpus_equal(current->mm->cpu_vm_mask,
70 if (!cpus_equal(current->mm->cpu_vm_mask, *mask)) 68 cpumask_of_cpu(smp_processor_id())))
71 smp_call_function(flush_ldt, current->mm, 1); 69 smp_call_function(flush_ldt, current->mm, 1);
72 preempt_enable(); 70 preempt_enable();
73#else 71#else
diff --git a/arch/x86/kernel/machine_kexec_32.c b/arch/x86/kernel/machine_kexec_32.c
index 8864230d55af..0732adba05ca 100644
--- a/arch/x86/kernel/machine_kexec_32.c
+++ b/arch/x86/kernel/machine_kexec_32.c
@@ -12,6 +12,7 @@
12#include <linux/init.h> 12#include <linux/init.h>
13#include <linux/numa.h> 13#include <linux/numa.h>
14#include <linux/ftrace.h> 14#include <linux/ftrace.h>
15#include <linux/suspend.h>
15 16
16#include <asm/pgtable.h> 17#include <asm/pgtable.h>
17#include <asm/pgalloc.h> 18#include <asm/pgalloc.h>
@@ -22,6 +23,7 @@
22#include <asm/cpufeature.h> 23#include <asm/cpufeature.h>
23#include <asm/desc.h> 24#include <asm/desc.h>
24#include <asm/system.h> 25#include <asm/system.h>
26#include <asm/cacheflush.h>
25 27
26#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE))) 28#define PAGE_ALIGNED __attribute__ ((__aligned__(PAGE_SIZE)))
27static u32 kexec_pgd[1024] PAGE_ALIGNED; 29static u32 kexec_pgd[1024] PAGE_ALIGNED;
@@ -77,7 +79,7 @@ static void load_segments(void)
77/* 79/*
78 * A architecture hook called to validate the 80 * A architecture hook called to validate the
79 * proposed image and prepare the control pages 81 * proposed image and prepare the control pages
80 * as needed. The pages for KEXEC_CONTROL_CODE_SIZE 82 * as needed. The pages for KEXEC_CONTROL_PAGE_SIZE
81 * have been allocated, but the segments have yet 83 * have been allocated, but the segments have yet
82 * been copied into the kernel. 84 * been copied into the kernel.
83 * 85 *
@@ -85,10 +87,12 @@ static void load_segments(void)
85 * reboot code buffer to allow us to avoid allocations 87 * reboot code buffer to allow us to avoid allocations
86 * later. 88 * later.
87 * 89 *
88 * Currently nothing. 90 * Make control page executable.
89 */ 91 */
90int machine_kexec_prepare(struct kimage *image) 92int machine_kexec_prepare(struct kimage *image)
91{ 93{
94 if (nx_enabled)
95 set_pages_x(image->control_code_page, 1);
92 return 0; 96 return 0;
93} 97}
94 98
@@ -98,27 +102,54 @@ int machine_kexec_prepare(struct kimage *image)
98 */ 102 */
99void machine_kexec_cleanup(struct kimage *image) 103void machine_kexec_cleanup(struct kimage *image)
100{ 104{
105 if (nx_enabled)
106 set_pages_nx(image->control_code_page, 1);
101} 107}
102 108
103/* 109/*
104 * Do not allocate memory (or fail in any way) in machine_kexec(). 110 * Do not allocate memory (or fail in any way) in machine_kexec().
105 * We are past the point of no return, committed to rebooting now. 111 * We are past the point of no return, committed to rebooting now.
106 */ 112 */
107NORET_TYPE void machine_kexec(struct kimage *image) 113void machine_kexec(struct kimage *image)
108{ 114{
109 unsigned long page_list[PAGES_NR]; 115 unsigned long page_list[PAGES_NR];
110 void *control_page; 116 void *control_page;
117 int save_ftrace_enabled;
118 asmlinkage unsigned long
119 (*relocate_kernel_ptr)(unsigned long indirection_page,
120 unsigned long control_page,
121 unsigned long start_address,
122 unsigned int has_pae,
123 unsigned int preserve_context);
124
125#ifdef CONFIG_KEXEC_JUMP
126 if (kexec_image->preserve_context)
127 save_processor_state();
128#endif
111 129
112 tracer_disable(); 130 save_ftrace_enabled = __ftrace_enabled_save();
113 131
114 /* Interrupts aren't acceptable while we reboot */ 132 /* Interrupts aren't acceptable while we reboot */
115 local_irq_disable(); 133 local_irq_disable();
116 134
135 if (image->preserve_context) {
136#ifdef CONFIG_X86_IO_APIC
137 /* We need to put APICs in legacy mode so that we can
138 * get timer interrupts in second kernel. kexec/kdump
139 * paths already have calls to disable_IO_APIC() in
140 * one form or other. kexec jump path also need
141 * one.
142 */
143 disable_IO_APIC();
144#endif
145 }
146
117 control_page = page_address(image->control_code_page); 147 control_page = page_address(image->control_code_page);
118 memcpy(control_page, relocate_kernel, PAGE_SIZE); 148 memcpy(control_page, relocate_kernel, KEXEC_CONTROL_CODE_MAX_SIZE);
119 149
150 relocate_kernel_ptr = control_page;
120 page_list[PA_CONTROL_PAGE] = __pa(control_page); 151 page_list[PA_CONTROL_PAGE] = __pa(control_page);
121 page_list[VA_CONTROL_PAGE] = (unsigned long)relocate_kernel; 152 page_list[VA_CONTROL_PAGE] = (unsigned long)control_page;
122 page_list[PA_PGD] = __pa(kexec_pgd); 153 page_list[PA_PGD] = __pa(kexec_pgd);
123 page_list[VA_PGD] = (unsigned long)kexec_pgd; 154 page_list[VA_PGD] = (unsigned long)kexec_pgd;
124#ifdef CONFIG_X86_PAE 155#ifdef CONFIG_X86_PAE
@@ -131,6 +162,7 @@ NORET_TYPE void machine_kexec(struct kimage *image)
131 page_list[VA_PTE_0] = (unsigned long)kexec_pte0; 162 page_list[VA_PTE_0] = (unsigned long)kexec_pte0;
132 page_list[PA_PTE_1] = __pa(kexec_pte1); 163 page_list[PA_PTE_1] = __pa(kexec_pte1);
133 page_list[VA_PTE_1] = (unsigned long)kexec_pte1; 164 page_list[VA_PTE_1] = (unsigned long)kexec_pte1;
165 page_list[PA_SWAP_PAGE] = (page_to_pfn(image->swap_page) << PAGE_SHIFT);
134 166
135 /* The segment registers are funny things, they have both a 167 /* The segment registers are funny things, they have both a
136 * visible and an invisible part. Whenever the visible part is 168 * visible and an invisible part. Whenever the visible part is
@@ -149,8 +181,17 @@ NORET_TYPE void machine_kexec(struct kimage *image)
149 set_idt(phys_to_virt(0),0); 181 set_idt(phys_to_virt(0),0);
150 182
151 /* now call it */ 183 /* now call it */
152 relocate_kernel((unsigned long)image->head, (unsigned long)page_list, 184 image->start = relocate_kernel_ptr((unsigned long)image->head,
153 image->start, cpu_has_pae); 185 (unsigned long)page_list,
186 image->start, cpu_has_pae,
187 image->preserve_context);
188
189#ifdef CONFIG_KEXEC_JUMP
190 if (kexec_image->preserve_context)
191 restore_processor_state();
192#endif
193
194 __ftrace_enabled_restore(save_ftrace_enabled);
154} 195}
155 196
156void arch_crash_save_vmcoreinfo(void) 197void arch_crash_save_vmcoreinfo(void)
diff --git a/arch/x86/kernel/machine_kexec_64.c b/arch/x86/kernel/machine_kexec_64.c
index 9dd9262693a3..c43caa3a91f3 100644
--- a/arch/x86/kernel/machine_kexec_64.c
+++ b/arch/x86/kernel/machine_kexec_64.c
@@ -181,7 +181,7 @@ void machine_kexec_cleanup(struct kimage *image)
181 * Do not allocate memory (or fail in any way) in machine_kexec(). 181 * Do not allocate memory (or fail in any way) in machine_kexec().
182 * We are past the point of no return, committed to rebooting now. 182 * We are past the point of no return, committed to rebooting now.
183 */ 183 */
184NORET_TYPE void machine_kexec(struct kimage *image) 184void machine_kexec(struct kimage *image)
185{ 185{
186 unsigned long page_list[PAGES_NR]; 186 unsigned long page_list[PAGES_NR];
187 void *control_page; 187 void *control_page;
diff --git a/arch/x86/kernel/mfgpt_32.c b/arch/x86/kernel/mfgpt_32.c
index 07c0f828f488..3b599518c322 100644
--- a/arch/x86/kernel/mfgpt_32.c
+++ b/arch/x86/kernel/mfgpt_32.c
@@ -33,6 +33,8 @@
33#include <linux/module.h> 33#include <linux/module.h>
34#include <asm/geode.h> 34#include <asm/geode.h>
35 35
36#define MFGPT_DEFAULT_IRQ 7
37
36static struct mfgpt_timer_t { 38static struct mfgpt_timer_t {
37 unsigned int avail:1; 39 unsigned int avail:1;
38} mfgpt_timers[MFGPT_MAX_TIMERS]; 40} mfgpt_timers[MFGPT_MAX_TIMERS];
@@ -157,29 +159,48 @@ int geode_mfgpt_toggle_event(int timer, int cmp, int event, int enable)
157} 159}
158EXPORT_SYMBOL_GPL(geode_mfgpt_toggle_event); 160EXPORT_SYMBOL_GPL(geode_mfgpt_toggle_event);
159 161
160int geode_mfgpt_set_irq(int timer, int cmp, int irq, int enable) 162int geode_mfgpt_set_irq(int timer, int cmp, int *irq, int enable)
161{ 163{
162 u32 val, dummy; 164 u32 zsel, lpc, dummy;
163 int offset; 165 int shift;
164 166
165 if (timer < 0 || timer >= MFGPT_MAX_TIMERS) 167 if (timer < 0 || timer >= MFGPT_MAX_TIMERS)
166 return -EIO; 168 return -EIO;
167 169
168 if (geode_mfgpt_toggle_event(timer, cmp, MFGPT_EVENT_IRQ, enable)) 170 /*
171 * Unfortunately, MFGPTs come in pairs sharing their IRQ lines. If VSA
172 * is using the same CMP of the timer's Siamese twin, the IRQ is set to
173 * 2, and we mustn't use nor change it.
174 * XXX: Likewise, 2 Linux drivers might clash if the 2nd overwrites the
175 * IRQ of the 1st. This can only happen if forcing an IRQ, calling this
176 * with *irq==0 is safe. Currently there _are_ no 2 drivers.
