aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel
diff options
context:
space:
mode:
authorIngo Molnar <mingo@elte.hu>2009-04-21 04:46:39 -0400
committerIngo Molnar <mingo@elte.hu>2009-04-21 04:46:54 -0400
commit8ecee4620e76aae418bfa0e8cc830e92cb559bbb (patch)
tree49a36784c0a26c8494a37087e37502101013b35d /arch/x86/kernel
parent6424fb38667fffbbb1b90be0ffd9a0c540db6a4b (diff)
parenta939b96cccdb65df80a52447ec8e4a6d79c56dbb (diff)
Merge branch 'linus' into x86/mm
Merge reason: refresh the topic: there's been 290 non-merges commits upstream to arch/x86 alone. Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'arch/x86/kernel')
-rw-r--r--arch/x86/kernel/Makefile7
-rw-r--r--arch/x86/kernel/acpi/boot.c65
-rw-r--r--arch/x86/kernel/alternative.c29
-rw-r--r--arch/x86/kernel/amd_iommu.c33
-rw-r--r--arch/x86/kernel/apic/apic.c76
-rw-r--r--arch/x86/kernel/apic/apic_flat_64.c22
-rw-r--r--arch/x86/kernel/apic/bigsmp_32.c16
-rw-r--r--arch/x86/kernel/apic/es7000_32.c17
-rw-r--r--arch/x86/kernel/apic/io_apic.c162
-rw-r--r--arch/x86/kernel/apic/nmi.c11
-rw-r--r--arch/x86/kernel/apic/numaq_32.c11
-rw-r--r--arch/x86/kernel/apic/probe_32.c3
-rw-r--r--arch/x86/kernel/apic/summit_32.c21
-rw-r--r--arch/x86/kernel/apic/x2apic_uv_x.c25
-rw-r--r--arch/x86/kernel/apm_32.c263
-rw-r--r--arch/x86/kernel/asm-offsets_32.c1
-rw-r--r--arch/x86/kernel/asm-offsets_64.c1
-rw-r--r--arch/x86/kernel/bios_uv.c3
-rw-r--r--arch/x86/kernel/cpu/addon_cpuid_features.c1
-rw-r--r--arch/x86/kernel/cpu/common.c13
-rw-r--r--arch/x86/kernel/cpu/cpu.h20
-rw-r--r--[-rwxr-xr-x]arch/x86/kernel/cpu/cpu_debug.c0
-rw-r--r--arch/x86/kernel/cpu/cpufreq/Kconfig19
-rw-r--r--arch/x86/kernel/cpu/cpufreq/Makefile8
-rw-r--r--arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c133
-rw-r--r--arch/x86/kernel/cpu/cpufreq/cpufreq-nforce2.c54
-rw-r--r--arch/x86/kernel/cpu/cpufreq/e_powersaver.c21
-rw-r--r--arch/x86/kernel/cpu/cpufreq/elanfreq.c6
-rw-r--r--arch/x86/kernel/cpu/cpufreq/gx-suspmod.c105
-rw-r--r--arch/x86/kernel/cpu/cpufreq/longhaul.c198
-rw-r--r--arch/x86/kernel/cpu/cpufreq/longhaul.h12
-rw-r--r--arch/x86/kernel/cpu/cpufreq/longrun.c25
-rw-r--r--arch/x86/kernel/cpu/cpufreq/p4-clockmod.c74
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k6.c44
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k7.c239
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k8.c399
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k8.h5
-rw-r--r--arch/x86/kernel/cpu/cpufreq/sc520_freq.c30
-rw-r--r--arch/x86/kernel/cpu/cpufreq/speedstep-ich.c72
-rw-r--r--arch/x86/kernel/cpu/cpufreq/speedstep-lib.c163
-rw-r--r--arch/x86/kernel/cpu/cpufreq/speedstep-lib.h18
-rw-r--r--arch/x86/kernel/cpu/cpufreq/speedstep-smi.c166
-rw-r--r--arch/x86/kernel/cpu/intel.c8
-rw-r--r--arch/x86/kernel/cpu/intel_cacheinfo.c24
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_64.c10
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_amd_64.c6
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_intel_64.c2
-rw-r--r--arch/x86/kernel/cpu/mtrr/generic.c3
-rw-r--r--arch/x86/kernel/cpu/mtrr/if.c10
-rw-r--r--arch/x86/kernel/cpu/proc.c6
-rw-r--r--arch/x86/kernel/dumpstack.c7
-rw-r--r--arch/x86/kernel/e820.c10
-rw-r--r--arch/x86/kernel/ftrace.c268
-rw-r--r--arch/x86/kernel/hpet.c80
-rw-r--r--arch/x86/kernel/irq.c25
-rw-r--r--arch/x86/kernel/irqinit_32.c2
-rw-r--r--arch/x86/kernel/irqinit_64.c1
-rw-r--r--arch/x86/kernel/kprobes.c17
-rw-r--r--arch/x86/kernel/mfgpt_32.c1
-rw-r--r--arch/x86/kernel/microcode_amd.c43
-rw-r--r--arch/x86/kernel/microcode_core.c125
-rw-r--r--arch/x86/kernel/microcode_intel.c83
-rw-r--r--arch/x86/kernel/mpparse.c7
-rw-r--r--arch/x86/kernel/pci-calgary_64.c38
-rw-r--r--arch/x86/kernel/pci-dma.c25
-rw-r--r--arch/x86/kernel/pci-gart_64.c34
-rw-r--r--arch/x86/kernel/pci-nommu.c23
-rw-r--r--arch/x86/kernel/pci-swiotlb.c (renamed from arch/x86/kernel/pci-swiotlb_64.c)19
-rw-r--r--arch/x86/kernel/process.c25
-rw-r--r--arch/x86/kernel/process_32.c2
-rw-r--r--arch/x86/kernel/process_64.c2
-rw-r--r--arch/x86/kernel/ptrace.c10
-rw-r--r--arch/x86/kernel/quirks.c3
-rw-r--r--arch/x86/kernel/reboot.c8
-rw-r--r--arch/x86/kernel/setup.c1
-rw-r--r--arch/x86/kernel/setup_percpu.c16
-rw-r--r--arch/x86/kernel/signal.c2
-rw-r--r--arch/x86/kernel/smpboot.c70
-rw-r--r--arch/x86/kernel/syscall_table_32.S2
-rw-r--r--arch/x86/kernel/time_64.c2
-rw-r--r--arch/x86/kernel/tlb_uv.c198
-rw-r--r--arch/x86/kernel/tsc.c12
-rw-r--r--arch/x86/kernel/uv_sysfs.c4
-rw-r--r--arch/x86/kernel/vmiclock_32.c1
-rw-r--r--arch/x86/kernel/vsmp_64.c12
-rw-r--r--arch/x86/kernel/xsave.c2
86 files changed, 2252 insertions, 1588 deletions
diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 339ce35648e6..145cce75cda7 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -66,11 +66,11 @@ obj-$(CONFIG_X86_MPPARSE) += mpparse.o
66obj-y += apic/ 66obj-y += apic/
67obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o 67obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o
68obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o 68obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o
69obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o 69obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o
70obj-$(CONFIG_FTRACE_SYSCALLS) += ftrace.o
70obj-$(CONFIG_KEXEC) += machine_kexec_$(BITS).o 71obj-$(CONFIG_KEXEC) += machine_kexec_$(BITS).o
71obj-$(CONFIG_KEXEC) += relocate_kernel_$(BITS).o crash.o 72obj-$(CONFIG_KEXEC) += relocate_kernel_$(BITS).o crash.o
72obj-$(CONFIG_CRASH_DUMP) += crash_dump_$(BITS).o 73obj-$(CONFIG_CRASH_DUMP) += crash_dump_$(BITS).o
73obj-$(CONFIG_X86_VSMP) += vsmp_64.o
74obj-$(CONFIG_KPROBES) += kprobes.o 74obj-$(CONFIG_KPROBES) += kprobes.o
75obj-$(CONFIG_MODULES) += module_$(BITS).o 75obj-$(CONFIG_MODULES) += module_$(BITS).o
76obj-$(CONFIG_EFI) += efi.o efi_$(BITS).o efi_stub_$(BITS).o 76obj-$(CONFIG_EFI) += efi.o efi_$(BITS).o efi_stub_$(BITS).o
@@ -106,7 +106,7 @@ obj-$(CONFIG_MICROCODE) += microcode.o
106 106
107obj-$(CONFIG_X86_CHECK_BIOS_CORRUPTION) += check.o 107obj-$(CONFIG_X86_CHECK_BIOS_CORRUPTION) += check.o
108 108
109obj-$(CONFIG_SWIOTLB) += pci-swiotlb_64.o # NB rename without _64 109obj-$(CONFIG_SWIOTLB) += pci-swiotlb.o
110 110
111### 111###
112# 64 bit specific files 112# 64 bit specific files
@@ -120,4 +120,5 @@ ifeq ($(CONFIG_X86_64),y)
120 obj-$(CONFIG_AMD_IOMMU) += amd_iommu_init.o amd_iommu.o 120 obj-$(CONFIG_AMD_IOMMU) += amd_iommu_init.o amd_iommu.o
121 121
122 obj-$(CONFIG_PCI_MMCONFIG) += mmconf-fam10h_64.o 122 obj-$(CONFIG_PCI_MMCONFIG) += mmconf-fam10h_64.o
123 obj-y += vsmp_64.o
123endif 124endif
diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index a18eb7ce2236..723989d7f802 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -230,6 +230,35 @@ static void __cpuinit acpi_register_lapic(int id, u8 enabled)
230} 230}
231 231
232static int __init 232static int __init
233acpi_parse_x2apic(struct acpi_subtable_header *header, const unsigned long end)
234{
235 struct acpi_madt_local_x2apic *processor = NULL;
236
237 processor = (struct acpi_madt_local_x2apic *)header;
238
239 if (BAD_MADT_ENTRY(processor, end))
240 return -EINVAL;
241
242 acpi_table_print_madt_entry(header);
243
244#ifdef CONFIG_X86_X2APIC
245 /*
246 * We need to register disabled CPU as well to permit
247 * counting disabled CPUs. This allows us to size
248 * cpus_possible_map more accurately, to permit
249 * to not preallocating memory for all NR_CPUS
250 * when we use CPU hotplug.
251 */
252 acpi_register_lapic(processor->local_apic_id, /* APIC ID */
253 processor->lapic_flags & ACPI_MADT_ENABLED);
254#else
255 printk(KERN_WARNING PREFIX "x2apic entry ignored\n");
256#endif
257
258 return 0;
259}
260
261static int __init
233acpi_parse_lapic(struct acpi_subtable_header * header, const unsigned long end) 262acpi_parse_lapic(struct acpi_subtable_header * header, const unsigned long end)
234{ 263{
235 struct acpi_madt_local_apic *processor = NULL; 264 struct acpi_madt_local_apic *processor = NULL;
@@ -289,6 +318,25 @@ acpi_parse_lapic_addr_ovr(struct acpi_subtable_header * header,
289} 318}
290 319
291static int __init 320static int __init
321acpi_parse_x2apic_nmi(struct acpi_subtable_header *header,
322 const unsigned long end)
323{
324 struct acpi_madt_local_x2apic_nmi *x2apic_nmi = NULL;
325
326 x2apic_nmi = (struct acpi_madt_local_x2apic_nmi *)header;
327
328 if (BAD_MADT_ENTRY(x2apic_nmi, end))
329 return -EINVAL;
330
331 acpi_table_print_madt_entry(header);
332
333 if (x2apic_nmi->lint != 1)
334 printk(KERN_WARNING PREFIX "NMI not connected to LINT 1!\n");
335
336 return 0;
337}
338
339static int __init
292acpi_parse_lapic_nmi(struct acpi_subtable_header * header, const unsigned long end) 340acpi_parse_lapic_nmi(struct acpi_subtable_header * header, const unsigned long end)
293{ 341{
294 struct acpi_madt_local_apic_nmi *lapic_nmi = NULL; 342 struct acpi_madt_local_apic_nmi *lapic_nmi = NULL;
@@ -793,6 +841,7 @@ static int __init early_acpi_parse_madt_lapic_addr_ovr(void)
793static int __init acpi_parse_madt_lapic_entries(void) 841static int __init acpi_parse_madt_lapic_entries(void)
794{ 842{
795 int count; 843 int count;
844 int x2count = 0;
796 845
797 if (!cpu_has_apic) 846 if (!cpu_has_apic)
798 return -ENODEV; 847 return -ENODEV;
@@ -816,22 +865,28 @@ static int __init acpi_parse_madt_lapic_entries(void)
816 count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_SAPIC, 865 count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_SAPIC,
817 acpi_parse_sapic, MAX_APICS); 866 acpi_parse_sapic, MAX_APICS);
818 867
819 if (!count) 868 if (!count) {
869 x2count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_X2APIC,
870 acpi_parse_x2apic, MAX_APICS);
820 count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC, 871 count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC,
821 acpi_parse_lapic, MAX_APICS); 872 acpi_parse_lapic, MAX_APICS);
822 if (!count) { 873 }
874 if (!count && !x2count) {
823 printk(KERN_ERR PREFIX "No LAPIC entries present\n"); 875 printk(KERN_ERR PREFIX "No LAPIC entries present\n");
824 /* TBD: Cleanup to allow fallback to MPS */ 876 /* TBD: Cleanup to allow fallback to MPS */
825 return -ENODEV; 877 return -ENODEV;
826 } else if (count < 0) { 878 } else if (count < 0 || x2count < 0) {
827 printk(KERN_ERR PREFIX "Error parsing LAPIC entry\n"); 879 printk(KERN_ERR PREFIX "Error parsing LAPIC entry\n");
828 /* TBD: Cleanup to allow fallback to MPS */ 880 /* TBD: Cleanup to allow fallback to MPS */
829 return count; 881 return count;
830 } 882 }
831 883
884 x2count =
885 acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_X2APIC_NMI,
886 acpi_parse_x2apic_nmi, 0);
832 count = 887 count =
833 acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC_NMI, acpi_parse_lapic_nmi, 0); 888 acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC_NMI, acpi_parse_lapic_nmi, 0);
834 if (count < 0) { 889 if (count < 0 || x2count < 0) {
835 printk(KERN_ERR PREFIX "Error parsing LAPIC NMI entry\n"); 890 printk(KERN_ERR PREFIX "Error parsing LAPIC NMI entry\n");
836 /* TBD: Cleanup to allow fallback to MPS */ 891 /* TBD: Cleanup to allow fallback to MPS */
837 return count; 892 return count;
@@ -1470,7 +1525,7 @@ static int __init dmi_ignore_irq0_timer_override(const struct dmi_system_id *d)
1470 1525
1471/* 1526/*
1472 * If your system is blacklisted here, but you find that acpi=force 1527 * If your system is blacklisted here, but you find that acpi=force
1473 * works for you, please contact acpi-devel@sourceforge.net 1528 * works for you, please contact linux-acpi@vger.kernel.org
1474 */ 1529 */
1475static struct dmi_system_id __initdata acpi_dmi_table[] = { 1530static struct dmi_system_id __initdata acpi_dmi_table[] = {
1476 /* 1531 /*
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 4c80f1557433..f57658702571 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -5,6 +5,7 @@
5#include <linux/kprobes.h> 5#include <linux/kprobes.h>
6#include <linux/mm.h> 6#include <linux/mm.h>
7#include <linux/vmalloc.h> 7#include <linux/vmalloc.h>
8#include <linux/memory.h>
8#include <asm/alternative.h> 9#include <asm/alternative.h>
9#include <asm/sections.h> 10#include <asm/sections.h>
10#include <asm/pgtable.h> 11#include <asm/pgtable.h>
@@ -12,7 +13,9 @@
12#include <asm/nmi.h> 13#include <asm/nmi.h>
13#include <asm/vsyscall.h> 14#include <asm/vsyscall.h>
14#include <asm/cacheflush.h> 15#include <asm/cacheflush.h>
16#include <asm/tlbflush.h>
15#include <asm/io.h> 17#include <asm/io.h>
18#include <asm/fixmap.h>
16 19
17#define MAX_PATCH_LEN (255-1) 20#define MAX_PATCH_LEN (255-1)
18 21
@@ -226,6 +229,7 @@ static void alternatives_smp_lock(u8 **start, u8 **end, u8 *text, u8 *text_end)
226{ 229{
227 u8 **ptr; 230 u8 **ptr;
228 231
232 mutex_lock(&text_mutex);
229 for (ptr = start; ptr < end; ptr++) { 233 for (ptr = start; ptr < end; ptr++) {
230 if (*ptr < text) 234 if (*ptr < text)
231 continue; 235 continue;
@@ -234,6 +238,7 @@ static void alternatives_smp_lock(u8 **start, u8 **end, u8 *text, u8 *text_end)
234 /* turn DS segment override prefix into lock prefix */ 238 /* turn DS segment override prefix into lock prefix */
235 text_poke(*ptr, ((unsigned char []){0xf0}), 1); 239 text_poke(*ptr, ((unsigned char []){0xf0}), 1);
236 }; 240 };
241 mutex_unlock(&text_mutex);
237} 242}
238 243
239static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end) 244static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end)
@@ -243,6 +248,7 @@ static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end
243 if (noreplace_smp) 248 if (noreplace_smp)
244 return; 249 return;
245 250
251 mutex_lock(&text_mutex);
246 for (ptr = start; ptr < end; ptr++) { 252 for (ptr = start; ptr < end; ptr++) {
247 if (*ptr < text) 253 if (*ptr < text)
248 continue; 254 continue;
@@ -251,6 +257,7 @@ static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end
251 /* turn lock prefix into DS segment override prefix */ 257 /* turn lock prefix into DS segment override prefix */
252 text_poke(*ptr, ((unsigned char []){0x3E}), 1); 258 text_poke(*ptr, ((unsigned char []){0x3E}), 1);
253 }; 259 };
260 mutex_unlock(&text_mutex);
254} 261}
255 262
256struct smp_alt_module { 263struct smp_alt_module {
@@ -500,15 +507,16 @@ void *text_poke_early(void *addr, const void *opcode, size_t len)
500 * It means the size must be writable atomically and the address must be aligned 507 * It means the size must be writable atomically and the address must be aligned
501 * in a way that permits an atomic write. It also makes sure we fit on a single 508 * in a way that permits an atomic write. It also makes sure we fit on a single
502 * page. 509 * page.
510 *
511 * Note: Must be called under text_mutex.
503 */ 512 */
504void *__kprobes text_poke(void *addr, const void *opcode, size_t len) 513void *__kprobes text_poke(void *addr, const void *opcode, size_t len)
505{ 514{
515 unsigned long flags;
506 char *vaddr; 516 char *vaddr;
507 int nr_pages = 2;
508 struct page *pages[2]; 517 struct page *pages[2];
509 int i; 518 int i;
510 519
511 might_sleep();
512 if (!core_kernel_text((unsigned long)addr)) { 520 if (!core_kernel_text((unsigned long)addr)) {
513 pages[0] = vmalloc_to_page(addr); 521 pages[0] = vmalloc_to_page(addr);
514 pages[1] = vmalloc_to_page(addr + PAGE_SIZE); 522 pages[1] = vmalloc_to_page(addr + PAGE_SIZE);
@@ -518,18 +526,21 @@ void *__kprobes text_poke(void *addr, const void *opcode, size_t len)
518 pages[1] = virt_to_page(addr + PAGE_SIZE); 526 pages[1] = virt_to_page(addr + PAGE_SIZE);
519 } 527 }
520 BUG_ON(!pages[0]); 528 BUG_ON(!pages[0]);
521 if (!pages[1]) 529 local_irq_save(flags);
522 nr_pages = 1; 530 set_fixmap(FIX_TEXT_POKE0, page_to_phys(pages[0]));
523 vaddr = vmap(pages, nr_pages, VM_MAP, PAGE_KERNEL); 531 if (pages[1])
524 BUG_ON(!vaddr); 532 set_fixmap(FIX_TEXT_POKE1, page_to_phys(pages[1]));
525 local_irq_disable(); 533 vaddr = (char *)fix_to_virt(FIX_TEXT_POKE0);
526 memcpy(&vaddr[(unsigned long)addr & ~PAGE_MASK], opcode, len); 534 memcpy(&vaddr[(unsigned long)addr & ~PAGE_MASK], opcode, len);
527 local_irq_enable(); 535 clear_fixmap(FIX_TEXT_POKE0);
528 vunmap(vaddr); 536 if (pages[1])
537 clear_fixmap(FIX_TEXT_POKE1);
538 local_flush_tlb();
529 sync_core(); 539 sync_core();
530 /* Could also do a CLFLUSH here to speed up CPU recovery; but 540 /* Could also do a CLFLUSH here to speed up CPU recovery; but
531 that causes hangs on some VIA CPUs. */ 541 that causes hangs on some VIA CPUs. */
532 for (i = 0; i < len; i++) 542 for (i = 0; i < len; i++)
533 BUG_ON(((char *)addr)[i] != ((char *)opcode)[i]); 543 BUG_ON(((char *)addr)[i] != ((char *)opcode)[i]);
544 local_irq_restore(flags);
534 return addr; 545 return addr;
535} 546}
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 5113c080f0c4..a97db99dad52 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -22,10 +22,9 @@
22#include <linux/bitops.h> 22#include <linux/bitops.h>
23#include <linux/debugfs.h> 23#include <linux/debugfs.h>
24#include <linux/scatterlist.h> 24#include <linux/scatterlist.h>
25#include <linux/dma-mapping.h>
25#include <linux/iommu-helper.h> 26#include <linux/iommu-helper.h>
26#ifdef CONFIG_IOMMU_API
27#include <linux/iommu.h> 27#include <linux/iommu.h>
28#endif
29#include <asm/proto.h> 28#include <asm/proto.h>
30#include <asm/iommu.h> 29#include <asm/iommu.h>
31#include <asm/gart.h> 30#include <asm/gart.h>
@@ -1297,8 +1296,10 @@ static void __unmap_single(struct amd_iommu *iommu,
1297/* 1296/*
1298 * The exported map_single function for dma_ops. 1297 * The exported map_single function for dma_ops.
1299 */ 1298 */
1300static dma_addr_t map_single(struct device *dev, phys_addr_t paddr, 1299static dma_addr_t map_page(struct device *dev, struct page *page,
1301 size_t size, int dir) 1300 unsigned long offset, size_t size,
1301 enum dma_data_direction dir,
1302 struct dma_attrs *attrs)
1302{ 1303{
1303 unsigned long flags; 1304 unsigned long flags;
1304 struct amd_iommu *iommu; 1305 struct amd_iommu *iommu;
@@ -1306,6 +1307,7 @@ static dma_addr_t map_single(struct device *dev, phys_addr_t paddr,
1306 u16 devid; 1307 u16 devid;
1307 dma_addr_t addr; 1308 dma_addr_t addr;
1308 u64 dma_mask; 1309 u64 dma_mask;
1310 phys_addr_t paddr = page_to_phys(page) + offset;
1309 1311
1310 INC_STATS_COUNTER(cnt_map_single); 1312 INC_STATS_COUNTER(cnt_map_single);
1311 1313
@@ -1340,8 +1342,8 @@ out:
1340/* 1342/*
1341 * The exported unmap_single function for dma_ops. 1343 * The exported unmap_single function for dma_ops.
1342 */ 1344 */
1343static void unmap_single(struct device *dev, dma_addr_t dma_addr, 1345static void unmap_page(struct device *dev, dma_addr_t dma_addr, size_t size,
1344 size_t size, int dir) 1346 enum dma_data_direction dir, struct dma_attrs *attrs)
1345{ 1347{
1346 unsigned long flags; 1348 unsigned long flags;
1347 struct amd_iommu *iommu; 1349 struct amd_iommu *iommu;
@@ -1390,7 +1392,8 @@ static int map_sg_no_iommu(struct device *dev, struct scatterlist *sglist,
1390 * lists). 1392 * lists).
1391 */ 1393 */
1392static int map_sg(struct device *dev, struct scatterlist *sglist, 1394static int map_sg(struct device *dev, struct scatterlist *sglist,
1393 int nelems, int dir) 1395 int nelems, enum dma_data_direction dir,
1396 struct dma_attrs *attrs)
1394{ 1397{
1395 unsigned long flags; 1398 unsigned long flags;
1396 struct amd_iommu *iommu; 1399 struct amd_iommu *iommu;
@@ -1457,7 +1460,8 @@ unmap:
1457 * lists). 1460 * lists).
1458 */ 1461 */
1459static void unmap_sg(struct device *dev, struct scatterlist *sglist, 1462static void unmap_sg(struct device *dev, struct scatterlist *sglist,
1460 int nelems, int dir) 1463 int nelems, enum dma_data_direction dir,
1464 struct dma_attrs *attrs)
1461{ 1465{
1462 unsigned long flags; 1466 unsigned long flags;
1463 struct amd_iommu *iommu; 1467 struct amd_iommu *iommu;
@@ -1644,11 +1648,11 @@ static void prealloc_protection_domains(void)
1644 } 1648 }
1645} 1649}
1646 1650
1647static struct dma_mapping_ops amd_iommu_dma_ops = { 1651static struct dma_map_ops amd_iommu_dma_ops = {
1648 .alloc_coherent = alloc_coherent, 1652 .alloc_coherent = alloc_coherent,
1649 .free_coherent = free_coherent, 1653 .free_coherent = free_coherent,
1650 .map_single = map_single, 1654 .map_page = map_page,
1651 .unmap_single = unmap_single, 1655 .unmap_page = unmap_page,
1652 .map_sg = map_sg, 1656 .map_sg = map_sg,
1653 .unmap_sg = unmap_sg, 1657 .unmap_sg = unmap_sg,
1654 .dma_supported = amd_iommu_dma_supported, 1658 .dma_supported = amd_iommu_dma_supported,
@@ -1924,6 +1928,12 @@ static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
1924 return paddr; 1928 return paddr;
1925} 1929}
1926 1930
1931static int amd_iommu_domain_has_cap(struct iommu_domain *domain,
1932 unsigned long cap)
1933{
1934 return 0;
1935}
1936
1927static struct iommu_ops amd_iommu_ops = { 1937static struct iommu_ops amd_iommu_ops = {
1928 .domain_init = amd_iommu_domain_init, 1938 .domain_init = amd_iommu_domain_init,
1929 .domain_destroy = amd_iommu_domain_destroy, 1939 .domain_destroy = amd_iommu_domain_destroy,
@@ -1932,5 +1942,6 @@ static struct iommu_ops amd_iommu_ops = {
1932 .map = amd_iommu_map_range, 1942 .map = amd_iommu_map_range,
1933 .unmap = amd_iommu_unmap_range, 1943 .unmap = amd_iommu_unmap_range,
1934 .iova_to_phys = amd_iommu_iova_to_phys, 1944 .iova_to_phys = amd_iommu_iova_to_phys,
1945 .domain_has_cap = amd_iommu_domain_has_cap,
1935}; 1946};
1936 1947
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 85eb8e100818..f2870920f246 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -431,6 +431,12 @@ static void __cpuinit setup_APIC_timer(void)
431{ 431{
432 struct clock_event_device *levt = &__get_cpu_var(lapic_events); 432 struct clock_event_device *levt = &__get_cpu_var(lapic_events);
433 433
434 if (cpu_has(&current_cpu_data, X86_FEATURE_ARAT)) {
435 lapic_clockevent.features &= ~CLOCK_EVT_FEAT_C3STOP;
436 /* Make LAPIC timer preferrable over percpu HPET */
437 lapic_clockevent.rating = 150;
438 }
439
434 memcpy(levt, &lapic_clockevent, sizeof(*levt)); 440 memcpy(levt, &lapic_clockevent, sizeof(*levt));
435 levt->cpumask = cpumask_of(smp_processor_id()); 441 levt->cpumask = cpumask_of(smp_processor_id());
436 442
@@ -1304,6 +1310,7 @@ void __init enable_IR_x2apic(void)
1304#ifdef CONFIG_INTR_REMAP 1310#ifdef CONFIG_INTR_REMAP
1305 int ret; 1311 int ret;
1306 unsigned long flags; 1312 unsigned long flags;
1313 struct IO_APIC_route_entry **ioapic_entries = NULL;
1307 1314
1308 if (!cpu_has_x2apic) 1315 if (!cpu_has_x2apic)
1309 return; 1316 return;
@@ -1334,17 +1341,23 @@ void __init enable_IR_x2apic(void)
1334 return; 1341 return;
1335 } 1342 }
1336 1343
1337 ret = save_IO_APIC_setup(); 1344 ioapic_entries = alloc_ioapic_entries();
1345 if (!ioapic_entries) {
1346 pr_info("Allocate ioapic_entries failed: %d\n", ret);
1347 goto end;
1348 }
1349
1350 ret = save_IO_APIC_setup(ioapic_entries);
1338 if (ret) { 1351 if (ret) {
1339 pr_info("Saving IO-APIC state failed: %d\n", ret); 1352 pr_info("Saving IO-APIC state failed: %d\n", ret);
1340 goto end; 1353 goto end;
1341 } 1354 }
1342 1355
1343 local_irq_save(flags); 1356 local_irq_save(flags);
1344 mask_IO_APIC_setup(); 1357 mask_IO_APIC_setup(ioapic_entries);
1345 mask_8259A(); 1358 mask_8259A();
1346 1359
1347 ret = enable_intr_remapping(1); 1360 ret = enable_intr_remapping(EIM_32BIT_APIC_ID);
1348 1361
1349 if (ret && x2apic_preenabled) { 1362 if (ret && x2apic_preenabled) {
1350 local_irq_restore(flags); 1363 local_irq_restore(flags);
@@ -1364,9 +1377,9 @@ end_restore:
1364 /* 1377 /*
1365 * IR enabling failed 1378 * IR enabling failed
1366 */ 1379 */
1367 restore_IO_APIC_setup(); 1380 restore_IO_APIC_setup(ioapic_entries);
1368 else 1381 else
1369 reinit_intr_remapped_IO_APIC(x2apic_preenabled); 1382 reinit_intr_remapped_IO_APIC(x2apic_preenabled, ioapic_entries);
1370 1383
1371 unmask_8259A(); 1384 unmask_8259A();
1372 local_irq_restore(flags); 1385 local_irq_restore(flags);
@@ -1379,6 +1392,8 @@ end:
1379 pr_info("Enabled Interrupt-remapping\n"); 1392 pr_info("Enabled Interrupt-remapping\n");
1380 } else 1393 } else
1381 pr_err("Failed to enable Interrupt-remapping and x2apic\n"); 1394 pr_err("Failed to enable Interrupt-remapping and x2apic\n");
1395 if (ioapic_entries)
1396 free_ioapic_entries(ioapic_entries);
1382#else 1397#else
1383 if (!cpu_has_x2apic) 1398 if (!cpu_has_x2apic)
1384 return; 1399 return;
@@ -1954,6 +1969,10 @@ static int lapic_suspend(struct sys_device *dev, pm_message_t state)
1954 1969
1955 local_irq_save(flags); 1970 local_irq_save(flags);
1956 disable_local_APIC(); 1971 disable_local_APIC();
1972#ifdef CONFIG_INTR_REMAP
1973 if (intr_remapping_enabled)
1974 disable_intr_remapping();
1975#endif
1957 local_irq_restore(flags); 1976 local_irq_restore(flags);
1958 return 0; 1977 return 0;
1959} 1978}
@@ -1964,15 +1983,41 @@ static int lapic_resume(struct sys_device *dev)
1964 unsigned long flags; 1983 unsigned long flags;
1965 int maxlvt; 1984 int maxlvt;
1966 1985
1986#ifdef CONFIG_INTR_REMAP
1987 int ret;
1988 struct IO_APIC_route_entry **ioapic_entries = NULL;
1989
1967 if (!apic_pm_state.active) 1990 if (!apic_pm_state.active)
1968 return 0; 1991 return 0;
1969 1992
1970 maxlvt = lapic_get_maxlvt();
1971
1972 local_irq_save(flags); 1993 local_irq_save(flags);
1994 if (x2apic) {
1995 ioapic_entries = alloc_ioapic_entries();
1996 if (!ioapic_entries) {
1997 WARN(1, "Alloc ioapic_entries in lapic resume failed.");
1998 return -ENOMEM;
1999 }
2000
2001 ret = save_IO_APIC_setup(ioapic_entries);
2002 if (ret) {
2003 WARN(1, "Saving IO-APIC state failed: %d\n", ret);
2004 free_ioapic_entries(ioapic_entries);
2005 return ret;
2006 }
2007
2008 mask_IO_APIC_setup(ioapic_entries);
2009 mask_8259A();
2010 enable_x2apic();
2011 }
2012#else
2013 if (!apic_pm_state.active)
2014 return 0;
1973 2015
2016 local_irq_save(flags);
1974 if (x2apic) 2017 if (x2apic)
1975 enable_x2apic(); 2018 enable_x2apic();
2019#endif
2020
1976 else { 2021 else {
1977 /* 2022 /*
1978 * Make sure the APICBASE points to the right address 2023 * Make sure the APICBASE points to the right address
@@ -1986,6 +2031,7 @@ static int lapic_resume(struct sys_device *dev)
1986 wrmsr(MSR_IA32_APICBASE, l, h); 2031 wrmsr(MSR_IA32_APICBASE, l, h);
1987 } 2032 }
1988 2033
2034 maxlvt = lapic_get_maxlvt();
1989 apic_write(APIC_LVTERR, ERROR_APIC_VECTOR | APIC_LVT_MASKED); 2035 apic_write(APIC_LVTERR, ERROR_APIC_VECTOR | APIC_LVT_MASKED);
1990 apic_write(APIC_ID, apic_pm_state.apic_id); 2036 apic_write(APIC_ID, apic_pm_state.apic_id);
1991 apic_write(APIC_DFR, apic_pm_state.apic_dfr); 2037 apic_write(APIC_DFR, apic_pm_state.apic_dfr);
@@ -2009,8 +2055,20 @@ static int lapic_resume(struct sys_device *dev)
2009 apic_write(APIC_ESR, 0); 2055 apic_write(APIC_ESR, 0);
2010 apic_read(APIC_ESR); 2056 apic_read(APIC_ESR);
2011 2057
2058#ifdef CONFIG_INTR_REMAP
2059 if (intr_remapping_enabled)
2060 reenable_intr_remapping(EIM_32BIT_APIC_ID);
2061
2062 if (x2apic) {
2063 unmask_8259A();
2064 restore_IO_APIC_setup(ioapic_entries);
2065 free_ioapic_entries(ioapic_entries);
2066 }
2067#endif
2068
2012 local_irq_restore(flags); 2069 local_irq_restore(flags);
2013 2070
2071
2014 return 0; 2072 return 0;
2015} 2073}
2016 2074
@@ -2048,7 +2106,9 @@ static int __init init_lapic_sysfs(void)
2048 error = sysdev_register(&device_lapic); 2106 error = sysdev_register(&device_lapic);
2049 return error; 2107 return error;
2050} 2108}
2051device_initcall(init_lapic_sysfs); 2109
2110/* local apic needs to resume before other devices access its registers. */
2111core_initcall(init_lapic_sysfs);
2052 2112
2053#else /* CONFIG_PM */ 2113#else /* CONFIG_PM */
2054 2114
diff --git a/arch/x86/kernel/apic/apic_flat_64.c b/arch/x86/kernel/apic/apic_flat_64.c
index f933822dba18..306e5e88fb6f 100644
--- a/arch/x86/kernel/apic/apic_flat_64.c
+++ b/arch/x86/kernel/apic/apic_flat_64.c
@@ -159,20 +159,6 @@ static int flat_apic_id_registered(void)
159 return physid_isset(read_xapic_id(), phys_cpu_present_map); 159 return physid_isset(read_xapic_id(), phys_cpu_present_map);
160} 160}
161 161
162static unsigned int flat_cpu_mask_to_apicid(const struct cpumask *cpumask)
163{
164 return cpumask_bits(cpumask)[0] & APIC_ALL_CPUS;
165}
166
167static unsigned int flat_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
168 const struct cpumask *andmask)
169{
170 unsigned long mask1 = cpumask_bits(cpumask)[0] & APIC_ALL_CPUS;
171 unsigned long mask2 = cpumask_bits(andmask)[0] & APIC_ALL_CPUS;
172
173 return mask1 & mask2;
174}
175
176static int flat_phys_pkg_id(int initial_apic_id, int index_msb) 162static int flat_phys_pkg_id(int initial_apic_id, int index_msb)
177{ 163{
178 return hard_smp_processor_id() >> index_msb; 164 return hard_smp_processor_id() >> index_msb;
@@ -213,8 +199,8 @@ struct apic apic_flat = {
213 .set_apic_id = set_apic_id, 199 .set_apic_id = set_apic_id,
214 .apic_id_mask = 0xFFu << 24, 200 .apic_id_mask = 0xFFu << 24,
215 201
216 .cpu_mask_to_apicid = flat_cpu_mask_to_apicid, 202 .cpu_mask_to_apicid = default_cpu_mask_to_apicid,
217 .cpu_mask_to_apicid_and = flat_cpu_mask_to_apicid_and, 203 .cpu_mask_to_apicid_and = default_cpu_mask_to_apicid_and,
218 204
219 .send_IPI_mask = flat_send_IPI_mask, 205 .send_IPI_mask = flat_send_IPI_mask,
220 .send_IPI_mask_allbutself = flat_send_IPI_mask_allbutself, 206 .send_IPI_mask_allbutself = flat_send_IPI_mask_allbutself,
@@ -226,7 +212,7 @@ struct apic apic_flat = {
226 .trampoline_phys_high = DEFAULT_TRAMPOLINE_PHYS_HIGH, 212 .trampoline_phys_high = DEFAULT_TRAMPOLINE_PHYS_HIGH,
227 .wait_for_init_deassert = NULL, 213 .wait_for_init_deassert = NULL,
228 .smp_callin_clear_local_apic = NULL, 214 .smp_callin_clear_local_apic = NULL,
229 .inquire_remote_apic = NULL, 215 .inquire_remote_apic = default_inquire_remote_apic,
230 216
231 .read = native_apic_mem_read, 217 .read = native_apic_mem_read,
232 .write = native_apic_mem_write, 218 .write = native_apic_mem_write,
@@ -376,7 +362,7 @@ struct apic apic_physflat = {
376 .trampoline_phys_high = DEFAULT_TRAMPOLINE_PHYS_HIGH, 362 .trampoline_phys_high = DEFAULT_TRAMPOLINE_PHYS_HIGH,
377 .wait_for_init_deassert = NULL, 363 .wait_for_init_deassert = NULL,
378 .smp_callin_clear_local_apic = NULL, 364 .smp_callin_clear_local_apic = NULL,
379 .inquire_remote_apic = NULL, 365 .inquire_remote_apic = default_inquire_remote_apic,
380 366
381 .read = native_apic_mem_read, 367 .read = native_apic_mem_read,
382 .write = native_apic_mem_write, 368 .write = native_apic_mem_write,
diff --git a/arch/x86/kernel/apic/bigsmp_32.c b/arch/x86/kernel/apic/bigsmp_32.c
index d806ecaa948f..676cdac385c0 100644
--- a/arch/x86/kernel/apic/bigsmp_32.c
+++ b/arch/x86/kernel/apic/bigsmp_32.c
@@ -26,12 +26,12 @@ static int bigsmp_apic_id_registered(void)
26 return 1; 26 return 1;
27} 27}
28 28
29static const cpumask_t *bigsmp_target_cpus(void) 29static const struct cpumask *bigsmp_target_cpus(void)
30{ 30{
31#ifdef CONFIG_SMP 31#ifdef CONFIG_SMP
32 return &cpu_online_map; 32 return cpu_online_mask;
33#else 33#else
34 return &cpumask_of_cpu(0); 34 return cpumask_of(0);
35#endif 35#endif
36} 36}
37 37
@@ -118,9 +118,9 @@ static int bigsmp_check_phys_apicid_present(int boot_cpu_physical_apicid)
118} 118}
119 119
120/* As we are using single CPU as destination, pick only one CPU here */ 120/* As we are using single CPU as destination, pick only one CPU here */
121static unsigned int bigsmp_cpu_mask_to_apicid(const cpumask_t *cpumask) 121static unsigned int bigsmp_cpu_mask_to_apicid(const struct cpumask *cpumask)
122{ 122{
123 return bigsmp_cpu_to_logical_apicid(first_cpu(*cpumask)); 123 return bigsmp_cpu_to_logical_apicid(cpumask_first(cpumask));
124} 124}
125 125
126static unsigned int bigsmp_cpu_mask_to_apicid_and(const struct cpumask *cpumask, 126static unsigned int bigsmp_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
@@ -188,10 +188,10 @@ static const struct dmi_system_id bigsmp_dmi_table[] = {
188 { } /* NULL entry stops DMI scanning */ 188 { } /* NULL entry stops DMI scanning */
189}; 189};
190 190
191static void bigsmp_vector_allocation_domain(int cpu, cpumask_t *retmask) 191static void bigsmp_vector_allocation_domain(int cpu, struct cpumask *retmask)
192{ 192{
193 cpus_clear(*retmask); 193 cpumask_clear(retmask);
194 cpu_set(cpu, *retmask); 194 cpumask_set_cpu(cpu, retmask);
195} 195}
196 196
197static int probe_bigsmp(void) 197static int probe_bigsmp(void)
diff --git a/arch/x86/kernel/apic/es7000_32.c b/arch/x86/kernel/apic/es7000_32.c
index 19588f2770ee..1c11b819f245 100644
--- a/arch/x86/kernel/apic/es7000_32.c
+++ b/arch/x86/kernel/apic/es7000_32.c
@@ -410,7 +410,7 @@ static void es7000_enable_apic_mode(void)
410 WARN(1, "Command failed, status = %x\n", mip_status); 410 WARN(1, "Command failed, status = %x\n", mip_status);
411} 411}
412 412
413static void es7000_vector_allocation_domain(int cpu, cpumask_t *retmask) 413static void es7000_vector_allocation_domain(int cpu, struct cpumask *retmask)
414{ 414{
415 /* Careful. Some cpus do not strictly honor the set of cpus 415 /* Careful. Some cpus do not strictly honor the set of cpus
416 * specified in the interrupt destination when using lowest 416 * specified in the interrupt destination when using lowest
@@ -420,7 +420,8 @@ static void es7000_vector_allocation_domain(int cpu, cpumask_t *retmask)
420 * deliver interrupts to the wrong hyperthread when only one 420 * deliver interrupts to the wrong hyperthread when only one
421 * hyperthread was specified in the interrupt desitination. 421 * hyperthread was specified in the interrupt desitination.
422 */ 422 */
423 *retmask = (cpumask_t){ { [0] = APIC_ALL_CPUS, } }; 423 cpumask_clear(retmask);
424 cpumask_bits(retmask)[0] = APIC_ALL_CPUS;
424} 425}
425 426
426 427
@@ -455,14 +456,14 @@ static int es7000_apic_id_registered(void)
455 return 1; 456 return 1;
456} 457}
457 458
458static const cpumask_t *target_cpus_cluster(void) 459static const struct cpumask *target_cpus_cluster(void)
459{ 460{
460 return &CPU_MASK_ALL; 461 return cpu_all_mask;
461} 462}
462 463
463static const cpumask_t *es7000_target_cpus(void) 464static const struct cpumask *es7000_target_cpus(void)
464{ 465{
465 return &cpumask_of_cpu(smp_processor_id()); 466 return cpumask_of(smp_processor_id());
466} 467}
467 468
468static unsigned long 469static unsigned long
@@ -517,7 +518,7 @@ static void es7000_setup_apic_routing(void)
517 "Enabling APIC mode: %s. Using %d I/O APICs, target cpus %lx\n", 518 "Enabling APIC mode: %s. Using %d I/O APICs, target cpus %lx\n",
518 (apic_version[apic] == 0x14) ? 519 (apic_version[apic] == 0x14) ?
519 "Physical Cluster" : "Logical Cluster", 520 "Physical Cluster" : "Logical Cluster",
520 nr_ioapics, cpus_addr(*es7000_target_cpus())[0]); 521 nr_ioapics, cpumask_bits(es7000_target_cpus())[0]);
521} 522}
522 523
523static int es7000_apicid_to_node(int logical_apicid) 524static int es7000_apicid_to_node(int logical_apicid)
@@ -572,7 +573,7 @@ static int es7000_check_phys_apicid_present(int cpu_physical_apicid)
572 return 1; 573 return 1;
573} 574}
574 575
575static unsigned int es7000_cpu_mask_to_apicid(const cpumask_t *cpumask) 576static unsigned int es7000_cpu_mask_to_apicid(const struct cpumask *cpumask)
576{ 577{
577 unsigned int round = 0; 578 unsigned int round = 0;
578 int cpu, uninitialized_var(apicid); 579 int cpu, uninitialized_var(apicid);
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index 86827d85488a..a2789e42e162 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -592,10 +592,12 @@ set_desc_affinity(struct irq_desc *desc, const struct cpumask *mask)
592 if (assign_irq_vector(irq, cfg, mask)) 592 if (assign_irq_vector(irq, cfg, mask))
593 return BAD_APICID; 593 return BAD_APICID;
594 594
595 cpumask_and(desc->affinity, cfg->domain, mask); 595 /* check that before desc->addinity get updated */
596 set_extra_move_desc(desc, mask); 596 set_extra_move_desc(desc, mask);
597 597
598 return apic->cpu_mask_to_apicid_and(desc->affinity, cpu_online_mask); 598 cpumask_copy(desc->affinity, mask);
599
600 return apic->cpu_mask_to_apicid_and(desc->affinity, cfg->domain);
599} 601}
600 602
601static void 603static void
@@ -849,63 +851,74 @@ __setup("pirq=", ioapic_pirq_setup);
849#endif /* CONFIG_X86_32 */ 851#endif /* CONFIG_X86_32 */
850 852
851#ifdef CONFIG_INTR_REMAP 853#ifdef CONFIG_INTR_REMAP
852/* I/O APIC RTE contents at the OS boot up */ 854struct IO_APIC_route_entry **alloc_ioapic_entries(void)
853static struct IO_APIC_route_entry *early_ioapic_entries[MAX_IO_APICS]; 855{
856 int apic;
857 struct IO_APIC_route_entry **ioapic_entries;
858
859 ioapic_entries = kzalloc(sizeof(*ioapic_entries) * nr_ioapics,
860 GFP_ATOMIC);
861 if (!ioapic_entries)
862 return 0;
863
864 for (apic = 0; apic < nr_ioapics; apic++) {
865 ioapic_entries[apic] =
866 kzalloc(sizeof(struct IO_APIC_route_entry) *
867 nr_ioapic_registers[apic], GFP_ATOMIC);
868 if (!ioapic_entries[apic])
869 goto nomem;
870 }
871
872 return ioapic_entries;
873
874nomem:
875 while (--apic >= 0)
876 kfree(ioapic_entries[apic]);
877 kfree(ioapic_entries);
878
879 return 0;
880}
854 881
855/* 882/*
856 * Saves all the IO-APIC RTE's 883 * Saves all the IO-APIC RTE's
857 */ 884 */
858int save_IO_APIC_setup(void) 885int save_IO_APIC_setup(struct IO_APIC_route_entry **ioapic_entries)
859{ 886{
860 union IO_APIC_reg_01 reg_01;
861 unsigned long flags;
862 int apic, pin; 887 int apic, pin;
863 888
864 /* 889 if (!ioapic_entries)
865 * The number of IO-APIC IRQ registers (== #pins): 890 return -ENOMEM;
866 */
867 for (apic = 0; apic < nr_ioapics; apic++) {
868 spin_lock_irqsave(&ioapic_lock, flags);
869 reg_01.raw = io_apic_read(apic, 1);
870 spin_unlock_irqrestore(&ioapic_lock, flags);
871 nr_ioapic_registers[apic] = reg_01.bits.entries+1;
872 }
873 891
874 for (apic = 0; apic < nr_ioapics; apic++) { 892 for (apic = 0; apic < nr_ioapics; apic++) {
875 early_ioapic_entries[apic] = 893 if (!ioapic_entries[apic])
876 kzalloc(sizeof(struct IO_APIC_route_entry) * 894 return -ENOMEM;
877 nr_ioapic_registers[apic], GFP_KERNEL);
878 if (!early_ioapic_entries[apic])
879 goto nomem;
880 }
881 895
882 for (apic = 0; apic < nr_ioapics; apic++)
883 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) 896 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++)
884 early_ioapic_entries[apic][pin] = 897 ioapic_entries[apic][pin] =
885 ioapic_read_entry(apic, pin); 898 ioapic_read_entry(apic, pin);
899 }
886 900
887 return 0; 901 return 0;
888
889nomem:
890 while (apic >= 0)
891 kfree(early_ioapic_entries[apic--]);
892 memset(early_ioapic_entries, 0,
893 ARRAY_SIZE(early_ioapic_entries));
894
895 return -ENOMEM;
896} 902}
897 903
898void mask_IO_APIC_setup(void) 904/*
905 * Mask all IO APIC entries.
906 */
907void mask_IO_APIC_setup(struct IO_APIC_route_entry **ioapic_entries)
899{ 908{
900 int apic, pin; 909 int apic, pin;
901 910
911 if (!ioapic_entries)
912 return;
913
902 for (apic = 0; apic < nr_ioapics; apic++) { 914 for (apic = 0; apic < nr_ioapics; apic++) {
903 if (!early_ioapic_entries[apic]) 915 if (!ioapic_entries[apic])
904 break; 916 break;
917
905 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) { 918 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
906 struct IO_APIC_route_entry entry; 919 struct IO_APIC_route_entry entry;
907 920
908 entry = early_ioapic_entries[apic][pin]; 921 entry = ioapic_entries[apic][pin];
909 if (!entry.mask) { 922 if (!entry.mask) {
910 entry.mask = 1; 923 entry.mask = 1;
911 ioapic_write_entry(apic, pin, entry); 924 ioapic_write_entry(apic, pin, entry);
@@ -914,22 +927,30 @@ void mask_IO_APIC_setup(void)
914 } 927 }
915} 928}
916 929
917void restore_IO_APIC_setup(void) 930/*
931 * Restore IO APIC entries which was saved in ioapic_entries.
932 */
933int restore_IO_APIC_setup(struct IO_APIC_route_entry **ioapic_entries)
918{ 934{
919 int apic, pin; 935 int apic, pin;
920 936
937 if (!ioapic_entries)
938 return -ENOMEM;
939
921 for (apic = 0; apic < nr_ioapics; apic++) { 940 for (apic = 0; apic < nr_ioapics; apic++) {
922 if (!early_ioapic_entries[apic]) 941 if (!ioapic_entries[apic])
923 break; 942 return -ENOMEM;
943
924 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) 944 for (pin = 0; pin < nr_ioapic_registers[apic]; pin++)
925 ioapic_write_entry(apic, pin, 945 ioapic_write_entry(apic, pin,
926 early_ioapic_entries[apic][pin]); 946 ioapic_entries[apic][pin]);
927 kfree(early_ioapic_entries[apic]);
928 early_ioapic_entries[apic] = NULL;
929 } 947 }
948 return 0;
930} 949}
931 950
932void reinit_intr_remapped_IO_APIC(int intr_remapping) 951void reinit_intr_remapped_IO_APIC(int intr_remapping,
952 struct IO_APIC_route_entry **ioapic_entries)
953
933{ 954{
934 /* 955 /*
935 * for now plain restore of previous settings. 956 * for now plain restore of previous settings.
@@ -938,7 +959,17 @@ void reinit_intr_remapped_IO_APIC(int intr_remapping)
938 * table entries. for now, do a plain restore, and wait for 959 * table entries. for now, do a plain restore, and wait for
939 * the setup_IO_APIC_irqs() to do proper initialization. 960 * the setup_IO_APIC_irqs() to do proper initialization.
940 */ 961 */
941 restore_IO_APIC_setup(); 962 restore_IO_APIC_setup(ioapic_entries);
963}
964
965void free_ioapic_entries(struct IO_APIC_route_entry **ioapic_entries)
966{
967 int apic;
968
969 for (apic = 0; apic < nr_ioapics; apic++)
970 kfree(ioapic_entries[apic]);
971
972 kfree(ioapic_entries);
942} 973}
943#endif 974#endif
944 975
@@ -1428,7 +1459,6 @@ void __setup_vector_irq(int cpu)
1428 1459
1429static struct irq_chip ioapic_chip; 1460static struct irq_chip ioapic_chip;
1430static struct irq_chip ir_ioapic_chip; 1461static struct irq_chip ir_ioapic_chip;
1431static struct irq_chip msi_ir_chip;
1432 1462
1433#define IOAPIC_AUTO -1 1463#define IOAPIC_AUTO -1
1434#define IOAPIC_EDGE 0 1464#define IOAPIC_EDGE 0
@@ -2494,7 +2524,6 @@ static void irq_complete_move(struct irq_desc **descp)
2494static inline void irq_complete_move(struct irq_desc **descp) {} 2524static inline void irq_complete_move(struct irq_desc **descp) {}
2495#endif 2525#endif
2496 2526
2497#ifdef CONFIG_INTR_REMAP
2498static void __eoi_ioapic_irq(unsigned int irq, struct irq_cfg *cfg) 2527static void __eoi_ioapic_irq(unsigned int irq, struct irq_cfg *cfg)
2499{ 2528{
2500 int apic, pin; 2529 int apic, pin;
@@ -2528,6 +2557,7 @@ eoi_ioapic_irq(struct irq_desc *desc)
2528 spin_unlock_irqrestore(&ioapic_lock, flags); 2557 spin_unlock_irqrestore(&ioapic_lock, flags);
2529} 2558}
2530 2559
2560#ifdef CONFIG_X86_X2APIC
2531static void ack_x2apic_level(unsigned int irq) 2561static void ack_x2apic_level(unsigned int irq)
2532{ 2562{
2533 struct irq_desc *desc = irq_to_desc(irq); 2563 struct irq_desc *desc = irq_to_desc(irq);
@@ -2539,7 +2569,6 @@ static void ack_x2apic_edge(unsigned int irq)
2539{ 2569{
2540 ack_x2APIC_irq(); 2570 ack_x2APIC_irq();
2541} 2571}
2542
2543#endif 2572#endif
2544 2573
2545static void ack_apic_edge(unsigned int irq) 2574static void ack_apic_edge(unsigned int irq)
@@ -2605,6 +2634,9 @@ static void ack_apic_level(unsigned int irq)
2605 */ 2634 */
2606 ack_APIC_irq(); 2635 ack_APIC_irq();
2607 2636
2637 if (irq_remapped(irq))
2638 eoi_ioapic_irq(desc);
2639
2608 /* Now we can move and renable the irq */ 2640 /* Now we can move and renable the irq */
2609 if (unlikely(do_unmask_irq)) { 2641 if (unlikely(do_unmask_irq)) {
2610 /* Only migrate the irq if the ack has been received. 2642 /* Only migrate the irq if the ack has been received.
@@ -2650,6 +2682,26 @@ static void ack_apic_level(unsigned int irq)
2650#endif 2682#endif
2651} 2683}
2652 2684
2685#ifdef CONFIG_INTR_REMAP
2686static void ir_ack_apic_edge(unsigned int irq)
2687{
2688#ifdef CONFIG_X86_X2APIC
2689 if (x2apic_enabled())
2690 return ack_x2apic_edge(irq);
2691#endif
2692 return ack_apic_edge(irq);
2693}
2694
2695static void ir_ack_apic_level(unsigned int irq)
2696{
2697#ifdef CONFIG_X86_X2APIC
2698 if (x2apic_enabled())
2699 return ack_x2apic_level(irq);
2700#endif
2701 return ack_apic_level(irq);
2702}
2703#endif /* CONFIG_INTR_REMAP */
2704
2653static struct irq_chip ioapic_chip __read_mostly = { 2705static struct irq_chip ioapic_chip __read_mostly = {
2654 .name = "IO-APIC", 2706 .name = "IO-APIC",
2655 .startup = startup_ioapic_irq, 2707 .startup = startup_ioapic_irq,
@@ -2663,20 +2715,20 @@ static struct irq_chip ioapic_chip __read_mostly = {
2663 .retrigger = ioapic_retrigger_irq, 2715 .retrigger = ioapic_retrigger_irq,
2664}; 2716};
2665 2717
2666#ifdef CONFIG_INTR_REMAP
2667static struct irq_chip ir_ioapic_chip __read_mostly = { 2718static struct irq_chip ir_ioapic_chip __read_mostly = {
2668 .name = "IR-IO-APIC", 2719 .name = "IR-IO-APIC",
2669 .startup = startup_ioapic_irq, 2720 .startup = startup_ioapic_irq,
2670 .mask = mask_IO_APIC_irq, 2721 .mask = mask_IO_APIC_irq,
2671 .unmask = unmask_IO_APIC_irq, 2722 .unmask = unmask_IO_APIC_irq,
2672 .ack = ack_x2apic_edge, 2723#ifdef CONFIG_INTR_REMAP
2673 .eoi = ack_x2apic_level, 2724 .ack = ir_ack_apic_edge,
2725 .eoi = ir_ack_apic_level,
2674#ifdef CONFIG_SMP 2726#ifdef CONFIG_SMP
2675 .set_affinity = set_ir_ioapic_affinity_irq, 2727 .set_affinity = set_ir_ioapic_affinity_irq,
2676#endif 2728#endif
2729#endif
2677 .retrigger = ioapic_retrigger_irq, 2730 .retrigger = ioapic_retrigger_irq,
2678}; 2731};
2679#endif
2680 2732
2681static inline void init_IO_APIC_traps(void) 2733static inline void init_IO_APIC_traps(void)
2682{ 2734{
@@ -3391,18 +3443,18 @@ static struct irq_chip msi_chip = {
3391 .retrigger = ioapic_retrigger_irq, 3443 .retrigger = ioapic_retrigger_irq,
3392}; 3444};
3393 3445
3394#ifdef CONFIG_INTR_REMAP
3395static struct irq_chip msi_ir_chip = { 3446static struct irq_chip msi_ir_chip = {
3396 .name = "IR-PCI-MSI", 3447 .name = "IR-PCI-MSI",
3397 .unmask = unmask_msi_irq, 3448 .unmask = unmask_msi_irq,
3398 .mask = mask_msi_irq, 3449 .mask = mask_msi_irq,
3399 .ack = ack_x2apic_edge, 3450#ifdef CONFIG_INTR_REMAP
3451 .ack = ir_ack_apic_edge,
3400#ifdef CONFIG_SMP 3452#ifdef CONFIG_SMP
3401 .set_affinity = ir_set_msi_irq_affinity, 3453 .set_affinity = ir_set_msi_irq_affinity,
3402#endif 3454#endif
3455#endif
3403 .retrigger = ioapic_retrigger_irq, 3456 .retrigger = ioapic_retrigger_irq,
3404}; 3457};
3405#endif
3406 3458
3407/* 3459/*
3408 * Map the PCI dev to the corresponding remapping hardware unit 3460 * Map the PCI dev to the corresponding remapping hardware unit
@@ -3467,6 +3519,10 @@ int arch_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
3467 struct intel_iommu *iommu = NULL; 3519 struct intel_iommu *iommu = NULL;
3468 int index = 0; 3520 int index = 0;
3469 3521
3522 /* x86 doesn't support multiple MSI yet */
3523 if (type == PCI_CAP_ID_MSI && nvec > 1)
3524 return 1;
3525
3470 irq_want = nr_irqs_gsi; 3526 irq_want = nr_irqs_gsi;
3471 sub_handle = 0; 3527 sub_handle = 0;
3472 list_for_each_entry(msidesc, &dev->msi_list, list) { 3528 list_for_each_entry(msidesc, &dev->msi_list, list) {
diff --git a/arch/x86/kernel/apic/nmi.c b/arch/x86/kernel/apic/nmi.c
index bdfad80c3cf1..d6bd62407152 100644
--- a/arch/x86/kernel/apic/nmi.c
+++ b/arch/x86/kernel/apic/nmi.c
@@ -39,7 +39,7 @@
39int unknown_nmi_panic; 39int unknown_nmi_panic;
40int nmi_watchdog_enabled; 40int nmi_watchdog_enabled;
41 41
42static cpumask_t backtrace_mask = CPU_MASK_NONE; 42static cpumask_var_t backtrace_mask;
43 43
44/* nmi_active: 44/* nmi_active:
45 * >0: the lapic NMI watchdog is active, but can be disabled 45 * >0: the lapic NMI watchdog is active, but can be disabled
@@ -138,6 +138,7 @@ int __init check_nmi_watchdog(void)
138 if (!prev_nmi_count) 138 if (!prev_nmi_count)
139 goto error; 139 goto error;
140 140
141 alloc_cpumask_var(&backtrace_mask, GFP_KERNEL);
141 printk(KERN_INFO "Testing NMI watchdog ... "); 142 printk(KERN_INFO "Testing NMI watchdog ... ");
142 143
143#ifdef CONFIG_SMP 144#ifdef CONFIG_SMP
@@ -413,14 +414,14 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason)
413 touched = 1; 414 touched = 1;
414 } 415 }
415 416
416 if (cpu_isset(cpu, backtrace_mask)) { 417 if (cpumask_test_cpu(cpu, backtrace_mask)) {
417 static DEFINE_SPINLOCK(lock); /* Serialise the printks */ 418 static DEFINE_SPINLOCK(lock); /* Serialise the printks */
418 419
419 spin_lock(&lock); 420 spin_lock(&lock);
420 printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); 421 printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu);
421 dump_stack(); 422 dump_stack();
422 spin_unlock(&lock); 423 spin_unlock(&lock);
423 cpu_clear(cpu, backtrace_mask); 424 cpumask_clear_cpu(cpu, backtrace_mask);
424 } 425 }
425 426
426 /* Could check oops_in_progress here too, but it's safer not to */ 427 /* Could check oops_in_progress here too, but it's safer not to */
@@ -554,10 +555,10 @@ void __trigger_all_cpu_backtrace(void)
554{ 555{
555 int i; 556 int i;
556 557
557 backtrace_mask = cpu_online_map; 558 cpumask_copy(backtrace_mask, cpu_online_mask);
558 /* Wait for up to 10 seconds for all CPUs to do the backtrace */ 559 /* Wait for up to 10 seconds for all CPUs to do the backtrace */
559 for (i = 0; i < 10 * 1000; i++) { 560 for (i = 0; i < 10 * 1000; i++) {
560 if (cpus_empty(backtrace_mask)) 561 if (cpumask_empty(backtrace_mask))
561 break; 562 break;
562 mdelay(1); 563 mdelay(1);
563 } 564 }
diff --git a/arch/x86/kernel/apic/numaq_32.c b/arch/x86/kernel/apic/numaq_32.c
index ba2fc6465534..533e59c6fc82 100644
--- a/arch/x86/kernel/apic/numaq_32.c
+++ b/arch/x86/kernel/apic/numaq_32.c
@@ -334,9 +334,9 @@ static inline void numaq_smp_callin_clear_local_apic(void)
334 clear_local_APIC(); 334 clear_local_APIC();
335} 335}
336 336
337static inline const cpumask_t *numaq_target_cpus(void) 337static inline const struct cpumask *numaq_target_cpus(void)
338{ 338{
339 return &CPU_MASK_ALL; 339 return cpu_all_mask;
340} 340}
341 341
342static inline unsigned long 342static inline unsigned long
@@ -427,7 +427,7 @@ static inline int numaq_check_phys_apicid_present(int boot_cpu_physical_apicid)
427 * We use physical apicids here, not logical, so just return the default 427 * We use physical apicids here, not logical, so just return the default
428 * physical broadcast to stop people from breaking us 428 * physical broadcast to stop people from breaking us
429 */ 429 */
430static inline unsigned int numaq_cpu_mask_to_apicid(const cpumask_t *cpumask) 430static unsigned int numaq_cpu_mask_to_apicid(const struct cpumask *cpumask)
431{ 431{
432 return 0x0F; 432 return 0x0F;
433} 433}
@@ -462,7 +462,7 @@ static int probe_numaq(void)
462 return found_numaq; 462 return found_numaq;
463} 463}
464 464
465static void numaq_vector_allocation_domain(int cpu, cpumask_t *retmask) 465static void numaq_vector_allocation_domain(int cpu, struct cpumask *retmask)
466{ 466{
467 /* Careful. Some cpus do not strictly honor the set of cpus 467 /* Careful. Some cpus do not strictly honor the set of cpus
468 * specified in the interrupt destination when using lowest 468 * specified in the interrupt destination when using lowest
@@ -472,7 +472,8 @@ static void numaq_vector_allocation_domain(int cpu, cpumask_t *retmask)
472 * deliver interrupts to the wrong hyperthread when only one 472 * deliver interrupts to the wrong hyperthread when only one
473 * hyperthread was specified in the interrupt desitination. 473 * hyperthread was specified in the interrupt desitination.
474 */ 474 */
475 *retmask = (cpumask_t){ { [0] = APIC_ALL_CPUS, } }; 475 cpumask_clear(retmask);
476 cpumask_bits(retmask)[0] = APIC_ALL_CPUS;
476} 477}
477 478
478static void numaq_setup_portio_remap(void) 479static void numaq_setup_portio_remap(void)
diff --git a/arch/x86/kernel/apic/probe_32.c b/arch/x86/kernel/apic/probe_32.c
index 141c99a1c264..01eda2ac65e4 100644
--- a/arch/x86/kernel/apic/probe_32.c
+++ b/arch/x86/kernel/apic/probe_32.c
@@ -83,7 +83,8 @@ static void default_vector_allocation_domain(int cpu, struct cpumask *retmask)
83 * deliver interrupts to the wrong hyperthread when only one 83 * deliver interrupts to the wrong hyperthread when only one
84 * hyperthread was specified in the interrupt desitination. 84 * hyperthread was specified in the interrupt desitination.
85 */ 85 */
86 *retmask = (cpumask_t) { { [0] = APIC_ALL_CPUS } }; 86 cpumask_clear(retmask);
87 cpumask_bits(retmask)[0] = APIC_ALL_CPUS;
87} 88}
88 89
89/* should be called last. */ 90/* should be called last. */
diff --git a/arch/x86/kernel/apic/summit_32.c b/arch/x86/kernel/apic/summit_32.c
index aac52fa873ff..9cfe1f415d81 100644
--- a/arch/x86/kernel/apic/summit_32.c
+++ b/arch/x86/kernel/apic/summit_32.c
@@ -53,23 +53,19 @@ static unsigned summit_get_apic_id(unsigned long x)
53 return (x >> 24) & 0xFF; 53 return (x >> 24) & 0xFF;
54} 54}
55 55
56static inline void summit_send_IPI_mask(const cpumask_t *mask, int vector) 56static inline void summit_send_IPI_mask(const struct cpumask *mask, int vector)
57{ 57{
58 default_send_IPI_mask_sequence_logical(mask, vector); 58 default_send_IPI_mask_sequence_logical(mask, vector);
59} 59}
60 60
61static void summit_send_IPI_allbutself(int vector) 61static void summit_send_IPI_allbutself(int vector)
62{ 62{
63 cpumask_t mask = cpu_online_map; 63 default_send_IPI_mask_allbutself_logical(cpu_online_mask, vector);
64 cpu_clear(smp_processor_id(), mask);
65
66 if (!cpus_empty(mask))
67 summit_send_IPI_mask(&mask, vector);
68} 64}
69 65
70static void summit_send_IPI_all(int vector) 66static void summit_send_IPI_all(int vector)
71{ 67{
72 summit_send_IPI_mask(&cpu_online_map, vector); 68 summit_send_IPI_mask(cpu_online_mask, vector);
73} 69}
74 70
75#include <asm/tsc.h> 71#include <asm/tsc.h>
@@ -186,13 +182,13 @@ static inline int is_WPEG(struct rio_detail *rio){
186 182
187#define SUMMIT_APIC_DFR_VALUE (APIC_DFR_CLUSTER) 183#define SUMMIT_APIC_DFR_VALUE (APIC_DFR_CLUSTER)
188 184
189static const cpumask_t *summit_target_cpus(void) 185static const struct cpumask *summit_target_cpus(void)
190{ 186{
191 /* CPU_MASK_ALL (0xff) has undefined behaviour with 187 /* CPU_MASK_ALL (0xff) has undefined behaviour with
192 * dest_LowestPrio mode logical clustered apic interrupt routing 188 * dest_LowestPrio mode logical clustered apic interrupt routing
193 * Just start on cpu 0. IRQ balancing will spread load 189 * Just start on cpu 0. IRQ balancing will spread load
194 */ 190 */
195 return &cpumask_of_cpu(0); 191 return cpumask_of(0);
196} 192}
197 193
198static unsigned long summit_check_apicid_used(physid_mask_t bitmap, int apicid) 194static unsigned long summit_check_apicid_used(physid_mask_t bitmap, int apicid)
@@ -289,7 +285,7 @@ static int summit_check_phys_apicid_present(int boot_cpu_physical_apicid)
289 return 1; 285 return 1;
290} 286}
291 287
292static unsigned int summit_cpu_mask_to_apicid(const cpumask_t *cpumask) 288static unsigned int summit_cpu_mask_to_apicid(const struct cpumask *cpumask)
293{ 289{
294 unsigned int round = 0; 290 unsigned int round = 0;
295 int cpu, apicid = 0; 291 int cpu, apicid = 0;
@@ -346,7 +342,7 @@ static int probe_summit(void)
346 return 0; 342 return 0;
347} 343}
348 344
349static void summit_vector_allocation_domain(int cpu, cpumask_t *retmask) 345static void summit_vector_allocation_domain(int cpu, struct cpumask *retmask)
350{ 346{
351 /* Careful. Some cpus do not strictly honor the set of cpus 347 /* Careful. Some cpus do not strictly honor the set of cpus
352 * specified in the interrupt destination when using lowest 348 * specified in the interrupt destination when using lowest
@@ -356,7 +352,8 @@ static void summit_vector_allocation_domain(int cpu, cpumask_t *retmask)
356 * deliver interrupts to the wrong hyperthread when only one 352 * deliver interrupts to the wrong hyperthread when only one
357 * hyperthread was specified in the interrupt desitination. 353 * hyperthread was specified in the interrupt desitination.
358 */ 354 */
359 *retmask = (cpumask_t){ { [0] = APIC_ALL_CPUS, } }; 355 cpumask_clear(retmask);
356 cpumask_bits(retmask)[0] = APIC_ALL_CPUS;
360} 357}
361 358
362#ifdef CONFIG_X86_SUMMIT_NUMA 359#ifdef CONFIG_X86_SUMMIT_NUMA
diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c
index 1bd6da1f8fad..de1a50af807b 100644
--- a/arch/x86/kernel/apic/x2apic_uv_x.c
+++ b/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -118,17 +118,12 @@ static int uv_wakeup_secondary(int phys_apicid, unsigned long start_rip)
118 118
119static void uv_send_IPI_one(int cpu, int vector) 119static void uv_send_IPI_one(int cpu, int vector)
120{ 120{
121 unsigned long val, apicid; 121 unsigned long apicid;
122 int pnode; 122 int pnode;
123 123
124 apicid = per_cpu(x86_cpu_to_apicid, cpu); 124 apicid = per_cpu(x86_cpu_to_apicid, cpu);
125 pnode = uv_apicid_to_pnode(apicid); 125 pnode = uv_apicid_to_pnode(apicid);
126 126 uv_hub_send_ipi(pnode, apicid, vector);
127 val = (1UL << UVH_IPI_INT_SEND_SHFT) |
128 (apicid << UVH_IPI_INT_APIC_ID_SHFT) |
129 (vector << UVH_IPI_INT_VECTOR_SHFT);
130
131 uv_write_global_mmr64(pnode, UVH_IPI_INT, val);
132} 127}
133 128
134static void uv_send_IPI_mask(const struct cpumask *mask, int vector) 129static void uv_send_IPI_mask(const struct cpumask *mask, int vector)
@@ -554,7 +549,8 @@ void __init uv_system_init(void)
554 unsigned long gnode_upper, lowmem_redir_base, lowmem_redir_size; 549 unsigned long gnode_upper, lowmem_redir_base, lowmem_redir_size;
555 int bytes, nid, cpu, lcpu, pnode, blade, i, j, m_val, n_val; 550 int bytes, nid, cpu, lcpu, pnode, blade, i, j, m_val, n_val;
556 int max_pnode = 0; 551 int max_pnode = 0;
557 unsigned long mmr_base, present; 552 unsigned long mmr_base, present, paddr;
553 unsigned short pnode_mask;
558 554
559 map_low_mmrs(); 555 map_low_mmrs();
560 556
@@ -597,6 +593,7 @@ void __init uv_system_init(void)
597 } 593 }
598 } 594 }
599 595
596 pnode_mask = (1 << n_val) - 1;
600 node_id.v = uv_read_local_mmr(UVH_NODE_ID); 597 node_id.v = uv_read_local_mmr(UVH_NODE_ID);
601 gnode_upper = (((unsigned long)node_id.s.node_id) & 598 gnode_upper = (((unsigned long)node_id.s.node_id) &
602 ~((1 << n_val) - 1)) << m_val; 599 ~((1 << n_val) - 1)) << m_val;
@@ -620,7 +617,7 @@ void __init uv_system_init(void)
620 uv_cpu_hub_info(cpu)->numa_blade_id = blade; 617 uv_cpu_hub_info(cpu)->numa_blade_id = blade;
621 uv_cpu_hub_info(cpu)->blade_processor_id = lcpu; 618 uv_cpu_hub_info(cpu)->blade_processor_id = lcpu;
622 uv_cpu_hub_info(cpu)->pnode = pnode; 619 uv_cpu_hub_info(cpu)->pnode = pnode;
623 uv_cpu_hub_info(cpu)->pnode_mask = (1 << n_val) - 1; 620 uv_cpu_hub_info(cpu)->pnode_mask = pnode_mask;
624 uv_cpu_hub_info(cpu)->gpa_mask = (1 << (m_val + n_val)) - 1; 621 uv_cpu_hub_info(cpu)->gpa_mask = (1 << (m_val + n_val)) - 1;
625 uv_cpu_hub_info(cpu)->gnode_upper = gnode_upper; 622 uv_cpu_hub_info(cpu)->gnode_upper = gnode_upper;
626 uv_cpu_hub_info(cpu)->global_mmr_base = mmr_base; 623 uv_cpu_hub_info(cpu)->global_mmr_base = mmr_base;
@@ -636,6 +633,16 @@ void __init uv_system_init(void)
636 lcpu, blade); 633 lcpu, blade);
637 } 634 }
638 635
636 /* Add blade/pnode info for nodes without cpus */
637 for_each_online_node(nid) {
638 if (uv_node_to_blade[nid] >= 0)
639 continue;
640 paddr = node_start_pfn(nid) << PAGE_SHIFT;
641 pnode = (paddr >> m_val) & pnode_mask;
642 blade = boot_pnode_to_blade(pnode);
643 uv_node_to_blade[nid] = blade;
644 }
645
639 map_gru_high(max_pnode); 646 map_gru_high(max_pnode);
640 map_mmr_high(max_pnode); 647 map_mmr_high(max_pnode);
641 map_config_high(max_pnode); 648 map_config_high(max_pnode);
diff --git a/arch/x86/kernel/apm_32.c b/arch/x86/kernel/apm_32.c
index 10033fe718e0..49e0939bac42 100644
--- a/arch/x86/kernel/apm_32.c
+++ b/arch/x86/kernel/apm_32.c
@@ -466,7 +466,7 @@ static const lookup_t error_table[] = {
466 * @err: APM BIOS return code 466 * @err: APM BIOS return code
467 * 467 *
468 * Write a meaningful log entry to the kernel log in the event of 468 * Write a meaningful log entry to the kernel log in the event of
469 * an APM error. 469 * an APM error. Note that this also handles (negative) kernel errors.
470 */ 470 */
471 471
472static void apm_error(char *str, int err) 472static void apm_error(char *str, int err)
@@ -478,43 +478,14 @@ static void apm_error(char *str, int err)
478 break; 478 break;
479 if (i < ERROR_COUNT) 479 if (i < ERROR_COUNT)
480 printk(KERN_NOTICE "apm: %s: %s\n", str, error_table[i].msg); 480 printk(KERN_NOTICE "apm: %s: %s\n", str, error_table[i].msg);
481 else if (err < 0)
482 printk(KERN_NOTICE "apm: %s: linux error code %i\n", str, err);
481 else 483 else
482 printk(KERN_NOTICE "apm: %s: unknown error code %#2.2x\n", 484 printk(KERN_NOTICE "apm: %s: unknown error code %#2.2x\n",
483 str, err); 485 str, err);
484} 486}
485 487
486/* 488/*
487 * Lock APM functionality to physical CPU 0
488 */
489
490#ifdef CONFIG_SMP
491
492static cpumask_t apm_save_cpus(void)
493{
494 cpumask_t x = current->cpus_allowed;
495 /* Some bioses don't like being called from CPU != 0 */
496 set_cpus_allowed(current, cpumask_of_cpu(0));
497 BUG_ON(smp_processor_id() != 0);
498 return x;
499}
500
501static inline void apm_restore_cpus(cpumask_t mask)
502{
503 set_cpus_allowed(current, mask);
504}
505
506#else
507
508/*
509 * No CPU lockdown needed on a uniprocessor
510 */
511
512#define apm_save_cpus() (current->cpus_allowed)
513#define apm_restore_cpus(x) (void)(x)
514
515#endif
516
517/*
518 * These are the actual BIOS calls. Depending on APM_ZERO_SEGS and 489 * These are the actual BIOS calls. Depending on APM_ZERO_SEGS and
519 * apm_info.allow_ints, we are being really paranoid here! Not only 490 * apm_info.allow_ints, we are being really paranoid here! Not only
520 * are interrupts disabled, but all the segment registers (except SS) 491 * are interrupts disabled, but all the segment registers (except SS)
@@ -568,16 +539,23 @@ static inline void apm_irq_restore(unsigned long flags)
568# define APM_DO_RESTORE_SEGS 539# define APM_DO_RESTORE_SEGS
569#endif 540#endif
570 541
542struct apm_bios_call {
543 u32 func;
544 /* In and out */
545 u32 ebx;
546 u32 ecx;
547 /* Out only */
548 u32 eax;
549 u32 edx;
550 u32 esi;
551
552 /* Error: -ENOMEM, or bits 8-15 of eax */
553 int err;
554};
555
571/** 556/**
572 * apm_bios_call - Make an APM BIOS 32bit call 557 * __apm_bios_call - Make an APM BIOS 32bit call
573 * @func: APM function to execute 558 * @_call: pointer to struct apm_bios_call.
574 * @ebx_in: EBX register for call entry
575 * @ecx_in: ECX register for call entry
576 * @eax: EAX register return
577 * @ebx: EBX register return
578 * @ecx: ECX register return
579 * @edx: EDX register return
580 * @esi: ESI register return
581 * 559 *
582 * Make an APM call using the 32bit protected mode interface. The 560 * Make an APM call using the 32bit protected mode interface. The
583 * caller is responsible for knowing if APM BIOS is configured and 561 * caller is responsible for knowing if APM BIOS is configured and
@@ -586,80 +564,142 @@ static inline void apm_irq_restore(unsigned long flags)
586 * flag is loaded into AL. If there is an error, then the error 564 * flag is loaded into AL. If there is an error, then the error
587 * code is returned in AH (bits 8-15 of eax) and this function 565 * code is returned in AH (bits 8-15 of eax) and this function
588 * returns non-zero. 566 * returns non-zero.
567 *
568 * Note: this makes the call on the current CPU.
589 */ 569 */
590 570static long __apm_bios_call(void *_call)
591static u8 apm_bios_call(u32 func, u32 ebx_in, u32 ecx_in,
592 u32 *eax, u32 *ebx, u32 *ecx, u32 *edx, u32 *esi)
593{ 571{
594 APM_DECL_SEGS 572 APM_DECL_SEGS
595 unsigned long flags; 573 unsigned long flags;
596 cpumask_t cpus;
597 int cpu; 574 int cpu;
598 struct desc_struct save_desc_40; 575 struct desc_struct save_desc_40;
599 struct desc_struct *gdt; 576 struct desc_struct *gdt;
600 577 struct apm_bios_call *call = _call;
601 cpus = apm_save_cpus();
602 578
603 cpu = get_cpu(); 579 cpu = get_cpu();
580 BUG_ON(cpu != 0);
604 gdt = get_cpu_gdt_table(cpu); 581 gdt = get_cpu_gdt_table(cpu);
605 save_desc_40 = gdt[0x40 / 8]; 582 save_desc_40 = gdt[0x40 / 8];
606 gdt[0x40 / 8] = bad_bios_desc; 583 gdt[0x40 / 8] = bad_bios_desc;
607 584
608 apm_irq_save(flags); 585 apm_irq_save(flags);
609 APM_DO_SAVE_SEGS; 586 APM_DO_SAVE_SEGS;
610 apm_bios_call_asm(func, ebx_in, ecx_in, eax, ebx, ecx, edx, esi); 587 apm_bios_call_asm(call->func, call->ebx, call->ecx,
588 &call->eax, &call->ebx, &call->ecx, &call->edx,
589 &call->esi);
611 APM_DO_RESTORE_SEGS; 590 APM_DO_RESTORE_SEGS;
612 apm_irq_restore(flags); 591 apm_irq_restore(flags);
613 gdt[0x40 / 8] = save_desc_40; 592 gdt[0x40 / 8] = save_desc_40;
614 put_cpu(); 593 put_cpu();
615 apm_restore_cpus(cpus);
616 594
617 return *eax & 0xff; 595 return call->eax & 0xff;
596}
597
598/* Run __apm_bios_call or __apm_bios_call_simple on CPU 0 */
599static int on_cpu0(long (*fn)(void *), struct apm_bios_call *call)
600{
601 int ret;
602
603 /* Don't bother with work_on_cpu in the common case, so we don't
604 * have to worry about OOM or overhead. */
605 if (get_cpu() == 0) {
606 ret = fn(call);
607 put_cpu();
608 } else {
609 put_cpu();
610 ret = work_on_cpu(0, fn, call);
611 }
612
613 /* work_on_cpu can fail with -ENOMEM */
614 if (ret < 0)
615 call->err = ret;
616 else
617 call->err = (call->eax >> 8) & 0xff;
618
619 return ret;
618} 620}
619 621
620/** 622/**
621 * apm_bios_call_simple - make a simple APM BIOS 32bit call 623 * apm_bios_call - Make an APM BIOS 32bit call (on CPU 0)
622 * @func: APM function to invoke 624 * @call: the apm_bios_call registers.
623 * @ebx_in: EBX register value for BIOS call 625 *
624 * @ecx_in: ECX register value for BIOS call 626 * If there is an error, it is returned in @call.err.
625 * @eax: EAX register on return from the BIOS call 627 */
628static int apm_bios_call(struct apm_bios_call *call)
629{
630 return on_cpu0(__apm_bios_call, call);
631}
632
633/**
634 * __apm_bios_call_simple - Make an APM BIOS 32bit call (on CPU 0)
635 * @_call: pointer to struct apm_bios_call.
626 * 636 *
627 * Make a BIOS call that returns one value only, or just status. 637 * Make a BIOS call that returns one value only, or just status.
628 * If there is an error, then the error code is returned in AH 638 * If there is an error, then the error code is returned in AH
629 * (bits 8-15 of eax) and this function returns non-zero. This is 639 * (bits 8-15 of eax) and this function returns non-zero (it can
630 * used for simpler BIOS operations. This call may hold interrupts 640 * also return -ENOMEM). This is used for simpler BIOS operations.
631 * off for a long time on some laptops. 641 * This call may hold interrupts off for a long time on some laptops.
642 *
643 * Note: this makes the call on the current CPU.
632 */ 644 */
633 645static long __apm_bios_call_simple(void *_call)
634static u8 apm_bios_call_simple(u32 func, u32 ebx_in, u32 ecx_in, u32 *eax)
635{ 646{
636 u8 error; 647 u8 error;
637 APM_DECL_SEGS 648 APM_DECL_SEGS
638 unsigned long flags; 649 unsigned long flags;
639 cpumask_t cpus;
640 int cpu; 650 int cpu;
641 struct desc_struct save_desc_40; 651 struct desc_struct save_desc_40;
642 struct desc_struct *gdt; 652 struct desc_struct *gdt;
643 653 struct apm_bios_call *call = _call;
644 cpus = apm_save_cpus();
645 654
646 cpu = get_cpu(); 655 cpu = get_cpu();
656 BUG_ON(cpu != 0);
647 gdt = get_cpu_gdt_table(cpu); 657 gdt = get_cpu_gdt_table(cpu);
648 save_desc_40 = gdt[0x40 / 8]; 658 save_desc_40 = gdt[0x40 / 8];
649 gdt[0x40 / 8] = bad_bios_desc; 659 gdt[0x40 / 8] = bad_bios_desc;
650 660
651 apm_irq_save(flags); 661 apm_irq_save(flags);
652 APM_DO_SAVE_SEGS; 662 APM_DO_SAVE_SEGS;
653 error = apm_bios_call_simple_asm(func, ebx_in, ecx_in, eax); 663 error = apm_bios_call_simple_asm(call->func, call->ebx, call->ecx,
664 &call->eax);
654 APM_DO_RESTORE_SEGS; 665 APM_DO_RESTORE_SEGS;
655 apm_irq_restore(flags); 666 apm_irq_restore(flags);
656 gdt[0x40 / 8] = save_desc_40; 667 gdt[0x40 / 8] = save_desc_40;
657 put_cpu(); 668 put_cpu();
658 apm_restore_cpus(cpus);
659 return error; 669 return error;
660} 670}
661 671
662/** 672/**
673 * apm_bios_call_simple - make a simple APM BIOS 32bit call
674 * @func: APM function to invoke
675 * @ebx_in: EBX register value for BIOS call
676 * @ecx_in: ECX register value for BIOS call
677 * @eax: EAX register on return from the BIOS call
678 * @err: bits
679 *
680 * Make a BIOS call that returns one value only, or just status.
681 * If there is an error, then the error code is returned in @err
682 * and this function returns non-zero. This is used for simpler
683 * BIOS operations. This call may hold interrupts off for a long
684 * time on some laptops.
685 */
686static int apm_bios_call_simple(u32 func, u32 ebx_in, u32 ecx_in, u32 *eax,
687 int *err)
688{
689 struct apm_bios_call call;
690 int ret;
691
692 call.func = func;
693 call.ebx = ebx_in;
694 call.ecx = ecx_in;
695
696 ret = on_cpu0(__apm_bios_call_simple, &call);
697 *eax = call.eax;
698 *err = call.err;
699 return ret;
700}
701
702/**
663 * apm_driver_version - APM driver version 703 * apm_driver_version - APM driver version
664 * @val: loaded with the APM version on return 704 * @val: loaded with the APM version on return
665 * 705 *
@@ -678,9 +718,10 @@ static u8 apm_bios_call_simple(u32 func, u32 ebx_in, u32 ecx_in, u32 *eax)
678static int apm_driver_version(u_short *val) 718static int apm_driver_version(u_short *val)
679{ 719{
680 u32 eax; 720 u32 eax;
721 int err;
681 722
682 if (apm_bios_call_simple(APM_FUNC_VERSION, 0, *val, &eax)) 723 if (apm_bios_call_simple(APM_FUNC_VERSION, 0, *val, &eax, &err))
683 return (eax >> 8) & 0xff; 724 return err;
684 *val = eax; 725 *val = eax;
685 return APM_SUCCESS; 726 return APM_SUCCESS;
686} 727}
@@ -701,22 +742,21 @@ static int apm_driver_version(u_short *val)
701 * that APM 1.2 is in use. If no messges are pending the value 0x80 742 * that APM 1.2 is in use. If no messges are pending the value 0x80
702 * is returned (No power management events pending). 743 * is returned (No power management events pending).
703 */ 744 */
704
705static int apm_get_event(apm_event_t *event, apm_eventinfo_t *info) 745static int apm_get_event(apm_event_t *event, apm_eventinfo_t *info)
706{ 746{
707 u32 eax; 747 struct apm_bios_call call;
708 u32 ebx;
709 u32 ecx;
710 u32 dummy;
711 748
712 if (apm_bios_call(APM_FUNC_GET_EVENT, 0, 0, &eax, &ebx, &ecx, 749 call.func = APM_FUNC_GET_EVENT;
713 &dummy, &dummy)) 750 call.ebx = call.ecx = 0;
714 return (eax >> 8) & 0xff; 751
715 *event = ebx; 752 if (apm_bios_call(&call))
753 return call.err;
754
755 *event = call.ebx;
716 if (apm_info.connection_version < 0x0102) 756 if (apm_info.connection_version < 0x0102)
717 *info = ~0; /* indicate info not valid */ 757 *info = ~0; /* indicate info not valid */
718 else 758 else
719 *info = ecx; 759 *info = call.ecx;
720 return APM_SUCCESS; 760 return APM_SUCCESS;
721} 761}
722 762
@@ -737,9 +777,10 @@ static int apm_get_event(apm_event_t *event, apm_eventinfo_t *info)
737static int set_power_state(u_short what, u_short state) 777static int set_power_state(u_short what, u_short state)
738{ 778{
739 u32 eax; 779 u32 eax;
780 int err;
740 781
741 if (apm_bios_call_simple(APM_FUNC_SET_STATE, what, state, &eax)) 782 if (apm_bios_call_simple(APM_FUNC_SET_STATE, what, state, &eax, &err))
742 return (eax >> 8) & 0xff; 783 return err;
743 return APM_SUCCESS; 784 return APM_SUCCESS;
744} 785}
745 786
@@ -770,6 +811,7 @@ static int apm_do_idle(void)
770 u8 ret = 0; 811 u8 ret = 0;
771 int idled = 0; 812 int idled = 0;
772 int polling; 813 int polling;
814 int err;
773 815
774 polling = !!(current_thread_info()->status & TS_POLLING); 816 polling = !!(current_thread_info()->status & TS_POLLING);
775 if (polling) { 817 if (polling) {
@@ -782,7 +824,7 @@ static int apm_do_idle(void)
782 } 824 }
783 if (!need_resched()) { 825 if (!need_resched()) {
784 idled = 1; 826 idled = 1;
785 ret = apm_bios_call_simple(APM_FUNC_IDLE, 0, 0, &eax); 827 ret = apm_bios_call_simple(APM_FUNC_IDLE, 0, 0, &eax, &err);
786 } 828 }
787 if (polling) 829 if (polling)
788 current_thread_info()->status |= TS_POLLING; 830 current_thread_info()->status |= TS_POLLING;
@@ -797,8 +839,7 @@ static int apm_do_idle(void)
797 * Only report the failure the first 5 times. 839 * Only report the failure the first 5 times.
798 */ 840 */
799 if (++t < 5) { 841 if (++t < 5) {
800 printk(KERN_DEBUG "apm_do_idle failed (%d)\n", 842 printk(KERN_DEBUG "apm_do_idle failed (%d)\n", err);
801 (eax >> 8) & 0xff);
802 t = jiffies; 843 t = jiffies;
803 } 844 }
804 return -1; 845 return -1;
@@ -816,9 +857,10 @@ static int apm_do_idle(void)
816static void apm_do_busy(void) 857static void apm_do_busy(void)
817{ 858{
818 u32 dummy; 859 u32 dummy;
860 int err;
819 861
820 if (clock_slowed || ALWAYS_CALL_BUSY) { 862 if (clock_slowed || ALWAYS_CALL_BUSY) {
821 (void)apm_bios_call_simple(APM_FUNC_BUSY, 0, 0, &dummy); 863 (void)apm_bios_call_simple(APM_FUNC_BUSY, 0, 0, &dummy, &err);
822 clock_slowed = 0; 864 clock_slowed = 0;
823 } 865 }
824} 866}
@@ -937,7 +979,7 @@ static void apm_power_off(void)
937 979
938 /* Some bioses don't like being called from CPU != 0 */ 980 /* Some bioses don't like being called from CPU != 0 */
939 if (apm_info.realmode_power_off) { 981 if (apm_info.realmode_power_off) {
940 (void)apm_save_cpus(); 982 set_cpus_allowed_ptr(current, cpumask_of(0));
941 machine_real_restart(po_bios_call, sizeof(po_bios_call)); 983 machine_real_restart(po_bios_call, sizeof(po_bios_call));
942 } else { 984 } else {
943 (void)set_system_power_state(APM_STATE_OFF); 985 (void)set_system_power_state(APM_STATE_OFF);
@@ -956,12 +998,13 @@ static void apm_power_off(void)
956static int apm_enable_power_management(int enable) 998static int apm_enable_power_management(int enable)
957{ 999{
958 u32 eax; 1000 u32 eax;
1001 int err;
959 1002
960 if ((enable == 0) && (apm_info.bios.flags & APM_BIOS_DISENGAGED)) 1003 if ((enable == 0) && (apm_info.bios.flags & APM_BIOS_DISENGAGED))
961 return APM_NOT_ENGAGED; 1004 return APM_NOT_ENGAGED;
962 if (apm_bios_call_simple(APM_FUNC_ENABLE_PM, APM_DEVICE_BALL, 1005 if (apm_bios_call_simple(APM_FUNC_ENABLE_PM, APM_DEVICE_BALL,
963 enable, &eax)) 1006 enable, &eax, &err))
964 return (eax >> 8) & 0xff; 1007 return err;
965 if (enable) 1008 if (enable)
966 apm_info.bios.flags &= ~APM_BIOS_DISABLED; 1009 apm_info.bios.flags &= ~APM_BIOS_DISABLED;
967 else 1010 else
@@ -986,24 +1029,23 @@ static int apm_enable_power_management(int enable)
986 1029
987static int apm_get_power_status(u_short *status, u_short *bat, u_short *life) 1030static int apm_get_power_status(u_short *status, u_short *bat, u_short *life)
988{ 1031{
989 u32 eax; 1032 struct apm_bios_call call;
990 u32 ebx; 1033
991 u32 ecx; 1034 call.func = APM_FUNC_GET_STATUS;
992 u32 edx; 1035 call.ebx = APM_DEVICE_ALL;
993 u32 dummy; 1036 call.ecx = 0;
994 1037
995 if (apm_info.get_power_status_broken) 1038 if (apm_info.get_power_status_broken)
996 return APM_32_UNSUPPORTED; 1039 return APM_32_UNSUPPORTED;
997 if (apm_bios_call(APM_FUNC_GET_STATUS, APM_DEVICE_ALL, 0, 1040 if (apm_bios_call(&call))
998 &eax, &ebx, &ecx, &edx, &dummy)) 1041 return call.err;
999 return (eax >> 8) & 0xff; 1042 *status = call.ebx;
1000 *status = ebx; 1043 *bat = call.ecx;
1001 *bat = ecx;
1002 if (apm_info.get_power_status_swabinminutes) { 1044 if (apm_info.get_power_status_swabinminutes) {
1003 *life = swab16((u16)edx); 1045 *life = swab16((u16)call.edx);
1004 *life |= 0x8000; 1046 *life |= 0x8000;
1005 } else 1047 } else
1006 *life = edx; 1048 *life = call.edx;
1007 return APM_SUCCESS; 1049 return APM_SUCCESS;
1008} 1050}
1009 1051
@@ -1048,12 +1090,14 @@ static int apm_get_battery_status(u_short which, u_short *status,
1048static int apm_engage_power_management(u_short device, int enable) 1090static int apm_engage_power_management(u_short device, int enable)
1049{ 1091{
1050 u32 eax; 1092 u32 eax;
1093 int err;
1051 1094
1052 if ((enable == 0) && (device == APM_DEVICE_ALL) 1095 if ((enable == 0) && (device == APM_DEVICE_ALL)
1053 && (apm_info.bios.flags & APM_BIOS_DISABLED)) 1096 && (apm_info.bios.flags & APM_BIOS_DISABLED))
1054 return APM_DISABLED; 1097 return APM_DISABLED;
1055 if (apm_bios_call_simple(APM_FUNC_ENGAGE_PM, device, enable, &eax)) 1098 if (apm_bios_call_simple(APM_FUNC_ENGAGE_PM, device, enable,
1056 return (eax >> 8) & 0xff; 1099 &eax, &err))
1100 return err;
1057 if (device == APM_DEVICE_ALL) { 1101 if (device == APM_DEVICE_ALL) {
1058 if (enable) 1102 if (enable)
1059 apm_info.bios.flags &= ~APM_BIOS_DISENGAGED; 1103 apm_info.bios.flags &= ~APM_BIOS_DISENGAGED;
@@ -1190,8 +1234,10 @@ static int suspend(int vetoable)
1190 struct apm_user *as; 1234 struct apm_user *as;
1191 1235
1192 device_suspend(PMSG_SUSPEND); 1236 device_suspend(PMSG_SUSPEND);
1193 local_irq_disable(); 1237
1194 device_power_down(PMSG_SUSPEND); 1238 device_power_down(PMSG_SUSPEND);
1239
1240 local_irq_disable();
1195 sysdev_suspend(PMSG_SUSPEND); 1241 sysdev_suspend(PMSG_SUSPEND);
1196 1242
1197 local_irq_enable(); 1243 local_irq_enable();
@@ -1209,9 +1255,12 @@ static int suspend(int vetoable)
1209 if (err != APM_SUCCESS) 1255 if (err != APM_SUCCESS)
1210 apm_error("suspend", err); 1256 apm_error("suspend", err);
1211 err = (err == APM_SUCCESS) ? 0 : -EIO; 1257 err = (err == APM_SUCCESS) ? 0 : -EIO;
1258
1212 sysdev_resume(); 1259 sysdev_resume();
1213 device_power_up(PMSG_RESUME);
1214 local_irq_enable(); 1260 local_irq_enable();
1261
1262 device_power_up(PMSG_RESUME);
1263
1215 device_resume(PMSG_RESUME); 1264 device_resume(PMSG_RESUME);
1216 queue_event(APM_NORMAL_RESUME, NULL); 1265 queue_event(APM_NORMAL_RESUME, NULL);
1217 spin_lock(&user_list_lock); 1266 spin_lock(&user_list_lock);
@@ -1228,8 +1277,9 @@ static void standby(void)
1228{ 1277{
1229 int err; 1278 int err;
1230 1279
1231 local_irq_disable();
1232 device_power_down(PMSG_SUSPEND); 1280 device_power_down(PMSG_SUSPEND);
1281
1282 local_irq_disable();
1233 sysdev_suspend(PMSG_SUSPEND); 1283 sysdev_suspend(PMSG_SUSPEND);
1234 local_irq_enable(); 1284 local_irq_enable();
1235 1285
@@ -1239,8 +1289,9 @@ static void standby(void)
1239 1289
1240 local_irq_disable(); 1290 local_irq_disable();
1241 sysdev_resume(); 1291 sysdev_resume();
1242 device_power_up(PMSG_RESUME);
1243 local_irq_enable(); 1292 local_irq_enable();
1293
1294 device_power_up(PMSG_RESUME);
1244} 1295}
1245 1296
1246static apm_event_t get_event(void) 1297static apm_event_t get_event(void)
@@ -1682,16 +1733,14 @@ static int apm(void *unused)
1682 char *power_stat; 1733 char *power_stat;
1683 char *bat_stat; 1734 char *bat_stat;
1684 1735
1685#ifdef CONFIG_SMP
1686 /* 2002/08/01 - WT 1736 /* 2002/08/01 - WT
1687 * This is to avoid random crashes at boot time during initialization 1737 * This is to avoid random crashes at boot time during initialization
1688 * on SMP systems in case of "apm=power-off" mode. Seen on ASUS A7M266D. 1738 * on SMP systems in case of "apm=power-off" mode. Seen on ASUS A7M266D.