177 */
178 rdmsr(MSR_PIC_ZSEL_LOW, zsel, dummy);
179 shift = ((cmp == MFGPT_CMP1 ? 0 : 4) + timer % 4) * 4;
180 if (((zsel >> shift) & 0xF) == 2)
169 return -EIO; 181 return -EIO;
170 182
171 rdmsr(MSR_PIC_ZSEL_LOW, val, dummy); 183 /* Choose IRQ: if none supplied, keep IRQ already set or use default */
184 if (!*irq)
185 *irq = (zsel >> shift) & 0xF;
186 if (!*irq)
187 *irq = MFGPT_DEFAULT_IRQ;
172 188
173 offset = (timer % 4) * 4; 189 /* Can't use IRQ if it's 0 (=disabled), 2, or routed to LPC */
174 190 if (*irq < 1 || *irq == 2 || *irq > 15)
175 val &= ~((0xF << offset) | (0xF << (offset + 16))); 191 return -EIO;
192 rdmsr(MSR_PIC_IRQM_LPC, lpc, dummy);
193 if (lpc & (1 << *irq))
194 return -EIO;
176 195
196 /* All chosen and checked - go for it */
197 if (geode_mfgpt_toggle_event(timer, cmp, MFGPT_EVENT_IRQ, enable))
198 return -EIO;
177 if (enable) { 199 if (enable) {
178 val |= (irq & 0x0F) << (offset); 200 zsel = (zsel & ~(0xF << shift)) | (*irq << shift);
179 val |= (irq & 0x0F) << (offset + 16); 201 wrmsr(MSR_PIC_ZSEL_LOW, zsel, dummy);
180 } 202 }
181 203
182 wrmsr(MSR_PIC_ZSEL_LOW, val, dummy);
183 return 0; 204 return 0;
184} 205}
185 206
@@ -242,7 +263,7 @@ EXPORT_SYMBOL_GPL(geode_mfgpt_alloc_timer);
242static unsigned int mfgpt_tick_mode = CLOCK_EVT_MODE_SHUTDOWN; 263static unsigned int mfgpt_tick_mode = CLOCK_EVT_MODE_SHUTDOWN;
243static u16 mfgpt_event_clock; 264static u16 mfgpt_event_clock;
244 265
245static int irq = 7; 266static int irq;
246static int __init mfgpt_setup(char *str) 267static int __init mfgpt_setup(char *str)
247{ 268{
248 get_option(&str, &irq); 269 get_option(&str, &irq);
@@ -346,7 +367,7 @@ int __init mfgpt_timer_setup(void)
346 mfgpt_event_clock = timer; 367 mfgpt_event_clock = timer;
347 368
348 /* Set up the IRQ on the MFGPT side */ 369 /* Set up the IRQ on the MFGPT side */
349 if (geode_mfgpt_setup_irq(mfgpt_event_clock, MFGPT_CMP2, irq)) { 370 if (geode_mfgpt_setup_irq(mfgpt_event_clock, MFGPT_CMP2, &irq)) {
350 printk(KERN_ERR "mfgpt-timer: Could not set up IRQ %d\n", irq); 371 printk(KERN_ERR "mfgpt-timer: Could not set up IRQ %d\n", irq);
351 return -EIO; 372 return -EIO;
352 } 373 }
@@ -374,13 +395,14 @@ int __init mfgpt_timer_setup(void)
374 &mfgpt_clockevent); 395 &mfgpt_clockevent);
375 396
376 printk(KERN_INFO 397 printk(KERN_INFO
377 "mfgpt-timer: registering the MFGPT timer as a clock event.\n"); 398 "mfgpt-timer: Registering MFGPT timer %d as a clock event, using IRQ %d\n",
399 timer, irq);
378 clockevents_register_device(&mfgpt_clockevent); 400 clockevents_register_device(&mfgpt_clockevent);
379 401
380 return 0; 402 return 0;
381 403
382err: 404err:
383 geode_mfgpt_release_irq(mfgpt_event_clock, MFGPT_CMP2, irq); 405 geode_mfgpt_release_irq(mfgpt_event_clock, MFGPT_CMP2, &irq);
384 printk(KERN_ERR 406 printk(KERN_ERR
385 "mfgpt-timer: Unable to set up the MFGPT clock source\n"); 407 "mfgpt-timer: Unable to set up the MFGPT clock source\n");
386 return -EIO; 408 return -EIO;
diff --git a/arch/x86/kernel/microcode.c b/arch/x86/kernel/microcode.c
index 6994c751590e..652fa5c38ebe 100644
--- a/arch/x86/kernel/microcode.c
+++ b/arch/x86/kernel/microcode.c
@@ -388,7 +388,6 @@ static int do_microcode_update (void)
388 void *new_mc = NULL; 388 void *new_mc = NULL;
389 int cpu; 389 int cpu;
390 cpumask_t old; 390 cpumask_t old;
391 cpumask_of_cpu_ptr_declare(newmask);
392 391
393 old = current->cpus_allowed; 392 old = current->cpus_allowed;
394 393
@@ -405,8 +404,7 @@ static int do_microcode_update (void)
405 404
406 if (!uci->valid) 405 if (!uci->valid)
407 continue; 406 continue;
408 cpumask_of_cpu_ptr_next(newmask, cpu); 407 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
409 set_cpus_allowed_ptr(current, newmask);
410 error = get_maching_microcode(new_mc, cpu); 408 error = get_maching_microcode(new_mc, cpu);
411 if (error < 0) 409 if (error < 0)
412 goto out; 410 goto out;
@@ -576,7 +574,6 @@ static int apply_microcode_check_cpu(int cpu)
576 struct cpuinfo_x86 *c = &cpu_data(cpu); 574 struct cpuinfo_x86 *c = &cpu_data(cpu);
577 struct ucode_cpu_info *uci = ucode_cpu_info + cpu; 575 struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
578 cpumask_t old; 576 cpumask_t old;
579 cpumask_of_cpu_ptr(newmask, cpu);
580 unsigned int val[2]; 577 unsigned int val[2];
581 int err = 0; 578 int err = 0;
582 579
@@ -585,7 +582,7 @@ static int apply_microcode_check_cpu(int cpu)
585 return 0; 582 return 0;
586 583
587 old = current->cpus_allowed; 584 old = current->cpus_allowed;
588 set_cpus_allowed_ptr(current, newmask); 585 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
589 586
590 /* Check if the microcode we have in memory matches the CPU */ 587 /* Check if the microcode we have in memory matches the CPU */
591 if (c->x86_vendor != X86_VENDOR_INTEL || c->x86 < 6 || 588 if (c->x86_vendor != X86_VENDOR_INTEL || c->x86 < 6 ||
@@ -623,12 +620,11 @@ static int apply_microcode_check_cpu(int cpu)
623static void microcode_init_cpu(int cpu, int resume) 620static void microcode_init_cpu(int cpu, int resume)
624{ 621{
625 cpumask_t old; 622 cpumask_t old;
626 cpumask_of_cpu_ptr(newmask, cpu);
627 struct ucode_cpu_info *uci = ucode_cpu_info + cpu; 623 struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
628 624
629 old = current->cpus_allowed; 625 old = current->cpus_allowed;
630 626
631 set_cpus_allowed_ptr(current, newmask); 627 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
632 mutex_lock(&microcode_mutex); 628 mutex_lock(&microcode_mutex);
633 collect_cpu_info(cpu); 629 collect_cpu_info(cpu);
634 if (uci->valid && system_state == SYSTEM_RUNNING && !resume) 630 if (uci->valid && system_state == SYSTEM_RUNNING && !resume)
@@ -661,13 +657,10 @@ static ssize_t reload_store(struct sys_device *dev,
661 if (end == buf) 657 if (end == buf)
662 return -EINVAL; 658 return -EINVAL;
663 if (val == 1) { 659 if (val == 1) {
664 cpumask_t old; 660 cpumask_t old = current->cpus_allowed;
665 cpumask_of_cpu_ptr(newmask, cpu);
666
667 old = current->cpus_allowed;
668 661
669 get_online_cpus(); 662 get_online_cpus();
670 set_cpus_allowed_ptr(current, newmask); 663 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
671 664
672 mutex_lock(&microcode_mutex); 665 mutex_lock(&microcode_mutex);
673 if (uci->valid) 666 if (uci->valid)
diff --git a/arch/x86/kernel/mmconf-fam10h_64.c b/arch/x86/kernel/mmconf-fam10h_64.c
index fdfdc550b366..efc2f361fe85 100644
--- a/arch/x86/kernel/mmconf-fam10h_64.c
+++ b/arch/x86/kernel/mmconf-fam10h_64.c
@@ -238,7 +238,7 @@ static struct dmi_system_id __devinitdata mmconf_dmi_table[] = {
238 {} 238 {}
239}; 239};
240 240
241void __init check_enable_amd_mmconf_dmi(void) 241void __cpuinit check_enable_amd_mmconf_dmi(void)
242{ 242{
243 dmi_check_system(mmconf_dmi_table); 243 dmi_check_system(mmconf_dmi_table);
244} 244}
diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c
index 6ae005ccaed8..b3fb430725cb 100644
--- a/arch/x86/kernel/mpparse.c
+++ b/arch/x86/kernel/mpparse.c
@@ -49,7 +49,7 @@ static int __init mpf_checksum(unsigned char *mp, int len)
49 return sum & 0xFF; 49 return sum & 0xFF;
50} 50}
51 51
52static void __cpuinit MP_processor_info(struct mpc_config_processor *m) 52static void __init MP_processor_info(struct mpc_config_processor *m)
53{ 53{
54 int apicid; 54 int apicid;
55 char *bootup_cpu = ""; 55 char *bootup_cpu = "";
@@ -83,7 +83,7 @@ static void __init MP_bus_info(struct mpc_config_bus *m)
83 if (x86_quirks->mpc_oem_bus_info) 83 if (x86_quirks->mpc_oem_bus_info)
84 x86_quirks->mpc_oem_bus_info(m, str); 84 x86_quirks->mpc_oem_bus_info(m, str);
85 else 85 else
86 printk(KERN_INFO "Bus #%d is %s\n", m->mpc_busid, str); 86 apic_printk(APIC_VERBOSE, "Bus #%d is %s\n", m->mpc_busid, str);
87 87
88#if MAX_MP_BUSSES < 256 88#if MAX_MP_BUSSES < 256
89 if (m->mpc_busid >= MAX_MP_BUSSES) { 89 if (m->mpc_busid >= MAX_MP_BUSSES) {
@@ -154,7 +154,7 @@ static void __init MP_ioapic_info(struct mpc_config_ioapic *m)
154 154
155static void print_MP_intsrc_info(struct mpc_config_intsrc *m) 155static void print_MP_intsrc_info(struct mpc_config_intsrc *m)
156{ 156{
157 printk(KERN_CONT "Int: type %d, pol %d, trig %d, bus %02x," 157 apic_printk(APIC_VERBOSE, "Int: type %d, pol %d, trig %d, bus %02x,"
158 " IRQ %02x, APIC ID %x, APIC INT %02x\n", 158 " IRQ %02x, APIC ID %x, APIC INT %02x\n",
159 m->mpc_irqtype, m->mpc_irqflag & 3, 159 m->mpc_irqtype, m->mpc_irqflag & 3,
160 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbus, 160 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbus,
@@ -163,7 +163,7 @@ static void print_MP_intsrc_info(struct mpc_config_intsrc *m)
163 163
164static void __init print_mp_irq_info(struct mp_config_intsrc *mp_irq) 164static void __init print_mp_irq_info(struct mp_config_intsrc *mp_irq)
165{ 165{
166 printk(KERN_CONT "Int: type %d, pol %d, trig %d, bus %02x," 166 apic_printk(APIC_VERBOSE, "Int: type %d, pol %d, trig %d, bus %02x,"
167 " IRQ %02x, APIC ID %x, APIC INT %02x\n", 167 " IRQ %02x, APIC ID %x, APIC INT %02x\n",
168 mp_irq->mp_irqtype, mp_irq->mp_irqflag & 3, 168 mp_irq->mp_irqtype, mp_irq->mp_irqflag & 3,
169 (mp_irq->mp_irqflag >> 2) & 3, mp_irq->mp_srcbus, 169 (mp_irq->mp_irqflag >> 2) & 3, mp_irq->mp_srcbus,
@@ -235,7 +235,7 @@ static void __init MP_intsrc_info(struct mpc_config_intsrc *m)
235 235
236static void __init MP_lintsrc_info(struct mpc_config_lintsrc *m) 236static void __init MP_lintsrc_info(struct mpc_config_lintsrc *m)
237{ 237{
238 printk(KERN_INFO "Lint: type %d, pol %d, trig %d, bus %02x," 238 apic_printk(APIC_VERBOSE, "Lint: type %d, pol %d, trig %d, bus %02x,"
239 " IRQ %02x, APIC ID %x, APIC LINT %02x\n", 239 " IRQ %02x, APIC ID %x, APIC LINT %02x\n",
240 m->mpc_irqtype, m->mpc_irqflag & 3, 240 m->mpc_irqtype, m->mpc_irqflag & 3,
241 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbusid, 241 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbusid,
@@ -484,7 +484,7 @@ static void __init construct_default_ioirq_mptable(int mpc_default_type)
484} 484}
485 485
486 486
487static void construct_ioapic_table(int mpc_default_type) 487static void __init construct_ioapic_table(int mpc_default_type)
488{ 488{
489 struct mpc_config_ioapic ioapic; 489 struct mpc_config_ioapic ioapic;
490 struct mpc_config_bus bus; 490 struct mpc_config_bus bus;
@@ -529,7 +529,7 @@ static void construct_ioapic_table(int mpc_default_type)
529 construct_default_ioirq_mptable(mpc_default_type); 529 construct_default_ioirq_mptable(mpc_default_type);
530} 530}
531#else 531#else
532static inline void construct_ioapic_table(int mpc_default_type) { } 532static inline void __init construct_ioapic_table(int mpc_default_type) { }
533#endif 533#endif
534 534
535static inline void __init construct_default_ISA_mptable(int mpc_default_type) 535static inline void __init construct_default_ISA_mptable(int mpc_default_type)
@@ -695,7 +695,8 @@ static int __init smp_scan_config(unsigned long base, unsigned long length,
695 unsigned int *bp = phys_to_virt(base); 695 unsigned int *bp = phys_to_virt(base);
696 struct intel_mp_floating *mpf; 696 struct intel_mp_floating *mpf;
697 697
698 printk(KERN_DEBUG "Scan SMP from %p for %ld bytes.\n", bp, length); 698 apic_printk(APIC_VERBOSE, "Scan SMP from %p for %ld bytes.\n",
699 bp, length);
699 BUILD_BUG_ON(sizeof(*mpf) != 16); 700 BUILD_BUG_ON(sizeof(*mpf) != 16);
700 701
701 while (length > 0) { 702 while (length > 0) {
diff --git a/arch/x86/kernel/msr.c b/arch/x86/kernel/msr.c
index 9fd809552447..2e2af5d18191 100644
--- a/arch/x86/kernel/msr.c
+++ b/arch/x86/kernel/msr.c
@@ -72,21 +72,28 @@ static ssize_t msr_read(struct file *file, char __user *buf,
72 u32 data[2]; 72 u32 data[2];
73 u32 reg = *ppos; 73 u32 reg = *ppos;
74 int cpu = iminor(file->f_path.dentry->d_inode); 74 int cpu = iminor(file->f_path.dentry->d_inode);
75 int err; 75 int err = 0;
76 ssize_t bytes = 0;
76 77
77 if (count % 8) 78 if (count % 8)
78 return -EINVAL; /* Invalid chunk size */ 79 return -EINVAL; /* Invalid chunk size */
79 80
80 for (; count; count -= 8) { 81 for (; count; count -= 8) {
81 err = rdmsr_safe_on_cpu(cpu, reg, &data[0], &data[1]); 82 err = rdmsr_safe_on_cpu(cpu, reg, &data[0], &data[1]);