1689 * Some bioses don't like being called from CPU != 0. 1739 * Some bioses don't like being called from CPU != 0.
1690 * Method suggested by Ingo Molnar. 1740 * Method suggested by Ingo Molnar.
1691 */ 1741 */
1692 set_cpus_allowed(current, cpumask_of_cpu(0)); 1742 set_cpus_allowed_ptr(current, cpumask_of(0));
1693 BUG_ON(smp_processor_id() != 0); 1743 BUG_ON(smp_processor_id() != 0);
1694#endif
1695 1744
1696 if (apm_info.connection_version == 0) { 1745 if (apm_info.connection_version == 0) {
1697 apm_info.connection_version = apm_info.bios.version; 1746 apm_info.connection_version = apm_info.bios.version;
diff --git a/arch/x86/kernel/asm-offsets_32.c b/arch/x86/kernel/asm-offsets_32.c
index fbf2f33e3080..5a6aa1c1162f 100644
--- a/arch/x86/kernel/asm-offsets_32.c
+++ b/arch/x86/kernel/asm-offsets_32.c
@@ -18,6 +18,7 @@
18#include <asm/thread_info.h> 18#include <asm/thread_info.h>
19#include <asm/bootparam.h> 19#include <asm/bootparam.h>
20#include <asm/elf.h> 20#include <asm/elf.h>
21#include <asm/suspend.h>
21 22
22#include <xen/interface/xen.h> 23#include <xen/interface/xen.h>
23 24
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index 8793ab33e2c1..e72f062fb4b5 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -16,6 +16,7 @@
16#include <asm/thread_info.h> 16#include <asm/thread_info.h>
17#include <asm/ia32.h> 17#include <asm/ia32.h>
18#include <asm/bootparam.h> 18#include <asm/bootparam.h>
19#include <asm/suspend.h>
19 20
20#include <xen/interface/xen.h> 21#include <xen/interface/xen.h>
21 22
diff --git a/arch/x86/kernel/bios_uv.c b/arch/x86/kernel/bios_uv.c
index f63882728d91..63a88e1f987d 100644
--- a/arch/x86/kernel/bios_uv.c
+++ b/arch/x86/kernel/bios_uv.c
@@ -182,7 +182,8 @@ void uv_bios_init(void)
182 memcpy(&uv_systab, tab, sizeof(struct uv_systab)); 182 memcpy(&uv_systab, tab, sizeof(struct uv_systab));
183 iounmap(tab); 183 iounmap(tab);
184 184
185 printk(KERN_INFO "EFI UV System Table Revision %d\n", tab->revision); 185 printk(KERN_INFO "EFI UV System Table Revision %d\n",
186 uv_systab.revision);
186} 187}
187#else /* !CONFIG_EFI */ 188#else /* !CONFIG_EFI */
188 189
diff --git a/arch/x86/kernel/cpu/addon_cpuid_features.c b/arch/x86/kernel/cpu/addon_cpuid_features.c
index 8220ae69849d..c965e5212714 100644
--- a/arch/x86/kernel/cpu/addon_cpuid_features.c
+++ b/arch/x86/kernel/cpu/addon_cpuid_features.c
@@ -31,6 +31,7 @@ void __cpuinit init_scattered_cpuid_features(struct cpuinfo_x86 *c)
31 31
32 static const struct cpuid_bit __cpuinitconst cpuid_bits[] = { 32 static const struct cpuid_bit __cpuinitconst cpuid_bits[] = {
33 { X86_FEATURE_IDA, CR_EAX, 1, 0x00000006 }, 33 { X86_FEATURE_IDA, CR_EAX, 1, 0x00000006 },
34 { X86_FEATURE_ARAT, CR_EAX, 2, 0x00000006 },
34 { 0, 0, 0, 0 } 35 { 0, 0, 0, 0 }
35 }; 36 };
36 37
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index e2962cc1e27b..c4f667896c28 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -41,8 +41,6 @@
41 41
42#include "cpu.h" 42#include "cpu.h"
43 43
44#ifdef CONFIG_X86_64
45
46/* all of these masks are initialized in setup_cpu_local_masks() */ 44/* all of these masks are initialized in setup_cpu_local_masks() */
47cpumask_var_t cpu_initialized_mask; 45cpumask_var_t cpu_initialized_mask;
48cpumask_var_t cpu_callout_mask; 46cpumask_var_t cpu_callout_mask;
@@ -60,16 +58,6 @@ void __init setup_cpu_local_masks(void)
60 alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask); 58 alloc_bootmem_cpumask_var(&cpu_sibling_setup_mask);
61} 59}
62 60
63#else /* CONFIG_X86_32 */
64
65cpumask_t cpu_sibling_setup_map;
66cpumask_t cpu_callout_map;
67cpumask_t cpu_initialized;
68cpumask_t cpu_callin_map;
69
70#endif /* CONFIG_X86_32 */
71
72
73static const struct cpu_dev *this_cpu __cpuinitdata; 61static const struct cpu_dev *this_cpu __cpuinitdata;
74 62
75DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = { 63DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = {
@@ -859,6 +847,7 @@ static void vgetcpu_set_mode(void)
859void __init identify_boot_cpu(void) 847void __init identify_boot_cpu(void)
860{ 848{
861 identify_cpu(&boot_cpu_data); 849 identify_cpu(&boot_cpu_data);
850 init_c1e_mask();
862#ifdef CONFIG_X86_32 851#ifdef CONFIG_X86_32
863 sysenter_setup(); 852 sysenter_setup();
864 enable_sep_cpu(); 853 enable_sep_cpu();
diff --git a/arch/x86/kernel/cpu/cpu.h b/arch/x86/kernel/cpu/cpu.h
index 9469ecb5aeb8..6de9a908e400 100644
--- a/arch/x86/kernel/cpu/cpu.h
+++ b/arch/x86/kernel/cpu/cpu.h
@@ -3,25 +3,25 @@
3#define ARCH_X86_CPU_H 3#define ARCH_X86_CPU_H
4 4
5struct cpu_model_info { 5struct cpu_model_info {
6 int vendor; 6 int vendor;
7 int family; 7 int family;
8 const char *model_names[16]; 8 const char *model_names[16];
9}; 9};
10 10
11/* attempt to consolidate cpu attributes */ 11/* attempt to consolidate cpu attributes */
12struct cpu_dev { 12struct cpu_dev {
13 const char * c_vendor; 13 const char *c_vendor;
14 14
15 /* some have two possibilities for cpuid string */ 15 /* some have two possibilities for cpuid string */
16 const char * c_ident[2]; 16 const char *c_ident[2];
17 17
18 struct cpu_model_info c_models[4]; 18 struct cpu_model_info c_models[4];
19 19
20 void (*c_early_init)(struct cpuinfo_x86 *c); 20 void (*c_early_init)(struct cpuinfo_x86 *);
21 void (*c_init)(struct cpuinfo_x86 * c); 21 void (*c_init)(struct cpuinfo_x86 *);
22 void (*c_identify)(struct cpuinfo_x86 * c); 22 void (*c_identify)(struct cpuinfo_x86 *);
23 unsigned int (*c_size_cache)(struct cpuinfo_x86 * c, unsigned int size); 23 unsigned int (*c_size_cache)(struct cpuinfo_x86 *, unsigned int);
24 int c_x86_vendor; 24 int c_x86_vendor;
25}; 25};
26 26
27#define cpu_dev_register(cpu_devX) \ 27#define cpu_dev_register(cpu_devX) \
diff --git a/arch/x86/kernel/cpu/cpu_debug.c b/arch/x86/kernel/cpu/cpu_debug.c
index 46e29ab96c6a..46e29ab96c6a 100755..100644
--- a/arch/x86/kernel/cpu/cpu_debug.c
+++ b/arch/x86/kernel/cpu/cpu_debug.c
diff --git a/arch/x86/kernel/cpu/cpufreq/Kconfig b/arch/x86/kernel/cpu/cpufreq/Kconfig
index 65792c2cc462..52c839875478 100644
--- a/arch/x86/kernel/cpu/cpufreq/Kconfig
+++ b/arch/x86/kernel/cpu/cpufreq/Kconfig
@@ -87,30 +87,15 @@ config X86_POWERNOW_K7_ACPI
87config X86_POWERNOW_K8 87config X86_POWERNOW_K8
88 tristate "AMD Opteron/Athlon64 PowerNow!" 88 tristate "AMD Opteron/Athlon64 PowerNow!"
89 select CPU_FREQ_TABLE 89 select CPU_FREQ_TABLE
90 depends on ACPI && ACPI_PROCESSOR
90 help 91 help
91 This adds the CPUFreq driver for mobile AMD Opteron/Athlon64 processors. 92 This adds the CPUFreq driver for K8/K10 Opteron/Athlon64 processors.
92 93
93 To compile this driver as a module, choose M here: the 94 To compile this driver as a module, choose M here: the
94 module will be called powernow-k8. 95 module will be called powernow-k8.
95 96
96 For details, take a look at <file:Documentation/cpu-freq/>. 97 For details, take a look at <file:Documentation/cpu-freq/>.
97 98
98 If in doubt, say N.
99
100config X86_POWERNOW_K8_ACPI
101 bool
102 prompt "ACPI Support" if X86_32
103 depends on ACPI && X86_POWERNOW_K8 && ACPI_PROCESSOR
104 depends on !(X86_POWERNOW_K8 = y && ACPI_PROCESSOR = m)
105 default y
106 help
107 This provides access to the K8s Processor Performance States via ACPI.
108 This driver is probably required for CPUFreq to work with multi-socket and
109 SMP systems. It is not required on at least some single-socket yet
110 multi-core systems, even if SMP is enabled.
111
112 It is safe to say Y here.
113
114config X86_GX_SUSPMOD 99config X86_GX_SUSPMOD
115 tristate "Cyrix MediaGX/NatSemi Geode Suspend Modulation" 100 tristate "Cyrix MediaGX/NatSemi Geode Suspend Modulation"
116 depends on X86_32 && PCI 101 depends on X86_32 && PCI
diff --git a/arch/x86/kernel/cpu/cpufreq/Makefile b/arch/x86/kernel/cpu/cpufreq/Makefile
index 560f7760dae5..509296df294d 100644
--- a/arch/x86/kernel/cpu/cpufreq/Makefile
+++ b/arch/x86/kernel/cpu/cpufreq/Makefile
@@ -1,6 +1,11 @@
1# Link order matters. K8 is preferred to ACPI because of firmware bugs in early
2# K8 systems. ACPI is preferred to all other hardware-specific drivers.
3# speedstep-* is preferred over p4-clockmod.
4
5obj-$(CONFIG_X86_POWERNOW_K8) += powernow-k8.o
6obj-$(CONFIG_X86_ACPI_CPUFREQ) += acpi-cpufreq.o
1obj-$(CONFIG_X86_POWERNOW_K6) += powernow-k6.o 7obj-$(CONFIG_X86_POWERNOW_K6) += powernow-k6.o
2obj-$(CONFIG_X86_POWERNOW_K7) += powernow-k7.o 8obj-$(CONFIG_X86_POWERNOW_K7) += powernow-k7.o
3obj-$(CONFIG_X86_POWERNOW_K8) += powernow-k8.o
4obj-$(CONFIG_X86_LONGHAUL) += longhaul.o 9obj-$(CONFIG_X86_LONGHAUL) += longhaul.o
5obj-$(CONFIG_X86_E_POWERSAVER) += e_powersaver.o 10obj-$(CONFIG_X86_E_POWERSAVER) += e_powersaver.o
6obj-$(CONFIG_ELAN_CPUFREQ) += elanfreq.o 11obj-$(CONFIG_ELAN_CPUFREQ) += elanfreq.o
@@ -10,7 +15,6 @@ obj-$(CONFIG_X86_GX_SUSPMOD) += gx-suspmod.o
10obj-$(CONFIG_X86_SPEEDSTEP_ICH) += speedstep-ich.o 15obj-$(CONFIG_X86_SPEEDSTEP_ICH) += speedstep-ich.o
11obj-$(CONFIG_X86_SPEEDSTEP_LIB) += speedstep-lib.o 16obj-$(CONFIG_X86_SPEEDSTEP_LIB) += speedstep-lib.o
12obj-$(CONFIG_X86_SPEEDSTEP_SMI) += speedstep-smi.o 17obj-$(CONFIG_X86_SPEEDSTEP_SMI) += speedstep-smi.o
13obj-$(CONFIG_X86_ACPI_CPUFREQ) += acpi-cpufreq.o
14obj-$(CONFIG_X86_SPEEDSTEP_CENTRINO) += speedstep-centrino.o 18obj-$(CONFIG_X86_SPEEDSTEP_CENTRINO) += speedstep-centrino.o
15obj-$(CONFIG_X86_P4_CLOCKMOD) += p4-clockmod.o 19obj-$(CONFIG_X86_P4_CLOCKMOD) += p4-clockmod.o
16obj-$(CONFIG_X86_CPUFREQ_NFORCE2) += cpufreq-nforce2.o 20obj-$(CONFIG_X86_CPUFREQ_NFORCE2) += cpufreq-nforce2.o
diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
index 22590cf688ae..ecdb682ab516 100644
--- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
+++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
@@ -1,5 +1,5 @@
1/* 1/*
2 * acpi-cpufreq.c - ACPI Processor P-States Driver ($Revision: 1.4 $) 2 * acpi-cpufreq.c - ACPI Processor P-States Driver
3 * 3 *
4 * Copyright (C) 2001, 2002 Andy Grover <andrew.grover@intel.com> 4 * Copyright (C) 2001, 2002 Andy Grover <andrew.grover@intel.com>
5 * Copyright (C) 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@intel.com> 5 * Copyright (C) 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@intel.com>
@@ -33,19 +33,21 @@
33#include <linux/cpufreq.h> 33#include <linux/cpufreq.h>
34#include <linux/compiler.h> 34#include <linux/compiler.h>
35#include <linux/dmi.h> 35#include <linux/dmi.h>
36#include <linux/ftrace.h> 36#include <trace/power.h>
37 37
38#include <linux/acpi.h> 38#include <linux/acpi.h>
39#include <linux/io.h>
40#include <linux/delay.h>
41#include <linux/uaccess.h>
42
39#include <acpi/processor.h> 43#include <acpi/processor.h>
40 44
41#include <asm/io.h>
42#include <asm/msr.h> 45#include <asm/msr.h>
43#include <asm/processor.h> 46#include <asm/processor.h>
44#include <asm/cpufeature.h> 47#include <asm/cpufeature.h>
45#include <asm/delay.h>
46#include <asm/uaccess.h>
47 48
48#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "acpi-cpufreq", msg) 49#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
50 "acpi-cpufreq", msg)
49 51
50MODULE_AUTHOR("Paul Diefenbaugh, Dominik Brodowski"); 52MODULE_AUTHOR("Paul Diefenbaugh, Dominik Brodowski");
51MODULE_DESCRIPTION("ACPI Processor P-States Driver"); 53MODULE_DESCRIPTION("ACPI Processor P-States Driver");
@@ -66,10 +68,13 @@ struct acpi_cpufreq_data {
66 unsigned int max_freq; 68 unsigned int max_freq;
67 unsigned int resume; 69 unsigned int resume;
68 unsigned int cpu_feature; 70 unsigned int cpu_feature;
71 u64 saved_aperf, saved_mperf;
69}; 72};
70 73
71static DEFINE_PER_CPU(struct acpi_cpufreq_data *, drv_data); 74static DEFINE_PER_CPU(struct acpi_cpufreq_data *, drv_data);
72 75
76DEFINE_TRACE(power_mark);
77
73/* acpi_perf_data is a pointer to percpu data. */ 78/* acpi_perf_data is a pointer to percpu data. */
74static struct acpi_processor_performance *acpi_perf_data; 79static struct acpi_processor_performance *acpi_perf_data;
75 80
@@ -95,7 +100,7 @@ static unsigned extract_io(u32 value, struct acpi_cpufreq_data *data)
95 100
96 perf = data->acpi_data; 101 perf = data->acpi_data;
97 102
98 for (i=0; i<perf->state_count; i++) { 103 for (i = 0; i < perf->state_count; i++) {
99 if (value == perf->states[i].status) 104 if (value == perf->states[i].status)
100 return data->freq_table[i].frequency; 105 return data->freq_table[i].frequency;
101 } 106 }
@@ -110,7 +115,7 @@ static unsigned extract_msr(u32 msr, struct acpi_cpufreq_data *data)
110 msr &= INTEL_MSR_RANGE; 115 msr &= INTEL_MSR_RANGE;
111 perf = data->acpi_data; 116 perf = data->acpi_data;
112 117
113 for (i=0; data->freq_table[i].frequency != CPUFREQ_TABLE_END; i++) { 118 for (i = 0; data->freq_table[i].frequency != CPUFREQ_TABLE_END; i++) {
114 if (msr == perf->states[data->freq_table[i].index].status) 119 if (msr == perf->states[data->freq_table[i].index].status)
115 return data->freq_table[i].frequency; 120 return data->freq_table[i].frequency;
116 } 121 }
@@ -138,19 +143,18 @@ struct io_addr {
138 u8 bit_width; 143 u8 bit_width;
139}; 144};
140 145
141typedef union {
142 struct msr_addr msr;
143 struct io_addr io;
144} drv_addr_union;
145
146struct drv_cmd { 146struct drv_cmd {
147 unsigned int type; 147 unsigned int type;
148 const struct cpumask *mask; 148 const struct cpumask *mask;
149 drv_addr_union addr; 149 union {
150 struct msr_addr msr;
151 struct io_addr io;
152 } addr;
150 u32 val; 153 u32 val;
151}; 154};
152 155
153static long do_drv_read(void *_cmd) 156/* Called via smp_call_function_single(), on the target CPU */
157static void do_drv_read(void *_cmd)
154{ 158{
155 struct drv_cmd *cmd = _cmd; 159 struct drv_cmd *cmd = _cmd;
156 u32 h; 160 u32 h;
@@ -167,10 +171,10 @@ static long do_drv_read(void *_cmd)
167 default: 171 default:
168 break; 172 break;
169 } 173 }
170 return 0;
171} 174}
172 175
173static long do_drv_write(void *_cmd) 176/* Called via smp_call_function_many(), on the target CPUs */
177static void do_drv_write(void *_cmd)
174{ 178{
175 struct drv_cmd *cmd = _cmd; 179 struct drv_cmd *cmd = _cmd;
176 u32 lo, hi; 180 u32 lo, hi;
@@ -189,23 +193,24 @@ static long do_drv_write(void *_cmd)
189 default: 193 default:
190 break; 194 break;
191 } 195 }
192 return 0;
193} 196}
194 197
195static void drv_read(struct drv_cmd *cmd) 198static void drv_read(struct drv_cmd *cmd)
196{ 199{
197 cmd->val = 0; 200 cmd->val = 0;
198 201
199 work_on_cpu(cpumask_any(cmd->mask), do_drv_read, cmd); 202 smp_call_function_single(cpumask_any(cmd->mask), do_drv_read, cmd, 1);
200} 203}
201 204
202static void drv_write(struct drv_cmd *cmd) 205static void drv_write(struct drv_cmd *cmd)
203{ 206{
204 unsigned int i; 207 int this_cpu;
205 208
206 for_each_cpu(i, cmd->mask) { 209 this_cpu = get_cpu();
207 work_on_cpu(i, do_drv_write, cmd); 210 if (cpumask_test_cpu(this_cpu, cmd->mask))
208 } 211 do_drv_write(cmd);
212 smp_call_function_many(cmd->mask, do_drv_write, cmd, 1);
213 put_cpu();
209} 214}
210 215
211static u32 get_cur_val(const struct cpumask *mask) 216static u32 get_cur_val(const struct cpumask *mask)
@@ -239,28 +244,23 @@ static u32 get_cur_val(const struct cpumask *mask)
239 return cmd.val; 244 return cmd.val;
240} 245}
241 246
242struct perf_cur { 247struct perf_pair {
243 union { 248 union {
244 struct { 249 struct {
245 u32 lo; 250 u32 lo;
246 u32 hi; 251 u32 hi;
247 } split; 252 } split;
248 u64 whole; 253 u64 whole;
249 } aperf_cur, mperf_cur; 254 } aperf, mperf;
250}; 255};
251 256
252 257/* Called via smp_call_function_single(), on the target CPU */
253static long read_measured_perf_ctrs(void *_cur) 258static void read_measured_perf_ctrs(void *_cur)
254{ 259{
255 struct perf_cur *cur = _cur; 260 struct perf_pair *cur = _cur;
256
257 rdmsr(MSR_IA32_APERF, cur->aperf_cur.split.lo, cur->aperf_cur.split.hi);
258 rdmsr(MSR_IA32_MPERF, cur->mperf_cur.split.lo, cur->mperf_cur.split.hi);
259 261
260 wrmsr(MSR_IA32_APERF, 0, 0); 262 rdmsr(MSR_IA32_APERF, cur->aperf.split.lo, cur->aperf.split.hi);
261 wrmsr(MSR_IA32_MPERF, 0, 0); 263 rdmsr(MSR_IA32_MPERF, cur->mperf.split.lo, cur->mperf.split.hi);
262
263 return 0;
264} 264}
265 265
266/* 266/*
@@ -279,52 +279,57 @@ static long read_measured_perf_ctrs(void *_cur)
279static unsigned int get_measured_perf(struct cpufreq_policy *policy, 279static unsigned int get_measured_perf(struct cpufreq_policy *policy,
280 unsigned int cpu) 280 unsigned int cpu)
281{ 281{
282 struct perf_cur cur; 282 struct perf_pair readin, cur;
283 unsigned int perf_percent; 283 unsigned int perf_percent;
284 unsigned int retval; 284 unsigned int retval;
285 285
286 if (!work_on_cpu(cpu, read_measured_perf_ctrs, &cur)) 286 if (smp_call_function_single(cpu, read_measured_perf_ctrs, &readin, 1))
287 return 0; 287 return 0;
288 288
289 cur.aperf.whole = readin.aperf.whole -
290 per_cpu(drv_data, cpu)->saved_aperf;
291 cur.mperf.whole = readin.mperf.whole -
292 per_cpu(drv_data, cpu)->saved_mperf;
293 per_cpu(drv_data, cpu)->saved_aperf = readin.aperf.whole;
294 per_cpu(drv_data, cpu)->saved_mperf = readin.mperf.whole;
295
289#ifdef __i386__ 296#ifdef __i386__
290 /* 297 /*
291 * We dont want to do 64 bit divide with 32 bit kernel 298 * We dont want to do 64 bit divide with 32 bit kernel
292 * Get an approximate value. Return failure in case we cannot get 299 * Get an approximate value. Return failure in case we cannot get
293 * an approximate value. 300 * an approximate value.
294 */ 301 */
295 if (unlikely(cur.aperf_cur.split.hi || cur.mperf_cur.split.hi)) { 302 if (unlikely(cur.aperf.split.hi || cur.mperf.split.hi)) {
296 int shift_count; 303 int shift_count;
297 u32 h; 304 u32 h;
298 305
299 h = max_t(u32, cur.aperf_cur.split.hi, cur.mperf_cur.split.hi); 306 h = max_t(u32, cur.aperf.split.hi, cur.mperf.split.hi);
300 shift_count = fls(h); 307 shift_count = fls(h);
301 308
302 cur.aperf_cur.whole >>= shift_count; 309 cur.aperf.whole >>= shift_count;
303 cur.mperf_cur.whole >>= shift_count; 310 cur.mperf.whole >>= shift_count;
304 } 311 }
305 312
306 if (((unsigned long)(-1) / 100) < cur.aperf_cur.split.lo) { 313 if (((unsigned long)(-1) / 100) < cur.aperf.split.lo) {
307 int shift_count = 7; 314 int shift_count = 7;
308 cur.aperf_cur.split.lo >>= shift_count; 315 cur.aperf.split.lo >>= shift_count;
309 cur.mperf_cur.split.lo >>= shift_count; 316 cur.mperf.split.lo >>= shift_count;
310 } 317 }
311 318
312 if (cur.aperf_cur.split.lo && cur.mperf_cur.split.lo) 319 if (cur.aperf.split.lo && cur.mperf.split.lo)
313 perf_percent = (cur.aperf_cur.split.lo * 100) / 320 perf_percent = (cur.aperf.split.lo * 100) / cur.mperf.split.lo;
314 cur.mperf_cur.split.lo;
315 else 321 else
316 perf_percent = 0; 322 perf_percent = 0;
317 323
318#else 324#else
319 if (unlikely(((unsigned long)(-1) / 100) < cur.aperf_cur.whole)) { 325 if (unlikely(((unsigned long)(-1) / 100) < cur.aperf.whole)) {
320 int shift_count = 7; 326 int shift_count = 7;
321 cur.aperf_cur.whole >>= shift_count; 327 cur.aperf.whole >>= shift_count;
322 cur.mperf_cur.whole >>= shift_count; 328 cur.mperf.whole >>= shift_count;
323 } 329 }
324 330
325 if (cur.aperf_cur.whole && cur.mperf_cur.whole) 331 if (cur.aperf.whole && cur.mperf.whole)
326 perf_percent = (cur.aperf_cur.whole * 100) / 332 perf_percent = (cur.aperf.whole * 100) / cur.mperf.whole;
327 cur.mperf_cur.whole;
328 else 333 else
329 perf_percent = 0; 334 perf_percent = 0;
330 335
@@ -369,7 +374,7 @@ static unsigned int check_freqs(const struct cpumask *mask, unsigned int freq,
369 unsigned int cur_freq; 374 unsigned int cur_freq;
370 unsigned int i; 375 unsigned int i;
371 376
372 for (i=0; i<100; i++) { 377 for (i = 0; i < 100; i++) {
373 cur_freq = extract_freq(get_cur_val(mask), data); 378 cur_freq = extract_freq(get_cur_val(mask), data);
374 if (cur_freq == freq) 379 if (cur_freq == freq)
375 return 1; 380 return 1;
@@ -494,7 +499,7 @@ acpi_cpufreq_guess_freq(struct acpi_cpufreq_data *data, unsigned int cpu)
494 unsigned long freq; 499 unsigned long freq;
495 unsigned long freqn = perf->states[0].core_frequency * 1000; 500 unsigned long freqn = perf->states[0].core_frequency * 1000;
496 501
497 for (i=0; i<(perf->state_count-1); i++) { 502 for (i = 0; i < (perf->state_count-1); i++) {
498 freq = freqn; 503 freq = freqn;
499 freqn = perf->states[i+1].core_frequency * 1000; 504 freqn = perf->states[i+1].core_frequency * 1000;
500 if ((2 * cpu_khz) > (freqn + freq)) { 505 if ((2 * cpu_khz) > (freqn + freq)) {
@@ -673,17 +678,29 @@ static int acpi_cpufreq_cpu_init(struct cpufreq_policy *policy)
673 678
674 /* detect transition latency */ 679 /* detect transition latency */
675 policy->cpuinfo.transition_latency = 0; 680 policy->cpuinfo.transition_latency = 0;
676 for (i=0; i<perf->state_count; i++) { 681 for (i = 0; i < perf->state_count; i++) {
677 if ((perf->states[i].transition_latency * 1000) > 682 if ((perf->states[i].transition_latency * 1000) >
678 policy->cpuinfo.transition_latency) 683 policy->cpuinfo.transition_latency)
679 policy->cpuinfo.transition_latency = 684 policy->cpuinfo.transition_latency =
680 perf->states[i].transition_latency * 1000; 685 perf->states[i].transition_latency * 1000;
681 } 686 }
682 687
688 /* Check for high latency (>20uS) from buggy BIOSes, like on T42 */
689 if (perf->control_register.space_id == ACPI_ADR_SPACE_FIXED_HARDWARE &&
690 policy->cpuinfo.transition_latency > 20 * 1000) {
691 static int print_once;
692 policy->cpuinfo.transition_latency = 20 * 1000;
693 if (!print_once) {
694 print_once = 1;
695 printk(KERN_INFO "Capping off P-state tranision latency"
696 " at 20 uS\n");
697 }
698 }
699
683 data->max_freq = perf->states[0].core_frequency * 1000; 700 data->max_freq = perf->states[0].core_frequency * 1000;
684 /* table init */ 701 /* table init */
685 for (i=0; i<perf->state_count; i++) { 702 for (i = 0; i < perf->state_count; i++) {
686 if (i>0 && perf->states[i].core_frequency >= 703 if (i > 0 && perf->states[i].core_frequency >=
687 data->freq_table[valid_states-1].frequency / 1000) 704 data->freq_table[valid_states-1].frequency / 1000)
688 continue; 705 continue;
689 706
diff --git a/arch/x86/kernel/cpu/cpufreq/cpufreq-nforce2.c b/arch/x86/kernel/cpu/cpufreq/cpufreq-nforce2.c
index 965ea52767ac..733093d60436 100644
--- a/arch/x86/kernel/cpu/cpufreq/cpufreq-nforce2.c
+++ b/arch/x86/kernel/cpu/cpufreq/cpufreq-nforce2.c
@@ -32,7 +32,7 @@
32 * nforce2_chipset: 32 * nforce2_chipset:
33 * FSB is changed using the chipset 33 * FSB is changed using the chipset
34 */ 34 */
35static struct pci_dev *nforce2_chipset_dev; 35static struct pci_dev *nforce2_dev;
36 36
37/* fid: 37/* fid:
38 * multiplier * 10 38 * multiplier * 10
@@ -56,7 +56,9 @@ MODULE_PARM_DESC(fid, "CPU multiplier to use (11.5 = 115)");
56MODULE_PARM_DESC(min_fsb, 56MODULE_PARM_DESC(min_fsb,
57 "Minimum FSB to use, if not defined: current FSB - 50"); 57 "Minimum FSB to use, if not defined: current FSB - 50");
58 58
59#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "cpufreq-nforce2", msg) 59#define PFX "cpufreq-nforce2: "
60#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
61 "cpufreq-nforce2", msg)
60 62
61/** 63/**
62 * nforce2_calc_fsb - calculate FSB 64 * nforce2_calc_fsb - calculate FSB
@@ -118,11 +120,11 @@ static void nforce2_write_pll(int pll)
118 int temp; 120 int temp;
119 121
120 /* Set the pll addr. to 0x00 */ 122 /* Set the pll addr. to 0x00 */
121 pci_write_config_dword(nforce2_chipset_dev, NFORCE2_PLLADR, 0); 123 pci_write_config_dword(nforce2_dev, NFORCE2_PLLADR, 0);
122 124
123 /* Now write the value in all 64 registers */ 125 /* Now write the value in all 64 registers */
124 for (temp = 0; temp <= 0x3f; temp++) 126 for (temp = 0; temp <= 0x3f; temp++)
125 pci_write_config_dword(nforce2_chipset_dev, NFORCE2_PLLREG, pll); 127 pci_write_config_dword(nforce2_dev, NFORCE2_PLLREG, pll);
126 128
127 return; 129 return;
128} 130}
@@ -139,8 +141,8 @@ static unsigned int nforce2_fsb_read(int bootfsb)
139 u32 fsb, temp = 0; 141 u32 fsb, temp = 0;
140 142
141 /* Get chipset boot FSB from subdevice 5 (FSB at boot-time) */ 143 /* Get chipset boot FSB from subdevice 5 (FSB at boot-time) */
142 nforce2_sub5 = pci_get_subsys(PCI_VENDOR_ID_NVIDIA, 144 nforce2_sub5 = pci_get_subsys(PCI_VENDOR_ID_NVIDIA, 0x01EF,
143 0x01EF, PCI_ANY_ID, PCI_ANY_ID, NULL); 145 PCI_ANY_ID, PCI_ANY_ID, NULL);
144 if (!nforce2_sub5) 146 if (!nforce2_sub5)
145 return 0; 147 return 0;
146 148
@@ -148,13 +150,13 @@ static unsigned int nforce2_fsb_read(int bootfsb)
148 fsb /= 1000000; 150 fsb /= 1000000;
149 151
150 /* Check if PLL register is already set */ 152 /* Check if PLL register is already set */
151 pci_read_config_byte(nforce2_chipset_dev, NFORCE2_PLLENABLE, (u8 *)&temp); 153 pci_read_config_byte(nforce2_dev, NFORCE2_PLLENABLE, (u8 *)&temp);
152 154
153 if (bootfsb || !temp) 155 if (bootfsb || !temp)
154 return fsb; 156 return fsb;
155 157
156 /* Use PLL register FSB value */ 158 /* Use PLL register FSB value */
157 pci_read_config_dword(nforce2_chipset_dev, NFORCE2_PLLREG, &temp); 159 pci_read_config_dword(nforce2_dev, NFORCE2_PLLREG, &temp);
158 fsb = nforce2_calc_fsb(temp); 160 fsb = nforce2_calc_fsb(temp);
159 161
160 return fsb; 162 return fsb;
@@ -174,18 +176,18 @@ static int nforce2_set_fsb(unsigned int fsb)
174 int pll = 0; 176 int pll = 0;
175 177
176 if ((fsb > max_fsb) || (fsb < NFORCE2_MIN_FSB)) { 178 if ((fsb > max_fsb) || (fsb < NFORCE2_MIN_FSB)) {
177 printk(KERN_ERR "cpufreq: FSB %d is out of range!\n", fsb); 179 printk(KERN_ERR PFX "FSB %d is out of range!\n", fsb);
178 return -EINVAL; 180 return -EINVAL;
179 } 181 }
180 182
181 tfsb = nforce2_fsb_read(0); 183 tfsb = nforce2_fsb_read(0);
182 if (!tfsb) { 184 if (!tfsb) {
183 printk(KERN_ERR "cpufreq: Error while reading the FSB\n"); 185 printk(KERN_ERR PFX "Error while reading the FSB\n");
184 return -EINVAL; 186 return -EINVAL;
185 } 187 }
186 188
187 /* First write? Then set actual value */ 189 /* First write? Then set actual value */
188 pci_read_config_byte(nforce2_chipset_dev, NFORCE2_PLLENABLE, (u8 *)&temp); 190 pci_read_config_byte(nforce2_dev, NFORCE2_PLLENABLE, (u8 *)&temp);
189 if (!temp) { 191 if (!temp) {
190 pll = nforce2_calc_pll(tfsb); 192 pll = nforce2_calc_pll(tfsb);
191 193
@@ -197,7 +199,7 @@ static int nforce2_set_fsb(unsigned int fsb)
197 199
198 /* Enable write access */ 200 /* Enable write access */
199 temp = 0x01; 201 temp = 0x01;
200 pci_write_config_byte(nforce2_chipset_dev, NFORCE2_PLLENABLE, (u8)temp); 202 pci_write_config_byte(nforce2_dev, NFORCE2_PLLENABLE, (u8)temp);
201 203
202 diff = tfsb - fsb; 204 diff = tfsb - fsb;
203 205
@@ -222,7 +224,7 @@ static int nforce2_set_fsb(unsigned int fsb)
222 } 224 }
223 225
224 temp = 0x40; 226 temp = 0x40;
225 pci_write_config_byte(nforce2_chipset_dev, NFORCE2_PLLADR, (u8)temp); 227 pci_write_config_byte(nforce2_dev, NFORCE2_PLLADR, (u8)temp);
226 228
227 return 0; 229 return 0;
228} 230}
@@ -244,7 +246,8 @@ static unsigned int nforce2_get(unsigned int cpu)
244 * nforce2_target - set a new CPUFreq policy 246 * nforce2_target - set a new CPUFreq policy
245 * @policy: new policy 247 * @policy: new policy
246 * @target_freq: the target frequency 248 * @target_freq: the target frequency
247 * @relation: how that frequency relates to achieved frequency (CPUFREQ_RELATION_L or CPUFREQ_RELATION_H) 249 * @relation: how that frequency relates to achieved frequency
250 * (CPUFREQ_RELATION_L or CPUFREQ_RELATION_H)
248 * 251 *
249 * Sets a new CPUFreq policy. 252 * Sets a new CPUFreq policy.
250 */ 253 */
@@ -276,7 +279,7 @@ static int nforce2_target(struct cpufreq_policy *policy,
276 /* local_irq_save(flags); */ 279 /* local_irq_save(flags); */
277 280
278 if (nforce2_set_fsb(target_fsb) < 0) 281 if (nforce2_set_fsb(target_fsb) < 0)
279 printk(KERN_ERR "cpufreq: Changing FSB to %d failed\n", 282 printk(KERN_ERR PFX "Changing FSB to %d failed\n",
280 target_fsb); 283 target_fsb);
281 else 284 else
282 dprintk("Changed FSB successfully to %d\n", 285 dprintk("Changed FSB successfully to %d\n",
@@ -327,8 +330,8 @@ static int nforce2_cpu_init(struct cpufreq_policy *policy)
327 /* FIX: Get FID from CPU */ 330 /* FIX: Get FID from CPU */
328 if (!fid) { 331 if (!fid) {
329 if (!cpu_khz) { 332 if (!cpu_khz) {
330 printk(KERN_WARNING 333 printk(KERN_WARNING PFX
331 "cpufreq: cpu_khz not set, can't calculate multiplier!\n"); 334 "cpu_khz not set, can't calculate multiplier!\n");
332 return -ENODEV; 335 return -ENODEV;
333 } 336 }
334 337
@@ -343,7 +346,7 @@ static int nforce2_cpu_init(struct cpufreq_policy *policy)
343 } 346 }
344 } 347 }
345 348
346 printk(KERN_INFO "cpufreq: FSB currently at %i MHz, FID %d.%d\n", fsb, 349 printk(KERN_INFO PFX "FSB currently at %i MHz, FID %d.%d\n", fsb,
347 fid / 10, fid % 10); 350 fid / 10, fid % 10);
348 351
349 /* Set maximum FSB to FSB at boot time */ 352 /* Set maximum FSB to FSB at boot time */
@@ -392,17 +395,18 @@ static struct cpufreq_driver nforce2_driver = {
392 */ 395 */
393static unsigned int nforce2_detect_chipset(void) 396static unsigned int nforce2_detect_chipset(void)
394{ 397{
395 nforce2_chipset_dev = pci_get_subsys(PCI_VENDOR_ID_NVIDIA, 398 nforce2_dev = pci_get_subsys(PCI_VENDOR_ID_NVIDIA,
396 PCI_DEVICE_ID_NVIDIA_NFORCE2, 399 PCI_DEVICE_ID_NVIDIA_NFORCE2,
397 PCI_ANY_ID, PCI_ANY_ID, NULL); 400 PCI_ANY_ID, PCI_ANY_ID, NULL);
398 401
399 if (nforce2_chipset_dev == NULL) 402 if (nforce2_dev == NULL)
400 return -ENODEV; 403 return -ENODEV;
401 404
402 printk(KERN_INFO "cpufreq: Detected nForce2 chipset revision %X\n", 405 printk(KERN_INFO PFX "Detected nForce2 chipset revision %X\n",
403 nforce2_chipset_dev->revision); 406 nforce2_dev->revision);
404 printk(KERN_INFO 407 printk(KERN_INFO PFX
405 "cpufreq: FSB changing is maybe unstable and can lead to crashes and data loss.\n"); 408 "FSB changing is maybe unstable and can lead to "
409 "crashes and data loss.\n");
406 410
407 return 0; 411 return 0;
408} 412}
@@ -420,7 +424,7 @@ static int __init nforce2_init(void)
420 424
421 /* detect chipset */ 425 /* detect chipset */
422 if (nforce2_detect_chipset()) { 426 if (nforce2_detect_chipset()) {
423 printk(KERN_ERR "cpufreq: No nForce2 chipset.\n"); 427 printk(KERN_INFO PFX "No nForce2 chipset.\n");
424 return -ENODEV; 428 return -ENODEV;
425 } 429 }
426 430
diff --git a/arch/x86/kernel/cpu/cpufreq/e_powersaver.c b/arch/x86/kernel/cpu/cpufreq/e_powersaver.c
index 41ab3f064cb1..35a257dd4bb7 100644
--- a/arch/x86/kernel/cpu/cpufreq/e_powersaver.c
+++ b/arch/x86/kernel/cpu/cpufreq/e_powersaver.c
@@ -12,12 +12,12 @@
12#include <linux/cpufreq.h> 12#include <linux/cpufreq.h>
13#include <linux/ioport.h> 13#include <linux/ioport.h>
14#include <linux/slab.h> 14#include <linux/slab.h>
15#include <linux/timex.h>
16#include <linux/io.h>
17#include <linux/delay.h>
15 18
16#include <asm/msr.h> 19#include <asm/msr.h>
17#include <asm/tsc.h> 20#include <asm/tsc.h>
18#include <asm/timex.h>
19#include <asm/io.h>
20#include <asm/delay.h>
21 21
22#define EPS_BRAND_C7M 0 22#define EPS_BRAND_C7M 0
23#define EPS_BRAND_C7 1 23#define EPS_BRAND_C7 1
@@ -184,7 +184,7 @@ static int eps_cpu_init(struct cpufreq_policy *policy)
184 break; 184 break;
185 } 185 }
186 186
187 switch(brand) { 187 switch (brand) {
188 case EPS_BRAND_C7M: 188 case EPS_BRAND_C7M:
189 printk(KERN_CONT "C7-M\n"); 189 printk(KERN_CONT "C7-M\n");
190 break; 190 break;
@@ -218,17 +218,20 @@ static int eps_cpu_init(struct cpufreq_policy *policy)
218 /* Print voltage and multiplier */ 218 /* Print voltage and multiplier */
219 rdmsr(MSR_IA32_PERF_STATUS, lo, hi); 219 rdmsr(MSR_IA32_PERF_STATUS, lo, hi);
220 current_voltage = lo & 0xff; 220 current_voltage = lo & 0xff;
221 printk(KERN_INFO "eps: Current voltage = %dmV\n", current_voltage * 16 + 700); 221 printk(KERN_INFO "eps: Current voltage = %dmV\n",
222 current_voltage * 16 + 700);
222 current_multiplier = (lo >> 8) & 0xff; 223 current_multiplier = (lo >> 8) & 0xff;
223 printk(KERN_INFO "eps: Current multiplier = %d\n", current_multiplier); 224 printk(KERN_INFO "eps: Current multiplier = %d\n", current_multiplier);
224 225
225 /* Print limits */ 226 /* Print limits */
226 max_voltage = hi & 0xff; 227 max_voltage = hi & 0xff;
227 printk(KERN_INFO "eps: Highest voltage = %dmV\n", max_voltage * 16 + 700); 228 printk(KERN_INFO "eps: Highest voltage = %dmV\n",
229 max_voltage * 16 + 700);
228 max_multiplier = (hi >> 8) & 0xff; 230 max_multiplier = (hi >> 8) & 0xff;
229 printk(KERN_INFO "eps: Highest multiplier = %d\n", max_multiplier); 231 printk(KERN_INFO "eps: Highest multiplier = %d\n", max_multiplier);
230 min_voltage = (hi >> 16) & 0xff; 232 min_voltage = (hi >> 16) & 0xff;
231 printk(KERN_INFO "eps: Lowest voltage = %dmV\n", min_voltage * 16 + 700); 233 printk(KERN_INFO "eps: Lowest voltage = %dmV\n",
234 min_voltage * 16 + 700);
232 min_multiplier = (hi >> 24) & 0xff; 235 min_multiplier = (hi >> 24) & 0xff;
233 printk(KERN_INFO "eps: Lowest multiplier = %d\n", min_multiplier); 236 printk(KERN_INFO "eps: Lowest multiplier = %d\n", min_multiplier);
234 237
@@ -318,7 +321,7 @@ static int eps_cpu_exit(struct cpufreq_policy *policy)
318 return 0; 321 return 0;
319} 322}
320 323
321static struct freq_attr* eps_attr[] = { 324static struct freq_attr *eps_attr[] = {
322 &cpufreq_freq_attr_scaling_available_freqs, 325 &cpufreq_freq_attr_scaling_available_freqs,
323 NULL, 326 NULL,
324}; 327};
@@ -356,7 +359,7 @@ static void __exit eps_exit(void)
356 cpufreq_unregister_driver(&eps_driver); 359 cpufreq_unregister_driver(&eps_driver);
357} 360}
358 361
359MODULE_AUTHOR("Rafa³ Bilski <rafalbilski@interia.pl>"); 362MODULE_AUTHOR("Rafal Bilski <rafalbilski@interia.pl>");
360MODULE_DESCRIPTION("Enhanced PowerSaver driver for VIA C7 CPU's."); 363MODULE_DESCRIPTION("Enhanced PowerSaver driver for VIA C7 CPU's.");
361MODULE_LICENSE("GPL"); 364MODULE_LICENSE("GPL");
362 365
diff --git a/arch/x86/kernel/cpu/cpufreq/elanfreq.c b/arch/x86/kernel/cpu/cpufreq/elanfreq.c
index fe613c93b366..006b278b0d5d 100644
--- a/arch/x86/kernel/cpu/cpufreq/elanfreq.c
+++ b/arch/x86/kernel/cpu/cpufreq/elanfreq.c
@@ -184,7 +184,8 @@ static int elanfreq_target(struct cpufreq_policy *policy,
184{ 184{
185 unsigned int newstate = 0; 185 unsigned int newstate = 0;
186 186
187 if (cpufreq_frequency_table_target(policy, &elanfreq_table[0], target_freq, relation, &newstate)) 187 if (cpufreq_frequency_table_target(policy, &elanfreq_table[0],
188 target_freq, relation, &newstate))
188 return -EINVAL; 189 return -EINVAL;
189 190
190 elanfreq_set_cpu_state(newstate); 191 elanfreq_set_cpu_state(newstate);
@@ -301,7 +302,8 @@ static void __exit elanfreq_exit(void)
301module_param(max_freq, int, 0444); 302module_param(max_freq, int, 0444);
302 303
303MODULE_LICENSE("GPL"); 304MODULE_LICENSE("GPL");
304MODULE_AUTHOR("Robert Schwebel <r.schwebel@pengutronix.de>, Sven Geggus <sven@geggus.net>"); 305MODULE_AUTHOR("Robert Schwebel <r.schwebel@pengutronix.de>, "
306 "Sven Geggus <sven@geggus.net>");
305MODULE_DESCRIPTION("cpufreq driver for AMD's Elan CPUs"); 307MODULE_DESCRIPTION("cpufreq driver for AMD's Elan CPUs");
306 308
307module_init(elanfreq_init); 309module_init(elanfreq_init);
diff --git a/arch/x86/kernel/cpu/cpufreq/gx-suspmod.c b/arch/x86/kernel/cpu/cpufreq/gx-suspmod.c
index 9d9eae82e60f..ac27ec2264d5 100644
--- a/arch/x86/kernel/cpu/cpufreq/gx-suspmod.c
+++ b/arch/x86/kernel/cpu/cpufreq/gx-suspmod.c
@@ -79,8 +79,9 @@
79#include <linux/smp.h> 79#include <linux/smp.h>
80#include <linux/cpufreq.h> 80#include <linux/cpufreq.h>
81#include <linux/pci.h> 81#include <linux/pci.h>
82#include <linux/errno.h>
83
82#include <asm/processor-cyrix.h> 84#include <asm/processor-cyrix.h>
83#include <asm/errno.h>
84 85
85/* PCI config registers, all at F0 */ 86/* PCI config registers, all at F0 */
86#define PCI_PMER1 0x80 /* power management enable register 1 */ 87#define PCI_PMER1 0x80 /* power management enable register 1 */
@@ -122,8 +123,8 @@ static struct gxfreq_params *gx_params;
122static int stock_freq; 123static int stock_freq;
123 124
124/* PCI bus clock - defaults to 30.000 if cpu_khz is not available */ 125/* PCI bus clock - defaults to 30.000 if cpu_khz is not available */
125static int pci_busclk = 0; 126static int pci_busclk;
126module_param (pci_busclk, int, 0444); 127module_param(pci_busclk, int, 0444);
127 128
128/* maximum duration for which the cpu may be suspended 129/* maximum duration for which the cpu may be suspended
129 * (32us * MAX_DURATION). If no parameter is given, this defaults 130 * (32us * MAX_DURATION). If no parameter is given, this defaults
@@ -132,7 +133,7 @@ module_param (pci_busclk, int, 0444);
132 * is suspended -- processing power is just 0.39% of what it used to be, 133 * is suspended -- processing power is just 0.39% of what it used to be,
133 * though. 781.25 kHz(!) for a 200 MHz processor -- wow. */ 134 * though. 781.25 kHz(!) for a 200 MHz processor -- wow. */
134static int max_duration = 255; 135static int max_duration = 255;
135module_param (max_duration, int, 0444); 136module_param(max_duration, int, 0444);
136 137
137/* For the default policy, we want at least some processing power 138/* For the default policy, we want at least some processing power
138 * - let's say 5%. (min = maxfreq / POLICY_MIN_DIV) 139 * - let's say 5%. (min = maxfreq / POLICY_MIN_DIV)
@@ -140,7 +141,8 @@ module_param (max_duration, int, 0444);
140#define POLICY_MIN_DIV 20 141#define POLICY_MIN_DIV 20
141 142
142 143
143#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "gx-suspmod", msg) 144#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
145 "gx-suspmod", msg)
144 146
145/** 147/**
146 * we can detect a core multipiler from dir0_lsb 148 * we can detect a core multipiler from dir0_lsb
@@ -166,12 +168,20 @@ static int gx_freq_mult[16] = {
166 * Low Level chipset interface * 168 * Low Level chipset interface *
167 ****************************************************************/ 169 ****************************************************************/
168static struct pci_device_id gx_chipset_tbl[] __initdata = { 170static struct pci_device_id gx_chipset_tbl[] __initdata = {
169 { PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5530_LEGACY, PCI_ANY_ID, PCI_ANY_ID }, 171 { PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5530_LEGACY,
170 { PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5520, PCI_ANY_ID, PCI_ANY_ID }, 172 PCI_ANY_ID, PCI_ANY_ID },
171 { PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5510, PCI_ANY_ID, PCI_ANY_ID }, 173 { PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5520,
174 PCI_ANY_ID, PCI_ANY_ID },
175 { PCI_VENDOR_ID_CYRIX, PCI_DEVICE_ID_CYRIX_5510,
176 PCI_ANY_ID, PCI_ANY_ID },
172 { 0, }, 177 { 0, },
173}; 178};
174 179
180static void gx_write_byte(int reg, int value)
181{
182 pci_write_config_byte(gx_params->cs55x0, reg, value);
183}
184
175/** 185/**
176 * gx_detect_chipset: 186 * gx_detect_chipset:
177 * 187 *
@@ -200,7 +210,8 @@ static __init struct pci_dev *gx_detect_chipset(void)
200/** 210/**
201 * gx_get_cpuspeed: 211 * gx_get_cpuspeed:
202 * 212 *
203 * Finds out at which efficient frequency the Cyrix MediaGX/NatSemi Geode CPU runs. 213 * Finds out at which efficient frequency the Cyrix MediaGX/NatSemi
214 * Geode CPU runs.
204 */ 215 */
205static unsigned int gx_get_cpuspeed(unsigned int cpu) 216static unsigned int gx_get_cpuspeed(unsigned int cpu)
206{ 217{
@@ -217,17 +228,18 @@ static unsigned int gx_get_cpuspeed(unsigned int cpu)
217 * 228 *
218 **/ 229 **/
219 230
220static unsigned int gx_validate_speed(unsigned int khz, u8 *on_duration, u8 *off_duration) 231static unsigned int gx_validate_speed(unsigned int khz, u8 *on_duration,
232 u8 *off_duration)
221{ 233{
222 unsigned int i; 234 unsigned int i;
223 u8 tmp_on, tmp_off; 235 u8 tmp_on, tmp_off;
224 int old_tmp_freq = stock_freq; 236 int old_tmp_freq = stock_freq;
225 int tmp_freq; 237 int tmp_freq;
226 238
227 *off_duration=1; 239 *off_duration = 1;
228 *on_duration=0; 240 *on_duration = 0;
229 241
230 for (i=max_duration; i>0; i--) { 242 for (i = max_duration; i > 0; i--) {
231 tmp_off = ((khz * i) / stock_freq) & 0xff; 243 tmp_off = ((khz * i) / stock_freq) & 0xff;
232 tmp_on = i - tmp_off; 244 tmp_on = i - tmp_off;
233 tmp_freq = (stock_freq * tmp_off) / i; 245 tmp_freq = (stock_freq * tmp_off) / i;
@@ -259,26 +271,34 @@ static void gx_set_cpuspeed(unsigned int khz)
259 freqs.cpu = 0; 271 freqs.cpu = 0;
260 freqs.old = gx_get_cpuspeed(0); 272 freqs.old = gx_get_cpuspeed(0);
261 273
262 new_khz = gx_validate_speed(khz, &gx_params->on_duration, &gx_params->off_duration); 274 new_khz = gx_validate_speed(khz, &gx_params->on_duration,
275 &gx_params->off_duration);
263 276
264 freqs.new = new_khz; 277 freqs.new = new_khz;
265 278
266 cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); 279 cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
267 local_irq_save(flags); 280 local_irq_save(flags);
268 281
269 if (new_khz != stock_freq) { /* if new khz == 100% of CPU speed, it is special case */ 282
283
284 if (new_khz != stock_freq) {
285 /* if new khz == 100% of CPU speed, it is special case */
270 switch (gx_params->cs55x0->device) { 286 switch (gx_params->cs55x0->device) {
271 case PCI_DEVICE_ID_CYRIX_5530_LEGACY: 287 case PCI_DEVICE_ID_CYRIX_5530_LEGACY:
272 pmer1 = gx_params->pci_pmer1 | IRQ_SPDUP | VID_SPDUP; 288 pmer1 = gx_params->pci_pmer1 | IRQ_SPDUP | VID_SPDUP;
273 /* FIXME: need to test other values -- Zwane,Miura */ 289 /* FIXME: need to test other values -- Zwane,Miura */
274 pci_write_config_byte(gx_params->cs55x0, PCI_IRQTC, 4); /* typical 2 to 4ms */ 290 /* typical 2 to 4ms */
275 pci_write_config_byte(gx_params->cs55x0, PCI_VIDTC, 100);/* typical 50 to 100ms */ 291 gx_write_byte(PCI_IRQTC, 4);
276 pci_write_config_byte(gx_params->cs55x0, PCI_PMER1, pmer1); 292 /* typical 50 to 100ms */
277 293 gx_write_byte(PCI_VIDTC, 100);
278 if (gx_params->cs55x0->revision < 0x10) { /* CS5530(rev 1.2, 1.3) */ 294 gx_write_byte(PCI_PMER1, pmer1);
279 suscfg = gx_params->pci_suscfg | SUSMOD; 295
280 } else { /* CS5530A,B.. */ 296 if (gx_params->cs55x0->revision < 0x10) {
281 suscfg = gx_params->pci_suscfg | SUSMOD | PWRSVE; 297 /* CS5530(rev 1.2, 1.3) */
298 suscfg = gx_params->pci_suscfg|SUSMOD;
299 } else {
300 /* CS5530A,B.. */
301 suscfg = gx_params->pci_suscfg|SUSMOD|PWRSVE;
282 } 302 }
283 break; 303 break;
284 case PCI_DEVICE_ID_CYRIX_5520: 304 case PCI_DEVICE_ID_CYRIX_5520:
@@ -294,13 +314,13 @@ static void gx_set_cpuspeed(unsigned int khz)
294 suscfg = gx_params->pci_suscfg & ~(SUSMOD); 314 suscfg = gx_params->pci_suscfg & ~(SUSMOD);
295 gx_params->off_duration = 0; 315 gx_params->off_duration = 0;
296 gx_params->on_duration = 0; 316 gx_params->on_duration = 0;
297 dprintk("suspend modulation disabled: cpu runs 100 percent speed.\n"); 317 dprintk("suspend modulation disabled: cpu runs 100%% speed.\n");
298 } 318 }
299 319
300 pci_write_config_byte(gx_params->cs55x0, PCI_MODOFF, gx_params->off_duration); 320 gx_write_byte(PCI_MODOFF, gx_params->off_duration);
301 pci_write_config_byte(gx_params->cs55x0, PCI_MODON, gx_params->on_duration); 321 gx_write_byte(PCI_MODON, gx_params->on_duration);
302 322
303 pci_write_config_byte(gx_params->cs55x0, PCI_SUSCFG, suscfg); 323 gx_write_byte(PCI_SUSCFG, suscfg);
304 pci_read_config_byte(gx_params->cs55x0, PCI_SUSCFG, &suscfg); 324 pci_read_config_byte(gx_params->cs55x0, PCI_SUSCFG, &suscfg);
305 325
306 local_irq_restore(flags); 326 local_irq_restore(flags);
@@ -334,7 +354,8 @@ static int cpufreq_gx_verify(struct cpufreq_policy *policy)
334 return -EINVAL; 354 return -EINVAL;
335 355
336 policy->cpu = 0; 356 policy->cpu = 0;
337 cpufreq_verify_within_limits(policy, (stock_freq / max_duration), stock_freq); 357 cpufreq_verify_within_limits(policy, (stock_freq / max_duration),
358 stock_freq);
338 359
339 /* it needs to be assured that at least one supported frequency is 360 /* it needs to be assured that at least one supported frequency is
340 * within policy->min and policy->max. If it is not, policy->max 361 * within policy->min and policy->max. If it is not, policy->max
@@ -354,7 +375,8 @@ static int cpufreq_gx_verify(struct cpufreq_policy *policy)
354 policy->max = tmp_freq; 375 policy->max = tmp_freq;
355 if (policy->max < policy->min) 376 if (policy->max < policy->min)
356 policy->max = policy->min; 377 policy->max = policy->min;
357 cpufreq_verify_within_limits(policy, (stock_freq / max_duration), stock_freq); 378 cpufreq_verify_within_limits(policy, (stock_freq / max_duration),
379 stock_freq);
358 380
359 return 0; 381 return 0;
360} 382}
@@ -398,18 +420,18 @@ static int cpufreq_gx_cpu_init(struct cpufreq_policy *policy)
398 return -ENODEV; 420 return -ENODEV;
399 421
400 /* determine maximum frequency */ 422 /* determine maximum frequency */
401 if (pci_busclk) { 423 if (pci_busclk)
402 maxfreq = pci_busclk * gx_freq_mult[getCx86(CX86_DIR1) & 0x0f]; 424 maxfreq = pci_busclk * gx_freq_mult[getCx86(CX86_DIR1) & 0x0f];
403 } else if (cpu_khz) { 425 else if (cpu_khz)
404 maxfreq = cpu_khz; 426 maxfreq = cpu_khz;
405 } else { 427 else
406 maxfreq = 30000 * gx_freq_mult[getCx86(CX86_DIR1) & 0x0f]; 428 maxfreq = 30000 * gx_freq_mult[getCx86(CX86_DIR1) & 0x0f];
407 } 429
408 stock_freq = maxfreq; 430 stock_freq = maxfreq;
409 curfreq = gx_get_cpuspeed(0); 431 curfreq = gx_get_cpuspeed(0);
410 432
411 dprintk("cpu max frequency is %d.\n", maxfreq); 433 dprintk("cpu max frequency is %d.\n", maxfreq);
412 dprintk("cpu current frequency is %dkHz.\n",curfreq); 434 dprintk("cpu current frequency is %dkHz.\n", curfreq);
413 435
414 /* setup basic struct for cpufreq API */ 436 /* setup basic struct for cpufreq API */
415 policy->cpu = 0; 437 policy->cpu = 0;
@@ -447,7 +469,8 @@ static int __init cpufreq_gx_init(void)
447 struct pci_dev *gx_pci; 469 struct pci_dev *gx_pci;
448 470
449 /* Test if we have the right hardware */ 471 /* Test if we have the right hardware */
450 if ((gx_pci = gx_detect_chipset()) == NULL) 472 gx_pci = gx_detect_chipset();
473 if (gx_pci == NULL)
451 return -ENODEV; 474 return -ENODEV;
452 475
453 /* check whether module parameters are sane */ 476 /* check whether module parameters are sane */
@@ -468,9 +491,11 @@ static int __init cpufreq_gx_init(void)
468 pci_read_config_byte(params->cs55x0, PCI_PMER1, &(params->pci_pmer1)); 491 pci_read_config_byte(params->cs55x0, PCI_PMER1, &(params->pci_pmer1));
469 pci_read_config_byte(params->cs55x0, PCI_PMER2, &(params->pci_pmer2)); 492 pci_read_config_byte(params->cs55x0, PCI_PMER2, &(params->pci_pmer2));
470 pci_read_config_byte(params->cs55x0, PCI_MODON, &(params->on_duration)); 493 pci_read_config_byte(params->cs55x0, PCI_MODON, &(params->on_duration));
471 pci_read_config_byte(params->cs55x0, PCI_MODOFF, &(params->off_duration)); 494 pci_read_config_byte(params->cs55x0, PCI_MODOFF,
495 &(params->off_duration));
472 496
473 if ((ret = cpufreq_register_driver(&gx_suspmod_driver))) { 497 ret = cpufreq_register_driver(&gx_suspmod_driver);
498 if (ret) {
474 kfree(params); 499 kfree(params);
475 return ret; /* register error! */ 500 return ret; /* register error! */
476 } 501 }
@@ -485,9 +510,9 @@ static void __exit cpufreq_gx_exit(void)
485 kfree(gx_params); 510 kfree(gx_params);
486} 511}
487 512
488MODULE_AUTHOR ("Hiroshi Miura <miura@da-cha.org>"); 513MODULE_AUTHOR("Hiroshi Miura <miura@da-cha.org>");
489MODULE_DESCRIPTION ("Cpufreq driver for Cyrix MediaGX and NatSemi Geode"); 514MODULE_DESCRIPTION("Cpufreq driver for Cyrix MediaGX and NatSemi Geode");
490MODULE_LICENSE ("GPL"); 515MODULE_LICENSE("GPL");
491 516
492module_init(cpufreq_gx_init); 517module_init(cpufreq_gx_init);
493module_exit(cpufreq_gx_exit); 518module_exit(cpufreq_gx_exit);
diff --git a/arch/x86/kernel/cpu/cpufreq/longhaul.c b/arch/x86/kernel/cpu/cpufreq/longhaul.c
index a4cff5d6e380..ce2ed3e4aad9 100644
--- a/arch/x86/kernel/cpu/cpufreq/longhaul.c
+++ b/arch/x86/kernel/cpu/cpufreq/longhaul.c
@@ -30,12 +30,11 @@
30#include <linux/slab.h> 30#include <linux/slab.h>
31#include <linux/string.h> 31#include <linux/string.h>
32#include <linux/delay.h> 32#include <linux/delay.h>
33#include <linux/timex.h>
34#include <linux/io.h>
35#include <linux/acpi.h>
33 36
34#include <asm/msr.h> 37#include <asm/msr.h>
35#include <asm/timex.h>
36#include <asm/io.h>
37#include <asm/acpi.h>
38#include <linux/acpi.h>
39#include <acpi/processor.h> 38#include <acpi/processor.h>
40 39
41#include "longhaul.h" 40#include "longhaul.h"
@@ -58,7 +57,7 @@
58#define USE_NORTHBRIDGE (1 << 2) 57#define USE_NORTHBRIDGE (1 << 2)
59 58
60static int cpu_model; 59static int cpu_model;
61static unsigned int numscales=16; 60static unsigned int numscales = 16;
62static unsigned int fsb; 61static unsigned int fsb;
63 62
64static const struct mV_pos *vrm_mV_table; 63static const struct mV_pos *vrm_mV_table;
@@ -67,8 +66,8 @@ static const unsigned char *mV_vrm_table;
67static unsigned int highest_speed, lowest_speed; /* kHz */ 66static unsigned int highest_speed, lowest_speed; /* kHz */
68static unsigned int minmult, maxmult; 67static unsigned int minmult, maxmult;
69static int can_scale_voltage; 68static int can_scale_voltage;
70static struct acpi_processor *pr = NULL; 69static struct acpi_processor *pr;
71static struct acpi_processor_cx *cx = NULL; 70static struct acpi_processor_cx *cx;
72static u32 acpi_regs_addr; 71static u32 acpi_regs_addr;
73static u8 longhaul_flags; 72static u8 longhaul_flags;
74static unsigned int longhaul_index; 73static unsigned int longhaul_index;
@@ -78,12 +77,13 @@ static int scale_voltage;
78static int disable_acpi_c3; 77static int disable_acpi_c3;
79static int revid_errata; 78static int revid_errata;
80 79
81#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "longhaul", msg) 80#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
81 "longhaul", msg)
82 82
83 83
84/* Clock ratios multiplied by 10 */ 84/* Clock ratios multiplied by 10 */
85static int clock_ratio[32]; 85static int mults[32];
86static int eblcr_table[32]; 86static int eblcr[32];
87static int longhaul_version; 87static int longhaul_version;
88static struct cpufreq_frequency_table *longhaul_table; 88static struct cpufreq_frequency_table *longhaul_table;
89 89
@@ -93,7 +93,7 @@ static char speedbuffer[8];
93static char *print_speed(int speed) 93static char *print_speed(int speed)
94{ 94{
95 if (speed < 1000) { 95 if (speed < 1000) {
96 snprintf(speedbuffer, sizeof(speedbuffer),"%dMHz", speed); 96 snprintf(speedbuffer, sizeof(speedbuffer), "%dMHz", speed);
97 return speedbuffer; 97 return speedbuffer;
98 } 98 }
99 99
@@ -122,27 +122,28 @@ static unsigned int calc_speed(int mult)
122 122
123static int longhaul_get_cpu_mult(void) 123static int longhaul_get_cpu_mult(void)
124{ 124{
125 unsigned long invalue=0,lo, hi; 125 unsigned long invalue = 0, lo, hi;
126 126
127 rdmsr (MSR_IA32_EBL_CR_POWERON, lo, hi); 127 rdmsr(MSR_IA32_EBL_CR_POWERON, lo, hi);
128 invalue = (lo & (1<<22|1<<23|1<<24|1<<25)) >>22; 128 invalue = (lo & (1<<22|1<<23|1<<24|1<<25))>>22;
129 if (longhaul_version==TYPE_LONGHAUL_V2 || longhaul_version==TYPE_POWERSAVER) { 129 if (longhaul_version == TYPE_LONGHAUL_V2 ||
130 longhaul_version == TYPE_POWERSAVER) {
130 if (lo & (1<<27)) 131 if (lo & (1<<27))
131 invalue+=16; 132 invalue += 16;
132 } 133 }
133 return eblcr_table[invalue]; 134 return eblcr[invalue];
134} 135}
135 136
136/* For processor with BCR2 MSR */ 137/* For processor with BCR2 MSR */
137 138
138static void do_longhaul1(unsigned int clock_ratio_index) 139static void do_longhaul1(unsigned int mults_index)
139{ 140{
140 union msr_bcr2 bcr2; 141 union msr_bcr2 bcr2;
141 142
142 rdmsrl(MSR_VIA_BCR2, bcr2.val); 143 rdmsrl(MSR_VIA_BCR2, bcr2.val);
143 /* Enable software clock multiplier */ 144 /* Enable software clock multiplier */
144 bcr2.bits.ESOFTBF = 1; 145 bcr2.bits.ESOFTBF = 1;
145 bcr2.bits.CLOCKMUL = clock_ratio_index & 0xff; 146 bcr2.bits.CLOCKMUL = mults_index & 0xff;
146 147
147 /* Sync to timer tick */ 148 /* Sync to timer tick */
148 safe_halt(); 149 safe_halt();
@@ -161,7 +162,7 @@ static void do_longhaul1(unsigned int clock_ratio_index)
161 162
162/* For processor with Longhaul MSR */ 163/* For processor with Longhaul MSR */
163 164
164static void do_powersaver(int cx_address, unsigned int clock_ratio_index, 165static void do_powersaver(int cx_address, unsigned int mults_index,
165 unsigned int dir) 166 unsigned int dir)
166{ 167{
167 union msr_longhaul longhaul; 168 union msr_longhaul longhaul;
@@ -173,11 +174,11 @@ static void do_powersaver(int cx_address, unsigned int clock_ratio_index,
173 longhaul.bits.RevisionKey = longhaul.bits.RevisionID; 174 longhaul.bits.RevisionKey = longhaul.bits.RevisionID;
174 else 175 else
175 longhaul.bits.RevisionKey = 0; 176 longhaul.bits.RevisionKey = 0;
176 longhaul.bits.SoftBusRatio = clock_ratio_index & 0xf; 177 longhaul.bits.SoftBusRatio = mults_index & 0xf;
177 longhaul.bits.SoftBusRatio4 = (clock_ratio_index & 0x10) >> 4; 178 longhaul.bits.SoftBusRatio4 = (mults_index & 0x10) >> 4;
178 /* Setup new voltage */ 179 /* Setup new voltage */
179 if (can_scale_voltage) 180 if (can_scale_voltage)
180 longhaul.bits.SoftVID = (clock_ratio_index >> 8) & 0x1f; 181 longhaul.bits.SoftVID = (mults_index >> 8) & 0x1f;
181 /* Sync to timer tick */ 182 /* Sync to timer tick */
182 safe_halt(); 183 safe_halt();
183 /* Raise voltage if necessary */ 184 /* Raise voltage if necessary */
@@ -240,14 +241,14 @@ static void do_powersaver(int cx_address, unsigned int clock_ratio_index,
240 241
241/** 242/**
242 * longhaul_set_cpu_frequency() 243 * longhaul_set_cpu_frequency()
243 * @clock_ratio_index : bitpattern of the new multiplier. 244 * @mults_index : bitpattern of the new multiplier.
244 * 245 *
245 * Sets a new clock ratio. 246 * Sets a new clock ratio.
246 */ 247 */
247 248
248static void longhaul_setstate(unsigned int table_index) 249static void longhaul_setstate(unsigned int table_index)
249{ 250{
250 unsigned int clock_ratio_index; 251 unsigned int mults_index;
251 int speed, mult; 252 int speed, mult;
252 struct cpufreq_freqs freqs; 253 struct cpufreq_freqs freqs;
253 unsigned long flags; 254 unsigned long flags;
@@ -256,9 +257,9 @@ static void longhaul_setstate(unsigned int table_index)
256 u32 bm_timeout = 1000; 257 u32 bm_timeout = 1000;
257 unsigned int dir = 0; 258 unsigned int dir = 0;
258 259
259 clock_ratio_index = longhaul_table[table_index].index; 260 mults_index = longhaul_table[table_index].index;
260 /* Safety precautions */ 261 /* Safety precautions */
261 mult = clock_ratio[clock_ratio_index & 0x1f]; 262 mult = mults[mults_index & 0x1f];
262 if (mult == -1) 263 if (mult == -1)
263 return; 264 return;
264 speed = calc_speed(mult); 265 speed = calc_speed(mult);
@@ -274,7 +275,7 @@ static void longhaul_setstate(unsigned int table_index)
274 275
275 cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); 276 cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
276 277
277 dprintk ("Setting to FSB:%dMHz Mult:%d.%dx (%s)\n", 278 dprintk("Setting to FSB:%dMHz Mult:%d.%dx (%s)\n",
278 fsb, mult/10, mult%10, print_speed(speed/1000)); 279 fsb, mult/10, mult%10, print_speed(speed/1000));
279retry_loop: 280retry_loop:
280 preempt_disable(); 281 preempt_disable();
@@ -282,8 +283,8 @@ retry_loop:
282 283
283 pic2_mask = inb(0xA1); 284 pic2_mask = inb(0xA1);
284 pic1_mask = inb(0x21); /* works on C3. save mask. */ 285 pic1_mask = inb(0x21); /* works on C3. save mask. */
285 outb(0xFF,0xA1); /* Overkill */ 286 outb(0xFF, 0xA1); /* Overkill */
286 outb(0xFE,0x21); /* TMR0 only */ 287 outb(0xFE, 0x21); /* TMR0 only */
287 288
288 /* Wait while PCI bus is busy. */ 289 /* Wait while PCI bus is busy. */
289 if (acpi_regs_addr && (longhaul_flags & USE_NORTHBRIDGE 290 if (acpi_regs_addr && (longhaul_flags & USE_NORTHBRIDGE
@@ -303,7 +304,7 @@ retry_loop:
303 outb(3, 0x22); 304 outb(3, 0x22);
304 } else if ((pr != NULL) && pr->flags.bm_control) { 305 } else if ((pr != NULL) && pr->flags.bm_control) {
305 /* Disable bus master arbitration */ 306 /* Disable bus master arbitration */
306 acpi_set_register(ACPI_BITREG_ARB_DISABLE, 1); 307 acpi_write_bit_register(ACPI_BITREG_ARB_DISABLE, 1);
307 } 308 }
308 switch (longhaul_version) { 309 switch (longhaul_version) {
309 310
@@ -312,7 +313,7 @@ retry_loop:
312 * Software controlled multipliers only. 313 * Software controlled multipliers only.
313 */ 314 */
314 case TYPE_LONGHAUL_V1: 315 case TYPE_LONGHAUL_V1:
315 do_longhaul1(clock_ratio_index); 316 do_longhaul1(mults_index);
316 break; 317 break;
317 318
318 /* 319 /*
@@ -326,10 +327,10 @@ retry_loop:
326 case TYPE_POWERSAVER: 327 case TYPE_POWERSAVER:
327 if (longhaul_flags & USE_ACPI_C3) { 328 if (longhaul_flags & USE_ACPI_C3) {
328 /* Don't allow wakeup */ 329 /* Don't allow wakeup */
329 acpi_set_register(ACPI_BITREG_BUS_MASTER_RLD, 0); 330 acpi_write_bit_register(ACPI_BITREG_BUS_MASTER_RLD, 0);
330 do_powersaver(cx->address, clock_ratio_index, dir); 331 do_powersaver(cx->address, mults_index, dir);
331 } else { 332 } else {
332 do_powersaver(0, clock_ratio_index, dir); 333 do_powersaver(0, mults_index, dir);
333 } 334 }
334 break; 335 break;
335 } 336 }
@@ -339,10 +340,10 @@ retry_loop:
339 outb(0, 0x22); 340 outb(0, 0x22);
340 } else if ((pr != NULL) && pr->flags.bm_control) { 341 } else if ((pr != NULL) && pr->flags.bm_control) {
341 /* Enable bus master arbitration */ 342 /* Enable bus master arbitration */
342 acpi_set_register(ACPI_BITREG_ARB_DISABLE, 0); 343 acpi_write_bit_register(ACPI_BITREG_ARB_DISABLE, 0);
343 } 344 }
344 outb(pic2_mask,0xA1); /* restore mask */ 345 outb(pic2_mask, 0xA1); /* restore mask */
345 outb(pic1_mask,0x21); 346 outb(pic1_mask, 0x21);
346 347
347 local_irq_restore(flags); 348 local_irq_restore(flags);
348 preempt_enable(); 349 preempt_enable();
@@ -392,7 +393,8 @@ retry_loop:
392 cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); 393 cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
393 394
394 if (!bm_timeout) 395 if (!bm_timeout)
395 printk(KERN_INFO PFX "Warning: Timeout while waiting for idle PCI bus.\n"); 396 printk(KERN_INFO PFX "Warning: Timeout while waiting for "
397 "idle PCI bus.\n");
396} 398}
397 399
398/* 400/*
@@ -458,31 +460,32 @@ static int __init longhaul_get_ranges(void)
458 break; 460 break;
459 } 461 }
460 462
461 dprintk ("MinMult:%d.%dx MaxMult:%d.%dx\n", 463 dprintk("MinMult:%d.%dx MaxMult:%d.%dx\n",
462 minmult/10, minmult%10, maxmult/10, maxmult%10); 464 minmult/10, minmult%10, maxmult/10, maxmult%10);
463 465
464 highest_speed = calc_speed(maxmult); 466 highest_speed = calc_speed(maxmult);
465 lowest_speed = calc_speed(minmult); 467 lowest_speed = calc_speed(minmult);
466 dprintk ("FSB:%dMHz Lowest speed: %s Highest speed:%s\n", fsb, 468 dprintk("FSB:%dMHz Lowest speed: %s Highest speed:%s\n", fsb,
467 print_speed(lowest_speed/1000), 469 print_speed(lowest_speed/1000),
468 print_speed(highest_speed/1000)); 470 print_speed(highest_speed/1000));
469 471
470 if (lowest_speed == highest_speed) { 472 if (lowest_speed == highest_speed) {
471 printk (KERN_INFO PFX "highestspeed == lowest, aborting.\n"); 473 printk(KERN_INFO PFX "highestspeed == lowest, aborting.\n");
472 return -EINVAL; 474 return -EINVAL;
473 } 475 }
474 if (lowest_speed > highest_speed) { 476 if (lowest_speed > highest_speed) {
475 printk (KERN_INFO PFX "nonsense! lowest (%d > %d) !\n", 477 printk(KERN_INFO PFX "nonsense! lowest (%d > %d) !\n",
476 lowest_speed, highest_speed); 478 lowest_speed, highest_speed);
477 return -EINVAL; 479 return -EINVAL;
478 } 480 }
479 481
480 longhaul_table = kmalloc((numscales + 1) * sizeof(struct cpufreq_frequency_table), GFP_KERNEL); 482 longhaul_table = kmalloc((numscales + 1) * sizeof(*longhaul_table),
481 if(!longhaul_table) 483 GFP_KERNEL);
484 if (!longhaul_table)
482 return -ENOMEM; 485 return -ENOMEM;
483 486
484 for (j = 0; j < numscales; j++) { 487 for (j = 0; j < numscales; j++) {
485 ratio = clock_ratio[j]; 488 ratio = mults[j];
486 if (ratio == -1) 489 if (ratio == -1)
487 continue; 490 continue;
488 if (ratio > maxmult || ratio < minmult) 491 if (ratio > maxmult || ratio < minmult)
@@ -507,13 +510,10 @@ static int __init longhaul_get_ranges(void)
507 } 510 }
508 } 511 }
509 if (min_i != j) { 512 if (min_i != j) {
510 unsigned int temp; 513 swap(longhaul_table[j].frequency,
511 temp = longhaul_table[j].frequency; 514 longhaul_table[min_i].frequency);
512 longhaul_table[j].frequency = longhaul_table[min_i].frequency; 515 swap(longhaul_table[j].index,
513 longhaul_table[min_i].frequency = temp; 516 longhaul_table[min_i].index);
514 temp = longhaul_table[j].index;
515 longhaul_table[j].index = longhaul_table[min_i].index;
516 longhaul_table[min_i].index = temp;
517 } 517 }
518 } 518 }
519 519
@@ -521,7 +521,7 @@ static int __init longhaul_get_ranges(void)
521 521
522 /* Find index we are running on */ 522 /* Find index we are running on */
523 for (j = 0; j < k; j++) { 523 for (j = 0; j < k; j++) {
524 if (clock_ratio[longhaul_table[j].index & 0x1f] == mult) { 524 if (mults[longhaul_table[j].index & 0x1f] == mult) {
525 longhaul_index = j; 525 longhaul_index = j;
526 break; 526 break;
527 } 527 }
@@ -559,20 +559,22 @@ static void __init longhaul_setup_voltagescaling(void)
559 maxvid = vrm_mV_table[longhaul.bits.MaximumVID]; 559 maxvid = vrm_mV_table[longhaul.bits.MaximumVID];
560 560
561 if (minvid.mV == 0 || maxvid.mV == 0 || minvid.mV > maxvid.mV) { 561 if (minvid.mV == 0 || maxvid.mV == 0 || minvid.mV > maxvid.mV) {
562 printk (KERN_INFO PFX "Bogus values Min:%d.%03d Max:%d.%03d. " 562 printk(KERN_INFO PFX "Bogus values Min:%d.%03d Max:%d.%03d. "
563 "Voltage scaling disabled.\n", 563 "Voltage scaling disabled.\n",
564 minvid.mV/1000, minvid.mV%1000, maxvid.mV/1000, maxvid.mV%1000); 564 minvid.mV/1000, minvid.mV%1000,
565 maxvid.mV/1000, maxvid.mV%1000);
565 return; 566 return;
566 } 567 }
567 568
568 if (minvid.mV == maxvid.mV) { 569 if (minvid.mV == maxvid.mV) {
569 printk (KERN_INFO PFX "Claims to support voltage scaling but min & max are " 570 printk(KERN_INFO PFX "Claims to support voltage scaling but "
570 "both %d.%03d. Voltage scaling disabled\n", 571 "min & max are both %d.%03d. "
572 "Voltage scaling disabled\n",
571 maxvid.mV/1000, maxvid.mV%1000); 573 maxvid.mV/1000, maxvid.mV%1000);
572 return; 574 return;
573 } 575 }
574 576
575 /* How many voltage steps */ 577 /* How many voltage steps*/
576 numvscales = maxvid.pos - minvid.pos + 1; 578 numvscales = maxvid.pos - minvid.pos + 1;
577 printk(KERN_INFO PFX 579 printk(KERN_INFO PFX
578 "Max VID=%d.%03d " 580 "Max VID=%d.%03d "
@@ -586,7 +588,7 @@ static void __init longhaul_setup_voltagescaling(void)
586 j = longhaul.bits.MinMHzBR; 588 j = longhaul.bits.MinMHzBR;
587 if (longhaul.bits.MinMHzBR4) 589 if (longhaul.bits.MinMHzBR4)
588 j += 16; 590 j += 16;
589 min_vid_speed = eblcr_table[j]; 591 min_vid_speed = eblcr[j];
590 if (min_vid_speed == -1) 592 if (min_vid_speed == -1)
591 return; 593 return;
592 switch (longhaul.bits.MinMHzFSB) { 594 switch (longhaul.bits.MinMHzFSB) {
@@ -617,7 +619,8 @@ static void __init longhaul_setup_voltagescaling(void)
617 pos = minvid.pos; 619 pos = minvid.pos;
618 longhaul_table[j].index |= mV_vrm_table[pos] << 8; 620 longhaul_table[j].index |= mV_vrm_table[pos] << 8;
619 vid = vrm_mV_table[mV_vrm_table[pos]]; 621 vid = vrm_mV_table[mV_vrm_table[pos]];
620 printk(KERN_INFO PFX "f: %d kHz, index: %d, vid: %d mV\n", speed, j, vid.mV); 622 printk(KERN_INFO PFX "f: %d kHz, index: %d, vid: %d mV\n",
623 speed, j, vid.mV);
621 j++; 624 j++;
622 } 625 }
623 626
@@ -640,7 +643,8 @@ static int longhaul_target(struct cpufreq_policy *policy,
640 unsigned int dir = 0; 643 unsigned int dir = 0;
641 u8 vid, current_vid; 644 u8 vid, current_vid;
642 645
643 if (cpufreq_frequency_table_target(policy, longhaul_table, target_freq, relation, &table_index)) 646 if (cpufreq_frequency_table_target(policy, longhaul_table, target_freq,
647 relation, &table_index))
644 return -EINVAL; 648 return -EINVAL;
645 649
646 /* Don't set same frequency again */ 650 /* Don't set same frequency again */
@@ -656,7 +660,8 @@ static int longhaul_target(struct cpufreq_policy *policy,
656 * this in hardware, C3 is old and we need to do this 660 * this in hardware, C3 is old and we need to do this
657 * in software. */ 661 * in software. */
658 i = longhaul_index; 662 i = longhaul_index;
659 current_vid = (longhaul_table[longhaul_index].index >> 8) & 0x1f; 663 current_vid = (longhaul_table[longhaul_index].index >> 8);
664 current_vid &= 0x1f;
660 if (table_index > longhaul_index) 665 if (table_index > longhaul_index)
661 dir = 1; 666 dir = 1;
662 while (i != table_index) { 667 while (i != table_index) {
@@ -691,9 +696,9 @@ static acpi_status longhaul_walk_callback(acpi_handle obj_handle,
691{ 696{
692 struct acpi_device *d; 697 struct acpi_device *d;
693 698
694 if ( acpi_bus_get_device(obj_handle, &d) ) { 699 if (acpi_bus_get_device(obj_handle, &d))
695 return 0; 700 return 0;
696 } 701
697 *return_value = acpi_driver_data(d); 702 *return_value = acpi_driver_data(d);
698 return 1; 703 return 1;
699} 704}
@@ -750,7 +755,7 @@ static int longhaul_setup_southbridge(void)
750 /* Find VT8235 southbridge */ 755 /* Find VT8235 southbridge */
751 dev = pci_get_device(PCI_VENDOR_ID_VIA, PCI_DEVICE_ID_VIA_8235, NULL); 756 dev = pci_get_device(PCI_VENDOR_ID_VIA, PCI_DEVICE_ID_VIA_8235, NULL);
752 if (dev == NULL) 757 if (dev == NULL)
753 /* Find VT8237 southbridge */ 758 /* Find VT8237 southbridge */
754 dev = pci_get_device(PCI_VENDOR_ID_VIA, 759 dev = pci_get_device(PCI_VENDOR_ID_VIA,
755 PCI_DEVICE_ID_VIA_8237, NULL); 760 PCI_DEVICE_ID_VIA_8237, NULL);
756 if (dev != NULL) { 761 if (dev != NULL) {
@@ -769,7 +774,8 @@ static int longhaul_setup_southbridge(void)
769 if (pci_cmd & 1 << 7) { 774 if (pci_cmd & 1 << 7) {
770 pci_read_config_dword(dev, 0x88, &acpi_regs_addr); 775 pci_read_config_dword(dev, 0x88, &acpi_regs_addr);
771 acpi_regs_addr &= 0xff00; 776 acpi_regs_addr &= 0xff00;
772 printk(KERN_INFO PFX "ACPI I/O at 0x%x\n", acpi_regs_addr); 777 printk(KERN_INFO PFX "ACPI I/O at 0x%x\n",
778 acpi_regs_addr);
773 } 779 }
774 780
775 pci_dev_put(dev); 781 pci_dev_put(dev);
@@ -781,7 +787,7 @@ static int longhaul_setup_southbridge(void)
781static int __init longhaul_cpu_init(struct cpufreq_policy *policy) 787static int __init longhaul_cpu_init(struct cpufreq_policy *policy)
782{ 788{
783 struct cpuinfo_x86 *c = &cpu_data(0); 789 struct cpuinfo_x86 *c = &cpu_data(0);
784 char *cpuname=NULL; 790 char *cpuname = NULL;
785 int ret; 791 int ret;
786 u32 lo, hi; 792 u32 lo, hi;
787 793
@@ -791,8 +797,8 @@ static int __init longhaul_cpu_init(struct cpufreq_policy *policy)
791 cpu_model = CPU_SAMUEL; 797 cpu_model = CPU_SAMUEL;
792 cpuname = "C3 'Samuel' [C5A]"; 798 cpuname = "C3 'Samuel' [C5A]";
793 longhaul_version = TYPE_LONGHAUL_V1; 799 longhaul_version = TYPE_LONGHAUL_V1;
794 memcpy (clock_ratio, samuel1_clock_ratio, sizeof(samuel1_clock_ratio)); 800 memcpy(mults, samuel1_mults, sizeof(samuel1_mults));
795 memcpy (eblcr_table, samuel1_eblcr, sizeof(samuel1_eblcr)); 801 memcpy(eblcr, samuel1_eblcr, sizeof(samuel1_eblcr));
796 break; 802 break;
797 803
798 case 7: 804 case 7:
@@ -803,10 +809,8 @@ static int __init longhaul_cpu_init(struct cpufreq_policy *policy)
803 cpuname = "C3 'Samuel 2' [C5B]"; 809 cpuname = "C3 'Samuel 2' [C5B]";
804 /* Note, this is not a typo, early Samuel2's had 810 /* Note, this is not a typo, early Samuel2's had
805 * Samuel1 ratios. */ 811 * Samuel1 ratios. */
806 memcpy(clock_ratio, samuel1_clock_ratio, 812 memcpy(mults, samuel1_mults, sizeof(samuel1_mults));
807 sizeof(samuel1_clock_ratio)); 813 memcpy(eblcr, samuel2_eblcr, sizeof(samuel2_eblcr));
808 memcpy(eblcr_table, samuel2_eblcr,
809 sizeof(samuel2_eblcr));
810 break; 814 break;
811 case 1 ... 15: 815 case 1 ... 15:
812 longhaul_version = TYPE_LONGHAUL_V1; 816 longhaul_version = TYPE_LONGHAUL_V1;
@@ -817,10 +821,8 @@ static int __init longhaul_cpu_init(struct cpufreq_policy *policy)
817 cpu_model = CPU_EZRA; 821 cpu_model = CPU_EZRA;
818 cpuname = "C3 'Ezra' [C5C]"; 822 cpuname = "C3 'Ezra' [C5C]";
819 } 823 }
820 memcpy(clock_ratio, ezra_clock_ratio, 824 memcpy(mults, ezra_mults, sizeof(ezra_mults));
821 sizeof(ezra_clock_ratio)); 825 memcpy(eblcr, ezra_eblcr, sizeof(ezra_eblcr));
822 memcpy(eblcr_table, ezra_eblcr,
823 sizeof(ezra_eblcr));
824 break; 826 break;
825 } 827 }
826 break; 828 break;
@@ -829,18 +831,16 @@ static int __init longhaul_cpu_init(struct cpufreq_policy *policy)
829 cpu_model = CPU_EZRA_T; 831 cpu_model = CPU_EZRA_T;
830 cpuname = "C3 'Ezra-T' [C5M]"; 832 cpuname = "C3 'Ezra-T' [C5M]";
831 longhaul_version = TYPE_POWERSAVER; 833 longhaul_version = TYPE_POWERSAVER;
832 numscales=32; 834 numscales = 32;
833 memcpy (clock_ratio, ezrat_clock_ratio, sizeof(ezrat_clock_ratio)); 835 memcpy(mults, ezrat_mults, sizeof(ezrat_mults));
834 memcpy (eblcr_table, ezrat_eblcr, sizeof(ezrat_eblcr)); 836 memcpy(eblcr, ezrat_eblcr, sizeof(ezrat_eblcr));
835 break; 837 break;
836 838
837 case 9: 839 case 9:
838 longhaul_version = TYPE_POWERSAVER; 840 longhaul_version = TYPE_POWERSAVER;
839 numscales = 32; 841 numscales = 32;
840 memcpy(clock_ratio, 842 memcpy(mults, nehemiah_mults, sizeof(nehemiah_mults));
841 nehemiah_clock_ratio, 843 memcpy(eblcr, nehemiah_eblcr, sizeof(nehemiah_eblcr));
842 sizeof(nehemiah_clock_ratio));
843 memcpy(eblcr_table, nehemiah_eblcr, sizeof(nehemiah_eblcr));
844 switch (c->x86_mask) { 844 switch (c->x86_mask) {
845 case 0 ... 1: 845 case 0 ... 1:
846 cpu_model = CPU_NEHEMIAH; 846 cpu_model = CPU_NEHEMIAH;
@@ -869,14 +869,14 @@ static int __init longhaul_cpu_init(struct cpufreq_policy *policy)
869 longhaul_version = TYPE_LONGHAUL_V1; 869 longhaul_version = TYPE_LONGHAUL_V1;
870 } 870 }
871 871
872 printk (KERN_INFO PFX "VIA %s CPU detected. ", cpuname); 872 printk(KERN_INFO PFX "VIA %s CPU detected. ", cpuname);
873 switch (longhaul_version) { 873 switch (longhaul_version) {
874 case TYPE_LONGHAUL_V1: 874 case TYPE_LONGHAUL_V1:
875 case TYPE_LONGHAUL_V2: 875 case TYPE_LONGHAUL_V2:
876 printk ("Longhaul v%d supported.\n", longhaul_version); 876 printk(KERN_CONT "Longhaul v%d supported.\n", longhaul_version);
877 break; 877 break;
878 case TYPE_POWERSAVER: 878 case TYPE_POWERSAVER:
879 printk ("Powersaver supported.\n"); 879 printk(KERN_CONT "Powersaver supported.\n");
880 break; 880 break;
881 }; 881 };
882 882
@@ -940,7 +940,7 @@ static int __devexit longhaul_cpu_exit(struct cpufreq_policy *policy)
940 return 0; 940 return 0;
941} 941}
942 942
943static struct freq_attr* longhaul_attr[] = { 943static struct freq_attr *longhaul_attr[] = {
944 &cpufreq_freq_attr_scaling_available_freqs, 944 &cpufreq_freq_attr_scaling_available_freqs,
945 NULL, 945 NULL,
946}; 946};
@@ -966,13 +966,15 @@ static int __init longhaul_init(void)
966 966
967#ifdef CONFIG_SMP 967#ifdef CONFIG_SMP
968 if (num_online_cpus() > 1) { 968 if (num_online_cpus() > 1) {
969 printk(KERN_ERR PFX "More than 1 CPU detected, longhaul disabled.\n"); 969 printk(KERN_ERR PFX "More than 1 CPU detected, "
970 "longhaul disabled.\n");
970 return -ENODEV; 971 return -ENODEV;
971 } 972 }
972#endif 973#endif
973#ifdef CONFIG_X86_IO_APIC 974#ifdef CONFIG_X86_IO_APIC
974 if (cpu_has_apic) { 975 if (cpu_has_apic) {
975 printk(KERN_ERR PFX "APIC detected. Longhaul is currently broken in this configuration.\n"); 976 printk(KERN_ERR PFX "APIC detected. Longhaul is currently "
977 "broken in this configuration.\n");
976 return -ENODEV; 978 return -ENODEV;
977 } 979 }
978#endif 980#endif
@@ -993,8 +995,8 @@ static void __exit longhaul_exit(void)
993{ 995{
994 int i; 996 int i;
995 997
996 for (i=0; i < numscales; i++) { 998 for (i = 0; i < numscales; i++) {
997 if (clock_ratio[i] == maxmult) { 999 if (mults[i] == maxmult) {
998 longhaul_setstate(i); 1000 longhaul_setstate(i);
999 break; 1001 break;
1000 } 1002 }
@@ -1007,11 +1009,11 @@ static void __exit longhaul_exit(void)
1007/* Even if BIOS is exporting ACPI C3 state, and it is used 1009/* Even if BIOS is exporting ACPI C3 state, and it is used
1008 * with success when CPU is idle, this state doesn't 1010 * with success when CPU is idle, this state doesn't
1009 * trigger frequency transition in some cases. */ 1011 * trigger frequency transition in some cases. */
1010module_param (disable_acpi_c3, int, 0644); 1012module_param(disable_acpi_c3, int, 0644);
1011MODULE_PARM_DESC(disable_acpi_c3, "Don't use ACPI C3 support"); 1013MODULE_PARM_DESC(disable_acpi_c3, "Don't use ACPI C3 support");
1012/* Change CPU voltage with frequency. Very usefull to save 1014/* Change CPU voltage with frequency. Very usefull to save
1013 * power, but most VIA C3 processors aren't supporting it. */ 1015 * power, but most VIA C3 processors aren't supporting it. */
1014module_param (scale_voltage, int, 0644); 1016module_param(scale_voltage, int, 0644);
1015MODULE_PARM_DESC(scale_voltage, "Scale voltage of processor"); 1017MODULE_PARM_DESC(scale_voltage, "Scale voltage of processor");
1016/* Force revision key to 0 for processors which doesn't 1018/* Force revision key to 0 for processors which doesn't
1017 * support voltage scaling, but are introducing itself as 1019 * support voltage scaling, but are introducing itself as
@@ -1019,9 +1021,9 @@ MODULE_PARM_DESC(scale_voltage, "Scale voltage of processor");
1019module_param(revid_errata, int, 0644); 1021module_param(revid_errata, int, 0644);
1020MODULE_PARM_DESC(revid_errata, "Ignore CPU Revision ID"); 1022MODULE_PARM_DESC(revid_errata, "Ignore CPU Revision ID");
1021 1023
1022MODULE_AUTHOR ("Dave Jones <davej@redhat.com>"); 1024MODULE_AUTHOR("Dave Jones <davej@redhat.com>");
1023MODULE_DESCRIPTION ("Longhaul driver for VIA Cyrix processors."); 1025MODULE_DESCRIPTION("Longhaul driver for VIA Cyrix processors.");
1024MODULE_LICENSE ("GPL"); 1026MODULE_LICENSE("GPL");
1025 1027
1026late_initcall(longhaul_init); 1028late_initcall(longhaul_init);
1027module_exit(longhaul_exit); 1029module_exit(longhaul_exit);
diff --git a/arch/x86/kernel/cpu/cpufreq/longhaul.h b/arch/x86/kernel/cpu/cpufreq/longhaul.h
index 4fcc320997df..e2360a469f79 100644
--- a/arch/x86/kernel/cpu/cpufreq/longhaul.h
+++ b/arch/x86/kernel/cpu/cpufreq/longhaul.h
@@ -49,14 +49,14 @@ union msr_longhaul {
49 49
50/* 50/*
51 * Clock ratio tables. Div/Mod by 10 to get ratio. 51 * Clock ratio tables. Div/Mod by 10 to get ratio.