82 if (err) 83 if (err) {
83 return -EIO; 84 if (err == -EFAULT) /* Fix idiotic error code */
84 if (copy_to_user(tmp, &data, 8)) 85 err = -EIO;
85 return -EFAULT; 86 break;
87 }
88 if (copy_to_user(tmp, &data, 8)) {
89 err = -EFAULT;
90 break;
91 }
86 tmp += 2; 92 tmp += 2;
93 bytes += 8;
87 } 94 }
88 95
89 return ((char __user *)tmp) - buf; 96 return bytes ? bytes : err;
90} 97}
91 98
92static ssize_t msr_write(struct file *file, const char __user *buf, 99static ssize_t msr_write(struct file *file, const char __user *buf,
@@ -96,21 +103,28 @@ static ssize_t msr_write(struct file *file, const char __user *buf,
96 u32 data[2]; 103 u32 data[2];
97 u32 reg = *ppos; 104 u32 reg = *ppos;
98 int cpu = iminor(file->f_path.dentry->d_inode); 105 int cpu = iminor(file->f_path.dentry->d_inode);
99 int err; 106 int err = 0;
107 ssize_t bytes = 0;
100 108
101 if (count % 8) 109 if (count % 8)
102 return -EINVAL; /* Invalid chunk size */ 110 return -EINVAL; /* Invalid chunk size */
103 111
104 for (; count; count -= 8) { 112 for (; count; count -= 8) {
105 if (copy_from_user(&data, tmp, 8)) 113 if (copy_from_user(&data, tmp, 8)) {
106 return -EFAULT; 114 err = -EFAULT;
115 break;
116 }
107 err = wrmsr_safe_on_cpu(cpu, reg, data[0], data[1]); 117 err = wrmsr_safe_on_cpu(cpu, reg, data[0], data[1]);
108 if (err) 118 if (err) {
109 return -EIO; 119 if (err == -EFAULT) /* Fix idiotic error code */
120 err = -EIO;
121 break;
122 }
110 tmp += 2; 123 tmp += 2;
124 bytes += 8;
111 } 125 }
112 126
113 return ((char __user *)tmp) - buf; 127 return bytes ? bytes : err;
114} 128}
115 129
116static int msr_open(struct inode *inode, struct file *file) 130static int msr_open(struct inode *inode, struct file *file)
@@ -131,7 +145,7 @@ static int msr_open(struct inode *inode, struct file *file)
131 ret = -EIO; /* MSR not supported */ 145 ret = -EIO; /* MSR not supported */
132out: 146out:
133 unlock_kernel(); 147 unlock_kernel();
134 return 0; 148 return ret;
135} 149}
136 150
137/* 151/*
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
index ac6d51222e7d..abb78a2cc4ad 100644
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -114,6 +114,23 @@ static __init void nmi_cpu_busy(void *data)
114} 114}
115#endif 115#endif
116 116
117static void report_broken_nmi(int cpu, int *prev_nmi_count)
118{
119 printk(KERN_CONT "\n");
120
121 printk(KERN_WARNING
122 "WARNING: CPU#%d: NMI appears to be stuck (%d->%d)!\n",
123 cpu, prev_nmi_count[cpu], get_nmi_count(cpu));
124
125 printk(KERN_WARNING
126 "Please report this to bugzilla.kernel.org,\n");
127 printk(KERN_WARNING
128 "and attach the output of the 'dmesg' command.\n");
129
130 per_cpu(wd_enabled, cpu) = 0;
131 atomic_dec(&nmi_active);
132}
133
117int __init check_nmi_watchdog(void) 134int __init check_nmi_watchdog(void)
118{ 135{
119 unsigned int *prev_nmi_count; 136 unsigned int *prev_nmi_count;
@@ -141,15 +158,8 @@ int __init check_nmi_watchdog(void)
141 for_each_online_cpu(cpu) { 158 for_each_online_cpu(cpu) {
142 if (!per_cpu(wd_enabled, cpu)) 159 if (!per_cpu(wd_enabled, cpu))
143 continue; 160 continue;
144 if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5) { 161 if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5)
145 printk(KERN_WARNING "WARNING: CPU#%d: NMI " 162 report_broken_nmi(cpu, prev_nmi_count);
146 "appears to be stuck (%d->%d)!\n",
147 cpu,
148 prev_nmi_count[cpu],
149 get_nmi_count(cpu));
150 per_cpu(wd_enabled, cpu) = 0;
151 atomic_dec(&nmi_active);
152 }
153 } 163 }
154 endflag = 1; 164 endflag = 1;
155 if (!atomic_read(&nmi_active)) { 165 if (!atomic_read(&nmi_active)) {
diff --git a/arch/x86/kernel/numaq_32.c b/arch/x86/kernel/numaq_32.c
index b8c45610b20a..eecc8c18f010 100644
--- a/arch/x86/kernel/numaq_32.c
+++ b/arch/x86/kernel/numaq_32.c
@@ -73,7 +73,7 @@ static void __init smp_dump_qct(void)
73} 73}
74 74
75 75
76void __init numaq_tsc_disable(void) 76void __cpuinit numaq_tsc_disable(void)
77{ 77{
78 if (!found_numaq) 78 if (!found_numaq)
79 return; 79 return;
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index 94da4d52d798..300da17e61cb 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -471,7 +471,7 @@ struct pv_lock_ops pv_lock_ops = {
471 .spin_unlock = __ticket_spin_unlock, 471 .spin_unlock = __ticket_spin_unlock,
472#endif 472#endif
473}; 473};
474EXPORT_SYMBOL_GPL(pv_lock_ops); 474EXPORT_SYMBOL(pv_lock_ops);
475 475
476EXPORT_SYMBOL_GPL(pv_time_ops); 476EXPORT_SYMBOL_GPL(pv_time_ops);
477EXPORT_SYMBOL (pv_cpu_ops); 477EXPORT_SYMBOL (pv_cpu_ops);
diff --git a/arch/x86/kernel/pci-calgary_64.c b/arch/x86/kernel/pci-calgary_64.c
index 151f2d171f7c..dcdac6c826e9 100644
--- a/arch/x86/kernel/pci-calgary_64.c
+++ b/arch/x86/kernel/pci-calgary_64.c
@@ -29,6 +29,7 @@
29#include <linux/mm.h> 29#include <linux/mm.h>
30#include <linux/spinlock.h> 30#include <linux/spinlock.h>
31#include <linux/string.h> 31#include <linux/string.h>
32#include <linux/crash_dump.h>
32#include <linux/dma-mapping.h> 33#include <linux/dma-mapping.h>
33#include <linux/bitops.h> 34#include <linux/bitops.h>
34#include <linux/pci_ids.h> 35#include <linux/pci_ids.h>
@@ -36,6 +37,7 @@
36#include <linux/delay.h> 37#include <linux/delay.h>
37#include <linux/scatterlist.h> 38#include <linux/scatterlist.h>
38#include <linux/iommu-helper.h> 39#include <linux/iommu-helper.h>
40
39#include <asm/iommu.h> 41#include <asm/iommu.h>
40#include <asm/calgary.h> 42#include <asm/calgary.h>
41#include <asm/tce.h> 43#include <asm/tce.h>
@@ -167,6 +169,8 @@ static void calgary_dump_error_regs(struct iommu_table *tbl);
167static void calioc2_handle_quirks(struct iommu_table *tbl, struct pci_dev *dev); 169static void calioc2_handle_quirks(struct iommu_table *tbl, struct pci_dev *dev);
168static void calioc2_tce_cache_blast(struct iommu_table *tbl); 170static void calioc2_tce_cache_blast(struct iommu_table *tbl);
169static void calioc2_dump_error_regs(struct iommu_table *tbl); 171static void calioc2_dump_error_regs(struct iommu_table *tbl);
172static void calgary_init_bitmap_from_tce_table(struct iommu_table *tbl);
173static void get_tce_space_from_tar(void);
170 174
171static struct cal_chipset_ops calgary_chip_ops = { 175static struct cal_chipset_ops calgary_chip_ops = {
172 .handle_quirks = calgary_handle_quirks, 176 .handle_quirks = calgary_handle_quirks,
@@ -339,9 +343,8 @@ static void iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr,
339 /* were we called with bad_dma_address? */ 343 /* were we called with bad_dma_address? */
340 badend = bad_dma_address + (EMERGENCY_PAGES * PAGE_SIZE); 344 badend = bad_dma_address + (EMERGENCY_PAGES * PAGE_SIZE);
341 if (unlikely((dma_addr >= bad_dma_address) && (dma_addr < badend))) { 345 if (unlikely((dma_addr >= bad_dma_address) && (dma_addr < badend))) {
342 printk(KERN_ERR "Calgary: driver tried unmapping bad DMA " 346 WARN(1, KERN_ERR "Calgary: driver tried unmapping bad DMA "
343 "address 0x%Lx\n", dma_addr); 347 "address 0x%Lx\n", dma_addr);
344 WARN_ON(1);
345 return; 348 return;
346 } 349 }
347 350
@@ -410,22 +413,6 @@ static void calgary_unmap_sg(struct device *dev,
410 } 413 }
411} 414}
412 415
413static int calgary_nontranslate_map_sg(struct device* dev,
414 struct scatterlist *sg, int nelems, int direction)
415{
416 struct scatterlist *s;
417 int i;
418
419 for_each_sg(sg, s, nelems, i) {
420 struct page *p = sg_page(s);
421
422 BUG_ON(!p);
423 s->dma_address = virt_to_bus(sg_virt(s));
424 s->dma_length = s->length;
425 }
426 return nelems;
427}
428
429static int calgary_map_sg(struct device *dev, struct scatterlist *sg, 416static int calgary_map_sg(struct device *dev, struct scatterlist *sg,
430 int nelems, int direction) 417 int nelems, int direction)
431{ 418{
@@ -436,9 +423,6 @@ static int calgary_map_sg(struct device *dev, struct scatterlist *sg,
436 unsigned long entry; 423 unsigned long entry;
437 int i; 424 int i;
438 425
439 if (!translation_enabled(tbl))
440 return calgary_nontranslate_map_sg(dev, sg, nelems, direction);
441
442 for_each_sg(sg, s, nelems, i) { 426 for_each_sg(sg, s, nelems, i) {
443 BUG_ON(!sg_page(s)); 427 BUG_ON(!sg_page(s));
444 428
@@ -474,7 +458,6 @@ error:
474static dma_addr_t calgary_map_single(struct device *dev, phys_addr_t paddr, 458static dma_addr_t calgary_map_single(struct device *dev, phys_addr_t paddr,
475 size_t size, int direction) 459 size_t size, int direction)
476{ 460{
477 dma_addr_t dma_handle = bad_dma_address;
478 void *vaddr = phys_to_virt(paddr); 461 void *vaddr = phys_to_virt(paddr);
479 unsigned long uaddr; 462 unsigned long uaddr;
480 unsigned int npages; 463 unsigned int npages;
@@ -483,12 +466,7 @@ static dma_addr_t calgary_map_single(struct device *dev, phys_addr_t paddr,
483 uaddr = (unsigned long)vaddr; 466 uaddr = (unsigned long)vaddr;
484 npages = num_dma_pages(uaddr, size); 467 npages = num_dma_pages(uaddr, size);
485 468
486 if (translation_enabled(tbl)) 469 return iommu_alloc(dev, tbl, vaddr, npages, direction);
487 dma_handle = iommu_alloc(dev, tbl, vaddr, npages, direction);
488 else
489 dma_handle = virt_to_bus(vaddr);
490
491 return dma_handle;
492} 470}
493 471
494static void calgary_unmap_single(struct device *dev, dma_addr_t dma_handle, 472static void calgary_unmap_single(struct device *dev, dma_addr_t dma_handle,
@@ -497,9 +475,6 @@ static void calgary_unmap_single(struct device *dev, dma_addr_t dma_handle,
497 struct iommu_table *tbl = find_iommu_table(dev); 475 struct iommu_table *tbl = find_iommu_table(dev);
498 unsigned int npages; 476 unsigned int npages;
499 477
500 if (!translation_enabled(tbl))
501 return;
502
503 npages = num_dma_pages(dma_handle, size); 478 npages = num_dma_pages(dma_handle, size);
504 iommu_free(tbl, dma_handle, npages); 479 iommu_free(tbl, dma_handle, npages);
505} 480}
@@ -522,18 +497,12 @@ static void* calgary_alloc_coherent(struct device *dev, size_t size,
522 goto error; 497 goto error;
523 memset(ret, 0, size); 498 memset(ret, 0, size);
524 499
525 if (translation_enabled(tbl)) { 500 /* set up tces to cover the allocated range */
526 /* set up tces to cover the allocated range */ 501 mapping = iommu_alloc(dev, tbl, ret, npages, DMA_BIDIRECTIONAL);
527 mapping = iommu_alloc(dev, tbl, ret, npages, DMA_BIDIRECTIONAL); 502 if (mapping == bad_dma_address)
528 if (mapping == bad_dma_address) 503 goto free;
529 goto free; 504 *dma_handle = mapping;
530
531 *dma_handle = mapping;
532 } else /* non translated slot */
533 *dma_handle = virt_to_bus(ret);
534
535 return ret; 505 return ret;
536
537free: 506free:
538 free_pages((unsigned long)ret, get_order(size)); 507 free_pages((unsigned long)ret, get_order(size));
539 ret = NULL; 508 ret = NULL;
@@ -541,7 +510,7 @@ error:
541 return ret; 510 return ret;
542} 511}
543 512
544static const struct dma_mapping_ops calgary_dma_ops = { 513static struct dma_mapping_ops calgary_dma_ops = {
545 .alloc_coherent = calgary_alloc_coherent, 514 .alloc_coherent = calgary_alloc_coherent,
546 .map_single = calgary_map_single, 515 .map_single = calgary_map_single,
547 .unmap_single = calgary_unmap_single, 516 .unmap_single = calgary_unmap_single,
@@ -830,7 +799,11 @@ static int __init calgary_setup_tar(struct pci_dev *dev, void __iomem *bbar)
830 799
831 tbl = pci_iommu(dev->bus); 800 tbl = pci_iommu(dev->bus);
832 tbl->it_base = (unsigned long)bus_info[dev->bus->number].tce_space; 801 tbl->it_base = (unsigned long)bus_info[dev->bus->number].tce_space;
833 tce_free(tbl, 0, tbl->it_size); 802
803 if (is_kdump_kernel())
804 calgary_init_bitmap_from_tce_table(tbl);
805 else
806 tce_free(tbl, 0, tbl->it_size);
834 807
835 if (is_calgary(dev->device)) 808 if (is_calgary(dev->device))
836 tbl->chip_ops = &calgary_chip_ops; 809 tbl->chip_ops = &calgary_chip_ops;
@@ -1209,6 +1182,10 @@ static int __init calgary_init(void)
1209 if (ret) 1182 if (ret)
1210 return ret; 1183 return ret;
1211 1184
1185 /* Purely for kdump kernel case */
1186 if (is_kdump_kernel())
1187 get_tce_space_from_tar();
1188
1212 do { 1189 do {
1213 dev = pci_get_device(PCI_VENDOR_ID_IBM, PCI_ANY_ID, dev); 1190 dev = pci_get_device(PCI_VENDOR_ID_IBM, PCI_ANY_ID, dev);
1214 if (!dev) 1191 if (!dev)
@@ -1230,6 +1207,16 @@ static int __init calgary_init(void)
1230 goto error; 1207 goto error;
1231 } while (1); 1208 } while (1);
1232 1209
1210 dev = NULL;
1211 for_each_pci_dev(dev) {
1212 struct iommu_table *tbl;
1213
1214 tbl = find_iommu_table(&dev->dev);
1215
1216 if (translation_enabled(tbl))
1217 dev->dev.archdata.dma_ops = &calgary_dma_ops;
1218 }
1219
1233 return ret; 1220 return ret;
1234 1221
1235error: 1222error:
@@ -1251,6 +1238,7 @@ error:
1251 calgary_disable_translation(dev); 1238 calgary_disable_translation(dev);
1252 calgary_free_bus(dev); 1239 calgary_free_bus(dev);
1253 pci_dev_put(dev); /* Undo calgary_init_one()'s pci_dev_get() */ 1240 pci_dev_put(dev); /* Undo calgary_init_one()'s pci_dev_get() */
1241 dev->dev.archdata.dma_ops = NULL;
1254 } while (1); 1242 } while (1);
1255 1243
1256 return ret; 1244 return ret;
@@ -1280,13 +1268,15 @@ static inline int __init determine_tce_table_size(u64 ram)
1280static int __init build_detail_arrays(void) 1268static int __init build_detail_arrays(void)
1281{ 1269{
1282 unsigned long ptr; 1270 unsigned long ptr;
1283 int i, scal_detail_size, rio_detail_size; 1271 unsigned numnodes, i;
1272 int scal_detail_size, rio_detail_size;
1284 1273
1285 if (rio_table_hdr->num_scal_dev > MAX_NUMNODES){ 1274 numnodes = rio_table_hdr->num_scal_dev;
1275 if (numnodes > MAX_NUMNODES){
1286 printk(KERN_WARNING 1276 printk(KERN_WARNING
1287 "Calgary: MAX_NUMNODES too low! Defined as %d, " 1277 "Calgary: MAX_NUMNODES too low! Defined as %d, "
1288 "but system has %d nodes.\n", 1278 "but system has %d nodes.\n",
1289 MAX_NUMNODES, rio_table_hdr->num_scal_dev); 1279 MAX_NUMNODES, numnodes);
1290 return -ENODEV; 1280 return -ENODEV;
1291 } 1281 }
1292 1282
@@ -1307,8 +1297,7 @@ static int __init build_detail_arrays(void)
1307 } 1297 }
1308 1298
1309 ptr = ((unsigned long)rio_table_hdr) + 3; 1299 ptr = ((unsigned long)rio_table_hdr) + 3;
1310 for (i = 0; i < rio_table_hdr->num_scal_dev; 1300 for (i = 0; i < numnodes; i++, ptr += scal_detail_size)
1311 i++, ptr += scal_detail_size)
1312 scal_devs[i] = (struct scal_detail *)ptr; 1301 scal_devs[i] = (struct scal_detail *)ptr;
1313 1302
1314 for (i = 0; i < rio_table_hdr->num_rio_dev; 1303 for (i = 0; i < rio_table_hdr->num_rio_dev;
@@ -1339,6 +1328,61 @@ static int __init calgary_bus_has_devices(int bus, unsigned short pci_dev)
1339 return (val != 0xffffffff); 1328 return (val != 0xffffffff);
1340} 1329}
1341 1330
1331/*
1332 * calgary_init_bitmap_from_tce_table():
1333 * Funtion for kdump case. In the second/kdump kernel initialize
1334 * the bitmap based on the tce table entries obtained from first kernel
1335 */
1336static void calgary_init_bitmap_from_tce_table(struct iommu_table *tbl)
1337{
1338 u64 *tp;
1339 unsigned int index;
1340 tp = ((u64 *)tbl->it_base);
1341 for (index = 0 ; index < tbl->it_size; index++) {
1342 if (*tp != 0x0)
1343 set_bit(index, tbl->it_map);
1344 tp++;
1345 }
1346}
1347
1348/*
1349 * get_tce_space_from_tar():
1350 * Function for kdump case. Get the tce tables from first kernel
1351 * by reading the contents of the base adress register of calgary iommu
1352 */
1353static void __init get_tce_space_from_tar(void)
1354{
1355 int bus;
1356 void __iomem *target;
1357 unsigned long tce_space;
1358
1359 for (bus = 0; bus < MAX_PHB_BUS_NUM; bus++) {
1360 struct calgary_bus_info *info = &bus_info[bus];
1361 unsigned short pci_device;
1362 u32 val;
1363
1364 val = read_pci_config(bus, 0, 0, 0);
1365 pci_device = (val & 0xFFFF0000) >> 16;
1366
1367 if (!is_cal_pci_dev(pci_device))
1368 continue;
1369 if (info->translation_disabled)
1370 continue;
1371
1372 if (calgary_bus_has_devices(bus, pci_device) ||
1373 translate_empty_slots) {
1374 target = calgary_reg(bus_info[bus].bbar,
1375 tar_offset(bus));
1376 tce_space = be64_to_cpu(readq(target));
1377 tce_space = tce_space & TAR_SW_BITS;
1378
1379 tce_space = tce_space & (~specified_table_size);
1380 info->tce_space = (u64 *)__va(tce_space);
1381 }
1382 }
1383 return;
1384}
1385
1342void __init detect_calgary(void) 1386void __init detect_calgary(void)
1343{ 1387{
1344 int bus; 1388 int bus;
@@ -1394,7 +1438,8 @@ void __init detect_calgary(void)
1394 return; 1438 return;
1395 } 1439 }
1396 1440
1397 specified_table_size = determine_tce_table_size(max_pfn * PAGE_SIZE); 1441 specified_table_size = determine_tce_table_size((is_kdump_kernel() ?
1442 saved_max_pfn : max_pfn) * PAGE_SIZE);
1398 1443
1399 for (bus = 0; bus < MAX_PHB_BUS_NUM; bus++) { 1444 for (bus = 0; bus < MAX_PHB_BUS_NUM; bus++) {
1400 struct calgary_bus_info *info = &bus_info[bus]; 1445 struct calgary_bus_info *info = &bus_info[bus];
@@ -1412,10 +1457,16 @@ void __init detect_calgary(void)
1412 1457
1413 if (calgary_bus_has_devices(bus, pci_device) || 1458 if (calgary_bus_has_devices(bus, pci_device) ||
1414 translate_empty_slots) { 1459 translate_empty_slots) {
1415 tbl = alloc_tce_table(); 1460 /*
1416 if (!tbl) 1461 * If it is kdump kernel, find and use tce tables
1417 goto cleanup; 1462 * from first kernel, else allocate tce tables here
1418 info->tce_space = tbl; 1463 */
1464 if (!is_kdump_kernel()) {
1465 tbl = alloc_tce_table();
1466 if (!tbl)
1467 goto cleanup;
1468 info->tce_space = tbl;
1469 }
1419 calgary_found = 1; 1470 calgary_found = 1;
1420 } 1471 }
1421 } 1472 }
@@ -1430,6 +1481,10 @@ void __init detect_calgary(void)
1430 printk(KERN_INFO "PCI-DMA: Calgary TCE table spec is %d, " 1481 printk(KERN_INFO "PCI-DMA: Calgary TCE table spec is %d, "
1431 "CONFIG_IOMMU_DEBUG is %s.\n", specified_table_size, 1482 "CONFIG_IOMMU_DEBUG is %s.\n", specified_table_size,
1432 debugging ? "enabled" : "disabled"); 1483 debugging ? "enabled" : "disabled");
1484
1485 /* swiotlb for devices that aren't behind the Calgary. */
1486 if (max_pfn > MAX_DMA32_PFN)
1487 swiotlb = 1;
1433 } 1488 }
1434 return; 1489 return;
1435 1490
@@ -1446,7 +1501,7 @@ int __init calgary_iommu_init(void)
1446{ 1501{
1447 int ret; 1502 int ret;
1448 1503
1449 if (no_iommu || swiotlb) 1504 if (no_iommu || (swiotlb && !calgary_detected))
1450 return -ENODEV; 1505 return -ENODEV;
1451 1506
1452 if (!calgary_detected) 1507 if (!calgary_detected)
@@ -1459,15 +1514,14 @@ int __init calgary_iommu_init(void)
1459 if (ret) { 1514 if (ret) {
1460 printk(KERN_ERR "PCI-DMA: Calgary init failed %d, " 1515 printk(KERN_ERR "PCI-DMA: Calgary init failed %d, "
1461 "falling back to no_iommu\n", ret); 1516 "falling back to no_iommu\n", ret);
1462 if (max_pfn > MAX_DMA32_PFN)
1463 printk(KERN_ERR "WARNING more than 4GB of memory, "
1464 "32bit PCI may malfunction.\n");
1465 return ret; 1517 return ret;
1466 } 1518 }
1467 1519
1468 force_iommu = 1; 1520 force_iommu = 1;
1469 bad_dma_address = 0x0; 1521 bad_dma_address = 0x0;
1470 dma_ops = &calgary_dma_ops; 1522 /* dma_ops is set to swiotlb or nommu */
1523 if (!dma_ops)
1524 dma_ops = &nommu_dma_ops;
1471 1525
1472 return 0; 1526 return 0;
1473} 1527}
diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c
index cbecb05551bb..87d4d6964ec2 100644
--- a/arch/x86/kernel/pci-dma.c
+++ b/arch/x86/kernel/pci-dma.c
@@ -11,7 +11,7 @@
11 11
12static int forbid_dac __read_mostly; 12static int forbid_dac __read_mostly;
13 13
14const struct dma_mapping_ops *dma_ops; 14struct dma_mapping_ops *dma_ops;
15EXPORT_SYMBOL(dma_ops); 15EXPORT_SYMBOL(dma_ops);
16 16
17static int iommu_sac_force __read_mostly; 17static int iommu_sac_force __read_mostly;
@@ -123,6 +123,14 @@ void __init pci_iommu_alloc(void)
123 123
124 pci_swiotlb_init(); 124 pci_swiotlb_init();
125} 125}
126
127unsigned long iommu_num_pages(unsigned long addr, unsigned long len)
128{
129 unsigned long size = roundup((addr & ~PAGE_MASK) + len, PAGE_SIZE);
130
131 return size >> PAGE_SHIFT;
132}
133EXPORT_SYMBOL(iommu_num_pages);
126#endif 134#endif
127 135
128/* 136/*
@@ -192,126 +200,10 @@ static __init int iommu_setup(char *p)
192} 200}
193early_param("iommu", iommu_setup); 201early_param("iommu", iommu_setup);
194 202
195#ifdef CONFIG_X86_32
196int dma_declare_coherent_memory(struct device *dev, dma_addr_t bus_addr,
197 dma_addr_t device_addr, size_t size, int flags)
198{
199 void __iomem *mem_base = NULL;
200 int pages = size >> PAGE_SHIFT;
201 int bitmap_size = BITS_TO_LONGS(pages) * sizeof(long);
202
203 if ((flags & (DMA_MEMORY_MAP | DMA_MEMORY_IO)) == 0)
204 goto out;
205 if (!size)
206 goto out;
207 if (dev->dma_mem)
208 goto out;
209
210 /* FIXME: this routine just ignores DMA_MEMORY_INCLUDES_CHILDREN */
211
212 mem_base = ioremap(bus_addr, size);
213 if (!mem_base)
214 goto out;
215
216 dev->dma_mem = kzalloc(sizeof(struct dma_coherent_mem), GFP_KERNEL);
217 if (!dev->dma_mem)
218 goto out;
219 dev->dma_mem->bitmap = kzalloc(bitmap_size, GFP_KERNEL);
220 if (!dev->dma_mem->bitmap)
221 goto free1_out;
222
223 dev->dma_mem->virt_base = mem_base;
224 dev->dma_mem->device_base = device_addr;
225 dev->dma_mem->size = pages;
226 dev->dma_mem->flags = flags;
227
228 if (flags & DMA_MEMORY_MAP)
229 return DMA_MEMORY_MAP;
230
231 return DMA_MEMORY_IO;
232
233 free1_out:
234 kfree(dev->dma_mem);
235 out:
236 if (mem_base)
237 iounmap(mem_base);
238 return 0;
239}
240EXPORT_SYMBOL(dma_declare_coherent_memory);
241
242void dma_release_declared_memory(struct device *dev)
243{
244 struct dma_coherent_mem *mem = dev->dma_mem;
245
246 if (!mem)
247 return;
248 dev->dma_mem = NULL;
249 iounmap(mem->virt_base);
250 kfree(mem->bitmap);
251 kfree(mem);
252}
253EXPORT_SYMBOL(dma_release_declared_memory);
254
255void *dma_mark_declared_memory_occupied(struct device *dev,
256 dma_addr_t device_addr, size_t size)
257{
258 struct dma_coherent_mem *mem = dev->dma_mem;
259 int pos, err;
260 int pages = (size + (device_addr & ~PAGE_MASK) + PAGE_SIZE - 1);
261
262 pages >>= PAGE_SHIFT;
263
264 if (!mem)
265 return ERR_PTR(-EINVAL);
266
267 pos = (device_addr - mem->device_base) >> PAGE_SHIFT;
268 err = bitmap_allocate_region(mem->bitmap, pos, get_order(pages));
269 if (err != 0)
270 return ERR_PTR(err);
271 return mem->virt_base + (pos << PAGE_SHIFT);
272}
273EXPORT_SYMBOL(dma_mark_declared_memory_occupied);
274
275static int dma_alloc_from_coherent_mem(struct device *dev, ssize_t size,
276 dma_addr_t *dma_handle, void **ret)
277{
278 struct dma_coherent_mem *mem = dev ? dev->dma_mem : NULL;
279 int order = get_order(size);
280
281 if (mem) {
282 int page = bitmap_find_free_region(mem->bitmap, mem->size,
283 order);
284 if (page >= 0) {
285 *dma_handle = mem->device_base + (page << PAGE_SHIFT);
286 *ret = mem->virt_base + (page << PAGE_SHIFT);
287 memset(*ret, 0, size);
288 }
289 if (mem->flags & DMA_MEMORY_EXCLUSIVE)
290 *ret = NULL;
291 }
292 return (mem != NULL);
293}
294
295static int dma_release_coherent(struct device *dev, int order, void *vaddr)
296{
297 struct dma_coherent_mem *mem = dev ? dev->dma_mem : NULL;
298
299 if (mem && vaddr >= mem->virt_base && vaddr <
300 (mem->virt_base + (mem->size << PAGE_SHIFT))) {
301 int page = (vaddr - mem->virt_base) >> PAGE_SHIFT;
302
303 bitmap_release_region(mem->bitmap, page, order);
304 return 1;
305 }
306 return 0;
307}
308#else
309#define dma_alloc_from_coherent_mem(dev, size, handle, ret) (0)
310#define dma_release_coherent(dev, order, vaddr) (0)
311#endif /* CONFIG_X86_32 */
312
313int dma_supported(struct device *dev, u64 mask) 203int dma_supported(struct device *dev, u64 mask)
314{ 204{
205 struct dma_mapping_ops *ops = get_dma_ops(dev);
206
315#ifdef CONFIG_PCI 207#ifdef CONFIG_PCI
316 if (mask > 0xffffffff && forbid_dac > 0) { 208 if (mask > 0xffffffff && forbid_dac > 0) {
317 dev_info(dev, "PCI: Disallowing DAC for device\n"); 209 dev_info(dev, "PCI: Disallowing DAC for device\n");
@@ -319,8 +211,8 @@ int dma_supported(struct device *dev, u64 mask)
319 } 211 }
320#endif 212#endif
321 213
322 if (dma_ops->dma_supported) 214 if (ops->dma_supported)
323 return dma_ops->dma_supported(dev, mask); 215 return ops->dma_supported(dev, mask);
324 216
325 /* Copied from i386. Doesn't make much sense, because it will 217 /* Copied from i386. Doesn't make much sense, because it will
326 only work for pci_alloc_coherent. 218 only work for pci_alloc_coherent.