52 * The eblcr ones specify the ratio read from the CPU. 52 * The eblcr values specify the ratio read from the CPU.
53 * The clock_ratio ones specify what to write to the CPU. 53 * The mults values specify what to write to the CPU.
54 */ 54 */
55 55
56/* 56/*
57 * VIA C3 Samuel 1 & Samuel 2 (stepping 0) 57 * VIA C3 Samuel 1 & Samuel 2 (stepping 0)
58 */ 58 */
59static const int __initdata samuel1_clock_ratio[16] = { 59static const int __initdata samuel1_mults[16] = {
60 -1, /* 0000 -> RESERVED */ 60 -1, /* 0000 -> RESERVED */
61 30, /* 0001 -> 3.0x */ 61 30, /* 0001 -> 3.0x */
62 40, /* 0010 -> 4.0x */ 62 40, /* 0010 -> 4.0x */
@@ -119,7 +119,7 @@ static const int __initdata samuel2_eblcr[16] = {
119/* 119/*
120 * VIA C3 Ezra 120 * VIA C3 Ezra
121 */ 121 */
122static const int __initdata ezra_clock_ratio[16] = { 122static const int __initdata ezra_mults[16] = {
123 100, /* 0000 -> 10.0x */ 123 100, /* 0000 -> 10.0x */
124 30, /* 0001 -> 3.0x */ 124 30, /* 0001 -> 3.0x */
125 40, /* 0010 -> 4.0x */ 125 40, /* 0010 -> 4.0x */
@@ -160,7 +160,7 @@ static const int __initdata ezra_eblcr[16] = {
160/* 160/*
161 * VIA C3 (Ezra-T) [C5M]. 161 * VIA C3 (Ezra-T) [C5M].
162 */ 162 */
163static const int __initdata ezrat_clock_ratio[32] = { 163static const int __initdata ezrat_mults[32] = {
164 100, /* 0000 -> 10.0x */ 164 100, /* 0000 -> 10.0x */
165 30, /* 0001 -> 3.0x */ 165 30, /* 0001 -> 3.0x */
166 40, /* 0010 -> 4.0x */ 166 40, /* 0010 -> 4.0x */
@@ -235,7 +235,7 @@ static const int __initdata ezrat_eblcr[32] = {
235/* 235/*
236 * VIA C3 Nehemiah */ 236 * VIA C3 Nehemiah */
237 237
238static const int __initdata nehemiah_clock_ratio[32] = { 238static const int __initdata nehemiah_mults[32] = {
239 100, /* 0000 -> 10.0x */ 239 100, /* 0000 -> 10.0x */
240 -1, /* 0001 -> 16.0x */ 240 -1, /* 0001 -> 16.0x */
241 40, /* 0010 -> 4.0x */ 241 40, /* 0010 -> 4.0x */
diff --git a/arch/x86/kernel/cpu/cpufreq/longrun.c b/arch/x86/kernel/cpu/cpufreq/longrun.c
index 777a7ff075de..da5f70fcb766 100644
--- a/arch/x86/kernel/cpu/cpufreq/longrun.c
+++ b/arch/x86/kernel/cpu/cpufreq/longrun.c
@@ -11,12 +11,13 @@
11#include <linux/init.h> 11#include <linux/init.h>
12#include <linux/slab.h> 12#include <linux/slab.h>
13#include <linux/cpufreq.h> 13#include <linux/cpufreq.h>
14#include <linux/timex.h>
14 15
15#include <asm/msr.h> 16#include <asm/msr.h>
16#include <asm/processor.h> 17#include <asm/processor.h>
17#include <asm/timex.h>
18 18
19#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "longrun", msg) 19#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
20 "longrun", msg)
20 21
21static struct cpufreq_driver longrun_driver; 22static struct cpufreq_driver longrun_driver;
22 23
@@ -51,7 +52,7 @@ static void __init longrun_get_policy(struct cpufreq_policy *policy)
51 msr_lo &= 0x0000007F; 52 msr_lo &= 0x0000007F;
52 msr_hi &= 0x0000007F; 53 msr_hi &= 0x0000007F;
53 54
54 if ( longrun_high_freq <= longrun_low_freq ) { 55 if (longrun_high_freq <= longrun_low_freq) {
55 /* Assume degenerate Longrun table */ 56 /* Assume degenerate Longrun table */
56 policy->min = policy->max = longrun_high_freq; 57 policy->min = policy->max = longrun_high_freq;
57 } else { 58 } else {
@@ -79,7 +80,7 @@ static int longrun_set_policy(struct cpufreq_policy *policy)
79 if (!policy) 80 if (!policy)
80 return -EINVAL; 81 return -EINVAL;
81 82
82 if ( longrun_high_freq <= longrun_low_freq ) { 83 if (longrun_high_freq <= longrun_low_freq) {
83 /* Assume degenerate Longrun table */ 84 /* Assume degenerate Longrun table */
84 pctg_lo = pctg_hi = 100; 85 pctg_lo = pctg_hi = 100;
85 } else { 86 } else {
@@ -152,7 +153,7 @@ static unsigned int longrun_get(unsigned int cpu)
152 cpuid(0x80860007, &eax, &ebx, &ecx, &edx); 153 cpuid(0x80860007, &eax, &ebx, &ecx, &edx);
153 dprintk("cpuid eax is %u\n", eax); 154 dprintk("cpuid eax is %u\n", eax);
154 155
155 return (eax * 1000); 156 return eax * 1000;
156} 157}
157 158
158/** 159/**
@@ -196,7 +197,8 @@ static unsigned int __init longrun_determine_freqs(unsigned int *low_freq,
196 rdmsr(MSR_TMTA_LRTI_VOLT_MHZ, msr_lo, msr_hi); 197 rdmsr(MSR_TMTA_LRTI_VOLT_MHZ, msr_lo, msr_hi);
197 *high_freq = msr_lo * 1000; /* to kHz */ 198 *high_freq = msr_lo * 1000; /* to kHz */
198 199
199 dprintk("longrun table interface told %u - %u kHz\n", *low_freq, *high_freq); 200 dprintk("longrun table interface told %u - %u kHz\n",
201 *low_freq, *high_freq);
200 202
201 if (*low_freq > *high_freq) 203 if (*low_freq > *high_freq)
202 *low_freq = *high_freq; 204 *low_freq = *high_freq;
@@ -219,7 +221,7 @@ static unsigned int __init longrun_determine_freqs(unsigned int *low_freq,
219 cpuid(0x80860007, &eax, &ebx, &ecx, &edx); 221 cpuid(0x80860007, &eax, &ebx, &ecx, &edx);
220 /* try decreasing in 10% steps, some processors react only 222 /* try decreasing in 10% steps, some processors react only
221 * on some barrier values */ 223 * on some barrier values */
222 for (try_hi = 80; try_hi > 0 && ecx > 90; try_hi -=10) { 224 for (try_hi = 80; try_hi > 0 && ecx > 90; try_hi -= 10) {
223 /* set to 0 to try_hi perf_pctg */ 225 /* set to 0 to try_hi perf_pctg */
224 msr_lo &= 0xFFFFFF80; 226 msr_lo &= 0xFFFFFF80;
225 msr_hi &= 0xFFFFFF80; 227 msr_hi &= 0xFFFFFF80;
@@ -236,7 +238,7 @@ static unsigned int __init longrun_determine_freqs(unsigned int *low_freq,
236 238
237 /* performance_pctg = (current_freq - low_freq)/(high_freq - low_freq) 239 /* performance_pctg = (current_freq - low_freq)/(high_freq - low_freq)
238 * eqals 240 * eqals
239 * low_freq * ( 1 - perf_pctg) = (cur_freq - high_freq * perf_pctg) 241 * low_freq * (1 - perf_pctg) = (cur_freq - high_freq * perf_pctg)
240 * 242 *
241 * high_freq * perf_pctg is stored tempoarily into "ebx". 243 * high_freq * perf_pctg is stored tempoarily into "ebx".
242 */ 244 */
@@ -317,9 +319,10 @@ static void __exit longrun_exit(void)
317} 319}
318 320
319 321
320MODULE_AUTHOR ("Dominik Brodowski <linux@brodo.de>"); 322MODULE_AUTHOR("Dominik Brodowski <linux@brodo.de>");
321MODULE_DESCRIPTION ("LongRun driver for Transmeta Crusoe and Efficeon processors."); 323MODULE_DESCRIPTION("LongRun driver for Transmeta Crusoe and "
322MODULE_LICENSE ("GPL"); 324 "Efficeon processors.");
325MODULE_LICENSE("GPL");
323 326
324module_init(longrun_init); 327module_init(longrun_init);
325module_exit(longrun_exit); 328module_exit(longrun_exit);
diff --git a/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c b/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c
index 3178c3acd97e..6ac55bd341ae 100644
--- a/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c
+++ b/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c
@@ -27,15 +27,17 @@
27#include <linux/cpufreq.h> 27#include <linux/cpufreq.h>
28#include <linux/slab.h> 28#include <linux/slab.h>
29#include <linux/cpumask.h> 29#include <linux/cpumask.h>
30#include <linux/timex.h>
30 31
31#include <asm/processor.h> 32#include <asm/processor.h>
32#include <asm/msr.h> 33#include <asm/msr.h>
33#include <asm/timex.h> 34#include <asm/timer.h>
34 35
35#include "speedstep-lib.h" 36#include "speedstep-lib.h"
36 37
37#define PFX "p4-clockmod: " 38#define PFX "p4-clockmod: "
38#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "p4-clockmod", msg) 39#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
40 "p4-clockmod", msg)
39 41
40/* 42/*
41 * Duty Cycle (3bits), note DC_DISABLE is not specified in 43 * Duty Cycle (3bits), note DC_DISABLE is not specified in
@@ -58,7 +60,8 @@ static int cpufreq_p4_setdc(unsigned int cpu, unsigned int newstate)
58{ 60{
59 u32 l, h; 61 u32 l, h;
60 62
61 if (!cpu_online(cpu) || (newstate > DC_DISABLE) || (newstate == DC_RESV)) 63 if (!cpu_online(cpu) ||
64 (newstate > DC_DISABLE) || (newstate == DC_RESV))
62 return -EINVAL; 65 return -EINVAL;
63 66
64 rdmsr_on_cpu(cpu, MSR_IA32_THERM_STATUS, &l, &h); 67 rdmsr_on_cpu(cpu, MSR_IA32_THERM_STATUS, &l, &h);
@@ -66,7 +69,8 @@ static int cpufreq_p4_setdc(unsigned int cpu, unsigned int newstate)
66 if (l & 0x01) 69 if (l & 0x01)
67 dprintk("CPU#%d currently thermal throttled\n", cpu); 70 dprintk("CPU#%d currently thermal throttled\n", cpu);
68 71
69 if (has_N44_O17_errata[cpu] && (newstate == DC_25PT || newstate == DC_DFLT)) 72 if (has_N44_O17_errata[cpu] &&
73 (newstate == DC_25PT || newstate == DC_DFLT))
70 newstate = DC_38PT; 74 newstate = DC_38PT;
71 75
72 rdmsr_on_cpu(cpu, MSR_IA32_THERM_CONTROL, &l, &h); 76 rdmsr_on_cpu(cpu, MSR_IA32_THERM_CONTROL, &l, &h);
@@ -112,7 +116,8 @@ static int cpufreq_p4_target(struct cpufreq_policy *policy,
112 struct cpufreq_freqs freqs; 116 struct cpufreq_freqs freqs;
113 int i; 117 int i;
114 118
115 if (cpufreq_frequency_table_target(policy, &p4clockmod_table[0], target_freq, relation, &newstate)) 119 if (cpufreq_frequency_table_target(policy, &p4clockmod_table[0],
120 target_freq, relation, &newstate))
116 return -EINVAL; 121 return -EINVAL;
117 122
118 freqs.old = cpufreq_p4_get(policy->cpu); 123 freqs.old = cpufreq_p4_get(policy->cpu);
@@ -127,7 +132,8 @@ static int cpufreq_p4_target(struct cpufreq_policy *policy,
127 cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); 132 cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
128 } 133 }
129 134
130 /* run on each logical CPU, see section 13.15.3 of IA32 Intel Architecture Software 135 /* run on each logical CPU,
136 * see section 13.15.3 of IA32 Intel Architecture Software
131 * Developer's Manual, Volume 3 137 * Developer's Manual, Volume 3
132 */ 138 */
133 for_each_cpu(i, policy->cpus) 139 for_each_cpu(i, policy->cpus)
@@ -153,28 +159,30 @@ static unsigned int cpufreq_p4_get_frequency(struct cpuinfo_x86 *c)
153{ 159{
154 if (c->x86 == 0x06) { 160 if (c->x86 == 0x06) {
155 if (cpu_has(c, X86_FEATURE_EST)) 161 if (cpu_has(c, X86_FEATURE_EST))
156 printk(KERN_WARNING PFX "Warning: EST-capable CPU detected. " 162 printk(KERN_WARNING PFX "Warning: EST-capable CPU "
157 "The acpi-cpufreq module offers voltage scaling" 163 "detected. The acpi-cpufreq module offers "
158 " in addition of frequency scaling. You should use " 164 "voltage scaling in addition of frequency "
159 "that instead of p4-clockmod, if possible.\n"); 165 "scaling. You should use that instead of "
166 "p4-clockmod, if possible.\n");
160 switch (c->x86_model) { 167 switch (c->x86_model) {
161 case 0x0E: /* Core */ 168 case 0x0E: /* Core */
162 case 0x0F: /* Core Duo */ 169 case 0x0F: /* Core Duo */
163 case 0x16: /* Celeron Core */ 170 case 0x16: /* Celeron Core */
164 p4clockmod_driver.flags |= CPUFREQ_CONST_LOOPS; 171 p4clockmod_driver.flags |= CPUFREQ_CONST_LOOPS;
165 return speedstep_get_processor_frequency(SPEEDSTEP_PROCESSOR_PCORE); 172 return speedstep_get_frequency(SPEEDSTEP_CPU_PCORE);
166 case 0x0D: /* Pentium M (Dothan) */ 173 case 0x0D: /* Pentium M (Dothan) */
167 p4clockmod_driver.flags |= CPUFREQ_CONST_LOOPS; 174 p4clockmod_driver.flags |= CPUFREQ_CONST_LOOPS;
168 /* fall through */ 175 /* fall through */
169 case 0x09: /* Pentium M (Banias) */ 176 case 0x09: /* Pentium M (Banias) */
170 return speedstep_get_processor_frequency(SPEEDSTEP_PROCESSOR_PM); 177 return speedstep_get_frequency(SPEEDSTEP_CPU_PM);
171 } 178 }
172 } 179 }
173 180
174 if (c->x86 != 0xF) { 181 if (c->x86 != 0xF) {
175 if (!cpu_has(c, X86_FEATURE_EST)) 182 if (!cpu_has(c, X86_FEATURE_EST))
176 printk(KERN_WARNING PFX "Unknown p4-clockmod-capable CPU. " 183 printk(KERN_WARNING PFX "Unknown CPU. "
177 "Please send an e-mail to <cpufreq@vger.kernel.org>\n"); 184 "Please send an e-mail to "
185 "<cpufreq@vger.kernel.org>\n");
178 return 0; 186 return 0;
179 } 187 }
180 188
@@ -182,16 +190,16 @@ static unsigned int cpufreq_p4_get_frequency(struct cpuinfo_x86 *c)
182 * throttling is active or not. */ 190 * throttling is active or not. */
183 p4clockmod_driver.flags |= CPUFREQ_CONST_LOOPS; 191 p4clockmod_driver.flags |= CPUFREQ_CONST_LOOPS;
184 192
185 if (speedstep_detect_processor() == SPEEDSTEP_PROCESSOR_P4M) { 193 if (speedstep_detect_processor() == SPEEDSTEP_CPU_P4M) {
186 printk(KERN_WARNING PFX "Warning: Pentium 4-M detected. " 194 printk(KERN_WARNING PFX "Warning: Pentium 4-M detected. "
187 "The speedstep-ich or acpi cpufreq modules offer " 195 "The speedstep-ich or acpi cpufreq modules offer "
188 "voltage scaling in addition of frequency scaling. " 196 "voltage scaling in addition of frequency scaling. "
189 "You should use either one instead of p4-clockmod, " 197 "You should use either one instead of p4-clockmod, "
190 "if possible.\n"); 198 "if possible.\n");
191 return speedstep_get_processor_frequency(SPEEDSTEP_PROCESSOR_P4M); 199 return speedstep_get_frequency(SPEEDSTEP_CPU_P4M);
192 } 200 }
193 201
194 return speedstep_get_processor_frequency(SPEEDSTEP_PROCESSOR_P4D); 202 return speedstep_get_frequency(SPEEDSTEP_CPU_P4D);
195} 203}
196 204
197 205
@@ -203,7 +211,7 @@ static int cpufreq_p4_cpu_init(struct cpufreq_policy *policy)
203 unsigned int i; 211 unsigned int i;
204 212
205#ifdef CONFIG_SMP 213#ifdef CONFIG_SMP
206 cpumask_copy(policy->cpus, &per_cpu(cpu_sibling_map, policy->cpu)); 214 cpumask_copy(policy->cpus, cpu_sibling_mask(policy->cpu));
207#endif 215#endif
208 216
209 /* Errata workaround */ 217 /* Errata workaround */
@@ -217,14 +225,20 @@ static int cpufreq_p4_cpu_init(struct cpufreq_policy *policy)
217 dprintk("has errata -- disabling low frequencies\n"); 225 dprintk("has errata -- disabling low frequencies\n");
218 } 226 }
219 227
228 if (speedstep_detect_processor() == SPEEDSTEP_CPU_P4D &&
229 c->x86_model < 2) {
230 /* switch to maximum frequency and measure result */
231 cpufreq_p4_setdc(policy->cpu, DC_DISABLE);
232 recalibrate_cpu_khz();
233 }
220 /* get max frequency */ 234 /* get max frequency */
221 stock_freq = cpufreq_p4_get_frequency(c); 235 stock_freq = cpufreq_p4_get_frequency(c);
222 if (!stock_freq) 236 if (!stock_freq)
223 return -EINVAL; 237 return -EINVAL;
224 238
225 /* table init */ 239 /* table init */
226 for (i=1; (p4clockmod_table[i].frequency != CPUFREQ_TABLE_END); i++) { 240 for (i = 1; (p4clockmod_table[i].frequency != CPUFREQ_TABLE_END); i++) {
227 if ((i<2) && (has_N44_O17_errata[policy->cpu])) 241 if ((i < 2) && (has_N44_O17_errata[policy->cpu]))
228 p4clockmod_table[i].frequency = CPUFREQ_ENTRY_INVALID; 242 p4clockmod_table[i].frequency = CPUFREQ_ENTRY_INVALID;
229 else 243 else
230 p4clockmod_table[i].frequency = (stock_freq * i)/8; 244 p4clockmod_table[i].frequency = (stock_freq * i)/8;
@@ -232,7 +246,10 @@ static int cpufreq_p4_cpu_init(struct cpufreq_policy *policy)
232 cpufreq_frequency_table_get_attr(p4clockmod_table, policy->cpu); 246 cpufreq_frequency_table_get_attr(p4clockmod_table, policy->cpu);
233 247
234 /* cpuinfo and default policy values */ 248 /* cpuinfo and default policy values */
235 policy->cpuinfo.transition_latency = 1000000; /* assumed */ 249
250 /* the transition latency is set to be 1 higher than the maximum
251 * transition latency of the ondemand governor */
252 policy->cpuinfo.transition_latency = 10000001;
236 policy->cur = stock_freq; 253 policy->cur = stock_freq;
237 254
238 return cpufreq_frequency_table_cpuinfo(policy, &p4clockmod_table[0]); 255 return cpufreq_frequency_table_cpuinfo(policy, &p4clockmod_table[0]);
@@ -258,12 +275,12 @@ static unsigned int cpufreq_p4_get(unsigned int cpu)
258 l = DC_DISABLE; 275 l = DC_DISABLE;
259 276
260 if (l != DC_DISABLE) 277 if (l != DC_DISABLE)
261 return (stock_freq * l / 8); 278 return stock_freq * l / 8;
262 279
263 return stock_freq; 280 return stock_freq;
264} 281}
265 282
266static struct freq_attr* p4clockmod_attr[] = { 283static struct freq_attr *p4clockmod_attr[] = {
267 &cpufreq_freq_attr_scaling_available_freqs, 284 &cpufreq_freq_attr_scaling_available_freqs,
268 NULL, 285 NULL,
269}; 286};
@@ -298,9 +315,10 @@ static int __init cpufreq_p4_init(void)
298 315
299 ret = cpufreq_register_driver(&p4clockmod_driver); 316 ret = cpufreq_register_driver(&p4clockmod_driver);
300 if (!ret) 317 if (!ret)
301 printk(KERN_INFO PFX "P4/Xeon(TM) CPU On-Demand Clock Modulation available\n"); 318 printk(KERN_INFO PFX "P4/Xeon(TM) CPU On-Demand Clock "
319 "Modulation available\n");
302 320
303 return (ret); 321 return ret;
304} 322}
305 323
306 324
@@ -310,9 +328,9 @@ static void __exit cpufreq_p4_exit(void)
310} 328}
311 329
312 330
313MODULE_AUTHOR ("Zwane Mwaikambo <zwane@commfireservices.com>"); 331MODULE_AUTHOR("Zwane Mwaikambo <zwane@commfireservices.com>");
314MODULE_DESCRIPTION ("cpufreq driver for Pentium(TM) 4/Xeon(TM)"); 332MODULE_DESCRIPTION("cpufreq driver for Pentium(TM) 4/Xeon(TM)");
315MODULE_LICENSE ("GPL"); 333MODULE_LICENSE("GPL");
316 334
317late_initcall(cpufreq_p4_init); 335late_initcall(cpufreq_p4_init);
318module_exit(cpufreq_p4_exit); 336module_exit(cpufreq_p4_exit);
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k6.c b/arch/x86/kernel/cpu/cpufreq/powernow-k6.c
index c1ac5790c63e..f10dea409f40 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k6.c
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k6.c
@@ -1,6 +1,7 @@
1/* 1/*
2 * This file was based upon code in Powertweak Linux (http://powertweak.sf.net) 2 * This file was based upon code in Powertweak Linux (http://powertweak.sf.net)
3 * (C) 2000-2003 Dave Jones, Arjan van de Ven, Janne Pänkälä, Dominik Brodowski. 3 * (C) 2000-2003 Dave Jones, Arjan van de Ven, Janne Pänkälä,
4 * Dominik Brodowski.
4 * 5 *
5 * Licensed under the terms of the GNU GPL License version 2. 6 * Licensed under the terms of the GNU GPL License version 2.
6 * 7 *
@@ -13,14 +14,15 @@
13#include <linux/cpufreq.h> 14#include <linux/cpufreq.h>
14#include <linux/ioport.h> 15#include <linux/ioport.h>
15#include <linux/slab.h> 16#include <linux/slab.h>
16
17#include <asm/msr.h>
18#include <linux/timex.h> 17#include <linux/timex.h>
19#include <linux/io.h> 18#include <linux/io.h>
20 19
20#include <asm/msr.h>
21
21#define POWERNOW_IOPORT 0xfff0 /* it doesn't matter where, as long 22#define POWERNOW_IOPORT 0xfff0 /* it doesn't matter where, as long
22 as it is unused */ 23 as it is unused */
23 24
25#define PFX "powernow-k6: "
24static unsigned int busfreq; /* FSB, in 10 kHz */ 26static unsigned int busfreq; /* FSB, in 10 kHz */
25static unsigned int max_multiplier; 27static unsigned int max_multiplier;
26 28
@@ -47,8 +49,8 @@ static struct cpufreq_frequency_table clock_ratio[] = {
47 */ 49 */
48static int powernow_k6_get_cpu_multiplier(void) 50static int powernow_k6_get_cpu_multiplier(void)
49{ 51{
50 u64 invalue = 0; 52 u64 invalue = 0;
51 u32 msrval; 53 u32 msrval;
52 54
53 msrval = POWERNOW_IOPORT + 0x1; 55 msrval = POWERNOW_IOPORT + 0x1;
54 wrmsr(MSR_K6_EPMR, msrval, 0); /* enable the PowerNow port */ 56 wrmsr(MSR_K6_EPMR, msrval, 0); /* enable the PowerNow port */
@@ -68,12 +70,12 @@ static int powernow_k6_get_cpu_multiplier(void)
68 */ 70 */
69static void powernow_k6_set_state(unsigned int best_i) 71static void powernow_k6_set_state(unsigned int best_i)
70{ 72{
71 unsigned long outvalue = 0, invalue = 0; 73 unsigned long outvalue = 0, invalue = 0;
72 unsigned long msrval; 74 unsigned long msrval;
73 struct cpufreq_freqs freqs; 75 struct cpufreq_freqs freqs;
74 76
75 if (clock_ratio[best_i].index > max_multiplier) { 77 if (clock_ratio[best_i].index > max_multiplier) {
76 printk(KERN_ERR "cpufreq: invalid target frequency\n"); 78 printk(KERN_ERR PFX "invalid target frequency\n");
77 return; 79 return;
78 } 80 }
79 81
@@ -119,7 +121,8 @@ static int powernow_k6_verify(struct cpufreq_policy *policy)
119 * powernow_k6_setpolicy - sets a new CPUFreq policy 121 * powernow_k6_setpolicy - sets a new CPUFreq policy
120 * @policy: new policy 122 * @policy: new policy
121 * @target_freq: the target frequency 123 * @target_freq: the target frequency
122 * @relation: how that frequency relates to achieved frequency (CPUFREQ_RELATION_L or CPUFREQ_RELATION_H) 124 * @relation: how that frequency relates to achieved frequency
125 * (CPUFREQ_RELATION_L or CPUFREQ_RELATION_H)
123 * 126 *
124 * sets a new CPUFreq policy 127 * sets a new CPUFreq policy
125 */ 128 */
@@ -127,9 +130,10 @@ static int powernow_k6_target(struct cpufreq_policy *policy,
127 unsigned int target_freq, 130 unsigned int target_freq,
128 unsigned int relation) 131 unsigned int relation)
129{ 132{
130 unsigned int newstate = 0; 133 unsigned int newstate = 0;
131 134
132 if (cpufreq_frequency_table_target(policy, &clock_ratio[0], target_freq, relation, &newstate)) 135 if (cpufreq_frequency_table_target(policy, &clock_ratio[0],
136 target_freq, relation, &newstate))
133 return -EINVAL; 137 return -EINVAL;
134 138
135 powernow_k6_set_state(newstate); 139 powernow_k6_set_state(newstate);
@@ -140,7 +144,7 @@ static int powernow_k6_target(struct cpufreq_policy *policy,
140 144
141static int powernow_k6_cpu_init(struct cpufreq_policy *policy) 145static int powernow_k6_cpu_init(struct cpufreq_policy *policy)
142{ 146{
143 unsigned int i; 147 unsigned int i, f;
144 int result; 148 int result;
145 149
146 if (policy->cpu != 0) 150 if (policy->cpu != 0)
@@ -152,10 +156,11 @@ static int powernow_k6_cpu_init(struct cpufreq_policy *policy)
152 156
153 /* table init */ 157 /* table init */
154 for (i = 0; (clock_ratio[i].frequency != CPUFREQ_TABLE_END); i++) { 158 for (i = 0; (clock_ratio[i].frequency != CPUFREQ_TABLE_END); i++) {
155 if (clock_ratio[i].index > max_multiplier) 159 f = clock_ratio[i].index;
160 if (f > max_multiplier)
156 clock_ratio[i].frequency = CPUFREQ_ENTRY_INVALID; 161 clock_ratio[i].frequency = CPUFREQ_ENTRY_INVALID;
157 else 162 else
158 clock_ratio[i].frequency = busfreq * clock_ratio[i].index; 163 clock_ratio[i].frequency = busfreq * f;
159 } 164 }
160 165
161 /* cpuinfo and default policy values */ 166 /* cpuinfo and default policy values */
@@ -185,7 +190,9 @@ static int powernow_k6_cpu_exit(struct cpufreq_policy *policy)
185 190
186static unsigned int powernow_k6_get(unsigned int cpu) 191static unsigned int powernow_k6_get(unsigned int cpu)
187{ 192{
188 return busfreq * powernow_k6_get_cpu_multiplier(); 193 unsigned int ret;
194 ret = (busfreq * powernow_k6_get_cpu_multiplier());
195 return ret;
189} 196}
190 197
191static struct freq_attr *powernow_k6_attr[] = { 198static struct freq_attr *powernow_k6_attr[] = {
@@ -221,7 +228,7 @@ static int __init powernow_k6_init(void)
221 return -ENODEV; 228 return -ENODEV;
222 229
223 if (!request_region(POWERNOW_IOPORT, 16, "PowerNow!")) { 230 if (!request_region(POWERNOW_IOPORT, 16, "PowerNow!")) {
224 printk("cpufreq: PowerNow IOPORT region already used.\n"); 231 printk(KERN_INFO PFX "PowerNow IOPORT region already used.\n");
225 return -EIO; 232 return -EIO;
226 } 233 }
227 234
@@ -246,7 +253,8 @@ static void __exit powernow_k6_exit(void)
246} 253}
247 254
248 255
249MODULE_AUTHOR("Arjan van de Ven, Dave Jones <davej@redhat.com>, Dominik Brodowski <linux@brodo.de>"); 256MODULE_AUTHOR("Arjan van de Ven, Dave Jones <davej@redhat.com>, "
257 "Dominik Brodowski <linux@brodo.de>");
250MODULE_DESCRIPTION("PowerNow! driver for AMD K6-2+ / K6-3+ processors."); 258MODULE_DESCRIPTION("PowerNow! driver for AMD K6-2+ / K6-3+ processors.");
251MODULE_LICENSE("GPL"); 259MODULE_LICENSE("GPL");
252 260
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k7.c b/arch/x86/kernel/cpu/cpufreq/powernow-k7.c
index 1b446d79a8fd..3c28ccd49742 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k7.c
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k7.c
@@ -6,10 +6,12 @@
6 * Licensed under the terms of the GNU GPL License version 2. 6 * Licensed under the terms of the GNU GPL License version 2.
7 * Based upon datasheets & sample CPUs kindly provided by AMD. 7 * Based upon datasheets & sample CPUs kindly provided by AMD.
8 * 8 *
9 * Errata 5: Processor may fail to execute a FID/VID change in presence of interrupt. 9 * Errata 5:
10 * - We cli/sti on stepping A0 CPUs around the FID/VID transition. 10 * CPU may fail to execute a FID/VID change in presence of interrupt.
11 * Errata 15: Processors with half frequency multipliers may hang upon wakeup from disconnect. 11 * - We cli/sti on stepping A0 CPUs around the FID/VID transition.
12 * - We disable half multipliers if ACPI is used on A0 stepping CPUs. 12 * Errata 15:
13 * CPU with half frequency multipliers may hang upon wakeup from disconnect.
14 * - We disable half multipliers if ACPI is used on A0 stepping CPUs.
13 */ 15 */
14 16
15#include <linux/kernel.h> 17#include <linux/kernel.h>
@@ -20,11 +22,11 @@
20#include <linux/slab.h> 22#include <linux/slab.h>
21#include <linux/string.h> 23#include <linux/string.h>
22#include <linux/dmi.h> 24#include <linux/dmi.h>
25#include <linux/timex.h>
26#include <linux/io.h>
23 27
28#include <asm/timer.h> /* Needed for recalibrate_cpu_khz() */
24#include <asm/msr.h> 29#include <asm/msr.h>
25#include <asm/timer.h>
26#include <asm/timex.h>
27#include <asm/io.h>
28#include <asm/system.h> 30#include <asm/system.h>
29 31
30#ifdef CONFIG_X86_POWERNOW_K7_ACPI 32#ifdef CONFIG_X86_POWERNOW_K7_ACPI
@@ -58,9 +60,9 @@ struct pst_s {
58union powernow_acpi_control_t { 60union powernow_acpi_control_t {
59 struct { 61 struct {
60 unsigned long fid:5, 62 unsigned long fid:5,
61 vid:5, 63 vid:5,
62 sgtc:20, 64 sgtc:20,
63 res1:2; 65 res1:2;
64 } bits; 66 } bits;
65 unsigned long val; 67 unsigned long val;
66}; 68};
@@ -94,14 +96,15 @@ static struct cpufreq_frequency_table *powernow_table;
94 96
95static unsigned int can_scale_bus; 97static unsigned int can_scale_bus;
96static unsigned int can_scale_vid; 98static unsigned int can_scale_vid;
97static unsigned int minimum_speed=-1; 99static unsigned int minimum_speed = -1;
98static unsigned int maximum_speed; 100static unsigned int maximum_speed;
99static unsigned int number_scales; 101static unsigned int number_scales;
100static unsigned int fsb; 102static unsigned int fsb;
101static unsigned int latency; 103static unsigned int latency;
102static char have_a0; 104static char have_a0;
103 105
104#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "powernow-k7", msg) 106#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
107 "powernow-k7", msg)
105 108
106static int check_fsb(unsigned int fsbspeed) 109static int check_fsb(unsigned int fsbspeed)
107{ 110{
@@ -109,7 +112,7 @@ static int check_fsb(unsigned int fsbspeed)
109 unsigned int f = fsb / 1000; 112 unsigned int f = fsb / 1000;
110 113
111 delta = (fsbspeed > f) ? fsbspeed - f : f - fsbspeed; 114 delta = (fsbspeed > f) ? fsbspeed - f : f - fsbspeed;
112 return (delta < 5); 115 return delta < 5;
113} 116}
114 117
115static int check_powernow(void) 118static int check_powernow(void)
@@ -117,24 +120,26 @@ static int check_powernow(void)
117 struct cpuinfo_x86 *c = &cpu_data(0); 120 struct cpuinfo_x86 *c = &cpu_data(0);
118 unsigned int maxei, eax, ebx, ecx, edx; 121 unsigned int maxei, eax, ebx, ecx, edx;
119 122
120 if ((c->x86_vendor != X86_VENDOR_AMD) || (c->x86 !=6)) { 123 if ((c->x86_vendor != X86_VENDOR_AMD) || (c->x86 != 6)) {
121#ifdef MODULE 124#ifdef MODULE
122 printk (KERN_INFO PFX "This module only works with AMD K7 CPUs\n"); 125 printk(KERN_INFO PFX "This module only works with "
126 "AMD K7 CPUs\n");
123#endif 127#endif
124 return 0; 128 return 0;
125 } 129 }
126 130
127 /* Get maximum capabilities */ 131 /* Get maximum capabilities */
128 maxei = cpuid_eax (0x80000000); 132 maxei = cpuid_eax(0x80000000);
129 if (maxei < 0x80000007) { /* Any powernow info ? */ 133 if (maxei < 0x80000007) { /* Any powernow info ? */
130#ifdef MODULE 134#ifdef MODULE
131 printk (KERN_INFO PFX "No powernow capabilities detected\n"); 135 printk(KERN_INFO PFX "No powernow capabilities detected\n");
132#endif 136#endif
133 return 0; 137 return 0;
134 } 138 }
135 139
136 if ((c->x86_model == 6) && (c->x86_mask == 0)) { 140 if ((c->x86_model == 6) && (c->x86_mask == 0)) {
137 printk (KERN_INFO PFX "K7 660[A0] core detected, enabling errata workarounds\n"); 141 printk(KERN_INFO PFX "K7 660[A0] core detected, "
142 "enabling errata workarounds\n");
138 have_a0 = 1; 143 have_a0 = 1;
139 } 144 }
140 145
@@ -144,37 +149,42 @@ static int check_powernow(void)
144 if (!(edx & (1 << 1 | 1 << 2))) 149 if (!(edx & (1 << 1 | 1 << 2)))
145 return 0; 150 return 0;
146 151
147 printk (KERN_INFO PFX "PowerNOW! Technology present. Can scale: "); 152 printk(KERN_INFO PFX "PowerNOW! Technology present. Can scale: ");
148 153
149 if (edx & 1 << 1) { 154 if (edx & 1 << 1) {
150 printk ("frequency"); 155 printk("frequency");
151 can_scale_bus=1; 156 can_scale_bus = 1;
152 } 157 }
153 158
154 if ((edx & (1 << 1 | 1 << 2)) == 0x6) 159 if ((edx & (1 << 1 | 1 << 2)) == 0x6)
155 printk (" and "); 160 printk(" and ");
156 161
157 if (edx & 1 << 2) { 162 if (edx & 1 << 2) {
158 printk ("voltage"); 163 printk("voltage");
159 can_scale_vid=1; 164 can_scale_vid = 1;
160 } 165 }
161 166
162 printk (".\n"); 167 printk(".\n");
163 return 1; 168 return 1;
164} 169}
165 170
171static void invalidate_entry(unsigned int entry)
172{
173 powernow_table[entry].frequency = CPUFREQ_ENTRY_INVALID;
174}
166 175
167static int get_ranges (unsigned char *pst) 176static int get_ranges(unsigned char *pst)
168{ 177{
169 unsigned int j; 178 unsigned int j;
170 unsigned int speed; 179 unsigned int speed;
171 u8 fid, vid; 180 u8 fid, vid;
172 181
173 powernow_table = kzalloc((sizeof(struct cpufreq_frequency_table) * (number_scales + 1)), GFP_KERNEL); 182 powernow_table = kzalloc((sizeof(struct cpufreq_frequency_table) *
183 (number_scales + 1)), GFP_KERNEL);
174 if (!powernow_table) 184 if (!powernow_table)
175 return -ENOMEM; 185 return -ENOMEM;
176 186
177 for (j=0 ; j < number_scales; j++) { 187 for (j = 0 ; j < number_scales; j++) {
178 fid = *pst++; 188 fid = *pst++;
179 189
180 powernow_table[j].frequency = (fsb * fid_codes[fid]) / 10; 190 powernow_table[j].frequency = (fsb * fid_codes[fid]) / 10;
@@ -182,10 +192,10 @@ static int get_ranges (unsigned char *pst)
182 192
183 speed = powernow_table[j].frequency; 193 speed = powernow_table[j].frequency;
184 194
185 if ((fid_codes[fid] % 10)==5) { 195 if ((fid_codes[fid] % 10) == 5) {
186#ifdef CONFIG_X86_POWERNOW_K7_ACPI 196#ifdef CONFIG_X86_POWERNOW_K7_ACPI
187 if (have_a0 == 1) 197 if (have_a0 == 1)
188 powernow_table[j].frequency = CPUFREQ_ENTRY_INVALID; 198 invalidate_entry(j);
189#endif 199#endif
190 } 200 }
191 201
@@ -197,7 +207,7 @@ static int get_ranges (unsigned char *pst)
197 vid = *pst++; 207 vid = *pst++;
198 powernow_table[j].index |= (vid << 8); /* upper 8 bits */ 208 powernow_table[j].index |= (vid << 8); /* upper 8 bits */
199 209
200 dprintk (" FID: 0x%x (%d.%dx [%dMHz]) " 210 dprintk(" FID: 0x%x (%d.%dx [%dMHz]) "
201 "VID: 0x%x (%d.%03dV)\n", fid, fid_codes[fid] / 10, 211 "VID: 0x%x (%d.%03dV)\n", fid, fid_codes[fid] / 10,
202 fid_codes[fid] % 10, speed/1000, vid, 212 fid_codes[fid] % 10, speed/1000, vid,
203 mobile_vid_table[vid]/1000, 213 mobile_vid_table[vid]/1000,
@@ -214,13 +224,13 @@ static void change_FID(int fid)
214{ 224{
215 union msr_fidvidctl fidvidctl; 225 union msr_fidvidctl fidvidctl;
216 226
217 rdmsrl (MSR_K7_FID_VID_CTL, fidvidctl.val); 227 rdmsrl(MSR_K7_FID_VID_CTL, fidvidctl.val);
218 if (fidvidctl.bits.FID != fid) { 228 if (fidvidctl.bits.FID != fid) {
219 fidvidctl.bits.SGTC = latency; 229 fidvidctl.bits.SGTC = latency;
220 fidvidctl.bits.FID = fid; 230 fidvidctl.bits.FID = fid;
221 fidvidctl.bits.VIDC = 0; 231 fidvidctl.bits.VIDC = 0;
222 fidvidctl.bits.FIDC = 1; 232 fidvidctl.bits.FIDC = 1;
223 wrmsrl (MSR_K7_FID_VID_CTL, fidvidctl.val); 233 wrmsrl(MSR_K7_FID_VID_CTL, fidvidctl.val);
224 } 234 }
225} 235}
226 236
@@ -229,18 +239,18 @@ static void change_VID(int vid)
229{ 239{
230 union msr_fidvidctl fidvidctl; 240 union msr_fidvidctl fidvidctl;
231 241
232 rdmsrl (MSR_K7_FID_VID_CTL, fidvidctl.val); 242 rdmsrl(MSR_K7_FID_VID_CTL, fidvidctl.val);
233 if (fidvidctl.bits.VID != vid) { 243 if (fidvidctl.bits.VID != vid) {
234 fidvidctl.bits.SGTC = latency; 244 fidvidctl.bits.SGTC = latency;
235 fidvidctl.bits.VID = vid; 245 fidvidctl.bits.VID = vid;
236 fidvidctl.bits.FIDC = 0; 246 fidvidctl.bits.FIDC = 0;
237 fidvidctl.bits.VIDC = 1; 247 fidvidctl.bits.VIDC = 1;
238 wrmsrl (MSR_K7_FID_VID_CTL, fidvidctl.val); 248 wrmsrl(MSR_K7_FID_VID_CTL, fidvidctl.val);
239 } 249 }
240} 250}
241 251
242 252
243static void change_speed (unsigned int index) 253static void change_speed(unsigned int index)
244{ 254{
245 u8 fid, vid; 255 u8 fid, vid;
246 struct cpufreq_freqs freqs; 256 struct cpufreq_freqs freqs;
@@ -257,7 +267,7 @@ static void change_speed (unsigned int index)
257 267
258 freqs.cpu = 0; 268 freqs.cpu = 0;
259 269
260 rdmsrl (MSR_K7_FID_VID_STATUS, fidvidstatus.val); 270 rdmsrl(MSR_K7_FID_VID_STATUS, fidvidstatus.val);
261 cfid = fidvidstatus.bits.CFID; 271 cfid = fidvidstatus.bits.CFID;
262 freqs.old = fsb * fid_codes[cfid] / 10; 272 freqs.old = fsb * fid_codes[cfid] / 10;
263 273
@@ -321,12 +331,14 @@ static int powernow_acpi_init(void)
321 goto err1; 331 goto err1;
322 } 332 }
323 333
324 if (acpi_processor_perf->control_register.space_id != ACPI_ADR_SPACE_FIXED_HARDWARE) { 334 if (acpi_processor_perf->control_register.space_id !=
335 ACPI_ADR_SPACE_FIXED_HARDWARE) {
325 retval = -ENODEV; 336 retval = -ENODEV;
326 goto err2; 337 goto err2;
327 } 338 }
328 339
329 if (acpi_processor_perf->status_register.space_id != ACPI_ADR_SPACE_FIXED_HARDWARE) { 340 if (acpi_processor_perf->status_register.space_id !=
341 ACPI_ADR_SPACE_FIXED_HARDWARE) {
330 retval = -ENODEV; 342 retval = -ENODEV;
331 goto err2; 343 goto err2;
332 } 344 }
@@ -338,7 +350,8 @@ static int powernow_acpi_init(void)
338 goto err2; 350 goto err2;
339 } 351 }
340 352
341 powernow_table = kzalloc((number_scales + 1) * (sizeof(struct cpufreq_frequency_table)), GFP_KERNEL); 353 powernow_table = kzalloc((sizeof(struct cpufreq_frequency_table) *
354 (number_scales + 1)), GFP_KERNEL);
342 if (!powernow_table) { 355 if (!powernow_table) {
343 retval = -ENOMEM; 356 retval = -ENOMEM;
344 goto err2; 357 goto err2;
@@ -352,7 +365,7 @@ static int powernow_acpi_init(void)
352 unsigned int speed, speed_mhz; 365 unsigned int speed, speed_mhz;
353 366
354 pc.val = (unsigned long) state->control; 367 pc.val = (unsigned long) state->control;
355 dprintk ("acpi: P%d: %d MHz %d mW %d uS control %08x SGTC %d\n", 368 dprintk("acpi: P%d: %d MHz %d mW %d uS control %08x SGTC %d\n",
356 i, 369 i,
357 (u32) state->core_frequency, 370 (u32) state->core_frequency,
358 (u32) state->power, 371 (u32) state->power,
@@ -381,12 +394,12 @@ static int powernow_acpi_init(void)
381 if (speed % 1000 > 0) 394 if (speed % 1000 > 0)
382 speed_mhz++; 395 speed_mhz++;
383 396
384 if ((fid_codes[fid] % 10)==5) { 397 if ((fid_codes[fid] % 10) == 5) {
385 if (have_a0 == 1) 398 if (have_a0 == 1)
386 powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID; 399 invalidate_entry(i);
387 } 400 }
388 401
389 dprintk (" FID: 0x%x (%d.%dx [%dMHz]) " 402 dprintk(" FID: 0x%x (%d.%dx [%dMHz]) "
390 "VID: 0x%x (%d.%03dV)\n", fid, fid_codes[fid] / 10, 403 "VID: 0x%x (%d.%03dV)\n", fid, fid_codes[fid] / 10,
391 fid_codes[fid] % 10, speed_mhz, vid, 404 fid_codes[fid] % 10, speed_mhz, vid,
392 mobile_vid_table[vid]/1000, 405 mobile_vid_table[vid]/1000,
@@ -422,7 +435,8 @@ err1:
422err05: 435err05:
423 kfree(acpi_processor_perf); 436 kfree(acpi_processor_perf);
424err0: 437err0:
425 printk(KERN_WARNING PFX "ACPI perflib can not be used in this platform\n"); 438 printk(KERN_WARNING PFX "ACPI perflib can not be used on "
439 "this platform\n");
426 acpi_processor_perf = NULL; 440 acpi_processor_perf = NULL;
427 return retval; 441 return retval;
428} 442}
@@ -435,7 +449,14 @@ static int powernow_acpi_init(void)
435} 449}
436#endif 450#endif
437 451
438static int powernow_decode_bios (int maxfid, int startvid) 452static void print_pst_entry(struct pst_s *pst, unsigned int j)
453{
454 dprintk("PST:%d (@%p)\n", j, pst);
455 dprintk(" cpuid: 0x%x fsb: %d maxFID: 0x%x startvid: 0x%x\n",
456 pst->cpuid, pst->fsbspeed, pst->maxfid, pst->startvid);
457}
458
459static int powernow_decode_bios(int maxfid, int startvid)
439{ 460{
440 struct psb_s *psb; 461 struct psb_s *psb;
441 struct pst_s *pst; 462 struct pst_s *pst;
@@ -446,61 +467,67 @@ static int powernow_decode_bios (int maxfid, int startvid)
446 467
447 etuple = cpuid_eax(0x80000001); 468 etuple = cpuid_eax(0x80000001);
448 469
449 for (i=0xC0000; i < 0xffff0 ; i+=16) { 470 for (i = 0xC0000; i < 0xffff0 ; i += 16) {
450 471
451 p = phys_to_virt(i); 472 p = phys_to_virt(i);
452 473
453 if (memcmp(p, "AMDK7PNOW!", 10) == 0){ 474 if (memcmp(p, "AMDK7PNOW!", 10) == 0) {
454 dprintk ("Found PSB header at %p\n", p); 475 dprintk("Found PSB header at %p\n", p);
455 psb = (struct psb_s *) p; 476 psb = (struct psb_s *) p;
456 dprintk ("Table version: 0x%x\n", psb->tableversion); 477 dprintk("Table version: 0x%x\n", psb->tableversion);
457 if (psb->tableversion != 0x12) { 478 if (psb->tableversion != 0x12) {
458 printk (KERN_INFO PFX "Sorry, only v1.2 tables supported right now\n"); 479 printk(KERN_INFO PFX "Sorry, only v1.2 tables"
480 " supported right now\n");
459 return -ENODEV; 481 return -ENODEV;
460 } 482 }
461 483
462 dprintk ("Flags: 0x%x\n", psb->flags); 484 dprintk("Flags: 0x%x\n", psb->flags);
463 if ((psb->flags & 1)==0) { 485 if ((psb->flags & 1) == 0)
464 dprintk ("Mobile voltage regulator\n"); 486 dprintk("Mobile voltage regulator\n");
465 } else { 487 else
466 dprintk ("Desktop voltage regulator\n"); 488 dprintk("Desktop voltage regulator\n");
467 }
468 489
469 latency = psb->settlingtime; 490 latency = psb->settlingtime;
470 if (latency < 100) { 491 if (latency < 100) {
471 printk(KERN_INFO PFX "BIOS set settling time to %d microseconds. " 492 printk(KERN_INFO PFX "BIOS set settling time "
472 "Should be at least 100. Correcting.\n", latency); 493 "to %d microseconds. "
494 "Should be at least 100. "
495 "Correcting.\n", latency);
473 latency = 100; 496 latency = 100;
474 } 497 }
475 dprintk ("Settling Time: %d microseconds.\n", psb->settlingtime); 498 dprintk("Settling Time: %d microseconds.\n",
476 dprintk ("Has %d PST tables. (Only dumping ones relevant to this CPU).\n", psb->numpst); 499 psb->settlingtime);
500 dprintk("Has %d PST tables. (Only dumping ones "
501 "relevant to this CPU).\n",
502 psb->numpst);
477 503
478 p += sizeof (struct psb_s); 504 p += sizeof(struct psb_s);
479 505
480 pst = (struct pst_s *) p; 506 pst = (struct pst_s *) p;
481 507
482 for (j=0; j<psb->numpst; j++) { 508 for (j = 0; j < psb->numpst; j++) {
483 pst = (struct pst_s *) p; 509 pst = (struct pst_s *) p;
484 number_scales = pst->numpstates; 510 number_scales = pst->numpstates;
485 511
486 if ((etuple == pst->cpuid) && check_fsb(pst->fsbspeed) && 512 if ((etuple == pst->cpuid) &&
487 (maxfid==pst->maxfid) && (startvid==pst->startvid)) 513 check_fsb(pst->fsbspeed) &&
488 { 514 (maxfid == pst->maxfid) &&
489 dprintk ("PST:%d (@%p)\n", j, pst); 515 (startvid == pst->startvid)) {
490 dprintk (" cpuid: 0x%x fsb: %d maxFID: 0x%x startvid: 0x%x\n", 516 print_pst_entry(pst, j);
491 pst->cpuid, pst->fsbspeed, pst->maxfid, pst->startvid); 517 p = (char *)pst + sizeof(struct pst_s);
492 518 ret = get_ranges(p);
493 ret = get_ranges ((char *) pst + sizeof (struct pst_s));
494 return ret; 519 return ret;
495 } else { 520 } else {
496 unsigned int k; 521 unsigned int k;
497 p = (char *) pst + sizeof (struct pst_s); 522 p = (char *)pst + sizeof(struct pst_s);
498 for (k=0; k<number_scales; k++) 523 for (k = 0; k < number_scales; k++)
499 p+=2; 524 p += 2;
500 } 525 }
501 } 526 }
502 printk (KERN_INFO PFX "No PST tables match this cpuid (0x%x)\n", etuple); 527 printk(KERN_INFO PFX "No PST tables match this cpuid "
503 printk (KERN_INFO PFX "This is indicative of a broken BIOS.\n"); 528 "(0x%x)\n", etuple);
529 printk(KERN_INFO PFX "This is indicative of a broken "
530 "BIOS.\n");
504 531
505 return -EINVAL; 532 return -EINVAL;
506 } 533 }
@@ -511,13 +538,14 @@ static int powernow_decode_bios (int maxfid, int startvid)
511} 538}
512 539
513 540
514static int powernow_target (struct cpufreq_policy *policy, 541static int powernow_target(struct cpufreq_policy *policy,
515 unsigned int target_freq, 542 unsigned int target_freq,
516 unsigned int relation) 543 unsigned int relation)
517{ 544{
518 unsigned int newstate; 545 unsigned int newstate;
519 546
520 if (cpufreq_frequency_table_target(policy, powernow_table, target_freq, relation, &newstate)) 547 if (cpufreq_frequency_table_target(policy, powernow_table, target_freq,
548 relation, &newstate))
521 return -EINVAL; 549 return -EINVAL;
522 550
523 change_speed(newstate); 551 change_speed(newstate);
@@ -526,7 +554,7 @@ static int powernow_target (struct cpufreq_policy *policy,
526} 554}
527 555
528 556
529static int powernow_verify (struct cpufreq_policy *policy) 557static int powernow_verify(struct cpufreq_policy *policy)
530{ 558{
531 return cpufreq_frequency_table_verify(policy, powernow_table); 559 return cpufreq_frequency_table_verify(policy, powernow_table);
532} 560}
@@ -566,18 +594,23 @@ static unsigned int powernow_get(unsigned int cpu)
566 594
567 if (cpu) 595 if (cpu)
568 return 0; 596 return 0;
569 rdmsrl (MSR_K7_FID_VID_STATUS, fidvidstatus.val); 597 rdmsrl(MSR_K7_FID_VID_STATUS, fidvidstatus.val);
570 cfid = fidvidstatus.bits.CFID; 598 cfid = fidvidstatus.bits.CFID;
571 599
572 return (fsb * fid_codes[cfid] / 10); 600 return fsb * fid_codes[cfid] / 10;
573} 601}
574 602
575 603
576static int __init acer_cpufreq_pst(const struct dmi_system_id *d) 604static int __init acer_cpufreq_pst(const struct dmi_system_id *d)
577{ 605{
578 printk(KERN_WARNING "%s laptop with broken PST tables in BIOS detected.\n", d->ident); 606 printk(KERN_WARNING PFX
579 printk(KERN_WARNING "You need to downgrade to 3A21 (09/09/2002), or try a newer BIOS than 3A71 (01/20/2003)\n"); 607 "%s laptop with broken PST tables in BIOS detected.\n",
580 printk(KERN_WARNING "cpufreq scaling has been disabled as a result of this.\n"); 608 d->ident);
609 printk(KERN_WARNING PFX
610 "You need to downgrade to 3A21 (09/09/2002), or try a newer "
611 "BIOS than 3A71 (01/20/2003)\n");
612 printk(KERN_WARNING PFX
613 "cpufreq scaling has been disabled as a result of this.\n");
581 return 0; 614 return 0;
582} 615}
583 616
@@ -598,7 +631,7 @@ static struct dmi_system_id __initdata powernow_dmi_table[] = {
598 { } 631 { }
599}; 632};
600 633
601static int __init powernow_cpu_init (struct cpufreq_policy *policy) 634static int __init powernow_cpu_init(struct cpufreq_policy *policy)
602{ 635{
603 union msr_fidvidstatus fidvidstatus; 636 union msr_fidvidstatus fidvidstatus;
604 int result; 637 int result;
@@ -606,7 +639,7 @@ static int __init powernow_cpu_init (struct cpufreq_policy *policy)
606 if (policy->cpu != 0) 639 if (policy->cpu != 0)
607 return -ENODEV; 640 return -ENODEV;
608 641
609 rdmsrl (MSR_K7_FID_VID_STATUS, fidvidstatus.val); 642 rdmsrl(MSR_K7_FID_VID_STATUS, fidvidstatus.val);
610 643
611 recalibrate_cpu_khz(); 644 recalibrate_cpu_khz();
612 645
@@ -618,19 +651,21 @@ static int __init powernow_cpu_init (struct cpufreq_policy *policy)
618 dprintk("FSB: %3dMHz\n", fsb/1000); 651 dprintk("FSB: %3dMHz\n", fsb/1000);
619 652
620 if (dmi_check_system(powernow_dmi_table) || acpi_force) { 653 if (dmi_check_system(powernow_dmi_table) || acpi_force) {
621 printk (KERN_INFO PFX "PSB/PST known to be broken. Trying ACPI instead\n"); 654 printk(KERN_INFO PFX "PSB/PST known to be broken. "
655 "Trying ACPI instead\n");
622 result = powernow_acpi_init(); 656 result = powernow_acpi_init();
623 } else { 657 } else {
624 result = powernow_decode_bios(fidvidstatus.bits.MFID, fidvidstatus.bits.SVID); 658 result = powernow_decode_bios(fidvidstatus.bits.MFID,
659 fidvidstatus.bits.SVID);
625 if (result) { 660 if (result) {
626 printk (KERN_INFO PFX "Trying ACPI perflib\n"); 661 printk(KERN_INFO PFX "Trying ACPI perflib\n");
627 maximum_speed = 0; 662 maximum_speed = 0;
628 minimum_speed = -1; 663 minimum_speed = -1;
629 latency = 0; 664 latency = 0;
630 result = powernow_acpi_init(); 665 result = powernow_acpi_init();
631 if (result) { 666 if (result) {
632 printk (KERN_INFO PFX "ACPI and legacy methods failed\n"); 667 printk(KERN_INFO PFX
633 printk (KERN_INFO PFX "See http://www.codemonkey.org.uk/projects/cpufreq/powernow-k7.html\n"); 668 "ACPI and legacy methods failed\n");
634 } 669 }
635 } else { 670 } else {
636 /* SGTC use the bus clock as timer */ 671 /* SGTC use the bus clock as timer */
@@ -642,10 +677,11 @@ static int __init powernow_cpu_init (struct cpufreq_policy *policy)
642 if (result) 677 if (result)
643 return result; 678 return result;
644 679
645 printk (KERN_INFO PFX "Minimum speed %d MHz. Maximum speed %d MHz.\n", 680 printk(KERN_INFO PFX "Minimum speed %d MHz. Maximum speed %d MHz.\n",
646 minimum_speed/1000, maximum_speed/1000); 681 minimum_speed/1000, maximum_speed/1000);
647 682
648 policy->cpuinfo.transition_latency = cpufreq_scale(2000000UL, fsb, latency); 683 policy->cpuinfo.transition_latency =
684 cpufreq_scale(2000000UL, fsb, latency);
649 685
650 policy->cur = powernow_get(0); 686 policy->cur = powernow_get(0);
651 687
@@ -654,7 +690,8 @@ static int __init powernow_cpu_init (struct cpufreq_policy *policy)
654 return cpufreq_frequency_table_cpuinfo(policy, powernow_table); 690 return cpufreq_frequency_table_cpuinfo(policy, powernow_table);
655} 691}
656 692
657static int powernow_cpu_exit (struct cpufreq_policy *policy) { 693static int powernow_cpu_exit(struct cpufreq_policy *policy)
694{
658 cpufreq_frequency_table_put_attr(policy->cpu); 695 cpufreq_frequency_table_put_attr(policy->cpu);
659 696
660#ifdef CONFIG_X86_POWERNOW_K7_ACPI 697#ifdef CONFIG_X86_POWERNOW_K7_ACPI
@@ -669,7 +706,7 @@ static int powernow_cpu_exit (struct cpufreq_policy *policy) {
669 return 0; 706 return 0;
670} 707}
671 708
672static struct freq_attr* powernow_table_attr[] = { 709static struct freq_attr *powernow_table_attr[] = {
673 &cpufreq_freq_attr_scaling_available_freqs, 710 &cpufreq_freq_attr_scaling_available_freqs,
674 NULL, 711 NULL,
675}; 712};
@@ -685,15 +722,15 @@ static struct cpufreq_driver powernow_driver = {
685 .attr = powernow_table_attr, 722 .attr = powernow_table_attr,
686}; 723};
687 724
688static int __init powernow_init (void) 725static int __init powernow_init(void)
689{ 726{
690 if (check_powernow()==0) 727 if (check_powernow() == 0)
691 return -ENODEV; 728 return -ENODEV;
692 return cpufreq_register_driver(&powernow_driver); 729 return cpufreq_register_driver(&powernow_driver);
693} 730}
694 731
695 732
696static void __exit powernow_exit (void) 733static void __exit powernow_exit(void)
697{ 734{
698 cpufreq_unregister_driver(&powernow_driver); 735 cpufreq_unregister_driver(&powernow_driver);
699} 736}
@@ -701,9 +738,9 @@ static void __exit powernow_exit (void)
701module_param(acpi_force, int, 0444); 738module_param(acpi_force, int, 0444);
702MODULE_PARM_DESC(acpi_force, "Force ACPI to be used."); 739MODULE_PARM_DESC(acpi_force, "Force ACPI to be used.");
703 740
704MODULE_AUTHOR ("Dave Jones <davej@redhat.com>"); 741MODULE_AUTHOR("Dave Jones <davej@redhat.com>");
705MODULE_DESCRIPTION ("Powernow driver for AMD K7 processors."); 742MODULE_DESCRIPTION("Powernow driver for AMD K7 processors.");
706MODULE_LICENSE ("GPL"); 743MODULE_LICENSE("GPL");
707 744
708late_initcall(powernow_init); 745late_initcall(powernow_init);
709module_exit(powernow_exit); 746module_exit(powernow_exit);
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
index 6428aa17b40e..4709ead2db52 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
@@ -33,16 +33,14 @@
33#include <linux/string.h> 33#include <linux/string.h>
34#include <linux/cpumask.h> 34#include <linux/cpumask.h>
35#include <linux/sched.h> /* for current / set_cpus_allowed() */ 35#include <linux/sched.h> /* for current / set_cpus_allowed() */
36#include <linux/io.h>
37#include <linux/delay.h>
36 38
37#include <asm/msr.h> 39#include <asm/msr.h>
38#include <asm/io.h>
39#include <asm/delay.h>
40 40
41#ifdef CONFIG_X86_POWERNOW_K8_ACPI
42#include <linux/acpi.h> 41#include <linux/acpi.h>
43#include <linux/mutex.h> 42#include <linux/mutex.h>
44#include <acpi/processor.h> 43#include <acpi/processor.h>
45#endif
46 44
47#define PFX "powernow-k8: " 45#define PFX "powernow-k8: "
48#define VERSION "version 2.20.00" 46#define VERSION "version 2.20.00"
@@ -56,7 +54,10 @@ static DEFINE_PER_CPU(struct powernow_k8_data *, powernow_data);
56static int cpu_family = CPU_OPTERON; 54static int cpu_family = CPU_OPTERON;
57 55
58#ifndef CONFIG_SMP 56#ifndef CONFIG_SMP
59DEFINE_PER_CPU(cpumask_t, cpu_core_map); 57static inline const struct cpumask *cpu_core_mask(int cpu)
58{
59 return cpumask_of(0);
60}
60#endif 61#endif
61 62
62/* Return a frequency in MHz, given an input fid */ 63/* Return a frequency in MHz, given an input fid */
@@ -71,7 +72,8 @@ static u32 find_khz_freq_from_fid(u32 fid)
71 return 1000 * find_freq_from_fid(fid); 72 return 1000 * find_freq_from_fid(fid);
72} 73}
73 74
74static u32 find_khz_freq_from_pstate(struct cpufreq_frequency_table *data, u32 pstate) 75static u32 find_khz_freq_from_pstate(struct cpufreq_frequency_table *data,
76 u32 pstate)
75{ 77{
76 return data[pstate].frequency; 78 return data[pstate].frequency;
77} 79}
@@ -186,7 +188,9 @@ static int write_new_fid(struct powernow_k8_data *data, u32 fid)
186 return 1; 188 return 1;
187 } 189 }
188 190
189 lo = fid | (data->currvid << MSR_C_LO_VID_SHIFT) | MSR_C_LO_INIT_FID_VID; 191 lo = fid;
192 lo |= (data->currvid << MSR_C_LO_VID_SHIFT);
193 lo |= MSR_C_LO_INIT_FID_VID;
190 194
191 dprintk("writing fid 0x%x, lo 0x%x, hi 0x%x\n", 195 dprintk("writing fid 0x%x, lo 0x%x, hi 0x%x\n",
192 fid, lo, data->plllock * PLL_LOCK_CONVERSION); 196 fid, lo, data->plllock * PLL_LOCK_CONVERSION);
@@ -194,7 +198,9 @@ static int write_new_fid(struct powernow_k8_data *data, u32 fid)
194 do { 198 do {
195 wrmsr(MSR_FIDVID_CTL, lo, data->plllock * PLL_LOCK_CONVERSION); 199 wrmsr(MSR_FIDVID_CTL, lo, data->plllock * PLL_LOCK_CONVERSION);
196 if (i++ > 100) { 200 if (i++ > 100) {
197 printk(KERN_ERR PFX "Hardware error - pending bit very stuck - no further pstate changes possible\n"); 201 printk(KERN_ERR PFX
202 "Hardware error - pending bit very stuck - "
203 "no further pstate changes possible\n");
198 return 1; 204 return 1;
199 } 205 }
200 } while (query_current_values_with_pending_wait(data)); 206 } while (query_current_values_with_pending_wait(data));
@@ -202,14 +208,16 @@ static int write_new_fid(struct powernow_k8_data *data, u32 fid)
202 count_off_irt(data); 208 count_off_irt(data);
203 209
204 if (savevid != data->currvid) { 210 if (savevid != data->currvid) {
205 printk(KERN_ERR PFX "vid change on fid trans, old 0x%x, new 0x%x\n", 211 printk(KERN_ERR PFX
206 savevid, data->currvid); 212 "vid change on fid trans, old 0x%x, new 0x%x\n",
213 savevid, data->currvid);
207 return 1; 214 return 1;
208 } 215 }
209 216
210 if (fid != data->currfid) { 217 if (fid != data->currfid) {
211 printk(KERN_ERR PFX "fid trans failed, fid 0x%x, curr 0x%x\n", fid, 218 printk(KERN_ERR PFX
212 data->currfid); 219 "fid trans failed, fid 0x%x, curr 0x%x\n", fid,
220 data->currfid);
213 return 1; 221 return 1;
214 } 222 }
215 223
@@ -228,7 +236,9 @@ static int write_new_vid(struct powernow_k8_data *data, u32 vid)
228 return 1; 236 return 1;
229 } 237 }
230 238
231 lo = data->currfid | (vid << MSR_C_LO_VID_SHIFT) | MSR_C_LO_INIT_FID_VID; 239 lo = data->currfid;
240 lo |= (vid << MSR_C_LO_VID_SHIFT);
241 lo |= MSR_C_LO_INIT_FID_VID;
232 242
233 dprintk("writing vid 0x%x, lo 0x%x, hi 0x%x\n", 243 dprintk("writing vid 0x%x, lo 0x%x, hi 0x%x\n",
234 vid, lo, STOP_GRANT_5NS); 244 vid, lo, STOP_GRANT_5NS);
@@ -236,20 +246,24 @@ static int write_new_vid(struct powernow_k8_data *data, u32 vid)
236 do { 246 do {
237 wrmsr(MSR_FIDVID_CTL, lo, STOP_GRANT_5NS); 247 wrmsr(MSR_FIDVID_CTL, lo, STOP_GRANT_5NS);
238 if (i++ > 100) { 248 if (i++ > 100) {
239 printk(KERN_ERR PFX "internal error - pending bit very stuck - no further pstate changes possible\n"); 249 printk(KERN_ERR PFX "internal error - pending bit "
250 "very stuck - no further pstate "
251 "changes possible\n");
240 return 1; 252 return 1;
241 } 253 }
242 } while (query_current_values_with_pending_wait(data)); 254 } while (query_current_values_with_pending_wait(data));
243 255
244 if (savefid != data->currfid) { 256 if (savefid != data->currfid) {
245 printk(KERN_ERR PFX "fid changed on vid trans, old 0x%x new 0x%x\n", 257 printk(KERN_ERR PFX "fid changed on vid trans, old "
258 "0x%x new 0x%x\n",
246 savefid, data->currfid); 259 savefid, data->currfid);
247 return 1; 260 return 1;
248 } 261 }
249 262
250 if (vid != data->currvid) { 263 if (vid != data->currvid) {
251 printk(KERN_ERR PFX "vid trans failed, vid 0x%x, curr 0x%x\n", vid, 264 printk(KERN_ERR PFX "vid trans failed, vid 0x%x, "
252 data->currvid); 265 "curr 0x%x\n",
266 vid, data->currvid);
253 return 1; 267 return 1;
254 } 268 }
255 269
@@ -261,7 +275,8 @@ static int write_new_vid(struct powernow_k8_data *data, u32 vid)
261 * Decreasing vid codes represent increasing voltages: 275 * Decreasing vid codes represent increasing voltages:
262 * vid of 0 is 1.550V, vid of 0x1e is 0.800V, vid of VID_OFF is off. 276 * vid of 0 is 1.550V, vid of 0x1e is 0.800V, vid of VID_OFF is off.
263 */ 277 */
264static int decrease_vid_code_by_step(struct powernow_k8_data *data, u32 reqvid, u32 step) 278static int decrease_vid_code_by_step(struct powernow_k8_data *data,
279 u32 reqvid, u32 step)
265{ 280{
266 if ((data->currvid - reqvid) > step) 281 if ((data->currvid - reqvid) > step)
267 reqvid = data->currvid - step; 282 reqvid = data->currvid - step;
@@ -283,7 +298,8 @@ static int transition_pstate(struct powernow_k8_data *data, u32 pstate)
283} 298}
284 299
285/* Change Opteron/Athlon64 fid and vid, by the 3 phases. */ 300/* Change Opteron/Athlon64 fid and vid, by the 3 phases. */
286static int transition_fid_vid(struct powernow_k8_data *data, u32 reqfid, u32 reqvid) 301static int transition_fid_vid(struct powernow_k8_data *data,
302 u32 reqfid, u32 reqvid)
287{ 303{
288 if (core_voltage_pre_transition(data, reqvid)) 304 if (core_voltage_pre_transition(data, reqvid))
289 return 1; 305 return 1;
@@ -298,7 +314,8 @@ static int transition_fid_vid(struct powernow_k8_data *data, u32 reqfid, u32 req
298 return 1; 314 return 1;
299 315
300 if ((reqfid != data->currfid) || (reqvid != data->currvid)) { 316 if ((reqfid != data->currfid) || (reqvid != data->currvid)) {
301 printk(KERN_ERR PFX "failed (cpu%d): req 0x%x 0x%x, curr 0x%x 0x%x\n", 317 printk(KERN_ERR PFX "failed (cpu%d): req 0x%x 0x%x, "
318 "curr 0x%x 0x%x\n",
302 smp_processor_id(), 319 smp_processor_id(),
303 reqfid, reqvid, data->currfid, data->currvid); 320 reqfid, reqvid, data->currfid, data->currvid);
304 return 1; 321 return 1;
@@ -311,13 +328,15 @@ static int transition_fid_vid(struct powernow_k8_data *data, u32 reqfid, u32 req
311} 328}
312 329
313/* Phase 1 - core voltage transition ... setup voltage */ 330/* Phase 1 - core voltage transition ... setup voltage */
314static int core_voltage_pre_transition(struct powernow_k8_data *data, u32 reqvid) 331static int core_voltage_pre_transition(struct powernow_k8_data *data,
332 u32 reqvid)
315{ 333{
316 u32 rvosteps = data->rvo; 334 u32 rvosteps = data->rvo;
317 u32 savefid = data->currfid; 335 u32 savefid = data->currfid;
318 u32 maxvid, lo; 336 u32 maxvid, lo;
319 337
320 dprintk("ph1 (cpu%d): start, currfid 0x%x, currvid 0x%x, reqvid 0x%x, rvo 0x%x\n", 338 dprintk("ph1 (cpu%d): start, currfid 0x%x, currvid 0x%x, "
339 "reqvid 0x%x, rvo 0x%x\n",
321 smp_processor_id(), 340 smp_processor_id(),
322 data->currfid, data->currvid, reqvid, data->rvo); 341 data->currfid, data->currvid, reqvid, data->rvo);
323 342
@@ -340,7 +359,7 @@ static int core_voltage_pre_transition(struct powernow_k8_data *data, u32 reqvid
340 } else { 359 } else {
341 dprintk("ph1: changing vid for rvo, req 0x%x\n", 360 dprintk("ph1: changing vid for rvo, req 0x%x\n",
342 data->currvid - 1); 361 data->currvid - 1);
343 if (decrease_vid_code_by_step(data, data->currvid - 1, 1)) 362 if (decrease_vid_code_by_step(data, data->currvid-1, 1))
344 return 1; 363 return 1;
345 rvosteps--; 364 rvosteps--;
346 } 365 }
@@ -350,7 +369,8 @@ static int core_voltage_pre_transition(struct powernow_k8_data *data, u32 reqvid
350 return 1; 369 return 1;
351 370
352 if (savefid != data->currfid) { 371 if (savefid != data->currfid) {
353 printk(KERN_ERR PFX "ph1 err, currfid changed 0x%x\n", data->currfid); 372 printk(KERN_ERR PFX "ph1 err, currfid changed 0x%x\n",
373 data->currfid);
354 return 1; 374 return 1;
355 } 375 }
356 376
@@ -363,20 +383,24 @@ static int core_voltage_pre_transition(struct powernow_k8_data *data, u32 reqvid
363/* Phase 2 - core frequency transition */ 383/* Phase 2 - core frequency transition */
364static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid) 384static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid)
365{ 385{
366 u32 vcoreqfid, vcocurrfid, vcofiddiff, fid_interval, savevid = data->currvid; 386 u32 vcoreqfid, vcocurrfid, vcofiddiff;
387 u32 fid_interval, savevid = data->currvid;
367 388
368 if ((reqfid < HI_FID_TABLE_BOTTOM) && (data->currfid < HI_FID_TABLE_BOTTOM)) { 389 if ((reqfid < HI_FID_TABLE_BOTTOM) &&
369 printk(KERN_ERR PFX "ph2: illegal lo-lo transition 0x%x 0x%x\n", 390 (data->currfid < HI_FID_TABLE_BOTTOM)) {
370 reqfid, data->currfid); 391 printk(KERN_ERR PFX "ph2: illegal lo-lo transition "
392 "0x%x 0x%x\n", reqfid, data->currfid);
371 return 1; 393 return 1;
372 } 394 }
373 395
374 if (data->currfid == reqfid) { 396 if (data->currfid == reqfid) {
375 printk(KERN_ERR PFX "ph2 null fid transition 0x%x\n", data->currfid); 397 printk(KERN_ERR PFX "ph2 null fid transition 0x%x\n",
398 data->currfid);
376 return 0; 399 return 0;
377 } 400 }
378 401
379 dprintk("ph2 (cpu%d): starting, currfid 0x%x, currvid 0x%x, reqfid 0x%x\n", 402 dprintk("ph2 (cpu%d): starting, currfid 0x%x, currvid 0x%x, "
403 "reqfid 0x%x\n",
380 smp_processor_id(), 404 smp_processor_id(),
381 data->currfid, data->currvid, reqfid); 405 data->currfid, data->currvid, reqfid);
382 406
@@ -390,14 +414,14 @@ static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid)
390 414
391 if (reqfid > data->currfid) { 415 if (reqfid > data->currfid) {
392 if (data->currfid > LO_FID_TABLE_TOP) { 416 if (data->currfid > LO_FID_TABLE_TOP) {
393 if (write_new_fid(data, data->currfid + fid_interval)) { 417 if (write_new_fid(data,
418 data->currfid + fid_interval))
394 return 1; 419 return 1;
395 }
396 } else { 420 } else {
397 if (write_new_fid 421 if (write_new_fid
398 (data, 2 + convert_fid_to_vco_fid(data->currfid))) { 422 (data,
423 2 + convert_fid_to_vco_fid(data->currfid)))
399 return 1; 424 return 1;
400 }
401 } 425 }
402 } else { 426 } else {
403 if (write_new_fid(data, data->currfid - fid_interval)) 427 if (write_new_fid(data, data->currfid - fid_interval))
@@ -417,7 +441,8 @@ static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid)
417 441
418 if (data->currfid != reqfid) { 442 if (data->currfid != reqfid) {
419 printk(KERN_ERR PFX 443 printk(KERN_ERR PFX
420 "ph2: mismatch, failed fid transition, curr 0x%x, req 0x%x\n", 444 "ph2: mismatch, failed fid transition, "
445 "curr 0x%x, req 0x%x\n",
421 data->currfid, reqfid); 446 data->currfid, reqfid);
422 return 1; 447 return 1;
423 } 448 }
@@ -435,7 +460,8 @@ static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid)
435} 460}
436 461
437/* Phase 3 - core voltage transition flow ... jump to the final vid. */ 462/* Phase 3 - core voltage transition flow ... jump to the final vid. */
438static int core_voltage_post_transition(struct powernow_k8_data *data, u32 reqvid) 463static int core_voltage_post_transition(struct powernow_k8_data *data,
464 u32 reqvid)
439{ 465{
440 u32 savefid = data->currfid; 466 u32 savefid = data->currfid;
441 u32 savereqvid = reqvid; 467 u32 savereqvid = reqvid;
@@ -457,7 +483,8 @@ static int core_voltage_post_transition(struct powernow_k8_data *data, u32 reqvi
457 483
458 if (data->currvid != reqvid) { 484 if (data->currvid != reqvid) {
459 printk(KERN_ERR PFX 485 printk(KERN_ERR PFX
460 "ph3: failed vid transition\n, req 0x%x, curr 0x%x", 486 "ph3: failed vid transition\n, "
487 "req 0x%x, curr 0x%x",
461 reqvid, data->currvid); 488 reqvid, data->currvid);
462 return 1; 489 return 1;
463 } 490 }
@@ -508,7 +535,8 @@ static int check_supported_cpu(unsigned int cpu)
508 if ((eax & CPUID_XFAM) == CPUID_XFAM_K8) { 535 if ((eax & CPUID_XFAM) == CPUID_XFAM_K8) {
509 if (((eax & CPUID_USE_XFAM_XMOD) != CPUID_USE_XFAM_XMOD) || 536 if (((eax & CPUID_USE_XFAM_XMOD) != CPUID_USE_XFAM_XMOD) ||
510 ((eax & CPUID_XMOD) > CPUID_XMOD_REV_MASK)) { 537 ((eax & CPUID_XMOD) > CPUID_XMOD_REV_MASK)) {
511 printk(KERN_INFO PFX "Processor cpuid %x not supported\n", eax); 538 printk(KERN_INFO PFX
539 "Processor cpuid %x not supported\n", eax);
512 goto out; 540 goto out;
513 } 541 }
514 542
@@ -520,8 +548,10 @@ static int check_supported_cpu(unsigned int cpu)
520 } 548 }
521 549
522 cpuid(CPUID_FREQ_VOLT_CAPABILITIES, &eax, &ebx, &ecx, &edx); 550 cpuid(CPUID_FREQ_VOLT_CAPABILITIES, &eax, &ebx, &ecx, &edx);
523 if ((edx & P_STATE_TRANSITION_CAPABLE) != P_STATE_TRANSITION_CAPABLE) { 551 if ((edx & P_STATE_TRANSITION_CAPABLE)
524 printk(KERN_INFO PFX "Power state transitions not supported\n"); 552 != P_STATE_TRANSITION_CAPABLE) {
553 printk(KERN_INFO PFX
554 "Power state transitions not supported\n");
525 goto out; 555 goto out;
526 } 556 }
527 } else { /* must be a HW Pstate capable processor */ 557 } else { /* must be a HW Pstate capable processor */
@@ -539,7 +569,8 @@ out:
539 return rc; 569 return rc;
540} 570}
541 571
542static int check_pst_table(struct powernow_k8_data *data, struct pst_s *pst, u8 maxvid) 572static int check_pst_table(struct powernow_k8_data *data, struct pst_s *pst,
573 u8 maxvid)
543{ 574{
544 unsigned int j; 575 unsigned int j;
545 u8 lastfid = 0xff; 576 u8 lastfid = 0xff;
@@ -550,12 +581,14 @@ static int check_pst_table(struct powernow_k8_data *data, struct pst_s *pst, u8
550 j, pst[j].vid); 581 j, pst[j].vid);
551 return -EINVAL; 582 return -EINVAL;
552 } 583 }
553 if (pst[j].vid < data->rvo) { /* vid + rvo >= 0 */ 584 if (pst[j].vid < data->rvo) {
585 /* vid + rvo >= 0 */
554 printk(KERN_ERR FW_BUG PFX "0 vid exceeded with pstate" 586 printk(KERN_ERR FW_BUG PFX "0 vid exceeded with pstate"
555 " %d\n", j); 587 " %d\n", j);
556 return -ENODEV; 588 return -ENODEV;
557 } 589 }
558 if (pst[j].vid < maxvid + data->rvo) { /* vid + rvo >= maxvid */ 590 if (pst[j].vid < maxvid + data->rvo) {
591 /* vid + rvo >= maxvid */
559 printk(KERN_ERR FW_BUG PFX "maxvid exceeded with pstate" 592 printk(KERN_ERR FW_BUG PFX "maxvid exceeded with pstate"
560 " %d\n", j); 593 " %d\n", j);
561 return -ENODEV; 594 return -ENODEV;
@@ -579,23 +612,31 @@ static int check_pst_table(struct powernow_k8_data *data, struct pst_s *pst, u8
579 return -EINVAL; 612 return -EINVAL;
580 } 613 }
581 if (lastfid > LO_FID_TABLE_TOP) 614 if (lastfid > LO_FID_TABLE_TOP)
582 printk(KERN_INFO FW_BUG PFX "first fid not from lo freq table\n"); 615 printk(KERN_INFO FW_BUG PFX
616 "first fid not from lo freq table\n");
583 617
584 return 0; 618 return 0;
585} 619}
586 620
621static void invalidate_entry(struct powernow_k8_data *data, unsigned int entry)
622{
623 data->powernow_table[entry].frequency = CPUFREQ_ENTRY_INVALID;
624}
625
587static void print_basics(struct powernow_k8_data *data) 626static void print_basics(struct powernow_k8_data *data)
588{ 627{
589 int j; 628 int j;
590 for (j = 0; j < data->numps; j++) { 629 for (j = 0; j < data->numps; j++) {
591 if (data->powernow_table[j].frequency != CPUFREQ_ENTRY_INVALID) { 630 if (data->powernow_table[j].frequency !=
631 CPUFREQ_ENTRY_INVALID) {
592 if (cpu_family == CPU_HW_PSTATE) { 632 if (cpu_family == CPU_HW_PSTATE) {
593 printk(KERN_INFO PFX " %d : pstate %d (%d MHz)\n", 633 printk(KERN_INFO PFX
594 j, 634 " %d : pstate %d (%d MHz)\n", j,
595 data->powernow_table[j].index, 635 data->powernow_table[j].index,
596 data->powernow_table[j].frequency/1000); 636 data->powernow_table[j].frequency/1000);
597 } else { 637 } else {
598 printk(KERN_INFO PFX " %d : fid 0x%x (%d MHz), vid 0x%x\n", 638 printk(KERN_INFO PFX
639 " %d : fid 0x%x (%d MHz), vid 0x%x\n",
599 j, 640 j,
600 data->powernow_table[j].index & 0xff, 641 data->powernow_table[j].index & 0xff,
601 data->powernow_table[j].frequency/1000, 642 data->powernow_table[j].frequency/1000,
@@ -604,20 +645,25 @@ static void print_basics(struct powernow_k8_data *data)
604 } 645 }
605 } 646 }
606 if (data->batps) 647 if (data->batps)
607 printk(KERN_INFO PFX "Only %d pstates on battery\n", data->batps); 648 printk(KERN_INFO PFX "Only %d pstates on battery\n",
649 data->batps);
608} 650}
609 651
610static int fill_powernow_table(struct powernow_k8_data *data, struct pst_s *pst, u8 maxvid) 652static int fill_powernow_table(struct powernow_k8_data *data,
653 struct pst_s *pst, u8 maxvid)
611{ 654{
612 struct cpufreq_frequency_table *powernow_table; 655 struct cpufreq_frequency_table *powernow_table;
613 unsigned int j; 656 unsigned int j;
614 657
615 if (data->batps) { /* use ACPI support to get full speed on mains power */ 658 if (data->batps) {
616 printk(KERN_WARNING PFX "Only %d pstates usable (use ACPI driver for full range\n", data->batps); 659 /* use ACPI support to get full speed on mains power */
660 printk(KERN_WARNING PFX
661 "Only %d pstates usable (use ACPI driver for full "
662 "range\n", data->batps);
617 data->numps = data->batps; 663 data->numps = data->batps;
618 } 664 }
619 665
620 for ( j=1; j<data->numps; j++ ) { 666 for (j = 1; j < data->numps; j++) {
621 if (pst[j-1].fid >= pst[j].fid) { 667 if (pst[j-1].fid >= pst[j].fid) {
622 printk(KERN_ERR PFX "PST out of sequence\n"); 668 printk(KERN_ERR PFX "PST out of sequence\n");
623 return -EINVAL; 669 return -EINVAL;
@@ -640,9 +686,11 @@ static int fill_powernow_table(struct powernow_k8_data *data, struct pst_s *pst,
640 } 686 }
641 687
642 for (j = 0; j < data->numps; j++) { 688 for (j = 0; j < data->numps; j++) {
689 int freq;
643 powernow_table[j].index = pst[j].fid; /* lower 8 bits */ 690 powernow_table[j].index = pst[j].fid; /* lower 8 bits */
644 powernow_table[j].index |= (pst[j].vid << 8); /* upper 8 bits */ 691 powernow_table[j].index |= (pst[j].vid << 8); /* upper 8 bits */
645 powernow_table[j].frequency = find_khz_freq_from_fid(pst[j].fid); 692 freq = find_khz_freq_from_fid(pst[j].fid);
693 powernow_table[j].frequency = freq;
646 } 694 }
647 powernow_table[data->numps].frequency = CPUFREQ_TABLE_END; 695 powernow_table[data->numps].frequency = CPUFREQ_TABLE_END;
648 powernow_table[data->numps].index = 0; 696 powernow_table[data->numps].index = 0;
@@ -654,11 +702,12 @@ static int fill_powernow_table(struct powernow_k8_data *data, struct pst_s *pst,
654 702
655 dprintk("cfid 0x%x, cvid 0x%x\n", data->currfid, data->currvid); 703 dprintk("cfid 0x%x, cvid 0x%x\n", data->currfid, data->currvid);
656 data->powernow_table = powernow_table; 704 data->powernow_table = powernow_table;
657 if (first_cpu(per_cpu(cpu_core_map, data->cpu)) == data->cpu) 705 if (cpumask_first(cpu_core_mask(data->cpu)) == data->cpu)
658 print_basics(data); 706 print_basics(data);
659 707
660 for (j = 0; j < data->numps; j++) 708 for (j = 0; j < data->numps; j++)
661 if ((pst[j].fid==data->currfid) && (pst[j].vid==data->currvid)) 709 if ((pst[j].fid == data->currfid) &&
710 (pst[j].vid == data->currvid))
662 return 0; 711 return 0;
663 712
664 dprintk("currfid/vid do not match PST, ignoring\n"); 713 dprintk("currfid/vid do not match PST, ignoring\n");
@@ -698,7 +747,8 @@ static int find_psb_table(struct powernow_k8_data *data)
698 } 747 }
699 748
700 data->vstable = psb->vstable; 749 data->vstable = psb->vstable;
701 dprintk("voltage stabilization time: %d(*20us)\n", data->vstable); 750 dprintk("voltage stabilization time: %d(*20us)\n",
751 data->vstable);
702 752
703 dprintk("flags2: 0x%x\n", psb->flags2); 753 dprintk("flags2: 0x%x\n", psb->flags2);
704 data->rvo = psb->flags2 & 3; 754 data->rvo = psb->flags2 & 3;
@@ -713,11 +763,12 @@ static int find_psb_table(struct powernow_k8_data *data)
713 763
714 dprintk("numpst: 0x%x\n", psb->num_tables); 764 dprintk("numpst: 0x%x\n", psb->num_tables);
715 cpst = psb->num_tables; 765 cpst = psb->num_tables;
716 if ((psb->cpuid == 0x00000fc0) || (psb->cpuid == 0x00000fe0) ){ 766 if ((psb->cpuid == 0x00000fc0) ||
767 (psb->cpuid == 0x00000fe0)) {
717 thiscpuid = cpuid_eax(CPUID_PROCESSOR_SIGNATURE); 768 thiscpuid = cpuid_eax(CPUID_PROCESSOR_SIGNATURE);
718 if ((thiscpuid == 0x00000fc0) || (thiscpuid == 0x00000fe0) ) { 769 if ((thiscpuid == 0x00000fc0) ||
770 (thiscpuid == 0x00000fe0))
719 cpst = 1; 771 cpst = 1;
720 }
721 } 772 }
722 if (cpst != 1) { 773 if (cpst != 1) {
723 printk(KERN_ERR FW_BUG PFX "numpst must be 1\n"); 774 printk(KERN_ERR FW_BUG PFX "numpst must be 1\n");
@@ -732,7 +783,8 @@ static int find_psb_table(struct powernow_k8_data *data)
732 783
733 data->numps = psb->numps; 784 data->numps = psb->numps;
734 dprintk("numpstates: 0x%x\n", data->numps); 785 dprintk("numpstates: 0x%x\n", data->numps);
735 return fill_powernow_table(data, (struct pst_s *)(psb+1), maxvid); 786 return fill_powernow_table(data,
787 (struct pst_s *)(psb+1), maxvid);
736 } 788 }
737 /* 789 /*
738 * If you see this message, complain to BIOS manufacturer. If 790 * If you see this message, complain to BIOS manufacturer. If
@@ -745,28 +797,31 @@ static int find_psb_table(struct powernow_k8_data *data)
745 * BIOS and Kernel Developer's Guide, which is available on 797 * BIOS and Kernel Developer's Guide, which is available on
746 * www.amd.com 798 * www.amd.com
747 */ 799 */
748 printk(KERN_ERR PFX "BIOS error - no PSB or ACPI _PSS objects\n"); 800 printk(KERN_ERR FW_BUG PFX "No PSB or ACPI _PSS objects\n");
749 return -ENODEV; 801 return -ENODEV;
750} 802}
751 803
752#ifdef CONFIG_X86_POWERNOW_K8_ACPI 804static void powernow_k8_acpi_pst_values(struct powernow_k8_data *data,
753static void powernow_k8_acpi_pst_values(struct powernow_k8_data *data, unsigned int index) 805 unsigned int index)
754{ 806{
807 acpi_integer control;
808
755 if (!data->acpi_data.state_count || (cpu_family == CPU_HW_PSTATE)) 809 if (!data->acpi_data.state_count || (cpu_family == CPU_HW_PSTATE))
756 return; 810 return;
757 811
758 data->irt = (data->acpi_data.states[index].control >> IRT_SHIFT) & IRT_MASK; 812 control = data->acpi_data.states[index].control; data->irt = (control
759 data->rvo = (data->acpi_data.states[index].control >> RVO_SHIFT) & RVO_MASK; 813 >> IRT_SHIFT) & IRT_MASK; data->rvo = (control >>
760 data->exttype = (data->acpi_data.states[index].control >> EXT_TYPE_SHIFT) & EXT_TYPE_MASK; 814 RVO_SHIFT) & RVO_MASK; data->exttype = (control
761 data->plllock = (data->acpi_data.states[index].control >> PLL_L_SHIFT) & PLL_L_MASK; 815 >> EXT_TYPE_SHIFT) & EXT_TYPE_MASK;
762 data->vidmvs = 1 << ((data->acpi_data.states[index].control >> MVS_SHIFT) & MVS_MASK); 816 data->plllock = (control >> PLL_L_SHIFT) & PLL_L_MASK; data->vidmvs = 1
763 data->vstable = (data->acpi_data.states[index].control >> VST_SHIFT) & VST_MASK; 817 << ((control >> MVS_SHIFT) & MVS_MASK); data->vstable =
764} 818 (control >> VST_SHIFT) & VST_MASK; }
765 819
766static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data) 820static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data)
767{ 821{
768 struct cpufreq_frequency_table *powernow_table; 822 struct cpufreq_frequency_table *powernow_table;
769 int ret_val = -ENODEV; 823 int ret_val = -ENODEV;
824 acpi_integer space_id;
770 825
771 if (acpi_processor_register_performance(&data->acpi_data, data->cpu)) { 826 if (acpi_processor_register_performance(&data->acpi_data, data->cpu)) {
772 dprintk("register performance failed: bad ACPI data\n"); 827 dprintk("register performance failed: bad ACPI data\n");
@@ -779,11 +834,12 @@ static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data)
779 goto err_out; 834 goto err_out;
780 } 835 }
781 836
782 if ((data->acpi_data.control_register.space_id != ACPI_ADR_SPACE_FIXED_HARDWARE) || 837 space_id = data->acpi_data.control_register.space_id;
783 (data->acpi_data.status_register.space_id != ACPI_ADR_SPACE_FIXED_HARDWARE)) { 838 if ((space_id != ACPI_ADR_SPACE_FIXED_HARDWARE) ||
839 (space_id != ACPI_ADR_SPACE_FIXED_HARDWARE)) {
784 dprintk("Invalid control/status registers (%x - %x)\n", 840 dprintk("Invalid control/status registers (%x - %x)\n",
785 data->acpi_data.control_register.space_id, 841 data->acpi_data.control_register.space_id,
786 data->acpi_data.status_register.space_id); 842 space_id);
787 goto err_out; 843 goto err_out;
788 } 844 }
789 845
@@ -802,13 +858,14 @@ static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data)
802 if (ret_val) 858 if (ret_val)
803 goto err_out_mem; 859 goto err_out_mem;
804 860
805 powernow_table[data->acpi_data.state_count].frequency = CPUFREQ_TABLE_END; 861 powernow_table[data->acpi_data.state_count].frequency =
862 CPUFREQ_TABLE_END;
806 powernow_table[data->acpi_data.state_count].index = 0; 863 powernow_table[data->acpi_data.state_count].index = 0;
807 data->powernow_table = powernow_table; 864 data->powernow_table = powernow_table;
808 865
809 /* fill in data */ 866 /* fill in data */
810 data->numps = data->acpi_data.state_count; 867 data->numps = data->acpi_data.state_count;
811 if (first_cpu(per_cpu(cpu_core_map, data->cpu)) == data->cpu) 868 if (cpumask_first(cpu_core_mask(data->cpu)) == data->cpu)
812 print_basics(data); 869 print_basics(data);
813 powernow_k8_acpi_pst_values(data, 0); 870 powernow_k8_acpi_pst_values(data, 0);
814 871
@@ -830,13 +887,15 @@ err_out_mem:
830err_out: 887err_out:
831 acpi_processor_unregister_performance(&data->acpi_data, data->cpu); 888 acpi_processor_unregister_performance(&data->acpi_data, data->cpu);
832 889
833 /* data->acpi_data.state_count informs us at ->exit() whether ACPI was used */ 890 /* data->acpi_data.state_count informs us at ->exit()
891 * whether ACPI was used */
834 data->acpi_data.state_count = 0; 892 data->acpi_data.state_count = 0;
835 893
836 return ret_val; 894 return ret_val;
837} 895}
838 896
839static int fill_powernow_table_pstate(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table) 897static int fill_powernow_table_pstate(struct powernow_k8_data *data,
898 struct cpufreq_frequency_table *powernow_table)
840{ 899{
841 int i; 900 int i;
842 u32 hi = 0, lo = 0; 901 u32 hi = 0, lo = 0;
@@ -848,84 +907,101 @@ static int fill_powernow_table_pstate(struct powernow_k8_data *data, struct cpuf
848 907
849 index = data->acpi_data.states[i].control & HW_PSTATE_MASK; 908 index = data->acpi_data.states[i].control & HW_PSTATE_MASK;
850 if (index > data->max_hw_pstate) { 909 if (index > data->max_hw_pstate) {
851 printk(KERN_ERR PFX "invalid pstate %d - bad value %d.\n", i, index); 910 printk(KERN_ERR PFX "invalid pstate %d - "
852 printk(KERN_ERR PFX "Please report to BIOS manufacturer\n"); 911 "bad value %d.\n", i, index);
853 powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID; 912 printk(KERN_ERR PFX "Please report to BIOS "
913 "manufacturer\n");
914 invalidate_entry(data, i);
854 continue; 915 continue;
855 } 916 }
856 rdmsr(MSR_PSTATE_DEF_BASE + index, lo, hi); 917 rdmsr(MSR_PSTATE_DEF_BASE + index, lo, hi);
857 if (!(hi & HW_PSTATE_VALID_MASK)) { 918 if (!(hi & HW_PSTATE_VALID_MASK)) {
858 dprintk("invalid pstate %d, ignoring\n", index); 919 dprintk("invalid pstate %d, ignoring\n", index);
859 powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID; 920 invalidate_entry(data, i);
860 continue; 921 continue;
861 } 922 }
862 923
863 powernow_table[i].index = index; 924 powernow_table[i].index = index;
864 925
865 powernow_table[i].frequency = data->acpi_data.states[i].core_frequency * 1000; 926 powernow_table[i].frequency =
927 data->acpi_data.states[i].core_frequency * 1000;
866 } 928 }
867 return 0; 929 return 0;
868} 930}
869 931
870static int fill_powernow_table_fidvid(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table) 932static int fill_powernow_table_fidvid(struct powernow_k8_data *data,
933 struct cpufreq_frequency_table *powernow_table)
871{ 934{
872 int i; 935 int i;
873 int cntlofreq = 0; 936 int cntlofreq = 0;
937
874 for (i = 0; i < data->acpi_data.state_count; i++) { 938 for (i = 0; i < data->acpi_data.state_count; i++) {
875 u32 fid; 939 u32 fid;
876 u32 vid; 940 u32 vid;
941 u32 freq, index;
942 acpi_integer status, control;
877 943
878 if (data->exttype) { 944 if (data->exttype) {
879 fid = data->acpi_data.states[i].status & EXT_FID_MASK; 945 status = data->acpi_data.states[i].status;
880 vid = (data->acpi_data.states[i].status >> VID_SHIFT) & EXT_VID_MASK; 946 fid = status & EXT_FID_MASK;
947 vid = (status >> VID_SHIFT) & EXT_VID_MASK;
881 } else { 948 } else {
882 fid = data->acpi_data.states[i].control & FID_MASK; 949 control = data->acpi_data.states[i].control;
883 vid = (data->acpi_data.states[i].control >> VID_SHIFT) & VID_MASK; 950 fid = control & FID_MASK;
951 vid = (control >> VID_SHIFT) & VID_MASK;
884 } 952 }
885 953
886 dprintk(" %d : fid 0x%x, vid 0x%x\n", i, fid, vid); 954 dprintk(" %d : fid 0x%x, vid 0x%x\n", i, fid, vid);
887 955
888 powernow_table[i].index = fid; /* lower 8 bits */ 956 index = fid | (vid<<8);
889 powernow_table[i].index |= (vid << 8); /* upper 8 bits */ 957 powernow_table[i].index = index;
890 powernow_table[i].frequency = find_khz_freq_from_fid(fid); 958
959 freq = find_khz_freq_from_fid(fid);
960 powernow_table[i].frequency = freq;
891 961
892 /* verify frequency is OK */ 962 /* verify frequency is OK */
893 if ((powernow_table[i].frequency > (MAX_FREQ * 1000)) || 963 if ((freq > (MAX_FREQ * 1000)) || (freq < (MIN_FREQ * 1000))) {
894 (powernow_table[i].frequency < (MIN_FREQ * 1000))) { 964 dprintk("invalid freq %u kHz, ignoring\n", freq);
895 dprintk("invalid freq %u kHz, ignoring\n", powernow_table[i].frequency); 965 invalidate_entry(data, i);
896 powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
897 continue; 966 continue;
898 } 967 }
899 968
900 /* verify voltage is OK - BIOSs are using "off" to indicate invalid */ 969 /* verify voltage is OK -
970 * BIOSs are using "off" to indicate invalid */
901 if (vid == VID_OFF) { 971 if (vid == VID_OFF) {
902 dprintk("invalid vid %u, ignoring\n", vid); 972 dprintk("invalid vid %u, ignoring\n", vid);
903 powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID; 973 invalidate_entry(data, i);
904 continue; 974 continue;
905 } 975 }
906 976
907 /* verify only 1 entry from the lo frequency table */ 977 /* verify only 1 entry from the lo frequency table */
908 if (fid < HI_FID_TABLE_BOTTOM) { 978 if (fid < HI_FID_TABLE_BOTTOM) {
909 if (cntlofreq) { 979 if (cntlofreq) {
910 /* if both entries are the same, ignore this one ... */ 980 /* if both entries are the same,
911 if ((powernow_table[i].frequency != powernow_table[cntlofreq].frequency) || 981 * ignore this one ... */
912 (powernow_table[i].index != powernow_table[cntlofreq].index)) { 982 if ((freq != powernow_table[cntlofreq].frequency) ||
913 printk(KERN_ERR PFX "Too many lo freq table entries\n"); 983 (index != powernow_table[cntlofreq].index)) {
984 printk(KERN_ERR PFX
985 "Too many lo freq table "
986 "entries\n");
914 return 1; 987 return 1;
915 } 988 }
916 989
917 dprintk("double low frequency table entry, ignoring it.\n"); 990 dprintk("double low frequency table entry, "
918 powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID; 991 "ignoring it.\n");
992 invalidate_entry(data, i);
919 continue; 993 continue;
920 } else 994 } else
921 cntlofreq = i; 995 cntlofreq = i;
922 } 996 }
923 997
924 if (powernow_table[i].frequency != (data->acpi_data.states[i].core_frequency * 1000)) { 998 if (freq != (data->acpi_data.states[i].core_frequency * 1000)) {
925 printk(KERN_INFO PFX "invalid freq entries %u kHz vs. %u kHz\n", 999 printk(KERN_INFO PFX "invalid freq entries "
926 powernow_table[i].frequency, 1000 "%u kHz vs. %u kHz\n", freq,
927 (unsigned int) (data->acpi_data.states[i].core_frequency * 1000)); 1001 (unsigned int)
928 powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID; 1002 (data->acpi_data.states[i].core_frequency
1003 * 1000));
1004 invalidate_entry(data, i);
929 continue; 1005 continue;
930 } 1006 }
931 } 1007 }
@@ -935,7 +1011,8 @@ static int fill_powernow_table_fidvid(struct powernow_k8_data *data, struct cpuf
935static void powernow_k8_cpu_exit_acpi(struct powernow_k8_data *data) 1011static void powernow_k8_cpu_exit_acpi(struct powernow_k8_data *data)
936{ 1012{
937 if (data->acpi_data.state_count) 1013 if (data->acpi_data.state_count)
938 acpi_processor_unregister_performance(&data->acpi_data, data->cpu); 1014 acpi_processor_unregister_performance(&data->acpi_data,
1015 data->cpu);
939 free_cpumask_var(data->acpi_data.shared_cpu_map); 1016 free_cpumask_var(data->acpi_data.shared_cpu_map);
940} 1017}
941 1018
@@ -953,15 +1030,9 @@ static int get_transition_latency(struct powernow_k8_data *data)
953 return 1000 * max_latency; 1030 return 1000 * max_latency;
954} 1031}
955 1032
956#else
957static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data) { return -ENODEV; }
958static void powernow_k8_cpu_exit_acpi(struct powernow_k8_data *data) { return; }
959static void powernow_k8_acpi_pst_values(struct powernow_k8_data *data, unsigned int index) { return; }
960static int get_transition_latency(struct powernow_k8_data *data) { return 0; }
961#endif /* CONFIG_X86_POWERNOW_K8_ACPI */
962
963/* Take a frequency, and issue the fid/vid transition command */ 1033/* Take a frequency, and issue the fid/vid transition command */
964static int transition_frequency_fidvid(struct powernow_k8_data *data, unsigned int index) 1034static int transition_frequency_fidvid(struct powernow_k8_data *data,
1035 unsigned int index)
965{ 1036{
966 u32 fid = 0; 1037 u32 fid = 0;
967 u32 vid = 0; 1038 u32 vid = 0;
@@ -989,7 +1060,8 @@ static int transition_frequency_fidvid(struct powernow_k8_data *data, unsigned i
989 return 0; 1060 return 0;
990 } 1061 }
991 1062
992 if ((fid < HI_FID_TABLE_BOTTOM) && (data->currfid < HI_FID_TABLE_BOTTOM)) { 1063 if ((fid < HI_FID_TABLE_BOTTOM) &&
1064 (data->currfid < HI_FID_TABLE_BOTTOM)) {
993 printk(KERN_ERR PFX 1065 printk(KERN_ERR PFX
994 "ignoring illegal change in lo freq table-%x to 0x%x\n", 1066 "ignoring illegal change in lo freq table-%x to 0x%x\n",
995 data->currfid, fid); 1067 data->currfid, fid);
@@ -1017,7 +1089,8 @@ static int transition_frequency_fidvid(struct powernow_k8_data *data, unsigned i
1017} 1089}
1018 1090
1019/* Take a frequency, and issue the hardware pstate transition command */ 1091/* Take a frequency, and issue the hardware pstate transition command */
1020static int transition_frequency_pstate(struct powernow_k8_data *data, unsigned int index) 1092static int transition_frequency_pstate(struct powernow_k8_data *data,
1093 unsigned int index)
1021{ 1094{
1022 u32 pstate = 0; 1095 u32 pstate = 0;
1023 int res, i; 1096 int res, i;
@@ -1029,7 +1102,8 @@ static int transition_frequency_pstate(struct powernow_k8_data *data, unsigned i
1029 pstate = index & HW_PSTATE_MASK; 1102 pstate = index & HW_PSTATE_MASK;
1030 if (pstate > data->max_hw_pstate) 1103 if (pstate > data->max_hw_pstate)
1031 return 0; 1104 return 0;
1032 freqs.old = find_khz_freq_from_pstate(data->powernow_table, data->currpstate); 1105 freqs.old = find_khz_freq_from_pstate(data->powernow_table,
1106 data->currpstate);
1033 freqs.new = find_khz_freq_from_pstate(data->powernow_table, pstate); 1107 freqs.new = find_khz_freq_from_pstate(data->powernow_table, pstate);
1034 1108
1035 for_each_cpu_mask_nr(i, *(data->available_cores)) { 1109 for_each_cpu_mask_nr(i, *(data->available_cores)) {
@@ -1048,7 +1122,8 @@ static int transition_frequency_pstate(struct powernow_k8_data *data, unsigned i
1048} 1122}
1049 1123
1050/* Driver entry point to switch to the target frequency */ 1124/* Driver entry point to switch to the target frequency */
1051static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsigned relation) 1125static int powernowk8_target(struct cpufreq_policy *pol,
1126 unsigned targfreq, unsigned relation)
1052{ 1127{
1053 cpumask_t oldmask; 1128 cpumask_t oldmask;
1054 struct powernow_k8_data *data = per_cpu(powernow_data, pol->cpu); 1129 struct powernow_k8_data *data = per_cpu(powernow_data, pol->cpu);
@@ -1087,14 +1162,18 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi
1087 dprintk("targ: curr fid 0x%x, vid 0x%x\n", 1162 dprintk("targ: curr fid 0x%x, vid 0x%x\n",
1088 data->currfid, data->currvid); 1163 data->currfid, data->currvid);
1089 1164
1090 if ((checkvid != data->currvid) || (checkfid != data->currfid)) { 1165 if ((checkvid != data->currvid) ||
1166 (checkfid != data->currfid)) {
1091 printk(KERN_INFO PFX 1167 printk(KERN_INFO PFX
1092 "error - out of sync, fix 0x%x 0x%x, vid 0x%x 0x%x\n", 1168 "error - out of sync, fix 0x%x 0x%x, "
1093 checkfid, data->currfid, checkvid, data->currvid); 1169 "vid 0x%x 0x%x\n",
1170 checkfid, data->currfid,
1171 checkvid, data->currvid);
1094 } 1172 }
1095 } 1173 }
1096 1174
1097 if (cpufreq_frequency_table_target(pol, data->powernow_table, targfreq, relation, &newstate)) 1175 if (cpufreq_frequency_table_target(pol, data->powernow_table,
1176 targfreq, relation, &newstate))
1098 goto err_out; 1177 goto err_out;
1099 1178
1100 mutex_lock(&fidvid_mutex); 1179 mutex_lock(&fidvid_mutex);
@@ -1114,7 +1193,8 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi
1114 mutex_unlock(&fidvid_mutex); 1193 mutex_unlock(&fidvid_mutex);
1115 1194
1116 if (cpu_family == CPU_HW_PSTATE) 1195 if (cpu_family == CPU_HW_PSTATE)
1117 pol->cur = find_khz_freq_from_pstate(data->powernow_table, newstate); 1196 pol->cur = find_khz_freq_from_pstate(data->powernow_table,
1197 newstate);
1118 else 1198 else
1119 pol->cur = find_khz_freq_from_fid(data->currfid); 1199 pol->cur = find_khz_freq_from_fid(data->currfid);
1120 ret = 0; 1200 ret = 0;
@@ -1141,6 +1221,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1141 struct powernow_k8_data *data; 1221 struct powernow_k8_data *data;
1142 cpumask_t oldmask; 1222 cpumask_t oldmask;
1143 int rc; 1223 int rc;
1224 static int print_once;
1144 1225
1145 if (!cpu_online(pol->cpu)) 1226 if (!cpu_online(pol->cpu))
1146 return -ENODEV; 1227 return -ENODEV;
@@ -1163,33 +1244,31 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1163 * an UP version, and is deprecated by AMD. 1244 * an UP version, and is deprecated by AMD.