@@ -367,6 +259,7 @@ void *
367dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle, 259dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle,
368 gfp_t gfp) 260 gfp_t gfp)
369{ 261{
262 struct dma_mapping_ops *ops = get_dma_ops(dev);
370 void *memory = NULL; 263 void *memory = NULL;
371 struct page *page; 264 struct page *page;
372 unsigned long dma_mask = 0; 265 unsigned long dma_mask = 0;
@@ -376,7 +269,7 @@ dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle,
376 /* ignore region specifiers */ 269 /* ignore region specifiers */
377 gfp &= ~(__GFP_DMA | __GFP_HIGHMEM | __GFP_DMA32); 270 gfp &= ~(__GFP_DMA | __GFP_HIGHMEM | __GFP_DMA32);
378 271
379 if (dma_alloc_from_coherent_mem(dev, size, dma_handle, &memory)) 272 if (dma_alloc_from_coherent(dev, size, dma_handle, &memory))
380 return memory; 273 return memory;
381 274
382 if (!dev) { 275 if (!dev) {
@@ -435,8 +328,8 @@ dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle,
435 /* Let low level make its own zone decisions */ 328 /* Let low level make its own zone decisions */
436 gfp &= ~(GFP_DMA32|GFP_DMA); 329 gfp &= ~(GFP_DMA32|GFP_DMA);
437 330
438 if (dma_ops->alloc_coherent) 331 if (ops->alloc_coherent)
439 return dma_ops->alloc_coherent(dev, size, 332 return ops->alloc_coherent(dev, size,
440 dma_handle, gfp); 333 dma_handle, gfp);
441 return NULL; 334 return NULL;
442 } 335 }
@@ -448,14 +341,14 @@ dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle,
448 } 341 }
449 } 342 }
450 343
451 if (dma_ops->alloc_coherent) { 344 if (ops->alloc_coherent) {
452 free_pages((unsigned long)memory, get_order(size)); 345 free_pages((unsigned long)memory, get_order(size));
453 gfp &= ~(GFP_DMA|GFP_DMA32); 346 gfp &= ~(GFP_DMA|GFP_DMA32);
454 return dma_ops->alloc_coherent(dev, size, dma_handle, gfp); 347 return ops->alloc_coherent(dev, size, dma_handle, gfp);
455 } 348 }
456 349
457 if (dma_ops->map_simple) { 350 if (ops->map_simple) {
458 *dma_handle = dma_ops->map_simple(dev, virt_to_phys(memory), 351 *dma_handle = ops->map_simple(dev, virt_to_phys(memory),
459 size, 352 size,
460 PCI_DMA_BIDIRECTIONAL); 353 PCI_DMA_BIDIRECTIONAL);
461 if (*dma_handle != bad_dma_address) 354 if (*dma_handle != bad_dma_address)
@@ -477,12 +370,14 @@ EXPORT_SYMBOL(dma_alloc_coherent);
477void dma_free_coherent(struct device *dev, size_t size, 370void dma_free_coherent(struct device *dev, size_t size,
478 void *vaddr, dma_addr_t bus) 371 void *vaddr, dma_addr_t bus)
479{ 372{
373 struct dma_mapping_ops *ops = get_dma_ops(dev);
374
480 int order = get_order(size); 375 int order = get_order(size);
481 WARN_ON(irqs_disabled()); /* for portability */ 376 WARN_ON(irqs_disabled()); /* for portability */
482 if (dma_release_coherent(dev, order, vaddr)) 377 if (dma_release_from_coherent(dev, order, vaddr))
483 return; 378 return;
484 if (dma_ops->unmap_single) 379 if (ops->unmap_single)
485 dma_ops->unmap_single(dev, bus, size, 0); 380 ops->unmap_single(dev, bus, size, 0);
486 free_pages((unsigned long)vaddr, order); 381 free_pages((unsigned long)vaddr, order);
487} 382}
488EXPORT_SYMBOL(dma_free_coherent); 383EXPORT_SYMBOL(dma_free_coherent);
diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c
index df5f142657d2..49285f8fd4d5 100644
--- a/arch/x86/kernel/pci-gart_64.c
+++ b/arch/x86/kernel/pci-gart_64.c
@@ -67,9 +67,6 @@ static u32 gart_unmapped_entry;
67 (((x) & 0xfffff000) | (((x) >> 32) << 4) | GPTE_VALID | GPTE_COHERENT) 67 (((x) & 0xfffff000) | (((x) >> 32) << 4) | GPTE_VALID | GPTE_COHERENT)
68#define GPTE_DECODE(x) (((x) & 0xfffff000) | (((u64)(x) & 0xff0) << 28)) 68#define GPTE_DECODE(x) (((x) & 0xfffff000) | (((u64)(x) & 0xff0) << 28))
69 69
70#define to_pages(addr, size) \
71 (round_up(((addr) & ~PAGE_MASK) + (size), PAGE_SIZE) >> PAGE_SHIFT)
72
73#define EMERGENCY_PAGES 32 /* = 128KB */ 70#define EMERGENCY_PAGES 32 /* = 128KB */
74 71
75#ifdef CONFIG_AGP 72#ifdef CONFIG_AGP
@@ -241,7 +238,7 @@ nonforced_iommu(struct device *dev, unsigned long addr, size_t size)
241static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem, 238static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem,
242 size_t size, int dir) 239 size_t size, int dir)
243{ 240{
244 unsigned long npages = to_pages(phys_mem, size); 241 unsigned long npages = iommu_num_pages(phys_mem, size);
245 unsigned long iommu_page = alloc_iommu(dev, npages); 242 unsigned long iommu_page = alloc_iommu(dev, npages);
246 int i; 243 int i;
247 244
@@ -304,7 +301,7 @@ static void gart_unmap_single(struct device *dev, dma_addr_t dma_addr,
304 return; 301 return;
305 302
306 iommu_page = (dma_addr - iommu_bus_base)>>PAGE_SHIFT; 303 iommu_page = (dma_addr - iommu_bus_base)>>PAGE_SHIFT;
307 npages = to_pages(dma_addr, size); 304 npages = iommu_num_pages(dma_addr, size);
308 for (i = 0; i < npages; i++) { 305 for (i = 0; i < npages; i++) {
309 iommu_gatt_base[iommu_page + i] = gart_unmapped_entry; 306 iommu_gatt_base[iommu_page + i] = gart_unmapped_entry;
310 CLEAR_LEAK(iommu_page + i); 307 CLEAR_LEAK(iommu_page + i);
@@ -387,7 +384,7 @@ static int __dma_map_cont(struct device *dev, struct scatterlist *start,
387 } 384 }
388 385
389 addr = phys_addr; 386 addr = phys_addr;
390 pages = to_pages(s->offset, s->length); 387 pages = iommu_num_pages(s->offset, s->length);
391 while (pages--) { 388 while (pages--) {
392 iommu_gatt_base[iommu_page] = GPTE_ENCODE(addr); 389 iommu_gatt_base[iommu_page] = GPTE_ENCODE(addr);
393 SET_LEAK(iommu_page); 390 SET_LEAK(iommu_page);
@@ -470,7 +467,7 @@ gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
470 467
471 seg_size += s->length; 468 seg_size += s->length;
472 need = nextneed; 469 need = nextneed;
473 pages += to_pages(s->offset, s->length); 470 pages += iommu_num_pages(s->offset, s->length);
474 ps = s; 471 ps = s;
475 } 472 }
476 if (dma_map_cont(dev, start_sg, i - start, sgmap, pages, need) < 0) 473 if (dma_map_cont(dev, start_sg, i - start, sgmap, pages, need) < 0)
@@ -692,8 +689,7 @@ static __init int init_k8_gatt(struct agp_kern_info *info)
692 689
693extern int agp_amd64_init(void); 690extern int agp_amd64_init(void);
694 691
695static const struct dma_mapping_ops gart_dma_ops = { 692static struct dma_mapping_ops gart_dma_ops = {
696 .mapping_error = NULL,
697 .map_single = gart_map_single, 693 .map_single = gart_map_single,
698 .map_simple = gart_map_simple, 694 .map_simple = gart_map_simple,
699 .unmap_single = gart_unmap_single, 695 .unmap_single = gart_unmap_single,
diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c
index 792b9179eff3..3f91f71cdc3e 100644
--- a/arch/x86/kernel/pci-nommu.c
+++ b/arch/x86/kernel/pci-nommu.c
@@ -72,21 +72,9 @@ static int nommu_map_sg(struct device *hwdev, struct scatterlist *sg,
72 return nents; 72 return nents;
73} 73}
74 74
75/* Make sure we keep the same behaviour */ 75struct dma_mapping_ops nommu_dma_ops = {
76static int nommu_mapping_error(dma_addr_t dma_addr)
77{
78#ifdef CONFIG_X86_32
79 return 0;
80#else
81 return (dma_addr == bad_dma_address);
82#endif
83}
84
85
86const struct dma_mapping_ops nommu_dma_ops = {
87 .map_single = nommu_map_single, 76 .map_single = nommu_map_single,
88 .map_sg = nommu_map_sg, 77 .map_sg = nommu_map_sg,
89 .mapping_error = nommu_mapping_error,
90 .is_phys = 1, 78 .is_phys = 1,
91}; 79};
92 80
diff --git a/arch/x86/kernel/pci-swiotlb_64.c b/arch/x86/kernel/pci-swiotlb_64.c
index 20df839b9c20..c4ce0332759e 100644
--- a/arch/x86/kernel/pci-swiotlb_64.c
+++ b/arch/x86/kernel/pci-swiotlb_64.c
@@ -18,7 +18,7 @@ swiotlb_map_single_phys(struct device *hwdev, phys_addr_t paddr, size_t size,
18 return swiotlb_map_single(hwdev, phys_to_virt(paddr), size, direction); 18 return swiotlb_map_single(hwdev, phys_to_virt(paddr), size, direction);
19} 19}
20 20
21const struct dma_mapping_ops swiotlb_dma_ops = { 21struct dma_mapping_ops swiotlb_dma_ops = {
22 .mapping_error = swiotlb_dma_mapping_error, 22 .mapping_error = swiotlb_dma_mapping_error,
23 .alloc_coherent = swiotlb_alloc_coherent, 23 .alloc_coherent = swiotlb_alloc_coherent,
24 .free_coherent = swiotlb_free_coherent, 24 .free_coherent = swiotlb_free_coherent,
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 68f4ae2ac99a..ae584f87d06a 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -95,7 +95,6 @@ static inline void play_dead(void)
95{ 95{
96 /* This must be done before dead CPU ack */ 96 /* This must be done before dead CPU ack */
97 cpu_exit_clear(); 97 cpu_exit_clear();
98 wbinvd();
99 mb(); 98 mb();
100 /* Ack it */ 99 /* Ack it */
101 __get_cpu_var(cpu_state) = CPU_DEAD; 100 __get_cpu_var(cpu_state) = CPU_DEAD;
@@ -104,8 +103,8 @@ static inline void play_dead(void)
104 * With physical CPU hotplug, we should halt the cpu 103 * With physical CPU hotplug, we should halt the cpu
105 */ 104 */
106 local_irq_disable(); 105 local_irq_disable();
107 while (1) 106 /* mask all interrupts, flush any and all caches, and halt */
108 halt(); 107 wbinvd_halt();
109} 108}
110#else 109#else
111static inline void play_dead(void) 110static inline void play_dead(void)
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 91ffce47af8e..30ce4734edaa 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -93,14 +93,13 @@ DECLARE_PER_CPU(int, cpu_state);
93static inline void play_dead(void) 93static inline void play_dead(void)
94{ 94{
95 idle_task_exit(); 95 idle_task_exit();
96 wbinvd();
97 mb(); 96 mb();
98 /* Ack it */ 97 /* Ack it */
99 __get_cpu_var(cpu_state) = CPU_DEAD; 98 __get_cpu_var(cpu_state) = CPU_DEAD;
100 99
101 local_irq_disable(); 100 local_irq_disable();
102 while (1) 101 /* mask all interrupts, flush any and all caches, and halt */
103 halt(); 102 wbinvd_halt();
104} 103}
105#else 104#else
106static inline void play_dead(void) 105static inline void play_dead(void)
diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c
index 06a9f643817e..724adfc63cb9 100644
--- a/arch/x86/kernel/reboot.c
+++ b/arch/x86/kernel/reboot.c
@@ -414,25 +414,20 @@ void native_machine_shutdown(void)
414 414
415 /* The boot cpu is always logical cpu 0 */ 415 /* The boot cpu is always logical cpu 0 */
416 int reboot_cpu_id = 0; 416 int reboot_cpu_id = 0;
417 cpumask_of_cpu_ptr(newmask, reboot_cpu_id);
418 417
419#ifdef CONFIG_X86_32 418#ifdef CONFIG_X86_32
420 /* See if there has been given a command line override */ 419 /* See if there has been given a command line override */
421 if ((reboot_cpu != -1) && (reboot_cpu < NR_CPUS) && 420 if ((reboot_cpu != -1) && (reboot_cpu < NR_CPUS) &&
422 cpu_online(reboot_cpu)) { 421 cpu_online(reboot_cpu))
423 reboot_cpu_id = reboot_cpu; 422 reboot_cpu_id = reboot_cpu;
424 cpumask_of_cpu_ptr_next(newmask, reboot_cpu_id);
425 }
426#endif 423#endif
427 424
428 /* Make certain the cpu I'm about to reboot on is online */ 425 /* Make certain the cpu I'm about to reboot on is online */
429 if (!cpu_online(reboot_cpu_id)) { 426 if (!cpu_online(reboot_cpu_id))
430 reboot_cpu_id = smp_processor_id(); 427 reboot_cpu_id = smp_processor_id();
431 cpumask_of_cpu_ptr_next(newmask, reboot_cpu_id);
432 }
433 428
434 /* Make certain I only run on the appropriate processor */ 429 /* Make certain I only run on the appropriate processor */
435 set_cpus_allowed_ptr(current, newmask); 430 set_cpus_allowed_ptr(current, &cpumask_of_cpu(reboot_cpu_id));
436 431
437 /* O.K Now that I'm on the appropriate processor, 432 /* O.K Now that I'm on the appropriate processor,
438 * stop all of the others. 433 * stop all of the others.
diff --git a/arch/x86/kernel/relocate_kernel_32.S b/arch/x86/kernel/relocate_kernel_32.S
index c30fe25d470d..6f50664b2ba5 100644
--- a/arch/x86/kernel/relocate_kernel_32.S
+++ b/arch/x86/kernel/relocate_kernel_32.S
@@ -20,11 +20,45 @@
20#define PAGE_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY) 20#define PAGE_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
21#define PAE_PGD_ATTR (_PAGE_PRESENT) 21#define PAE_PGD_ATTR (_PAGE_PRESENT)
22 22
23/* control_page + KEXEC_CONTROL_CODE_MAX_SIZE
24 * ~ control_page + PAGE_SIZE are used as data storage and stack for
25 * jumping back
26 */
27#define DATA(offset) (KEXEC_CONTROL_CODE_MAX_SIZE+(offset))
28
29/* Minimal CPU state */
30#define ESP DATA(0x0)
31#define CR0 DATA(0x4)
32#define CR3 DATA(0x8)
33#define CR4 DATA(0xc)
34
35/* other data */
36#define CP_VA_CONTROL_PAGE DATA(0x10)
37#define CP_PA_PGD DATA(0x14)
38#define CP_PA_SWAP_PAGE DATA(0x18)
39#define CP_PA_BACKUP_PAGES_MAP DATA(0x1c)
40
23 .text 41 .text
24 .align PAGE_SIZE 42 .align PAGE_SIZE
25 .globl relocate_kernel 43 .globl relocate_kernel
26relocate_kernel: 44relocate_kernel:
27 movl 8(%esp), %ebp /* list of pages */ 45 /* Save the CPU context, used for jumping back */
46
47 pushl %ebx
48 pushl %esi
49 pushl %edi
50 pushl %ebp
51 pushf
52
53 movl 20+8(%esp), %ebp /* list of pages */
54 movl PTR(VA_CONTROL_PAGE)(%ebp), %edi
55 movl %esp, ESP(%edi)
56 movl %cr0, %eax
57 movl %eax, CR0(%edi)
58 movl %cr3, %eax
59 movl %eax, CR3(%edi)
60 movl %cr4, %eax
61 movl %eax, CR4(%edi)
28 62
29#ifdef CONFIG_X86_PAE 63#ifdef CONFIG_X86_PAE
30 /* map the control page at its virtual address */ 64 /* map the control page at its virtual address */
@@ -138,15 +172,25 @@ relocate_kernel:
138 172
139relocate_new_kernel: 173relocate_new_kernel:
140 /* read the arguments and say goodbye to the stack */ 174 /* read the arguments and say goodbye to the stack */
141 movl 4(%esp), %ebx /* page_list */ 175 movl 20+4(%esp), %ebx /* page_list */
142 movl 8(%esp), %ebp /* list of pages */ 176 movl 20+8(%esp), %ebp /* list of pages */
143 movl 12(%esp), %edx /* start address */ 177 movl 20+12(%esp), %edx /* start address */
144 movl 16(%esp), %ecx /* cpu_has_pae */ 178 movl 20+16(%esp), %ecx /* cpu_has_pae */
179 movl 20+20(%esp), %esi /* preserve_context */
145 180
146 /* zero out flags, and disable interrupts */ 181 /* zero out flags, and disable interrupts */
147 pushl $0 182 pushl $0
148 popfl 183 popfl
149 184
185 /* save some information for jumping back */
186 movl PTR(VA_CONTROL_PAGE)(%ebp), %edi
187 movl %edi, CP_VA_CONTROL_PAGE(%edi)
188 movl PTR(PA_PGD)(%ebp), %eax
189 movl %eax, CP_PA_PGD(%edi)
190 movl PTR(PA_SWAP_PAGE)(%ebp), %eax
191 movl %eax, CP_PA_SWAP_PAGE(%edi)
192 movl %ebx, CP_PA_BACKUP_PAGES_MAP(%edi)
193
150 /* get physical address of control page now */ 194 /* get physical address of control page now */
151 /* this is impossible after page table switch */ 195 /* this is impossible after page table switch */
152 movl PTR(PA_CONTROL_PAGE)(%ebp), %edi 196 movl PTR(PA_CONTROL_PAGE)(%ebp), %edi
@@ -197,8 +241,90 @@ identity_mapped:
197 xorl %eax, %eax 241 xorl %eax, %eax
198 movl %eax, %cr3 242 movl %eax, %cr3
199 243
244 movl CP_PA_SWAP_PAGE(%edi), %eax
245 pushl %eax
246 pushl %ebx
247 call swap_pages
248 addl $8, %esp
249
250 /* To be certain of avoiding problems with self-modifying code
251 * I need to execute a serializing instruction here.
252 * So I flush the TLB, it's handy, and not processor dependent.
253 */
254 xorl %eax, %eax
255 movl %eax, %cr3
256
257 /* set all of the registers to known values */
258 /* leave %esp alone */
259
260 testl %esi, %esi
261 jnz 1f
262 xorl %edi, %edi
263 xorl %eax, %eax
264 xorl %ebx, %ebx
265 xorl %ecx, %ecx
266 xorl %edx, %edx
267 xorl %esi, %esi
268 xorl %ebp, %ebp
269 ret
2701:
271 popl %edx
272 movl CP_PA_SWAP_PAGE(%edi), %esp
273 addl $PAGE_SIZE, %esp
2742:
275 call *%edx
276
277 /* get the re-entry point of the peer system */
278 movl 0(%esp), %ebp
279 call 1f
2801:
281 popl %ebx
282 subl $(1b - relocate_kernel), %ebx
283 movl CP_VA_CONTROL_PAGE(%ebx), %edi
284 lea PAGE_SIZE(%ebx), %esp
285 movl CP_PA_SWAP_PAGE(%ebx), %eax
286 movl CP_PA_BACKUP_PAGES_MAP(%ebx), %edx
287 pushl %eax
288 pushl %edx
289 call swap_pages
290 addl $8, %esp
291 movl CP_PA_PGD(%ebx), %eax
292 movl %eax, %cr3
293 movl %cr0, %eax
294 orl $(1<<31), %eax
295 movl %eax, %cr0
296 lea PAGE_SIZE(%edi), %esp
297 movl %edi, %eax
298 addl $(virtual_mapped - relocate_kernel), %eax
299 pushl %eax
300 ret
301
302virtual_mapped:
303 movl CR4(%edi), %eax
304 movl %eax, %cr4
305 movl CR3(%edi), %eax
306 movl %eax, %cr3
307 movl CR0(%edi), %eax
308 movl %eax, %cr0
309 movl ESP(%edi), %esp
310 movl %ebp, %eax
311
312 popf
313 popl %ebp
314 popl %edi
315 popl %esi
316 popl %ebx
317 ret
318
200 /* Do the copies */ 319 /* Do the copies */
201 movl %ebx, %ecx 320swap_pages:
321 movl 8(%esp), %edx
322 movl 4(%esp), %ecx
323 pushl %ebp
324 pushl %ebx
325 pushl %edi
326 pushl %esi
327 movl %ecx, %ebx
202 jmp 1f 328 jmp 1f
203 329
2040: /* top, read another word from the indirection page */ 3300: /* top, read another word from the indirection page */
@@ -226,27 +352,31 @@ identity_mapped:
226 movl %ecx, %esi /* For every source page do a copy */ 352 movl %ecx, %esi /* For every source page do a copy */
227 andl $0xfffff000, %esi 353 andl $0xfffff000, %esi
228 354
355 movl %edi, %eax
356 movl %esi, %ebp
357
358 movl %edx, %edi
229 movl $1024, %ecx 359 movl $1024, %ecx
230 rep ; movsl 360 rep ; movsl
231 jmp 0b
232
2333:
234 361
235 /* To be certain of avoiding problems with self-modifying code 362 movl %ebp, %edi
236 * I need to execute a serializing instruction here. 363 movl %eax, %esi
237 * So I flush the TLB, it's handy, and not processor dependent. 364 movl $1024, %ecx
238 */ 365 rep ; movsl
239 xorl %eax, %eax
240 movl %eax, %cr3
241 366
242 /* set all of the registers to known values */ 367 movl %eax, %edi
243 /* leave %esp alone */ 368 movl %edx, %esi
369 movl $1024, %ecx
370 rep ; movsl
244 371
245 xorl %eax, %eax 372 lea PAGE_SIZE(%ebp), %esi
246 xorl %ebx, %ebx 373 jmp 0b
247 xorl %ecx, %ecx 3743:
248 xorl %edx, %edx 375 popl %esi
249 xorl %esi, %esi 376 popl %edi
250 xorl %edi, %edi 377 popl %ebx
251 xorl %ebp, %ebp 378 popl %ebp
252 ret 379 ret
380
381 .globl kexec_control_code_size
382.set kexec_control_code_size, . - relocate_kernel
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index b520dae02bf4..9838f2539dfc 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -445,7 +445,7 @@ static void __init reserve_early_setup_data(void)
445 * @size: Size of the crashkernel memory to reserve. 445 * @size: Size of the crashkernel memory to reserve.
446 * Returns the base address on success, and -1ULL on failure. 446 * Returns the base address on success, and -1ULL on failure.
447 */ 447 */
448unsigned long long find_and_reserve_crashkernel(unsigned long long size) 448unsigned long long __init find_and_reserve_crashkernel(unsigned long long size)
449{ 449{
450 const unsigned long long alignment = 16<<20; /* 16M */ 450 const unsigned long long alignment = 16<<20; /* 16M */
451 unsigned long long start = 0LL; 451 unsigned long long start = 0LL;
@@ -670,6 +670,18 @@ void __init setup_arch(char **cmdline_p)
670 670
671 parse_early_param(); 671 parse_early_param();
672 672
673#ifdef CONFIG_X86_64
674 check_efer();
675#endif
676
677#if defined(CONFIG_VMI) && defined(CONFIG_X86_32)
678 /*
679 * Must be before kernel pagetables are setup
680 * or fixmap area is touched.
681 */
682 vmi_init();
683#endif
684
673 /* after early param, so could get panic from serial */ 685 /* after early param, so could get panic from serial */
674 reserve_early_setup_data(); 686 reserve_early_setup_data();
675 687
@@ -730,7 +742,6 @@ void __init setup_arch(char **cmdline_p)
730#else 742#else
731 num_physpages = max_pfn; 743 num_physpages = max_pfn;
732 744
733 check_efer();
734 745
735 /* How many end-of-memory variables you have, grandma! */ 746 /* How many end-of-memory variables you have, grandma! */
736 /* need this before calling reserve_initrd */ 747 /* need this before calling reserve_initrd */
@@ -788,10 +799,6 @@ void __init setup_arch(char **cmdline_p)
788 799
789 initmem_init(0, max_pfn); 800 initmem_init(0, max_pfn);
790 801
791#ifdef CONFIG_X86_64
792 dma32_reserve_bootmem();
793#endif
794
795#ifdef CONFIG_ACPI_SLEEP 802#ifdef CONFIG_ACPI_SLEEP
796 /* 803 /*
797 * Reserve low memory region for sleep support. 804 * Reserve low memory region for sleep support.
@@ -806,20 +813,21 @@ void __init setup_arch(char **cmdline_p)
806#endif 813#endif
807 reserve_crashkernel(); 814 reserve_crashkernel();
808 815
816#ifdef CONFIG_X86_64
817 /*
818 * dma32_reserve_bootmem() allocates bootmem which may conflict
819 * with the crashkernel command line, so do that after
820 * reserve_crashkernel()
821 */
822 dma32_reserve_bootmem();
823#endif
824
809 reserve_ibft_region(); 825 reserve_ibft_region();
810 826
811#ifdef CONFIG_KVM_CLOCK 827#ifdef CONFIG_KVM_CLOCK
812 kvmclock_init(); 828 kvmclock_init();
813#endif 829#endif
814 830
815#if defined(CONFIG_VMI) && defined(CONFIG_X86_32)
816 /*
817 * Must be after max_low_pfn is determined, and before kernel
818 * pagetables are setup.
819 */
820 vmi_init();
821#endif
822
823 paravirt_pagetable_setup_start(swapper_pg_dir); 831 paravirt_pagetable_setup_start(swapper_pg_dir);
824 paging_init(); 832 paging_init();
825 paravirt_pagetable_setup_done(swapper_pg_dir); 833 paravirt_pagetable_setup_done(swapper_pg_dir);
@@ -856,12 +864,6 @@ void __init setup_arch(char **cmdline_p)
856 init_apic_mappings(); 864 init_apic_mappings();
857 ioapic_init_mappings(); 865 ioapic_init_mappings();
858 866
859#if defined(CONFIG_SMP) && defined(CONFIG_X86_PC) && defined(CONFIG_X86_32)
860 if (def_to_bigsmp)
861 printk(KERN_WARNING "More than 8 CPUs detected and "
862 "CONFIG_X86_PC cannot handle it.\nUse "
863 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
864#endif
865 kvm_guest_init(); 867 kvm_guest_init();
866 868
867 e820_reserve_resources(); 869 e820_reserve_resources();
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index f7745f94c006..76e305e064f9 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -80,24 +80,6 @@ static void __init setup_per_cpu_maps(void)
80#endif 80#endif
81} 81}
82 82
83#ifdef CONFIG_HAVE_CPUMASK_OF_CPU_MAP
84cpumask_t *cpumask_of_cpu_map __read_mostly;
85EXPORT_SYMBOL(cpumask_of_cpu_map);
86
87/* requires nr_cpu_ids to be initialized */
88static void __init setup_cpumask_of_cpu(void)
89{
90 int i;
91
92 /* alloc_bootmem zeroes memory */
93 cpumask_of_cpu_map = alloc_bootmem_low(sizeof(cpumask_t) * nr_cpu_ids);
94 for (i = 0; i < nr_cpu_ids; i++)
95 cpu_set(i, cpumask_of_cpu_map[i]);
96}
97#else
98static inline void setup_cpumask_of_cpu(void) { }
99#endif
100
101#ifdef CONFIG_X86_32 83#ifdef CONFIG_X86_32
102/* 84/*
103 * Great future not-so-futuristic plan: make i386 and x86_64 do it 85 * Great future not-so-futuristic plan: make i386 and x86_64 do it
@@ -197,9 +179,6 @@ void __init setup_per_cpu_areas(void)
197 179
198 /* Setup node to cpumask map */ 180 /* Setup node to cpumask map */
199 setup_node_to_cpumask_map(); 181 setup_node_to_cpumask_map();
200
201 /* Setup cpumask_of_cpu map */
202 setup_cpumask_of_cpu();
203} 182}
204 183
205#endif 184#endif
diff --git a/arch/x86/kernel/signal_64.c b/arch/x86/kernel/signal_64.c
index b45ef8ddd651..ca316b5b742c 100644
--- a/arch/x86/kernel/signal_64.c
+++ b/arch/x86/kernel/signal_64.c
@@ -104,7 +104,16 @@ static inline int restore_i387(struct _fpstate __user *buf)
104 clts(); 104 clts();
105 task_thread_info(current)->status |= TS_USEDFPU; 105 task_thread_info(current)->status |= TS_USEDFPU;
106 } 106 }
107 return restore_fpu_checking((__force struct i387_fxsave_struct *)buf); 107 err = restore_fpu_checking((__force struct i387_fxsave_struct *)buf);
108 if (unlikely(err)) {
109 /*
110 * Encountered an error while doing the restore from the
111 * user buffer, clear the fpu state.