1164 */ 1245 */
1165 if (num_online_cpus() != 1) { 1246 if (num_online_cpus() != 1) {
1166#ifndef CONFIG_ACPI_PROCESSOR 1247 /*
1167 printk(KERN_ERR PFX "ACPI Processor support is required " 1248 * Replace this one with print_once as soon as such a
1168 "for SMP systems but is absent. Please load the " 1249 * thing gets introduced
1169 "ACPI Processor module before starting this " 1250 */
1170 "driver.\n"); 1251 if (!print_once) {
1171#else 1252 WARN_ONCE(1, KERN_ERR FW_BUG PFX "Your BIOS "
1172 printk(KERN_ERR FW_BUG PFX "Your BIOS does not provide" 1253 "does not provide ACPI _PSS objects "
1173 " ACPI _PSS objects in a way that Linux " 1254 "in a way that Linux understands. "
1174 "understands. Please report this to the Linux " 1255 "Please report this to the Linux ACPI"
1175 "ACPI maintainers and complain to your BIOS " 1256 " maintainers and complain to your "
1176 "vendor.\n"); 1257 "BIOS vendor.\n");
1177#endif 1258 print_once++;
1178 kfree(data); 1259 }
1179 return -ENODEV; 1260 goto err_out;
1180 } 1261 }
1181 if (pol->cpu != 0) { 1262 if (pol->cpu != 0) {
1182 printk(KERN_ERR FW_BUG PFX "No ACPI _PSS objects for " 1263 printk(KERN_ERR FW_BUG PFX "No ACPI _PSS objects for "
1183 "CPU other than CPU0. Complain to your BIOS " 1264 "CPU other than CPU0. Complain to your BIOS "
1184 "vendor.\n"); 1265 "vendor.\n");
1185 kfree(data); 1266 goto err_out;
1186 return -ENODEV;
1187 } 1267 }
1188 rc = find_psb_table(data); 1268 rc = find_psb_table(data);
1189 if (rc) { 1269 if (rc)
1190 kfree(data); 1270 goto err_out;
1191 return -ENODEV; 1271
1192 }
1193 /* Take a crude guess here. 1272 /* Take a crude guess here.
1194 * That guess was in microseconds, so multiply with 1000 */ 1273 * That guess was in microseconds, so multiply with 1000 */
1195 pol->cpuinfo.transition_latency = ( 1274 pol->cpuinfo.transition_latency = (
@@ -1204,16 +1283,16 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1204 1283
1205 if (smp_processor_id() != pol->cpu) { 1284 if (smp_processor_id() != pol->cpu) {
1206 printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu); 1285 printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu);
1207 goto err_out; 1286 goto err_out_unmask;
1208 } 1287 }
1209 1288
1210 if (pending_bit_stuck()) { 1289 if (pending_bit_stuck()) {
1211 printk(KERN_ERR PFX "failing init, change pending bit set\n"); 1290 printk(KERN_ERR PFX "failing init, change pending bit set\n");
1212 goto err_out; 1291 goto err_out_unmask;
1213 } 1292 }
1214 1293
1215 if (query_current_values_with_pending_wait(data)) 1294 if (query_current_values_with_pending_wait(data))
1216 goto err_out; 1295 goto err_out_unmask;
1217 1296
1218 if (cpu_family == CPU_OPTERON) 1297 if (cpu_family == CPU_OPTERON)
1219 fidvid_msr_init(); 1298 fidvid_msr_init();
@@ -1224,11 +1303,12 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1224 if (cpu_family == CPU_HW_PSTATE) 1303 if (cpu_family == CPU_HW_PSTATE)
1225 cpumask_copy(pol->cpus, cpumask_of(pol->cpu)); 1304 cpumask_copy(pol->cpus, cpumask_of(pol->cpu));
1226 else 1305 else
1227 cpumask_copy(pol->cpus, &per_cpu(cpu_core_map, pol->cpu)); 1306 cpumask_copy(pol->cpus, cpu_core_mask(pol->cpu));
1228 data->available_cores = pol->cpus; 1307 data->available_cores = pol->cpus;
1229 1308
1230 if (cpu_family == CPU_HW_PSTATE) 1309 if (cpu_family == CPU_HW_PSTATE)
1231 pol->cur = find_khz_freq_from_pstate(data->powernow_table, data->currpstate); 1310 pol->cur = find_khz_freq_from_pstate(data->powernow_table,
1311 data->currpstate);
1232 else 1312 else
1233 pol->cur = find_khz_freq_from_fid(data->currfid); 1313 pol->cur = find_khz_freq_from_fid(data->currfid);
1234 dprintk("policy current frequency %d kHz\n", pol->cur); 1314 dprintk("policy current frequency %d kHz\n", pol->cur);
@@ -1245,7 +1325,8 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1245 cpufreq_frequency_table_get_attr(data->powernow_table, pol->cpu); 1325 cpufreq_frequency_table_get_attr(data->powernow_table, pol->cpu);
1246 1326
1247 if (cpu_family == CPU_HW_PSTATE) 1327 if (cpu_family == CPU_HW_PSTATE)
1248 dprintk("cpu_init done, current pstate 0x%x\n", data->currpstate); 1328 dprintk("cpu_init done, current pstate 0x%x\n",
1329 data->currpstate);
1249 else 1330 else
1250 dprintk("cpu_init done, current fid 0x%x, vid 0x%x\n", 1331 dprintk("cpu_init done, current fid 0x%x, vid 0x%x\n",
1251 data->currfid, data->currvid); 1332 data->currfid, data->currvid);
@@ -1254,15 +1335,16 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1254 1335
1255 return 0; 1336 return 0;
1256 1337
1257err_out: 1338err_out_unmask:
1258 set_cpus_allowed_ptr(current, &oldmask); 1339 set_cpus_allowed_ptr(current, &oldmask);
1259 powernow_k8_cpu_exit_acpi(data); 1340 powernow_k8_cpu_exit_acpi(data);
1260 1341
1342err_out:
1261 kfree(data); 1343 kfree(data);
1262 return -ENODEV; 1344 return -ENODEV;
1263} 1345}
1264 1346
1265static int __devexit powernowk8_cpu_exit (struct cpufreq_policy *pol) 1347static int __devexit powernowk8_cpu_exit(struct cpufreq_policy *pol)
1266{ 1348{
1267 struct powernow_k8_data *data = per_cpu(powernow_data, pol->cpu); 1349 struct powernow_k8_data *data = per_cpu(powernow_data, pol->cpu);
1268 1350
@@ -1279,14 +1361,14 @@ static int __devexit powernowk8_cpu_exit (struct cpufreq_policy *pol)
1279 return 0; 1361 return 0;
1280} 1362}
1281 1363
1282static unsigned int powernowk8_get (unsigned int cpu) 1364static unsigned int powernowk8_get(unsigned int cpu)
1283{ 1365{
1284 struct powernow_k8_data *data; 1366 struct powernow_k8_data *data;
1285 cpumask_t oldmask = current->cpus_allowed; 1367 cpumask_t oldmask = current->cpus_allowed;
1286 unsigned int khz = 0; 1368 unsigned int khz = 0;
1287 unsigned int first; 1369 unsigned int first;
1288 1370
1289 first = first_cpu(per_cpu(cpu_core_map, cpu)); 1371 first = cpumask_first(cpu_core_mask(cpu));
1290 data = per_cpu(powernow_data, first); 1372 data = per_cpu(powernow_data, first);
1291 1373
1292 if (!data) 1374 if (!data)
@@ -1315,7 +1397,7 @@ out:
1315 return khz; 1397 return khz;
1316} 1398}
1317 1399
1318static struct freq_attr* powernow_k8_attr[] = { 1400static struct freq_attr *powernow_k8_attr[] = {
1319 &cpufreq_freq_attr_scaling_available_freqs, 1401 &cpufreq_freq_attr_scaling_available_freqs,
1320 NULL, 1402 NULL,
1321}; 1403};
@@ -1360,7 +1442,8 @@ static void __exit powernowk8_exit(void)
1360 cpufreq_unregister_driver(&cpufreq_amd64_driver); 1442 cpufreq_unregister_driver(&cpufreq_amd64_driver);
1361} 1443}
1362 1444
1363MODULE_AUTHOR("Paul Devriendt <paul.devriendt@amd.com> and Mark Langsdorf <mark.langsdorf@amd.com>"); 1445MODULE_AUTHOR("Paul Devriendt <paul.devriendt@amd.com> and "
1446 "Mark Langsdorf <mark.langsdorf@amd.com>");
1364MODULE_DESCRIPTION("AMD Athlon 64 and Opteron processor frequency driver."); 1447MODULE_DESCRIPTION("AMD Athlon 64 and Opteron processor frequency driver.");
1365MODULE_LICENSE("GPL"); 1448MODULE_LICENSE("GPL");
1366 1449
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.h b/arch/x86/kernel/cpu/cpufreq/powernow-k8.h
index 8ecc75b6c7c3..6c6698feade1 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.h
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.h
@@ -45,11 +45,10 @@ struct powernow_k8_data {
45 * frequency is in kHz */ 45 * frequency is in kHz */
46 struct cpufreq_frequency_table *powernow_table; 46 struct cpufreq_frequency_table *powernow_table;
47 47
48#ifdef CONFIG_X86_POWERNOW_K8_ACPI
49 /* the acpi table needs to be kept. it's only available if ACPI was 48 /* the acpi table needs to be kept. it's only available if ACPI was
50 * used to determine valid frequency/vid/fid states */ 49 * used to determine valid frequency/vid/fid states */
51 struct acpi_processor_performance acpi_data; 50 struct acpi_processor_performance acpi_data;
52#endif 51
53 /* we need to keep track of associated cores, but let cpufreq 52 /* we need to keep track of associated cores, but let cpufreq
54 * handle hotplug events - so just point at cpufreq pol->cpus 53 * handle hotplug events - so just point at cpufreq pol->cpus
55 * structure */ 54 * structure */
@@ -222,10 +221,8 @@ static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid);
222 221
223static void powernow_k8_acpi_pst_values(struct powernow_k8_data *data, unsigned int index); 222static void powernow_k8_acpi_pst_values(struct powernow_k8_data *data, unsigned int index);
224 223
225#ifdef CONFIG_X86_POWERNOW_K8_ACPI
226static int fill_powernow_table_pstate(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table); 224static int fill_powernow_table_pstate(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table);
227static int fill_powernow_table_fidvid(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table); 225static int fill_powernow_table_fidvid(struct powernow_k8_data *data, struct cpufreq_frequency_table *powernow_table);
228#endif
229 226
230#ifdef CONFIG_SMP 227#ifdef CONFIG_SMP
231static inline void define_siblings(int cpu, cpumask_t cpu_sharedcore_mask[]) 228static inline void define_siblings(int cpu, cpumask_t cpu_sharedcore_mask[])
diff --git a/arch/x86/kernel/cpu/cpufreq/sc520_freq.c b/arch/x86/kernel/cpu/cpufreq/sc520_freq.c
index 42da9bd677d6..435a996a613a 100644
--- a/arch/x86/kernel/cpu/cpufreq/sc520_freq.c
+++ b/arch/x86/kernel/cpu/cpufreq/sc520_freq.c
@@ -19,17 +19,19 @@
19 19
20#include <linux/delay.h> 20#include <linux/delay.h>
21#include <linux/cpufreq.h> 21#include <linux/cpufreq.h>
22#include <linux/timex.h>
23#include <linux/io.h>
22 24
23#include <asm/msr.h> 25#include <asm/msr.h>
24#include <asm/timex.h>
25#include <asm/io.h>
26 26
27#define MMCR_BASE 0xfffef000 /* The default base address */ 27#define MMCR_BASE 0xfffef000 /* The default base address */
28#define OFFS_CPUCTL 0x2 /* CPU Control Register */ 28#define OFFS_CPUCTL 0x2 /* CPU Control Register */
29 29
30static __u8 __iomem *cpuctl; 30static __u8 __iomem *cpuctl;
31 31
32#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "sc520_freq", msg) 32#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
33 "sc520_freq", msg)
34#define PFX "sc520_freq: "
33 35
34static struct cpufreq_frequency_table sc520_freq_table[] = { 36static struct cpufreq_frequency_table sc520_freq_table[] = {
35 {0x01, 100000}, 37 {0x01, 100000},
@@ -43,7 +45,8 @@ static unsigned int sc520_freq_get_cpu_frequency(unsigned int cpu)
43 45
44 switch (clockspeed_reg & 0x03) { 46 switch (clockspeed_reg & 0x03) {
45 default: 47 default:
46 printk(KERN_ERR "sc520_freq: error: cpuctl register has unexpected value %02x\n", clockspeed_reg); 48 printk(KERN_ERR PFX "error: cpuctl register has unexpected "
49 "value %02x\n", clockspeed_reg);
47 case 0x01: 50 case 0x01:
48 return 100000; 51 return 100000;
49 case 0x02: 52 case 0x02:
@@ -51,7 +54,7 @@ static unsigned int sc520_freq_get_cpu_frequency(unsigned int cpu)
51 } 54 }
52} 55}
53 56
54static void sc520_freq_set_cpu_state (unsigned int state) 57static void sc520_freq_set_cpu_state(unsigned int state)
55{ 58{
56 59
57 struct cpufreq_freqs freqs; 60 struct cpufreq_freqs freqs;
@@ -76,18 +79,19 @@ static void sc520_freq_set_cpu_state (unsigned int state)
76 cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); 79 cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
77}; 80};
78 81
79static int sc520_freq_verify (struct cpufreq_policy *policy) 82static int sc520_freq_verify(struct cpufreq_policy *policy)
80{ 83{
81 return cpufreq_frequency_table_verify(policy, &sc520_freq_table[0]); 84 return cpufreq_frequency_table_verify(policy, &sc520_freq_table[0]);
82} 85}
83 86
84static int sc520_freq_target (struct cpufreq_policy *policy, 87static int sc520_freq_target(struct cpufreq_policy *policy,
85 unsigned int target_freq, 88 unsigned int target_freq,
86 unsigned int relation) 89 unsigned int relation)
87{ 90{
88 unsigned int newstate = 0; 91 unsigned int newstate = 0;
89 92
90 if (cpufreq_frequency_table_target(policy, sc520_freq_table, target_freq, relation, &newstate)) 93 if (cpufreq_frequency_table_target(policy, sc520_freq_table,
94 target_freq, relation, &newstate))
91 return -EINVAL; 95 return -EINVAL;
92 96
93 sc520_freq_set_cpu_state(newstate); 97 sc520_freq_set_cpu_state(newstate);
@@ -116,7 +120,7 @@ static int sc520_freq_cpu_init(struct cpufreq_policy *policy)
116 120
117 result = cpufreq_frequency_table_cpuinfo(policy, sc520_freq_table); 121 result = cpufreq_frequency_table_cpuinfo(policy, sc520_freq_table);
118 if (result) 122 if (result)
119 return (result); 123 return result;
120 124
121 cpufreq_frequency_table_get_attr(sc520_freq_table, policy->cpu); 125 cpufreq_frequency_table_get_attr(sc520_freq_table, policy->cpu);
122 126
@@ -131,7 +135,7 @@ static int sc520_freq_cpu_exit(struct cpufreq_policy *policy)
131} 135}
132 136
133 137
134static struct freq_attr* sc520_freq_attr[] = { 138static struct freq_attr *sc520_freq_attr[] = {
135 &cpufreq_freq_attr_scaling_available_freqs, 139 &cpufreq_freq_attr_scaling_available_freqs,
136 NULL, 140 NULL,
137}; 141};
@@ -155,13 +159,13 @@ static int __init sc520_freq_init(void)
155 int err; 159 int err;
156 160
157 /* Test if we have the right hardware */ 161 /* Test if we have the right hardware */
158 if(c->x86_vendor != X86_VENDOR_AMD || 162 if (c->x86_vendor != X86_VENDOR_AMD ||
159 c->x86 != 4 || c->x86_model != 9) { 163 c->x86 != 4 || c->x86_model != 9) {
160 dprintk("no Elan SC520 processor found!\n"); 164 dprintk("no Elan SC520 processor found!\n");
161 return -ENODEV; 165 return -ENODEV;
162 } 166 }
163 cpuctl = ioremap((unsigned long)(MMCR_BASE + OFFS_CPUCTL), 1); 167 cpuctl = ioremap((unsigned long)(MMCR_BASE + OFFS_CPUCTL), 1);
164 if(!cpuctl) { 168 if (!cpuctl) {
165 printk(KERN_ERR "sc520_freq: error: failed to remap memory\n"); 169 printk(KERN_ERR "sc520_freq: error: failed to remap memory\n");
166 return -ENOMEM; 170 return -ENOMEM;
167 } 171 }
diff --git a/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c b/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c
index dedc1e98f168..016c1a4fa3fc 100644
--- a/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c
+++ b/arch/x86/kernel/cpu/cpufreq/speedstep-ich.c
@@ -39,7 +39,7 @@ static struct pci_dev *speedstep_chipset_dev;
39 39
40/* speedstep_processor 40/* speedstep_processor
41 */ 41 */
42static unsigned int speedstep_processor = 0; 42static unsigned int speedstep_processor;
43 43
44static u32 pmbase; 44static u32 pmbase;
45 45
@@ -54,7 +54,8 @@ static struct cpufreq_frequency_table speedstep_freqs[] = {
54}; 54};
55 55
56 56
57#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "speedstep-ich", msg) 57#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
58 "speedstep-ich", msg)
58 59
59 60
60/** 61/**
@@ -62,7 +63,7 @@ static struct cpufreq_frequency_table speedstep_freqs[] = {
62 * 63 *
63 * Returns: -ENODEV if no register could be found 64 * Returns: -ENODEV if no register could be found
64 */ 65 */
65static int speedstep_find_register (void) 66static int speedstep_find_register(void)
66{ 67{
67 if (!speedstep_chipset_dev) 68 if (!speedstep_chipset_dev)
68 return -ENODEV; 69 return -ENODEV;
@@ -90,7 +91,7 @@ static int speedstep_find_register (void)
90 * 91 *
91 * Tries to change the SpeedStep state. 92 * Tries to change the SpeedStep state.
92 */ 93 */
93static void speedstep_set_state (unsigned int state) 94static void speedstep_set_state(unsigned int state)
94{ 95{
95 u8 pm2_blk; 96 u8 pm2_blk;
96 u8 value; 97 u8 value;
@@ -133,11 +134,11 @@ static void speedstep_set_state (unsigned int state)
133 134
134 dprintk("read at pmbase 0x%x + 0x50 returned 0x%x\n", pmbase, value); 135 dprintk("read at pmbase 0x%x + 0x50 returned 0x%x\n", pmbase, value);
135 136
136 if (state == (value & 0x1)) { 137 if (state == (value & 0x1))
137 dprintk("change to %u MHz succeeded\n", (speedstep_get_processor_frequency(speedstep_processor) / 1000)); 138 dprintk("change to %u MHz succeeded\n",
138 } else { 139 speedstep_get_frequency(speedstep_processor) / 1000);
139 printk (KERN_ERR "cpufreq: change failed - I/O error\n"); 140 else
140 } 141 printk(KERN_ERR "cpufreq: change failed - I/O error\n");
141 142
142 return; 143 return;
143} 144}
@@ -149,7 +150,7 @@ static void speedstep_set_state (unsigned int state)
149 * Tries to activate the SpeedStep status and control registers. 150 * Tries to activate the SpeedStep status and control registers.
150 * Returns -EINVAL on an unsupported chipset, and zero on success. 151 * Returns -EINVAL on an unsupported chipset, and zero on success.
151 */ 152 */
152static int speedstep_activate (void) 153static int speedstep_activate(void)
153{ 154{
154 u16 value = 0; 155 u16 value = 0;
155 156
@@ -175,20 +176,18 @@ static int speedstep_activate (void)
175 * functions. Returns the SPEEDSTEP_CHIPSET_-number for the detected 176 * functions. Returns the SPEEDSTEP_CHIPSET_-number for the detected
176 * chipset, or zero on failure. 177 * chipset, or zero on failure.
177 */ 178 */
178static unsigned int speedstep_detect_chipset (void) 179static unsigned int speedstep_detect_chipset(void)
179{ 180{
180 speedstep_chipset_dev = pci_get_subsys(PCI_VENDOR_ID_INTEL, 181 speedstep_chipset_dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
181 PCI_DEVICE_ID_INTEL_82801DB_12, 182 PCI_DEVICE_ID_INTEL_82801DB_12,
182 PCI_ANY_ID, 183 PCI_ANY_ID, PCI_ANY_ID,
183 PCI_ANY_ID,
184 NULL); 184 NULL);
185 if (speedstep_chipset_dev) 185 if (speedstep_chipset_dev)
186 return 4; /* 4-M */ 186 return 4; /* 4-M */
187 187
188 speedstep_chipset_dev = pci_get_subsys(PCI_VENDOR_ID_INTEL, 188 speedstep_chipset_dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
189 PCI_DEVICE_ID_INTEL_82801CA_12, 189 PCI_DEVICE_ID_INTEL_82801CA_12,
190 PCI_ANY_ID, 190 PCI_ANY_ID, PCI_ANY_ID,
191 PCI_ANY_ID,
192 NULL); 191 NULL);
193 if (speedstep_chipset_dev) 192 if (speedstep_chipset_dev)
194 return 3; /* 3-M */ 193 return 3; /* 3-M */
@@ -196,8 +195,7 @@ static unsigned int speedstep_detect_chipset (void)
196 195
197 speedstep_chipset_dev = pci_get_subsys(PCI_VENDOR_ID_INTEL, 196 speedstep_chipset_dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
198 PCI_DEVICE_ID_INTEL_82801BA_10, 197 PCI_DEVICE_ID_INTEL_82801BA_10,
199 PCI_ANY_ID, 198 PCI_ANY_ID, PCI_ANY_ID,
200 PCI_ANY_ID,
201 NULL); 199 NULL);
202 if (speedstep_chipset_dev) { 200 if (speedstep_chipset_dev) {
203 /* speedstep.c causes lockups on Dell Inspirons 8000 and 201 /* speedstep.c causes lockups on Dell Inspirons 8000 and
@@ -208,8 +206,7 @@ static unsigned int speedstep_detect_chipset (void)
208 206
209 hostbridge = pci_get_subsys(PCI_VENDOR_ID_INTEL, 207 hostbridge = pci_get_subsys(PCI_VENDOR_ID_INTEL,
210 PCI_DEVICE_ID_INTEL_82815_MC, 208 PCI_DEVICE_ID_INTEL_82815_MC,
211 PCI_ANY_ID, 209 PCI_ANY_ID, PCI_ANY_ID,
212 PCI_ANY_ID,
213 NULL); 210 NULL);
214 211
215 if (!hostbridge) 212 if (!hostbridge)
@@ -236,7 +233,7 @@ static unsigned int _speedstep_get(const struct cpumask *cpus)
236 233
237 cpus_allowed = current->cpus_allowed; 234 cpus_allowed = current->cpus_allowed;
238 set_cpus_allowed_ptr(current, cpus); 235 set_cpus_allowed_ptr(current, cpus);
239 speed = speedstep_get_processor_frequency(speedstep_processor); 236 speed = speedstep_get_frequency(speedstep_processor);
240 set_cpus_allowed_ptr(current, &cpus_allowed); 237 set_cpus_allowed_ptr(current, &cpus_allowed);
241 dprintk("detected %u kHz as current frequency\n", speed); 238 dprintk("detected %u kHz as current frequency\n", speed);
242 return speed; 239 return speed;
@@ -251,11 +248,12 @@ static unsigned int speedstep_get(unsigned int cpu)
251 * speedstep_target - set a new CPUFreq policy 248 * speedstep_target - set a new CPUFreq policy
252 * @policy: new policy 249 * @policy: new policy
253 * @target_freq: the target frequency 250 * @target_freq: the target frequency
254 * @relation: how that frequency relates to achieved frequency (CPUFREQ_RELATION_L or CPUFREQ_RELATION_H) 251 * @relation: how that frequency relates to achieved frequency
252 * (CPUFREQ_RELATION_L or CPUFREQ_RELATION_H)
255 * 253 *
256 * Sets a new CPUFreq policy. 254 * Sets a new CPUFreq policy.
257 */ 255 */
258static int speedstep_target (struct cpufreq_policy *policy, 256static int speedstep_target(struct cpufreq_policy *policy,
259 unsigned int target_freq, 257 unsigned int target_freq,
260 unsigned int relation) 258 unsigned int relation)
261{ 259{
@@ -264,7 +262,8 @@ static int speedstep_target (struct cpufreq_policy *policy,
264 cpumask_t cpus_allowed; 262 cpumask_t cpus_allowed;
265 int i; 263 int i;
266 264
267 if (cpufreq_frequency_table_target(policy, &speedstep_freqs[0], target_freq, relation, &newstate)) 265 if (cpufreq_frequency_table_target(policy, &speedstep_freqs[0],
266 target_freq, relation, &newstate))
268 return -EINVAL; 267 return -EINVAL;
269 268
270 freqs.old = _speedstep_get(policy->cpus); 269 freqs.old = _speedstep_get(policy->cpus);
@@ -308,7 +307,7 @@ static int speedstep_target (struct cpufreq_policy *policy,
308 * Limit must be within speedstep_low_freq and speedstep_high_freq, with 307 * Limit must be within speedstep_low_freq and speedstep_high_freq, with
309 * at least one border included. 308 * at least one border included.
310 */ 309 */
311static int speedstep_verify (struct cpufreq_policy *policy) 310static int speedstep_verify(struct cpufreq_policy *policy)
312{ 311{
313 return cpufreq_frequency_table_verify(policy, &speedstep_freqs[0]); 312 return cpufreq_frequency_table_verify(policy, &speedstep_freqs[0]);
314} 313}
@@ -322,7 +321,7 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
322 321
323 /* only run on CPU to be set, or on its sibling */ 322 /* only run on CPU to be set, or on its sibling */
324#ifdef CONFIG_SMP 323#ifdef CONFIG_SMP
325 cpumask_copy(policy->cpus, &per_cpu(cpu_sibling_map, policy->cpu)); 324 cpumask_copy(policy->cpus, cpu_sibling_mask(policy->cpu));
326#endif 325#endif
327 326
328 cpus_allowed = current->cpus_allowed; 327 cpus_allowed = current->cpus_allowed;
@@ -344,7 +343,8 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
344 return -EIO; 343 return -EIO;
345 344
346 dprintk("currently at %s speed setting - %i MHz\n", 345 dprintk("currently at %s speed setting - %i MHz\n",
347 (speed == speedstep_freqs[SPEEDSTEP_LOW].frequency) ? "low" : "high", 346 (speed == speedstep_freqs[SPEEDSTEP_LOW].frequency)
347 ? "low" : "high",
348 (speed / 1000)); 348 (speed / 1000));
349 349
350 /* cpuinfo and default policy values */ 350 /* cpuinfo and default policy values */
@@ -352,9 +352,9 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
352 352
353 result = cpufreq_frequency_table_cpuinfo(policy, speedstep_freqs); 353 result = cpufreq_frequency_table_cpuinfo(policy, speedstep_freqs);
354 if (result) 354 if (result)
355 return (result); 355 return result;
356 356
357 cpufreq_frequency_table_get_attr(speedstep_freqs, policy->cpu); 357 cpufreq_frequency_table_get_attr(speedstep_freqs, policy->cpu);
358 358
359 return 0; 359 return 0;
360} 360}
@@ -366,7 +366,7 @@ static int speedstep_cpu_exit(struct cpufreq_policy *policy)
366 return 0; 366 return 0;
367} 367}
368 368
369static struct freq_attr* speedstep_attr[] = { 369static struct freq_attr *speedstep_attr[] = {
370 &cpufreq_freq_attr_scaling_available_freqs, 370 &cpufreq_freq_attr_scaling_available_freqs,
371 NULL, 371 NULL,
372}; 372};
@@ -396,13 +396,15 @@ static int __init speedstep_init(void)
396 /* detect processor */ 396 /* detect processor */
397 speedstep_processor = speedstep_detect_processor(); 397 speedstep_processor = speedstep_detect_processor();
398 if (!speedstep_processor) { 398 if (!speedstep_processor) {
399 dprintk("Intel(R) SpeedStep(TM) capable processor not found\n"); 399 dprintk("Intel(R) SpeedStep(TM) capable processor "
400 "not found\n");
400 return -ENODEV; 401 return -ENODEV;
401 } 402 }
402 403
403 /* detect chipset */ 404 /* detect chipset */
404 if (!speedstep_detect_chipset()) { 405 if (!speedstep_detect_chipset()) {
405 dprintk("Intel(R) SpeedStep(TM) for this chipset not (yet) available.\n"); 406 dprintk("Intel(R) SpeedStep(TM) for this chipset not "
407 "(yet) available.\n");
406 return -ENODEV; 408 return -ENODEV;
407 } 409 }
408 410
@@ -431,9 +433,11 @@ static void __exit speedstep_exit(void)
431} 433}
432 434
433 435
434MODULE_AUTHOR ("Dave Jones <davej@redhat.com>, Dominik Brodowski <linux@brodo.de>"); 436MODULE_AUTHOR("Dave Jones <davej@redhat.com>, "
435MODULE_DESCRIPTION ("Speedstep driver for Intel mobile processors on chipsets with ICH-M southbridges."); 437 "Dominik Brodowski <linux@brodo.de>");
436MODULE_LICENSE ("GPL"); 438MODULE_DESCRIPTION("Speedstep driver for Intel mobile processors on chipsets "
439 "with ICH-M southbridges.");
440MODULE_LICENSE("GPL");
437 441
438module_init(speedstep_init); 442module_init(speedstep_init);
439module_exit(speedstep_exit); 443module_exit(speedstep_exit);
diff --git a/arch/x86/kernel/cpu/cpufreq/speedstep-lib.c b/arch/x86/kernel/cpu/cpufreq/speedstep-lib.c
index cdac7d62369b..2e3c6862657b 100644
--- a/arch/x86/kernel/cpu/cpufreq/speedstep-lib.c
+++ b/arch/x86/kernel/cpu/cpufreq/speedstep-lib.c
@@ -16,12 +16,16 @@
16#include <linux/slab.h> 16#include <linux/slab.h>
17 17
18#include <asm/msr.h> 18#include <asm/msr.h>
19#include <asm/tsc.h>
19#include "speedstep-lib.h" 20#include "speedstep-lib.h"
20 21
21#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "speedstep-lib", msg) 22#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
23 "speedstep-lib", msg)
24
25#define PFX "speedstep-lib: "
22 26
23#ifdef CONFIG_X86_SPEEDSTEP_RELAXED_CAP_CHECK 27#ifdef CONFIG_X86_SPEEDSTEP_RELAXED_CAP_CHECK
24static int relaxed_check = 0; 28static int relaxed_check;
25#else 29#else
26#define relaxed_check 0 30#define relaxed_check 0
27#endif 31#endif
@@ -30,14 +34,14 @@ static int relaxed_check = 0;
30 * GET PROCESSOR CORE SPEED IN KHZ * 34 * GET PROCESSOR CORE SPEED IN KHZ *
31 *********************************************************************/ 35 *********************************************************************/
32 36
33static unsigned int pentium3_get_frequency (unsigned int processor) 37static unsigned int pentium3_get_frequency(unsigned int processor)
34{ 38{
35 /* See table 14 of p3_ds.pdf and table 22 of 29834003.pdf */ 39 /* See table 14 of p3_ds.pdf and table 22 of 29834003.pdf */
36 struct { 40 struct {
37 unsigned int ratio; /* Frequency Multiplier (x10) */ 41 unsigned int ratio; /* Frequency Multiplier (x10) */
38 u8 bitmap; /* power on configuration bits 42 u8 bitmap; /* power on configuration bits
39 [27, 25:22] (in MSR 0x2a) */ 43 [27, 25:22] (in MSR 0x2a) */
40 } msr_decode_mult [] = { 44 } msr_decode_mult[] = {
41 { 30, 0x01 }, 45 { 30, 0x01 },
42 { 35, 0x05 }, 46 { 35, 0x05 },
43 { 40, 0x02 }, 47 { 40, 0x02 },
@@ -52,7 +56,7 @@ static unsigned int pentium3_get_frequency (unsigned int processor)
52 { 85, 0x26 }, 56 { 85, 0x26 },
53 { 90, 0x20 }, 57 { 90, 0x20 },
54 { 100, 0x2b }, 58 { 100, 0x2b },
55 { 0, 0xff } /* error or unknown value */ 59 { 0, 0xff } /* error or unknown value */
56 }; 60 };
57 61
58 /* PIII(-M) FSB settings: see table b1-b of 24547206.pdf */ 62 /* PIII(-M) FSB settings: see table b1-b of 24547206.pdf */
@@ -60,7 +64,7 @@ static unsigned int pentium3_get_frequency (unsigned int processor)
60 unsigned int value; /* Front Side Bus speed in MHz */ 64 unsigned int value; /* Front Side Bus speed in MHz */
61 u8 bitmap; /* power on configuration bits [18: 19] 65 u8 bitmap; /* power on configuration bits [18: 19]
62 (in MSR 0x2a) */ 66 (in MSR 0x2a) */
63 } msr_decode_fsb [] = { 67 } msr_decode_fsb[] = {
64 { 66, 0x0 }, 68 { 66, 0x0 },
65 { 100, 0x2 }, 69 { 100, 0x2 },
66 { 133, 0x1 }, 70 { 133, 0x1 },
@@ -85,7 +89,7 @@ static unsigned int pentium3_get_frequency (unsigned int processor)
85 } 89 }
86 90
87 /* decode the multiplier */ 91 /* decode the multiplier */
88 if (processor == SPEEDSTEP_PROCESSOR_PIII_C_EARLY) { 92 if (processor == SPEEDSTEP_CPU_PIII_C_EARLY) {
89 dprintk("workaround for early PIIIs\n"); 93 dprintk("workaround for early PIIIs\n");
90 msr_lo &= 0x03c00000; 94 msr_lo &= 0x03c00000;
91 } else 95 } else
@@ -97,9 +101,10 @@ static unsigned int pentium3_get_frequency (unsigned int processor)
97 j++; 101 j++;
98 } 102 }
99 103
100 dprintk("speed is %u\n", (msr_decode_mult[j].ratio * msr_decode_fsb[i].value * 100)); 104 dprintk("speed is %u\n",
105 (msr_decode_mult[j].ratio * msr_decode_fsb[i].value * 100));
101 106
102 return (msr_decode_mult[j].ratio * msr_decode_fsb[i].value * 100); 107 return msr_decode_mult[j].ratio * msr_decode_fsb[i].value * 100;
103} 108}
104 109
105 110
@@ -112,20 +117,23 @@ static unsigned int pentiumM_get_frequency(void)
112 117
113 /* see table B-2 of 24547212.pdf */ 118 /* see table B-2 of 24547212.pdf */
114 if (msr_lo & 0x00040000) { 119 if (msr_lo & 0x00040000) {
115 printk(KERN_DEBUG "speedstep-lib: PM - invalid FSB: 0x%x 0x%x\n", msr_lo, msr_tmp); 120 printk(KERN_DEBUG PFX "PM - invalid FSB: 0x%x 0x%x\n",
121 msr_lo, msr_tmp);
116 return 0; 122 return 0;
117 } 123 }
118 124
119 msr_tmp = (msr_lo >> 22) & 0x1f; 125 msr_tmp = (msr_lo >> 22) & 0x1f;
120 dprintk("bits 22-26 are 0x%x, speed is %u\n", msr_tmp, (msr_tmp * 100 * 1000)); 126 dprintk("bits 22-26 are 0x%x, speed is %u\n",
127 msr_tmp, (msr_tmp * 100 * 1000));
121 128
122 return (msr_tmp * 100 * 1000); 129 return msr_tmp * 100 * 1000;
123} 130}
124 131
125static unsigned int pentium_core_get_frequency(void) 132static unsigned int pentium_core_get_frequency(void)
126{ 133{
127 u32 fsb = 0; 134 u32 fsb = 0;
128 u32 msr_lo, msr_tmp; 135 u32 msr_lo, msr_tmp;
136 int ret;
129 137
130 rdmsr(MSR_FSB_FREQ, msr_lo, msr_tmp); 138 rdmsr(MSR_FSB_FREQ, msr_lo, msr_tmp);
131 /* see table B-2 of 25366920.pdf */ 139 /* see table B-2 of 25366920.pdf */
@@ -153,12 +161,15 @@ static unsigned int pentium_core_get_frequency(void)
153 } 161 }
154 162
155 rdmsr(MSR_IA32_EBL_CR_POWERON, msr_lo, msr_tmp); 163 rdmsr(MSR_IA32_EBL_CR_POWERON, msr_lo, msr_tmp);
156 dprintk("PCORE - MSR_IA32_EBL_CR_POWERON: 0x%x 0x%x\n", msr_lo, msr_tmp); 164 dprintk("PCORE - MSR_IA32_EBL_CR_POWERON: 0x%x 0x%x\n",
165 msr_lo, msr_tmp);
157 166
158 msr_tmp = (msr_lo >> 22) & 0x1f; 167 msr_tmp = (msr_lo >> 22) & 0x1f;
159 dprintk("bits 22-26 are 0x%x, speed is %u\n", msr_tmp, (msr_tmp * fsb)); 168 dprintk("bits 22-26 are 0x%x, speed is %u\n",
169 msr_tmp, (msr_tmp * fsb));
160 170
161 return (msr_tmp * fsb); 171 ret = (msr_tmp * fsb);
172 return ret;
162} 173}
163 174
164 175
@@ -167,6 +178,16 @@ static unsigned int pentium4_get_frequency(void)
167 struct cpuinfo_x86 *c = &boot_cpu_data; 178 struct cpuinfo_x86 *c = &boot_cpu_data;
168 u32 msr_lo, msr_hi, mult; 179 u32 msr_lo, msr_hi, mult;
169 unsigned int fsb = 0; 180 unsigned int fsb = 0;
181 unsigned int ret;
182 u8 fsb_code;
183
184 /* Pentium 4 Model 0 and 1 do not have the Core Clock Frequency
185 * to System Bus Frequency Ratio Field in the Processor Frequency
186 * Configuration Register of the MSR. Therefore the current
187 * frequency cannot be calculated and has to be measured.