112 */
113 clear_fpu(tsk);
114 clear_used_math();
115 }
116 return err;
108} 117}
109 118
110/* 119/*
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 332512767f4f..7985c5b3f916 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -326,12 +326,16 @@ static void __cpuinit start_secondary(void *unused)
326 * for which cpus receive the IPI. Holding this 326 * for which cpus receive the IPI. Holding this
327 * lock helps us to not include this cpu in a currently in progress 327 * lock helps us to not include this cpu in a currently in progress
328 * smp_call_function(). 328 * smp_call_function().
329 *
330 * We need to hold vector_lock so there the set of online cpus
331 * does not change while we are assigning vectors to cpus. Holding
332 * this lock ensures we don't half assign or remove an irq from a cpu.
329 */ 333 */
330 ipi_call_lock_irq(); 334 ipi_call_lock_irq();
331#ifdef CONFIG_X86_IO_APIC 335 lock_vector_lock();
332 setup_vector_irq(smp_processor_id()); 336 __setup_vector_irq(smp_processor_id());
333#endif
334 cpu_set(smp_processor_id(), cpu_online_map); 337 cpu_set(smp_processor_id(), cpu_online_map);
338 unlock_vector_lock();
335 ipi_call_unlock_irq(); 339 ipi_call_unlock_irq();
336 per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE; 340 per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
337 341
@@ -752,6 +756,14 @@ static void __cpuinit do_fork_idle(struct work_struct *work)
752} 756}
753 757
754#ifdef CONFIG_X86_64 758#ifdef CONFIG_X86_64
759
760/* __ref because it's safe to call free_bootmem when after_bootmem == 0. */
761static void __ref free_bootmem_pda(struct x8664_pda *oldpda)
762{
763 if (!after_bootmem)
764 free_bootmem((unsigned long)oldpda, sizeof(*oldpda));
765}
766
755/* 767/*
756 * Allocate node local memory for the AP pda. 768 * Allocate node local memory for the AP pda.
757 * 769 *
@@ -780,8 +792,7 @@ int __cpuinit get_local_pda(int cpu)
780 792
781 if (oldpda) { 793 if (oldpda) {
782 memcpy(newpda, oldpda, size); 794 memcpy(newpda, oldpda, size);
783 if (!after_bootmem) 795 free_bootmem_pda(oldpda);
784 free_bootmem((unsigned long)oldpda, size);
785 } 796 }
786 797
787 newpda->in_bootmem = 0; 798 newpda->in_bootmem = 0;
@@ -1044,6 +1055,34 @@ static __init void disable_smp(void)
1044static int __init smp_sanity_check(unsigned max_cpus) 1055static int __init smp_sanity_check(unsigned max_cpus)
1045{ 1056{
1046 preempt_disable(); 1057 preempt_disable();
1058
1059#if defined(CONFIG_X86_PC) && defined(CONFIG_X86_32)
1060 if (def_to_bigsmp && nr_cpu_ids > 8) {
1061 unsigned int cpu;
1062 unsigned nr;
1063
1064 printk(KERN_WARNING
1065 "More than 8 CPUs detected - skipping them.\n"
1066 "Use CONFIG_X86_GENERICARCH and CONFIG_X86_BIGSMP.\n");
1067
1068 nr = 0;
1069 for_each_present_cpu(cpu) {
1070 if (nr >= 8)
1071 cpu_clear(cpu, cpu_present_map);
1072 nr++;
1073 }
1074
1075 nr = 0;
1076 for_each_possible_cpu(cpu) {
1077 if (nr >= 8)
1078 cpu_clear(cpu, cpu_possible_map);
1079 nr++;
1080 }
1081
1082 nr_cpu_ids = 8;
1083 }
1084#endif
1085
1047 if (!physid_isset(hard_smp_processor_id(), phys_cpu_present_map)) { 1086 if (!physid_isset(hard_smp_processor_id(), phys_cpu_present_map)) {
1048 printk(KERN_WARNING "weird, boot CPU (#%d) not listed" 1087 printk(KERN_WARNING "weird, boot CPU (#%d) not listed"
1049 "by the BIOS.\n", hard_smp_processor_id()); 1088 "by the BIOS.\n", hard_smp_processor_id());
@@ -1182,6 +1221,9 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
1182 printk(KERN_INFO "CPU%d: ", 0); 1221 printk(KERN_INFO "CPU%d: ", 0);
1183 print_cpu_info(&cpu_data(0)); 1222 print_cpu_info(&cpu_data(0));
1184 setup_boot_clock(); 1223 setup_boot_clock();
1224
1225 if (is_uv_system())
1226 uv_system_init();
1185out: 1227out:
1186 preempt_enable(); 1228 preempt_enable();
1187} 1229}
@@ -1336,7 +1378,9 @@ int __cpu_disable(void)
1336 remove_siblinginfo(cpu); 1378 remove_siblinginfo(cpu);
1337 1379
1338 /* It's now safe to remove this processor from the online map */ 1380 /* It's now safe to remove this processor from the online map */
1381 lock_vector_lock();
1339 remove_cpu_from_maps(cpu); 1382 remove_cpu_from_maps(cpu);
1383 unlock_vector_lock();
1340 fixup_irqs(cpu_online_map); 1384 fixup_irqs(cpu_online_map);
1341 return 0; 1385 return 0;
1342} 1386}
@@ -1370,17 +1414,3 @@ void __cpu_die(unsigned int cpu)
1370 BUG(); 1414 BUG();
1371} 1415}
1372#endif 1416#endif
1373
1374/*
1375 * If the BIOS enumerates physical processors before logical,
1376 * maxcpus=N at enumeration-time can be used to disable HT.
1377 */
1378static int __init parse_maxcpus(char *arg)
1379{
1380 extern unsigned int maxcpus;
1381
1382 if (arg)
1383 maxcpus = simple_strtoul(arg, NULL, 0);
1384 return 0;
1385}
1386early_param("maxcpus", parse_maxcpus);
diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c
index 99941b37eca0..397e309839dd 100644
--- a/arch/x86/kernel/smpcommon.c
+++ b/arch/x86/kernel/smpcommon.c
@@ -8,18 +8,21 @@
8DEFINE_PER_CPU(unsigned long, this_cpu_off); 8DEFINE_PER_CPU(unsigned long, this_cpu_off);
9EXPORT_PER_CPU_SYMBOL(this_cpu_off); 9EXPORT_PER_CPU_SYMBOL(this_cpu_off);
10 10
11/* Initialize the CPU's GDT. This is either the boot CPU doing itself 11/*
12 (still using the master per-cpu area), or a CPU doing it for a 12 * Initialize the CPU's GDT. This is either the boot CPU doing itself
13 secondary which will soon come up. */ 13 * (still using the master per-cpu area), or a CPU doing it for a
14 * secondary which will soon come up.
15 */
14__cpuinit void init_gdt(int cpu) 16__cpuinit void init_gdt(int cpu)
15{ 17{
16 struct desc_struct *gdt = get_cpu_gdt_table(cpu); 18 struct desc_struct gdt;
17 19
18 pack_descriptor(&gdt[GDT_ENTRY_PERCPU], 20 pack_descriptor(&gdt, __per_cpu_offset[cpu], 0xFFFFF,
19 __per_cpu_offset[cpu], 0xFFFFF,
20 0x2 | DESCTYPE_S, 0x8); 21 0x2 | DESCTYPE_S, 0x8);
22 gdt.s = 1;
21 23
22 gdt[GDT_ENTRY_PERCPU].s = 1; 24 write_gdt_entry(get_cpu_gdt_table(cpu),
25 GDT_ENTRY_PERCPU, &gdt, DESCTYPE_S);
23 26
24 per_cpu(this_cpu_off, cpu) = __per_cpu_offset[cpu]; 27 per_cpu(this_cpu_off, cpu) = __per_cpu_offset[cpu];
25 per_cpu(cpu_number, cpu) = cpu; 28 per_cpu(cpu_number, cpu) = cpu;
diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c
index d0fbb7712ab0..8b8c0d6640fa 100644
--- a/arch/x86/kernel/tlb_uv.c
+++ b/arch/x86/kernel/tlb_uv.c
@@ -17,6 +17,7 @@
17#include <asm/genapic.h> 17#include <asm/genapic.h>
18#include <asm/idle.h> 18#include <asm/idle.h>
19#include <asm/tsc.h> 19#include <asm/tsc.h>
20#include <asm/irq_vectors.h>
20 21
21#include <mach_apic.h> 22#include <mach_apic.h>
22 23
@@ -783,7 +784,7 @@ static int __init uv_bau_init(void)
783 uv_init_blade(blade, node, cur_cpu); 784 uv_init_blade(blade, node, cur_cpu);
784 cur_cpu += uv_blade_nr_possible_cpus(blade); 785 cur_cpu += uv_blade_nr_possible_cpus(blade);
785 } 786 }
786 set_intr_gate(UV_BAU_MESSAGE, uv_bau_message_intr1); 787 alloc_intr_gate(UV_BAU_MESSAGE, uv_bau_message_intr1);
787 uv_enable_timeouts(); 788 uv_enable_timeouts();
788 789
789 return 0; 790 return 0;
diff --git a/arch/x86/kernel/traps_64.c b/arch/x86/kernel/traps_64.c
index 3f18d73f420c..513caaca7115 100644
--- a/arch/x86/kernel/traps_64.c
+++ b/arch/x86/kernel/traps_64.c
@@ -1131,7 +1131,14 @@ asmlinkage void math_state_restore(void)
1131 } 1131 }
1132 1132
1133 clts(); /* Allow maths ops (or we recurse) */ 1133 clts(); /* Allow maths ops (or we recurse) */
1134 restore_fpu_checking(&me->thread.xstate->fxsave); 1134 /*
1135 * Paranoid restore. send a SIGSEGV if we fail to restore the state.
1136 */
1137 if (unlikely(restore_fpu_checking(&me->thread.xstate->fxsave))) {
1138 stts();
1139 force_sig(SIGSEGV, me);
1140 return;
1141 }
1135 task_thread_info(me)->status |= TS_USEDFPU; 1142 task_thread_info(me)->status |= TS_USEDFPU;
1136 me->fpu_counter++; 1143 me->fpu_counter++;
1137} 1144}
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 7603c0553909..8f98e9de1b82 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -104,7 +104,7 @@ __setup("notsc", notsc_setup);
104/* 104/*
105 * Read TSC and the reference counters. Take care of SMI disturbance 105 * Read TSC and the reference counters. Take care of SMI disturbance
106 */ 106 */
107static u64 __init tsc_read_refs(u64 *pm, u64 *hpet) 107static u64 tsc_read_refs(u64 *pm, u64 *hpet)
108{ 108{
109 u64 t1, t2; 109 u64 t1, t2;
110 int i; 110 int i;
@@ -122,80 +122,216 @@ static u64 __init tsc_read_refs(u64 *pm, u64 *hpet)
122 return ULLONG_MAX; 122 return ULLONG_MAX;
123} 123}
124 124
125/** 125/*
126 * native_calibrate_tsc - calibrate the tsc on boot 126 * Try to calibrate the TSC against the Programmable
127 * Interrupt Timer and return the frequency of the TSC
128 * in kHz.
129 *
130 * Return ULONG_MAX on failure to calibrate.
127 */ 131 */
128unsigned long native_calibrate_tsc(void) 132static unsigned long pit_calibrate_tsc(void)
129{ 133{
130 unsigned long flags; 134 u64 tsc, t1, t2, delta;
131 u64 tsc1, tsc2, tr1, tr2, delta, pm1, pm2, hpet1, hpet2; 135 unsigned long tscmin, tscmax;
132 int hpet = is_hpet_enabled(); 136 int pitcnt;
133 unsigned int tsc_khz_val = 0;
134
135 local_irq_save(flags);
136
137 tsc1 = tsc_read_refs(&pm1, hpet ? &hpet1 : NULL);
138 137
138 /* Set the Gate high, disable speaker */
139 outb((inb(0x61) & ~0x02) | 0x01, 0x61); 139 outb((inb(0x61) & ~0x02) | 0x01, 0x61);
140 140
141 /*
142 * Setup CTC channel 2* for mode 0, (interrupt on terminal
143 * count mode), binary count. Set the latch register to 50ms
144 * (LSB then MSB) to begin countdown.
145 */
141 outb(0xb0, 0x43); 146 outb(0xb0, 0x43);
142 outb((CLOCK_TICK_RATE / (1000 / 50)) & 0xff, 0x42); 147 outb((CLOCK_TICK_RATE / (1000 / 50)) & 0xff, 0x42);
143 outb((CLOCK_TICK_RATE / (1000 / 50)) >> 8, 0x42); 148 outb((CLOCK_TICK_RATE / (1000 / 50)) >> 8, 0x42);
144 tr1 = get_cycles();
145 while ((inb(0x61) & 0x20) == 0);
146 tr2 = get_cycles();
147 149
148 tsc2 = tsc_read_refs(&pm2, hpet ? &hpet2 : NULL); 150 tsc = t1 = t2 = get_cycles();
149 151
150 local_irq_restore(flags); 152 pitcnt = 0;
153 tscmax = 0;
154 tscmin = ULONG_MAX;
155 while ((inb(0x61) & 0x20) == 0) {
156 t2 = get_cycles();
157 delta = t2 - tsc;
158 tsc = t2;
159 if ((unsigned long) delta < tscmin)
160 tscmin = (unsigned int) delta;
161 if ((unsigned long) delta > tscmax)
162 tscmax = (unsigned int) delta;
163 pitcnt++;
164 }
151 165
152 /* 166 /*
153 * Preset the result with the raw and inaccurate PIT 167 * Sanity checks:
154 * calibration value 168 *
169 * If we were not able to read the PIT more than 5000
170 * times, then we have been hit by a massive SMI
171 *
172 * If the maximum is 10 times larger than the minimum,
173 * then we got hit by an SMI as well.
155 */ 174 */
156 delta = (tr2 - tr1); 175 if (pitcnt < 5000 || tscmax > 10 * tscmin)
176 return ULONG_MAX;
177
178 /* Calculate the PIT value */
179 delta = t2 - t1;
157 do_div(delta, 50); 180 do_div(delta, 50);
158 tsc_khz_val = delta; 181 return delta;
182}
183
184
185/**
186 * native_calibrate_tsc - calibrate the tsc on boot
187 */
188unsigned long native_calibrate_tsc(void)
189{
190 u64 tsc1, tsc2, delta, pm1, pm2, hpet1, hpet2;
191 unsigned long tsc_pit_min = ULONG_MAX, tsc_ref_min = ULONG_MAX;
192 unsigned long flags;
193 int hpet = is_hpet_enabled(), i;
194
195 /*
196 * Run 5 calibration loops to get the lowest frequency value
197 * (the best estimate). We use two different calibration modes
198 * here:
199 *
200 * 1) PIT loop. We set the PIT Channel 2 to oneshot mode and
201 * load a timeout of 50ms. We read the time right after we
202 * started the timer and wait until the PIT count down reaches
203 * zero. In each wait loop iteration we read the TSC and check
204 * the delta to the previous read. We keep track of the min
205 * and max values of that delta. The delta is mostly defined
206 * by the IO time of the PIT access, so we can detect when a
207 * SMI/SMM disturbance happend between the two reads. If the
208 * maximum time is significantly larger than the minimum time,
209 * then we discard the result and have another try.
210 *
211 * 2) Reference counter. If available we use the HPET or the
212 * PMTIMER as a reference to check the sanity of that value.
213 * We use separate TSC readouts and check inside of the
214 * reference read for a SMI/SMM disturbance. We dicard
215 * disturbed values here as well. We do that around the PIT
216 * calibration delay loop as we have to wait for a certain
217 * amount of time anyway.
218 */
219 for (i = 0; i < 5; i++) {
220 unsigned long tsc_pit_khz;
221
222 /*
223 * Read the start value and the reference count of
224 * hpet/pmtimer when available. Then do the PIT
225 * calibration, which will take at least 50ms, and
226 * read the end value.
227 */
228 local_irq_save(flags);
229 tsc1 = tsc_read_refs(&pm1, hpet ? &hpet1 : NULL);
230 tsc_pit_khz = pit_calibrate_tsc();
231 tsc2 = tsc_read_refs(&pm2, hpet ? &hpet2 : NULL);
232 local_irq_restore(flags);
233
234 /* Pick the lowest PIT TSC calibration so far */
235 tsc_pit_min = min(tsc_pit_min, tsc_pit_khz);
236
237 /* hpet or pmtimer available ? */
238 if (!hpet && !pm1 && !pm2)
239 continue;
240
241 /* Check, whether the sampling was disturbed by an SMI */
242 if (tsc1 == ULLONG_MAX || tsc2 == ULLONG_MAX)
243 continue;
244
245 tsc2 = (tsc2 - tsc1) * 1000000LL;
246
247 if (hpet) {
248 if (hpet2 < hpet1)
249 hpet2 += 0x100000000ULL;
250 hpet2 -= hpet1;
251 tsc1 = ((u64)hpet2 * hpet_readl(HPET_PERIOD));
252 do_div(tsc1, 1000000);
253 } else {
254 if (pm2 < pm1)
255 pm2 += (u64)ACPI_PM_OVRRUN;
256 pm2 -= pm1;
257 tsc1 = pm2 * 1000000000LL;
258 do_div(tsc1, PMTMR_TICKS_PER_SEC);
259 }
260
261 do_div(tsc2, tsc1);
262 tsc_ref_min = min(tsc_ref_min, (unsigned long) tsc2);
263 }
264
265 /*
266 * Now check the results.
267 */
268 if (tsc_pit_min == ULONG_MAX) {
269 /* PIT gave no useful value */
270 printk(KERN_WARNING "TSC: Unable to calibrate against PIT\n");
271
272 /* We don't have an alternative source, disable TSC */
273 if (!hpet && !pm1 && !pm2) {
274 printk("TSC: No reference (HPET/PMTIMER) available\n");
275 return 0;
276 }
277
278 /* The alternative source failed as well, disable TSC */
279 if (tsc_ref_min == ULONG_MAX) {
280 printk(KERN_WARNING "TSC: HPET/PMTIMER calibration "
281 "failed due to SMI disturbance.\n");
282 return 0;
283 }
284
285 /* Use the alternative source */
286 printk(KERN_INFO "TSC: using %s reference calibration\n",
287 hpet ? "HPET" : "PMTIMER");
288
289 return tsc_ref_min;
290 }
159 291
160 /* hpet or pmtimer available ? */ 292 /* We don't have an alternative source, use the PIT calibration value */
161 if (!hpet && !pm1 && !pm2) { 293 if (!hpet && !pm1 && !pm2) {
162 printk(KERN_INFO "TSC calibrated against PIT\n"); 294 printk(KERN_INFO "TSC: Using PIT calibration value\n");
163 goto out; 295 return tsc_pit_min;
164 } 296 }
165 297
166 /* Check, whether the sampling was disturbed by an SMI */ 298 /* The alternative source failed, use the PIT calibration value */
167 if (tsc1 == ULLONG_MAX || tsc2 == ULLONG_MAX) { 299 if (tsc_ref_min == ULONG_MAX) {
168 printk(KERN_WARNING "TSC calibration disturbed by SMI, " 300 printk(KERN_WARNING "TSC: HPET/PMTIMER calibration failed due "
169 "using PIT calibration result\n"); 301 "to SMI disturbance. Using PIT calibration\n");
170 goto out; 302 return tsc_pit_min;
171 } 303 }
172 304
173 tsc2 = (tsc2 - tsc1) * 1000000LL; 305 /* Check the reference deviation */
174 306 delta = ((u64) tsc_pit_min) * 100;
175 if (hpet) { 307 do_div(delta, tsc_ref_min);
176 printk(KERN_INFO "TSC calibrated against HPET\n"); 308
177 if (hpet2 < hpet1) 309 /*
178 hpet2 += 0x100000000ULL; 310 * If both calibration results are inside a 5% window, the we
179 hpet2 -= hpet1; 311 * use the lower frequency of those as it is probably the
180 tsc1 = ((u64)hpet2 * hpet_readl(HPET_PERIOD)); 312 * closest estimate.
181 do_div(tsc1, 1000000); 313 */
182 } else { 314 if (delta >= 95 && delta <= 105) {
183 printk(KERN_INFO "TSC calibrated against PM_TIMER\n"); 315 printk(KERN_INFO "TSC: PIT calibration confirmed by %s.\n",
184 if (pm2 < pm1) 316 hpet ? "HPET" : "PMTIMER");
185 pm2 += (u64)ACPI_PM_OVRRUN; 317 printk(KERN_INFO "TSC: using %s calibration value\n",
186 pm2 -= pm1; 318 tsc_pit_min <= tsc_ref_min ? "PIT" :
187 tsc1 = pm2 * 1000000000LL; 319 hpet ? "HPET" : "PMTIMER");
188 do_div(tsc1, PMTMR_TICKS_PER_SEC); 320 return tsc_pit_min <= tsc_ref_min ? tsc_pit_min : tsc_ref_min;
189 } 321 }
190 322
191 do_div(tsc2, tsc1); 323 printk(KERN_WARNING "TSC: PIT calibration deviates from %s: %lu %lu.\n",
192 tsc_khz_val = tsc2; 324 hpet ? "HPET" : "PMTIMER", tsc_pit_min, tsc_ref_min);
193 325
194out: 326 /*
195 return tsc_khz_val; 327 * The calibration values differ too much. In doubt, we use
328 * the PIT value as we know that there are PMTIMERs around
329 * running at double speed.
330 */
331 printk(KERN_INFO "TSC: Using PIT calibration value\n");
332 return tsc_pit_min;
196} 333}
197 334
198
199#ifdef CONFIG_X86_32 335#ifdef CONFIG_X86_32
200/* Only called from the Powernow K7 cpu freq driver */ 336/* Only called from the Powernow K7 cpu freq driver */
201int recalibrate_cpu_khz(void) 337int recalibrate_cpu_khz(void)
@@ -314,7 +450,7 @@ static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
314 mark_tsc_unstable("cpufreq changes"); 450 mark_tsc_unstable("cpufreq changes");
315 } 451 }
316 452
317 set_cyc2ns_scale(tsc_khz_ref, freq->cpu); 453 set_cyc2ns_scale(tsc_khz, freq->cpu);
318 454
319 return 0; 455 return 0;
320} 456}
@@ -325,6 +461,10 @@ static struct notifier_block time_cpufreq_notifier_block = {
325 461
326static int __init cpufreq_tsc(void) 462static int __init cpufreq_tsc(void)
327{ 463{
464 if (!cpu_has_tsc)
465 return 0;
466 if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
467 return 0;
328 cpufreq_register_notifier(&time_cpufreq_notifier_block, 468 cpufreq_register_notifier(&time_cpufreq_notifier_block,
329 CPUFREQ_TRANSITION_NOTIFIER); 469 CPUFREQ_TRANSITION_NOTIFIER);
330 return 0; 470 return 0;
diff --git a/arch/x86/kernel/tsc_sync.c b/arch/x86/kernel/tsc_sync.c
index 0577825cf89b..9ffb01c31c40 100644
--- a/arch/x86/kernel/tsc_sync.c
+++ b/arch/x86/kernel/tsc_sync.c
@@ -88,11 +88,9 @@ static __cpuinit void check_tsc_warp(void)
88 __raw_spin_unlock(&sync_lock); 88 __raw_spin_unlock(&sync_lock);
89 } 89 }
90 } 90 }
91 if (!(now-start)) { 91 WARN(!(now-start),
92 printk("Warning: zero tsc calibration delta: %Ld [max: %Ld]\n", 92 "Warning: zero tsc calibration delta: %Ld [max: %Ld]\n",
93 now-start, end-start); 93 now-start, end-start);
94 WARN_ON(1);
95 }
96} 94}
97 95
98/* 96/*
diff --git a/arch/x86/kernel/visws_quirks.c b/arch/x86/kernel/visws_quirks.c
index 41e01b145c48..594ef47f0a63 100644
--- a/arch/x86/kernel/visws_quirks.c
+++ b/arch/x86/kernel/visws_quirks.c
@@ -184,8 +184,6 @@ static int __init visws_get_smp_config(unsigned int early)
184 return 1; 184 return 1;
185} 185}
186 186
187extern unsigned int __cpuinitdata maxcpus;
188
189/* 187/*
190 * The Visual Workstation is Intel MP compliant in the hardware 188 * The Visual Workstation is Intel MP compliant in the hardware
191 * sense, but it doesn't have a BIOS(-configuration table). 189 * sense, but it doesn't have a BIOS(-configuration table).
@@ -244,8 +242,8 @@ static int __init visws_find_smp_config(unsigned int reserve)
244 ncpus = CO_CPU_MAX; 242 ncpus = CO_CPU_MAX;
245 } 243 }
246 244
247 if (ncpus > maxcpus) 245 if (ncpus > setup_max_cpus)
248 ncpus = maxcpus; 246 ncpus = setup_max_cpus;
249 247
250#ifdef CONFIG_X86_LOCAL_APIC 248#ifdef CONFIG_X86_LOCAL_APIC
251 smp_found_config = 1; 249 smp_found_config = 1;
diff --git a/arch/x86/kernel/vmi_32.c b/arch/x86/kernel/vmi_32.c
index 0a1b1a9d922d..6ca515d6db54 100644
--- a/arch/x86/kernel/vmi_32.c
+++ b/arch/x86/kernel/vmi_32.c
@@ -37,6 +37,7 @@
37#include <asm/timer.h> 37#include <asm/timer.h>
38#include <asm/vmi_time.h> 38#include <asm/vmi_time.h>
39#include <asm/kmap_types.h> 39#include <asm/kmap_types.h>
40#include <asm/setup.h>
40 41
41/* Convenient for calling VMI functions indirectly in the ROM */ 42/* Convenient for calling VMI functions indirectly in the ROM */
42typedef u32 __attribute__((regparm(1))) (VROMFUNC)(void); 43typedef u32 __attribute__((regparm(1))) (VROMFUNC)(void);
@@ -683,7 +684,7 @@ void vmi_bringup(void)
683{ 684{
684 /* We must establish the lowmem mapping for MMU ops to work */ 685 /* We must establish the lowmem mapping for MMU ops to work */
685 if (vmi_ops.set_linear_mapping) 686 if (vmi_ops.set_linear_mapping)
686 vmi_ops.set_linear_mapping(0, (void *)__PAGE_OFFSET, max_low_pfn, 0); 687 vmi_ops.set_linear_mapping(0, (void *)__PAGE_OFFSET, MAXMEM_PFN, 0);
687} 688}
688 689
689/* 690/*
diff --git a/arch/x86/kernel/vmlinux_32.lds.S b/arch/x86/kernel/vmlinux_32.lds.S
index cdb2363697d2..af5bdad84604 100644
--- a/arch/x86/kernel/vmlinux_32.lds.S
+++ b/arch/x86/kernel/vmlinux_32.lds.S
@@ -209,3 +209,11 @@ SECTIONS
209 209
210 DWARF_DEBUG 210 DWARF_DEBUG
211} 211}
212
213#ifdef CONFIG_KEXEC
214/* Link time checks */
215#include <asm/kexec.h>
216
217ASSERT(kexec_control_code_size <= KEXEC_CONTROL_CODE_MAX_SIZE,
218 "kexec control code size is too big")
219#endif
generated by cgit v1.2.2 (git 2.25.0) at 2025-08-12 15:48:37 -0400