188 */
189 if (c->x86_model < 2)
190 return cpu_khz;
170 191
171 rdmsr(0x2c, msr_lo, msr_hi); 192 rdmsr(0x2c, msr_lo, msr_hi);
172 193
@@ -177,62 +198,61 @@ static unsigned int pentium4_get_frequency(void)
177 * revision #12 in Table B-1: MSRs in the Pentium 4 and 198 * revision #12 in Table B-1: MSRs in the Pentium 4 and
178 * Intel Xeon Processors, on page B-4 and B-5. 199 * Intel Xeon Processors, on page B-4 and B-5.
179 */ 200 */
180 if (c->x86_model < 2) 201 fsb_code = (msr_lo >> 16) & 0x7;
202 switch (fsb_code) {
203 case 0:
181 fsb = 100 * 1000; 204 fsb = 100 * 1000;
182 else { 205 break;
183 u8 fsb_code = (msr_lo >> 16) & 0x7; 206 case 1:
184 switch (fsb_code) { 207 fsb = 13333 * 10;
185 case 0: 208 break;
186 fsb = 100 * 1000; 209 case 2:
187 break; 210 fsb = 200 * 1000;
188 case 1: 211 break;
189 fsb = 13333 * 10;
190 break;
191 case 2:
192 fsb = 200 * 1000;
193 break;
194 }
195 } 212 }
196 213
197 if (!fsb) 214 if (!fsb)
198 printk(KERN_DEBUG "speedstep-lib: couldn't detect FSB speed. Please send an e-mail to <linux@brodo.de>\n"); 215 printk(KERN_DEBUG PFX "couldn't detect FSB speed. "
216 "Please send an e-mail to <linux@brodo.de>\n");
199 217
200 /* Multiplier. */ 218 /* Multiplier. */
201 mult = msr_lo >> 24; 219 mult = msr_lo >> 24;
202 220
203 dprintk("P4 - FSB %u kHz; Multiplier %u; Speed %u kHz\n", fsb, mult, (fsb * mult)); 221 dprintk("P4 - FSB %u kHz; Multiplier %u; Speed %u kHz\n",
222 fsb, mult, (fsb * mult));
204 223
205 return (fsb * mult); 224 ret = (fsb * mult);
225 return ret;
206} 226}
207 227
208 228
209unsigned int speedstep_get_processor_frequency(unsigned int processor) 229unsigned int speedstep_get_frequency(unsigned int processor)
210{ 230{
211 switch (processor) { 231 switch (processor) {
212 case SPEEDSTEP_PROCESSOR_PCORE: 232 case SPEEDSTEP_CPU_PCORE:
213 return pentium_core_get_frequency(); 233 return pentium_core_get_frequency();
214 case SPEEDSTEP_PROCESSOR_PM: 234 case SPEEDSTEP_CPU_PM:
215 return pentiumM_get_frequency(); 235 return pentiumM_get_frequency();
216 case SPEEDSTEP_PROCESSOR_P4D: 236 case SPEEDSTEP_CPU_P4D:
217 case SPEEDSTEP_PROCESSOR_P4M: 237 case SPEEDSTEP_CPU_P4M:
218 return pentium4_get_frequency(); 238 return pentium4_get_frequency();
219 case SPEEDSTEP_PROCESSOR_PIII_T: 239 case SPEEDSTEP_CPU_PIII_T:
220 case SPEEDSTEP_PROCESSOR_PIII_C: 240 case SPEEDSTEP_CPU_PIII_C:
221 case SPEEDSTEP_PROCESSOR_PIII_C_EARLY: 241 case SPEEDSTEP_CPU_PIII_C_EARLY:
222 return pentium3_get_frequency(processor); 242 return pentium3_get_frequency(processor);
223 default: 243 default:
224 return 0; 244 return 0;
225 }; 245 };
226 return 0; 246 return 0;
227} 247}
228EXPORT_SYMBOL_GPL(speedstep_get_processor_frequency); 248EXPORT_SYMBOL_GPL(speedstep_get_frequency);
229 249
230 250
231/********************************************************************* 251/*********************************************************************
232 * DETECT SPEEDSTEP-CAPABLE PROCESSOR * 252 * DETECT SPEEDSTEP-CAPABLE PROCESSOR *
233 *********************************************************************/ 253 *********************************************************************/
234 254
235unsigned int speedstep_detect_processor (void) 255unsigned int speedstep_detect_processor(void)
236{ 256{
237 struct cpuinfo_x86 *c = &cpu_data(0); 257 struct cpuinfo_x86 *c = &cpu_data(0);
238 u32 ebx, msr_lo, msr_hi; 258 u32 ebx, msr_lo, msr_hi;
@@ -261,7 +281,7 @@ unsigned int speedstep_detect_processor (void)
261 * sample has ebx = 0x0f, production has 0x0e. 281 * sample has ebx = 0x0f, production has 0x0e.
262 */ 282 */
263 if ((ebx == 0x0e) || (ebx == 0x0f)) 283 if ((ebx == 0x0e) || (ebx == 0x0f))
264 return SPEEDSTEP_PROCESSOR_P4M; 284 return SPEEDSTEP_CPU_P4M;
265 break; 285 break;
266 case 7: 286 case 7:
267 /* 287 /*
@@ -272,7 +292,7 @@ unsigned int speedstep_detect_processor (void)
272 * samples are only of B-stepping... 292 * samples are only of B-stepping...
273 */ 293 */
274 if (ebx == 0x0e) 294 if (ebx == 0x0e)
275 return SPEEDSTEP_PROCESSOR_P4M; 295 return SPEEDSTEP_CPU_P4M;
276 break; 296 break;
277 case 9: 297 case 9:
278 /* 298 /*
@@ -288,10 +308,13 @@ unsigned int speedstep_detect_processor (void)
288 * M-P4-Ms may have either ebx=0xe or 0xf [see above] 308 * M-P4-Ms may have either ebx=0xe or 0xf [see above]
289 * M-P4/533 have either ebx=0xe or 0xf. [25317607.pdf] 309 * M-P4/533 have either ebx=0xe or 0xf. [25317607.pdf]
290 * also, M-P4M HTs have ebx=0x8, too 310 * also, M-P4M HTs have ebx=0x8, too
291 * For now, they are distinguished by the model_id string 311 * For now, they are distinguished by the model_id
312 * string
292 */ 313 */
293 if ((ebx == 0x0e) || (strstr(c->x86_model_id,"Mobile Intel(R) Pentium(R) 4") != NULL)) 314 if ((ebx == 0x0e) ||
294 return SPEEDSTEP_PROCESSOR_P4M; 315 (strstr(c->x86_model_id,
316 "Mobile Intel(R) Pentium(R) 4") != NULL))
317 return SPEEDSTEP_CPU_P4M;
295 break; 318 break;
296 default: 319 default:
297 break; 320 break;
@@ -301,7 +324,8 @@ unsigned int speedstep_detect_processor (void)
301 324
302 switch (c->x86_model) { 325 switch (c->x86_model) {
303 case 0x0B: /* Intel PIII [Tualatin] */ 326 case 0x0B: /* Intel PIII [Tualatin] */
304 /* cpuid_ebx(1) is 0x04 for desktop PIII, 0x06 for mobile PIII-M */ 327 /* cpuid_ebx(1) is 0x04 for desktop PIII,
328 * 0x06 for mobile PIII-M */
305 ebx = cpuid_ebx(0x00000001); 329 ebx = cpuid_ebx(0x00000001);
306 dprintk("ebx is %x\n", ebx); 330 dprintk("ebx is %x\n", ebx);
307 331
@@ -313,14 +337,15 @@ unsigned int speedstep_detect_processor (void)
313 /* So far all PIII-M processors support SpeedStep. See 337 /* So far all PIII-M processors support SpeedStep. See
314 * Intel's 24540640.pdf of June 2003 338 * Intel's 24540640.pdf of June 2003
315 */ 339 */
316 return SPEEDSTEP_PROCESSOR_PIII_T; 340 return SPEEDSTEP_CPU_PIII_T;
317 341
318 case 0x08: /* Intel PIII [Coppermine] */ 342 case 0x08: /* Intel PIII [Coppermine] */
319 343
320 /* all mobile PIII Coppermines have FSB 100 MHz 344 /* all mobile PIII Coppermines have FSB 100 MHz
321 * ==> sort out a few desktop PIIIs. */ 345 * ==> sort out a few desktop PIIIs. */
322 rdmsr(MSR_IA32_EBL_CR_POWERON, msr_lo, msr_hi); 346 rdmsr(MSR_IA32_EBL_CR_POWERON, msr_lo, msr_hi);
323 dprintk("Coppermine: MSR_IA32_EBL_CR_POWERON is 0x%x, 0x%x\n", msr_lo, msr_hi); 347 dprintk("Coppermine: MSR_IA32_EBL_CR_POWERON is 0x%x, 0x%x\n",
348 msr_lo, msr_hi);
324 msr_lo &= 0x00c0000; 349 msr_lo &= 0x00c0000;
325 if (msr_lo != 0x0080000) 350 if (msr_lo != 0x0080000)
326 return 0; 351 return 0;
@@ -332,13 +357,15 @@ unsigned int speedstep_detect_processor (void)
332 * bit 56 or 57 is set 357 * bit 56 or 57 is set
333 */ 358 */
334 rdmsr(MSR_IA32_PLATFORM_ID, msr_lo, msr_hi); 359 rdmsr(MSR_IA32_PLATFORM_ID, msr_lo, msr_hi);
335 dprintk("Coppermine: MSR_IA32_PLATFORM ID is 0x%x, 0x%x\n", msr_lo, msr_hi); 360 dprintk("Coppermine: MSR_IA32_PLATFORM ID is 0x%x, 0x%x\n",
336 if ((msr_hi & (1<<18)) && (relaxed_check ? 1 : (msr_hi & (3<<24)))) { 361 msr_lo, msr_hi);
362 if ((msr_hi & (1<<18)) &&
363 (relaxed_check ? 1 : (msr_hi & (3<<24)))) {
337 if (c->x86_mask == 0x01) { 364 if (c->x86_mask == 0x01) {
338 dprintk("early PIII version\n"); 365 dprintk("early PIII version\n");
339 return SPEEDSTEP_PROCESSOR_PIII_C_EARLY; 366 return SPEEDSTEP_CPU_PIII_C_EARLY;
340 } else 367 } else
341 return SPEEDSTEP_PROCESSOR_PIII_C; 368 return SPEEDSTEP_CPU_PIII_C;
342 } 369 }
343 370
344 default: 371 default:
@@ -369,7 +396,7 @@ unsigned int speedstep_get_freqs(unsigned int processor,
369 dprintk("trying to determine both speeds\n"); 396 dprintk("trying to determine both speeds\n");
370 397
371 /* get current speed */ 398 /* get current speed */
372 prev_speed = speedstep_get_processor_frequency(processor); 399 prev_speed = speedstep_get_frequency(processor);
373 if (!prev_speed) 400 if (!prev_speed)
374 return -EIO; 401 return -EIO;
375 402
@@ -379,7 +406,7 @@ unsigned int speedstep_get_freqs(unsigned int processor,
379 406
380 /* switch to low state */ 407 /* switch to low state */
381 set_state(SPEEDSTEP_LOW); 408 set_state(SPEEDSTEP_LOW);
382 *low_speed = speedstep_get_processor_frequency(processor); 409 *low_speed = speedstep_get_frequency(processor);
383 if (!*low_speed) { 410 if (!*low_speed) {
384 ret = -EIO; 411 ret = -EIO;
385 goto out; 412 goto out;
@@ -398,7 +425,7 @@ unsigned int speedstep_get_freqs(unsigned int processor,
398 if (transition_latency) 425 if (transition_latency)
399 do_gettimeofday(&tv2); 426 do_gettimeofday(&tv2);
400 427
401 *high_speed = speedstep_get_processor_frequency(processor); 428 *high_speed = speedstep_get_frequency(processor);
402 if (!*high_speed) { 429 if (!*high_speed) {
403 ret = -EIO; 430 ret = -EIO;
404 goto out; 431 goto out;
@@ -426,9 +453,12 @@ unsigned int speedstep_get_freqs(unsigned int processor,
426 /* check if the latency measurement is too high or too low 453 /* check if the latency measurement is too high or too low
427 * and set it to a safe value (500uSec) in that case 454 * and set it to a safe value (500uSec) in that case
428 */ 455 */
429 if (*transition_latency > 10000000 || *transition_latency < 50000) { 456 if (*transition_latency > 10000000 ||
430 printk (KERN_WARNING "speedstep: frequency transition measured seems out of " 457 *transition_latency < 50000) {
431 "range (%u nSec), falling back to a safe one of %u nSec.\n", 458 printk(KERN_WARNING PFX "frequency transition "
459 "measured seems out of range (%u "
460 "nSec), falling back to a safe one of"
461 "%u nSec.\n",
432 *transition_latency, 500000); 462 *transition_latency, 500000);
433 *transition_latency = 500000; 463 *transition_latency = 500000;
434 } 464 }
@@ -436,15 +466,16 @@ unsigned int speedstep_get_freqs(unsigned int processor,
436 466
437out: 467out:
438 local_irq_restore(flags); 468 local_irq_restore(flags);
439 return (ret); 469 return ret;
440} 470}
441EXPORT_SYMBOL_GPL(speedstep_get_freqs); 471EXPORT_SYMBOL_GPL(speedstep_get_freqs);
442 472
443#ifdef CONFIG_X86_SPEEDSTEP_RELAXED_CAP_CHECK 473#ifdef CONFIG_X86_SPEEDSTEP_RELAXED_CAP_CHECK
444module_param(relaxed_check, int, 0444); 474module_param(relaxed_check, int, 0444);
445MODULE_PARM_DESC(relaxed_check, "Don't do all checks for speedstep capability."); 475MODULE_PARM_DESC(relaxed_check,
476 "Don't do all checks for speedstep capability.");
446#endif 477#endif
447 478
448MODULE_AUTHOR ("Dominik Brodowski <linux@brodo.de>"); 479MODULE_AUTHOR("Dominik Brodowski <linux@brodo.de>");
449MODULE_DESCRIPTION ("Library for Intel SpeedStep 1 or 2 cpufreq drivers."); 480MODULE_DESCRIPTION("Library for Intel SpeedStep 1 or 2 cpufreq drivers.");
450MODULE_LICENSE ("GPL"); 481MODULE_LICENSE("GPL");
diff --git a/arch/x86/kernel/cpu/cpufreq/speedstep-lib.h b/arch/x86/kernel/cpu/cpufreq/speedstep-lib.h
index b11bcc608cac..2b6c04e5a304 100644
--- a/arch/x86/kernel/cpu/cpufreq/speedstep-lib.h
+++ b/arch/x86/kernel/cpu/cpufreq/speedstep-lib.h
@@ -12,17 +12,17 @@
12 12
13/* processors */ 13/* processors */
14 14
15#define SPEEDSTEP_PROCESSOR_PIII_C_EARLY 0x00000001 /* Coppermine core */ 15#define SPEEDSTEP_CPU_PIII_C_EARLY 0x00000001 /* Coppermine core */
16#define SPEEDSTEP_PROCESSOR_PIII_C 0x00000002 /* Coppermine core */ 16#define SPEEDSTEP_CPU_PIII_C 0x00000002 /* Coppermine core */
17#define SPEEDSTEP_PROCESSOR_PIII_T 0x00000003 /* Tualatin core */ 17#define SPEEDSTEP_CPU_PIII_T 0x00000003 /* Tualatin core */
18#define SPEEDSTEP_PROCESSOR_P4M 0x00000004 /* P4-M */ 18#define SPEEDSTEP_CPU_P4M 0x00000004 /* P4-M */
19 19
20/* the following processors are not speedstep-capable and are not auto-detected 20/* the following processors are not speedstep-capable and are not auto-detected
21 * in speedstep_detect_processor(). However, their speed can be detected using 21 * in speedstep_detect_processor(). However, their speed can be detected using
22 * the speedstep_get_processor_frequency() call. */ 22 * the speedstep_get_frequency() call. */
23#define SPEEDSTEP_PROCESSOR_PM 0xFFFFFF03 /* Pentium M */ 23#define SPEEDSTEP_CPU_PM 0xFFFFFF03 /* Pentium M */
24#define SPEEDSTEP_PROCESSOR_P4D 0xFFFFFF04 /* desktop P4 */ 24#define SPEEDSTEP_CPU_P4D 0xFFFFFF04 /* desktop P4 */
25#define SPEEDSTEP_PROCESSOR_PCORE 0xFFFFFF05 /* Core */ 25#define SPEEDSTEP_CPU_PCORE 0xFFFFFF05 /* Core */
26 26
27/* speedstep states -- only two of them */ 27/* speedstep states -- only two of them */
28 28
@@ -34,7 +34,7 @@
34extern unsigned int speedstep_detect_processor (void); 34extern unsigned int speedstep_detect_processor (void);
35 35
36/* detect the current speed (in khz) of the processor */ 36/* detect the current speed (in khz) of the processor */
37extern unsigned int speedstep_get_processor_frequency(unsigned int processor); 37extern unsigned int speedstep_get_frequency(unsigned int processor);
38 38
39 39
40/* detect the low and high speeds of the processor. The callback 40/* detect the low and high speeds of the processor. The callback
diff --git a/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c b/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c
index 8a85c93bd62a..befea088e4f5 100644
--- a/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c
+++ b/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c
@@ -19,8 +19,8 @@
19#include <linux/cpufreq.h> 19#include <linux/cpufreq.h>
20#include <linux/slab.h> 20#include <linux/slab.h>
21#include <linux/delay.h> 21#include <linux/delay.h>
22#include <linux/io.h>
22#include <asm/ist.h> 23#include <asm/ist.h>
23#include <asm/io.h>
24 24
25#include "speedstep-lib.h" 25#include "speedstep-lib.h"
26 26
@@ -30,12 +30,12 @@
30 * If user gives it, these are used. 30 * If user gives it, these are used.
31 * 31 *
32 */ 32 */
33static int smi_port = 0; 33static int smi_port;
34static int smi_cmd = 0; 34static int smi_cmd;
35static unsigned int smi_sig = 0; 35static unsigned int smi_sig;
36 36
37/* info about the processor */ 37/* info about the processor */
38static unsigned int speedstep_processor = 0; 38static unsigned int speedstep_processor;
39 39
40/* 40/*
41 * There are only two frequency states for each processor. Values 41 * There are only two frequency states for each processor. Values
@@ -56,12 +56,13 @@ static struct cpufreq_frequency_table speedstep_freqs[] = {
56 * of DMA activity going on? */ 56 * of DMA activity going on? */
57#define SMI_TRIES 5 57#define SMI_TRIES 5
58 58
59#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "speedstep-smi", msg) 59#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, \
60 "speedstep-smi", msg)
60 61
61/** 62/**
62 * speedstep_smi_ownership 63 * speedstep_smi_ownership
63 */ 64 */
64static int speedstep_smi_ownership (void) 65static int speedstep_smi_ownership(void)
65{ 66{
66 u32 command, result, magic, dummy; 67 u32 command, result, magic, dummy;
67 u32 function = GET_SPEEDSTEP_OWNER; 68 u32 function = GET_SPEEDSTEP_OWNER;
@@ -70,16 +71,18 @@ static int speedstep_smi_ownership (void)
70 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff); 71 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff);
71 magic = virt_to_phys(magic_data); 72 magic = virt_to_phys(magic_data);
72 73
73 dprintk("trying to obtain ownership with command %x at port %x\n", command, smi_port); 74 dprintk("trying to obtain ownership with command %x at port %x\n",
75 command, smi_port);
74 76
75 __asm__ __volatile__( 77 __asm__ __volatile__(
76 "push %%ebp\n" 78 "push %%ebp\n"
77 "out %%al, (%%dx)\n" 79 "out %%al, (%%dx)\n"
78 "pop %%ebp\n" 80 "pop %%ebp\n"
79 : "=D" (result), "=a" (dummy), "=b" (dummy), "=c" (dummy), "=d" (dummy), 81 : "=D" (result),
80 "=S" (dummy) 82 "=a" (dummy), "=b" (dummy), "=c" (dummy), "=d" (dummy),
83 "=S" (dummy)
81 : "a" (command), "b" (function), "c" (0), "d" (smi_port), 84 : "a" (command), "b" (function), "c" (0), "d" (smi_port),
82 "D" (0), "S" (magic) 85 "D" (0), "S" (magic)
83 : "memory" 86 : "memory"
84 ); 87 );
85 88
@@ -97,10 +100,10 @@ static int speedstep_smi_ownership (void)
97 * even hangs [cf. bugme.osdl.org # 1422] on earlier systems. Empirical testing 100 * even hangs [cf. bugme.osdl.org # 1422] on earlier systems. Empirical testing
98 * shows that the latter occurs if !(ist_info.event & 0xFFFF). 101 * shows that the latter occurs if !(ist_info.event & 0xFFFF).
99 */ 102 */
100static int speedstep_smi_get_freqs (unsigned int *low, unsigned int *high) 103static int speedstep_smi_get_freqs(unsigned int *low, unsigned int *high)
101{ 104{
102 u32 command, result = 0, edi, high_mhz, low_mhz, dummy; 105 u32 command, result = 0, edi, high_mhz, low_mhz, dummy;
103 u32 state=0; 106 u32 state = 0;
104 u32 function = GET_SPEEDSTEP_FREQS; 107 u32 function = GET_SPEEDSTEP_FREQS;
105 108
106 if (!(ist_info.event & 0xFFFF)) { 109 if (!(ist_info.event & 0xFFFF)) {
@@ -110,17 +113,25 @@ static int speedstep_smi_get_freqs (unsigned int *low, unsigned int *high)
110 113
111 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff); 114 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff);
112 115
113 dprintk("trying to determine frequencies with command %x at port %x\n", command, smi_port); 116 dprintk("trying to determine frequencies with command %x at port %x\n",
117 command, smi_port);
114 118
115 __asm__ __volatile__( 119 __asm__ __volatile__(
116 "push %%ebp\n" 120 "push %%ebp\n"
117 "out %%al, (%%dx)\n" 121 "out %%al, (%%dx)\n"
118 "pop %%ebp" 122 "pop %%ebp"
119 : "=a" (result), "=b" (high_mhz), "=c" (low_mhz), "=d" (state), "=D" (edi), "=S" (dummy) 123 : "=a" (result),
120 : "a" (command), "b" (function), "c" (state), "d" (smi_port), "S" (0), "D" (0) 124 "=b" (high_mhz),
125 "=c" (low_mhz),
126 "=d" (state), "=D" (edi), "=S" (dummy)
127 : "a" (command),
128 "b" (function),
129 "c" (state),
130 "d" (smi_port), "S" (0), "D" (0)
121 ); 131 );
122 132
123 dprintk("result %x, low_freq %u, high_freq %u\n", result, low_mhz, high_mhz); 133 dprintk("result %x, low_freq %u, high_freq %u\n",
134 result, low_mhz, high_mhz);
124 135
125 /* abort if results are obviously incorrect... */ 136 /* abort if results are obviously incorrect... */
126 if ((high_mhz + low_mhz) < 600) 137 if ((high_mhz + low_mhz) < 600)
@@ -137,26 +148,30 @@ static int speedstep_smi_get_freqs (unsigned int *low, unsigned int *high)
137 * @state: processor frequency state (SPEEDSTEP_LOW or SPEEDSTEP_HIGH) 148 * @state: processor frequency state (SPEEDSTEP_LOW or SPEEDSTEP_HIGH)
138 * 149 *
139 */ 150 */
140static int speedstep_get_state (void) 151static int speedstep_get_state(void)
141{ 152{
142 u32 function=GET_SPEEDSTEP_STATE; 153 u32 function = GET_SPEEDSTEP_STATE;
143 u32 result, state, edi, command, dummy; 154 u32 result, state, edi, command, dummy;
144 155
145 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff); 156 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff);
146 157
147 dprintk("trying to determine current setting with command %x at port %x\n", command, smi_port); 158 dprintk("trying to determine current setting with command %x "
159 "at port %x\n", command, smi_port);
148 160
149 __asm__ __volatile__( 161 __asm__ __volatile__(
150 "push %%ebp\n" 162 "push %%ebp\n"
151 "out %%al, (%%dx)\n" 163 "out %%al, (%%dx)\n"
152 "pop %%ebp\n" 164 "pop %%ebp\n"
153 : "=a" (result), "=b" (state), "=D" (edi), "=c" (dummy), "=d" (dummy), "=S" (dummy) 165 : "=a" (result),
154 : "a" (command), "b" (function), "c" (0), "d" (smi_port), "S" (0), "D" (0) 166 "=b" (state), "=D" (edi),
167 "=c" (dummy), "=d" (dummy), "=S" (dummy)
168 : "a" (command), "b" (function), "c" (0),
169 "d" (smi_port), "S" (0), "D" (0)
155 ); 170 );
156 171
157 dprintk("state is %x, result is %x\n", state, result); 172 dprintk("state is %x, result is %x\n", state, result);
158 173
159 return (state & 1); 174 return state & 1;
160} 175}
161 176
162 177
@@ -165,11 +180,11 @@ static int speedstep_get_state (void)
165 * @state: new processor frequency state (SPEEDSTEP_LOW or SPEEDSTEP_HIGH) 180 * @state: new processor frequency state (SPEEDSTEP_LOW or SPEEDSTEP_HIGH)
166 * 181 *
167 */ 182 */
168static void speedstep_set_state (unsigned int state) 183static void speedstep_set_state(unsigned int state)
169{ 184{
170 unsigned int result = 0, command, new_state, dummy; 185 unsigned int result = 0, command, new_state, dummy;
171 unsigned long flags; 186 unsigned long flags;
172 unsigned int function=SET_SPEEDSTEP_STATE; 187 unsigned int function = SET_SPEEDSTEP_STATE;
173 unsigned int retry = 0; 188 unsigned int retry = 0;
174 189
175 if (state > 0x1) 190 if (state > 0x1)
@@ -180,11 +195,14 @@ static void speedstep_set_state (unsigned int state)
180 195
181 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff); 196 command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff);
182 197
183 dprintk("trying to set frequency to state %u with command %x at port %x\n", state, command, smi_port); 198 dprintk("trying to set frequency to state %u "
199 "with command %x at port %x\n",
200 state, command, smi_port);
184 201
185 do { 202 do {
186 if (retry) { 203 if (retry) {
187 dprintk("retry %u, previous result %u, waiting...\n", retry, result); 204 dprintk("retry %u, previous result %u, waiting...\n",
205 retry, result);
188 mdelay(retry * 50); 206 mdelay(retry * 50);
189 } 207 }
190 retry++; 208 retry++;
@@ -192,20 +210,26 @@ static void speedstep_set_state (unsigned int state)
192 "push %%ebp\n" 210 "push %%ebp\n"
193 "out %%al, (%%dx)\n" 211 "out %%al, (%%dx)\n"
194 "pop %%ebp" 212 "pop %%ebp"
195 : "=b" (new_state), "=D" (result), "=c" (dummy), "=a" (dummy), 213 : "=b" (new_state), "=D" (result),
196 "=d" (dummy), "=S" (dummy) 214 "=c" (dummy), "=a" (dummy),
197 : "a" (command), "b" (function), "c" (state), "d" (smi_port), "S" (0), "D" (0) 215 "=d" (dummy), "=S" (dummy)
216 : "a" (command), "b" (function), "c" (state),
217 "d" (smi_port), "S" (0), "D" (0)
198 ); 218 );
199 } while ((new_state != state) && (retry <= SMI_TRIES)); 219 } while ((new_state != state) && (retry <= SMI_TRIES));
200 220
201 /* enable IRQs */ 221 /* enable IRQs */
202 local_irq_restore(flags); 222 local_irq_restore(flags);
203 223
204 if (new_state == state) { 224 if (new_state == state)
205 dprintk("change to %u MHz succeeded after %u tries with result %u\n", (speedstep_freqs[new_state].frequency / 1000), retry, result); 225 dprintk("change to %u MHz succeeded after %u tries "
206 } else { 226 "with result %u\n",
207 printk(KERN_ERR "cpufreq: change to state %u failed with new_state %u and result %u\n", state, new_state, result); 227 (speedstep_freqs[new_state].frequency / 1000),
208 } 228 retry, result);
229 else
230 printk(KERN_ERR "cpufreq: change to state %u "
231 "failed with new_state %u and result %u\n",
232 state, new_state, result);
209 233
210 return; 234 return;
211} 235}
@@ -219,13 +243,14 @@ static void speedstep_set_state (unsigned int state)
219 * 243 *
220 * Sets a new CPUFreq policy/freq. 244 * Sets a new CPUFreq policy/freq.
221 */ 245 */
222static int speedstep_target (struct cpufreq_policy *policy, 246static int speedstep_target(struct cpufreq_policy *policy,
223 unsigned int target_freq, unsigned int relation) 247 unsigned int target_freq, unsigned int relation)
224{ 248{
225 unsigned int newstate = 0; 249 unsigned int newstate = 0;
226 struct cpufreq_freqs freqs; 250 struct cpufreq_freqs freqs;
227 251
228 if (cpufreq_frequency_table_target(policy, &speedstep_freqs[0], target_freq, relation, &newstate)) 252 if (cpufreq_frequency_table_target(policy, &speedstep_freqs[0],
253 target_freq, relation, &newstate))
229 return -EINVAL; 254 return -EINVAL;
230 255
231 freqs.old = speedstep_freqs[speedstep_get_state()].frequency; 256 freqs.old = speedstep_freqs[speedstep_get_state()].frequency;
@@ -250,7 +275,7 @@ static int speedstep_target (struct cpufreq_policy *policy,
250 * Limit must be within speedstep_low_freq and speedstep_high_freq, with 275 * Limit must be within speedstep_low_freq and speedstep_high_freq, with
251 * at least one border included. 276 * at least one border included.
252 */ 277 */
253static int speedstep_verify (struct cpufreq_policy *policy) 278static int speedstep_verify(struct cpufreq_policy *policy)
254{ 279{
255 return cpufreq_frequency_table_verify(policy, &speedstep_freqs[0]); 280 return cpufreq_frequency_table_verify(policy, &speedstep_freqs[0]);
256} 281}
@@ -259,7 +284,8 @@ static int speedstep_verify (struct cpufreq_policy *policy)
259static int speedstep_cpu_init(struct cpufreq_policy *policy) 284static int speedstep_cpu_init(struct cpufreq_policy *policy)
260{ 285{
261 int result; 286 int result;
262 unsigned int speed,state; 287 unsigned int speed, state;
288 unsigned int *low, *high;
263 289
264 /* capability check */ 290 /* capability check */
265 if (policy->cpu != 0) 291 if (policy->cpu != 0)
@@ -272,19 +298,23 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
272 } 298 }
273 299
274 /* detect low and high frequency */ 300 /* detect low and high frequency */
275 result = speedstep_smi_get_freqs(&speedstep_freqs[SPEEDSTEP_LOW].frequency, 301 low = &speedstep_freqs[SPEEDSTEP_LOW].frequency;
276 &speedstep_freqs[SPEEDSTEP_HIGH].frequency); 302 high = &speedstep_freqs[SPEEDSTEP_HIGH].frequency;
303
304 result = speedstep_smi_get_freqs(low, high);
277 if (result) { 305 if (result) {
278 /* fall back to speedstep_lib.c dection mechanism: try both states out */ 306 /* fall back to speedstep_lib.c dection mechanism:
279 dprintk("could not detect low and high frequencies by SMI call.\n"); 307 * try both states out */
308 dprintk("could not detect low and high frequencies "
309 "by SMI call.\n");
280 result = speedstep_get_freqs(speedstep_processor, 310 result = speedstep_get_freqs(speedstep_processor,
281 &speedstep_freqs[SPEEDSTEP_LOW].frequency, 311 low, high,
282 &speedstep_freqs[SPEEDSTEP_HIGH].frequency,
283 NULL, 312 NULL,
284 &speedstep_set_state); 313 &speedstep_set_state);
285 314
286 if (result) { 315 if (result) {
287 dprintk("could not detect two different speeds -- aborting.\n"); 316 dprintk("could not detect two different speeds"
317 " -- aborting.\n");
288 return result; 318 return result;
289 } else 319 } else
290 dprintk("workaround worked.\n"); 320 dprintk("workaround worked.\n");
@@ -295,7 +325,8 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
295 speed = speedstep_freqs[state].frequency; 325 speed = speedstep_freqs[state].frequency;
296 326
297 dprintk("currently at %s speed setting - %i MHz\n", 327 dprintk("currently at %s speed setting - %i MHz\n",
298 (speed == speedstep_freqs[SPEEDSTEP_LOW].frequency) ? "low" : "high", 328 (speed == speedstep_freqs[SPEEDSTEP_LOW].frequency)
329 ? "low" : "high",
299 (speed / 1000)); 330 (speed / 1000));
300 331
301 /* cpuinfo and default policy values */ 332 /* cpuinfo and default policy values */
@@ -304,7 +335,7 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
304 335
305 result = cpufreq_frequency_table_cpuinfo(policy, speedstep_freqs); 336 result = cpufreq_frequency_table_cpuinfo(policy, speedstep_freqs);
306 if (result) 337 if (result)
307 return (result); 338 return result;
308 339
309 cpufreq_frequency_table_get_attr(speedstep_freqs, policy->cpu); 340 cpufreq_frequency_table_get_attr(speedstep_freqs, policy->cpu);
310 341
@@ -321,7 +352,7 @@ static unsigned int speedstep_get(unsigned int cpu)
321{ 352{
322 if (cpu) 353 if (cpu)
323 return -ENODEV; 354 return -ENODEV;
324 return speedstep_get_processor_frequency(speedstep_processor); 355 return speedstep_get_frequency(speedstep_processor);
325} 356}
326 357
327 358
@@ -335,7 +366,7 @@ static int speedstep_resume(struct cpufreq_policy *policy)
335 return result; 366 return result;
336} 367}
337 368
338static struct freq_attr* speedstep_attr[] = { 369static struct freq_attr *speedstep_attr[] = {
339 &cpufreq_freq_attr_scaling_available_freqs, 370 &cpufreq_freq_attr_scaling_available_freqs,
340 NULL, 371 NULL,
341}; 372};
@@ -364,21 +395,23 @@ static int __init speedstep_init(void)
364 speedstep_processor = speedstep_detect_processor(); 395 speedstep_processor = speedstep_detect_processor();
365 396
366 switch (speedstep_processor) { 397 switch (speedstep_processor) {
367 case SPEEDSTEP_PROCESSOR_PIII_T: 398 case SPEEDSTEP_CPU_PIII_T:
368 case SPEEDSTEP_PROCESSOR_PIII_C: 399 case SPEEDSTEP_CPU_PIII_C:
369 case SPEEDSTEP_PROCESSOR_PIII_C_EARLY: 400 case SPEEDSTEP_CPU_PIII_C_EARLY:
370 break; 401 break;
371 default: 402 default:
372 speedstep_processor = 0; 403 speedstep_processor = 0;
373 } 404 }
374 405
375 if (!speedstep_processor) { 406 if (!speedstep_processor) {
376 dprintk ("No supported Intel CPU detected.\n"); 407 dprintk("No supported Intel CPU detected.\n");
377 return -ENODEV; 408 return -ENODEV;
378 } 409 }
379 410
380 dprintk("signature:0x%.8lx, command:0x%.8lx, event:0x%.8lx, perf_level:0x%.8lx.\n", 411 dprintk("signature:0x%.8lx, command:0x%.8lx, "
381 ist_info.signature, ist_info.command, ist_info.event, ist_info.perf_level); 412 "event:0x%.8lx, perf_level:0x%.8lx.\n",
413 ist_info.signature, ist_info.command,
414 ist_info.event, ist_info.perf_level);
382 415
383 /* Error if no IST-SMI BIOS or no PARM 416 /* Error if no IST-SMI BIOS or no PARM
384 sig= 'ISGE' aka 'Intel Speedstep Gate E' */ 417 sig= 'ISGE' aka 'Intel Speedstep Gate E' */
@@ -416,17 +449,20 @@ static void __exit speedstep_exit(void)
416 cpufreq_unregister_driver(&speedstep_driver); 449 cpufreq_unregister_driver(&speedstep_driver);
417} 450}
418 451
419module_param(smi_port, int, 0444); 452module_param(smi_port, int, 0444);
420module_param(smi_cmd, int, 0444); 453module_param(smi_cmd, int, 0444);
421module_param(smi_sig, uint, 0444); 454module_param(smi_sig, uint, 0444);
422 455
423MODULE_PARM_DESC(smi_port, "Override the BIOS-given IST port with this value -- Intel's default setting is 0xb2"); 456MODULE_PARM_DESC(smi_port, "Override the BIOS-given IST port with this value "
424MODULE_PARM_DESC(smi_cmd, "Override the BIOS-given IST command with this value -- Intel's default setting is 0x82"); 457 "-- Intel's default setting is 0xb2");
425MODULE_PARM_DESC(smi_sig, "Set to 1 to fake the IST signature when using the SMI interface."); 458MODULE_PARM_DESC(smi_cmd, "Override the BIOS-given IST command with this value "
459 "-- Intel's default setting is 0x82");
460MODULE_PARM_DESC(smi_sig, "Set to 1 to fake the IST signature when using the "
461 "SMI interface.");
426 462
427MODULE_AUTHOR ("Hiroshi Miura"); 463MODULE_AUTHOR("Hiroshi Miura");
428MODULE_DESCRIPTION ("Speedstep driver for IST applet SMI interface."); 464MODULE_DESCRIPTION("Speedstep driver for IST applet SMI interface.");
429MODULE_LICENSE ("GPL"); 465MODULE_LICENSE("GPL");
430 466
431module_init(speedstep_init); 467module_init(speedstep_init);
432module_exit(speedstep_exit); 468module_exit(speedstep_exit);
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index b09d4eb52bb9..7437fa133c02 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -4,6 +4,7 @@
4#include <linux/string.h> 4#include <linux/string.h>
5#include <linux/bitops.h> 5#include <linux/bitops.h>
6#include <linux/smp.h> 6#include <linux/smp.h>
7#include <linux/sched.h>
7#include <linux/thread_info.h> 8#include <linux/thread_info.h>
8#include <linux/module.h> 9#include <linux/module.h>
9 10
@@ -61,11 +62,16 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
61 62
62 /* 63 /*
63 * c->x86_power is 8000_0007 edx. Bit 8 is TSC runs at constant rate 64 * c->x86_power is 8000_0007 edx. Bit 8 is TSC runs at constant rate
64 * with P/T states and does not stop in deep C-states 65 * with P/T states and does not stop in deep C-states.
66 *
67 * It is also reliable across cores and sockets. (but not across
68 * cabinets - we turn it off in that case explicitly.)
65 */ 69 */
66 if (c->x86_power & (1 << 8)) { 70 if (c->x86_power & (1 << 8)) {
67 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); 71 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
68 set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC); 72 set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
73 set_cpu_cap(c, X86_FEATURE_TSC_RELIABLE);
74 sched_clock_stable = 1;
69 } 75 }
70 76
71 /* 77 /*
diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c
index c471eb1a389c..483eda96e102 100644
--- a/arch/x86/kernel/cpu/intel_cacheinfo.c
+++ b/arch/x86/kernel/cpu/intel_cacheinfo.c
@@ -159,7 +159,7 @@ struct _cpuid4_info_regs {
159 unsigned long can_disable; 159 unsigned long can_disable;
160}; 160};
161 161
162#ifdef CONFIG_PCI 162#if defined(CONFIG_PCI) && defined(CONFIG_SYSFS)
163static struct pci_device_id k8_nb_id[] = { 163static struct pci_device_id k8_nb_id[] = {
164 { PCI_DEVICE(PCI_VENDOR_ID_AMD, 0x1103) }, 164 { PCI_DEVICE(PCI_VENDOR_ID_AMD, 0x1103) },
165 { PCI_DEVICE(PCI_VENDOR_ID_AMD, 0x1203) }, 165 { PCI_DEVICE(PCI_VENDOR_ID_AMD, 0x1203) },
@@ -324,15 +324,6 @@ __cpuinit cpuid4_cache_lookup_regs(int index,
324 return 0; 324 return 0;
325} 325}
326 326
327static int
328__cpuinit cpuid4_cache_lookup(int index, struct _cpuid4_info *this_leaf)
329{
330 struct _cpuid4_info_regs *leaf_regs =
331 (struct _cpuid4_info_regs *)this_leaf;
332
333 return cpuid4_cache_lookup_regs(index, leaf_regs);
334}
335
336static int __cpuinit find_num_cache_leaves(void) 327static int __cpuinit find_num_cache_leaves(void)
337{ 328{
338 unsigned int eax, ebx, ecx, edx; 329 unsigned int eax, ebx, ecx, edx;
@@ -508,6 +499,8 @@ unsigned int __cpuinit init_intel_cacheinfo(struct cpuinfo_x86 *c)
508 return l2; 499 return l2;
509} 500}
510 501
502#ifdef CONFIG_SYSFS
503
511/* pointer to _cpuid4_info array (for each cache leaf) */ 504/* pointer to _cpuid4_info array (for each cache leaf) */
512static DEFINE_PER_CPU(struct _cpuid4_info *, cpuid4_info); 505static DEFINE_PER_CPU(struct _cpuid4_info *, cpuid4_info);
513#define CPUID4_INFO_IDX(x, y) (&((per_cpu(cpuid4_info, x))[y])) 506#define CPUID4_INFO_IDX(x, y) (&((per_cpu(cpuid4_info, x))[y]))
@@ -571,6 +564,15 @@ static void __cpuinit free_cache_attributes(unsigned int cpu)
571 per_cpu(cpuid4_info, cpu) = NULL; 564 per_cpu(cpuid4_info, cpu) = NULL;
572} 565}
573 566
567static int
568__cpuinit cpuid4_cache_lookup(int index, struct _cpuid4_info *this_leaf)
569{
570 struct _cpuid4_info_regs *leaf_regs =
571 (struct _cpuid4_info_regs *)this_leaf;
572
573 return cpuid4_cache_lookup_regs(index, leaf_regs);
574}
575
574static void __cpuinit get_cpu_leaves(void *_retval) 576static void __cpuinit get_cpu_leaves(void *_retval)
575{ 577{
576 int j, *retval = _retval, cpu = smp_processor_id(); 578 int j, *retval = _retval, cpu = smp_processor_id();
@@ -612,8 +614,6 @@ static int __cpuinit detect_cache_attributes(unsigned int cpu)
612 return retval; 614 return retval;
613} 615}
614 616
615#ifdef CONFIG_SYSFS
616
617#include <linux/kobject.h> 617#include <linux/kobject.h>
618#include <linux/sysfs.h> 618#include <linux/sysfs.h>
619 619
diff --git a/arch/x86/kernel/cpu/mcheck/mce_64.c b/arch/x86/kernel/cpu/mcheck/mce_64.c
index ca14604611ec..863f89568b1a 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_64.c
@@ -990,7 +990,7 @@ static struct sysdev_attribute *mce_attributes[] = {
990 NULL 990 NULL
991}; 991};
992 992
993static cpumask_t mce_device_initialized = CPU_MASK_NONE; 993static cpumask_var_t mce_device_initialized;
994 994
995/* Per cpu sysdev init. All of the cpus still share the same ctl bank */ 995/* Per cpu sysdev init. All of the cpus still share the same ctl bank */
996static __cpuinit int mce_create_device(unsigned int cpu) 996static __cpuinit int mce_create_device(unsigned int cpu)
@@ -1021,7 +1021,7 @@ static __cpuinit int mce_create_device(unsigned int cpu)
1021 if (err) 1021 if (err)
1022 goto error2; 1022 goto error2;
1023 } 1023 }
1024 cpu_set(cpu, mce_device_initialized); 1024 cpumask_set_cpu(cpu, mce_device_initialized);
1025 1025
1026 return 0; 1026 return 0;
1027error2: 1027error2:
@@ -1043,7 +1043,7 @@ static __cpuinit void mce_remove_device(unsigned int cpu)
1043{ 1043{
1044 int i; 1044 int i;
1045 1045
1046 if (!cpu_isset(cpu, mce_device_initialized)) 1046 if (!cpumask_test_cpu(cpu, mce_device_initialized))
1047 return; 1047 return;
1048 1048
1049 for (i = 0; mce_attributes[i]; i++) 1049 for (i = 0; mce_attributes[i]; i++)
@@ -1053,7 +1053,7 @@ static __cpuinit void mce_remove_device(unsigned int cpu)
1053 sysdev_remove_file(&per_cpu(device_mce, cpu), 1053 sysdev_remove_file(&per_cpu(device_mce, cpu),
1054 &bank_attrs[i]); 1054 &bank_attrs[i]);
1055 sysdev_unregister(&per_cpu(device_mce,cpu)); 1055 sysdev_unregister(&per_cpu(device_mce,cpu));
1056 cpu_clear(cpu, mce_device_initialized); 1056 cpumask_clear_cpu(cpu, mce_device_initialized);
1057} 1057}
1058 1058
1059/* Make sure there are no machine checks on offlined CPUs. */ 1059/* Make sure there are no machine checks on offlined CPUs. */
@@ -1162,6 +1162,8 @@ static __init int mce_init_device(void)
1162 if (!mce_available(&boot_cpu_data)) 1162 if (!mce_available(&boot_cpu_data))
1163 return -EIO; 1163 return -EIO;
1164 1164
1165 alloc_cpumask_var(&mce_device_initialized, GFP_KERNEL);
1166
1165 err = mce_init_banks(); 1167 err = mce_init_banks();
1166 if (err) 1168 if (err)
1167 return err; 1169 return err;
diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
index 7d01be868870..56dde9c4bc96 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
@@ -485,7 +485,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
485 485
486#ifdef CONFIG_SMP 486#ifdef CONFIG_SMP
487 if (cpu_data(cpu).cpu_core_id && shared_bank[bank]) { /* symlink */ 487 if (cpu_data(cpu).cpu_core_id && shared_bank[bank]) { /* symlink */
488 i = cpumask_first(&per_cpu(cpu_core_map, cpu)); 488 i = cpumask_first(cpu_core_mask(cpu));
489 489
490 /* first core not up yet */ 490 /* first core not up yet */
491 if (cpu_data(i).cpu_core_id) 491 if (cpu_data(i).cpu_core_id)
@@ -505,7 +505,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
505 if (err) 505 if (err)
506 goto out; 506 goto out;
507 507
508 cpumask_copy(b->cpus, &per_cpu(cpu_core_map, cpu)); 508 cpumask_copy(b->cpus, cpu_core_mask(cpu));
509 per_cpu(threshold_banks, cpu)[bank] = b; 509 per_cpu(threshold_banks, cpu)[bank] = b;
510 goto out; 510 goto out;
511 } 511 }
@@ -529,7 +529,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
529#ifndef CONFIG_SMP 529#ifndef CONFIG_SMP
530 cpumask_setall(b->cpus); 530 cpumask_setall(b->cpus);
531#else 531#else
532 cpumask_copy(b->cpus, &per_cpu(cpu_core_map, cpu)); 532 cpumask_copy(b->cpus, cpu_core_mask(cpu));
533#endif 533#endif
534 534
535 per_cpu(threshold_banks, cpu)[bank] = b; 535 per_cpu(threshold_banks, cpu)[bank] = b;
diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c b/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
index 57df3d383470..d6b72df89d69 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
@@ -249,7 +249,7 @@ void cmci_rediscover(int dying)
249 for_each_online_cpu (cpu) { 249 for_each_online_cpu (cpu) {
250 if (cpu == dying) 250 if (cpu == dying)
251 continue; 251 continue;
252 if (set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu))) 252 if (set_cpus_allowed_ptr(current, cpumask_of(cpu)))
253 continue; 253 continue;
254 /* Recheck banks in case CPUs don't all have the same */ 254 /* Recheck banks in case CPUs don't all have the same */
255 if (cmci_supported(&banks)) 255 if (cmci_supported(&banks))
diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index 37f28fc7cf95..0b776c09aff3 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -462,9 +462,6 @@ static void generic_get_mtrr(unsigned int reg, unsigned long *base,
462 *base = base_hi << (32 - PAGE_SHIFT) | base_lo >> PAGE_SHIFT; 462 *base = base_hi << (32 - PAGE_SHIFT) | base_lo >> PAGE_SHIFT;
463 *type = base_lo & 0xff; 463 *type = base_lo & 0xff;
464 464
465 printk(KERN_DEBUG " get_mtrr: cpu%d reg%02d base=%010lx size=%010lx %s\n",
466 cpu, reg, *base, *size,
467 mtrr_attrib_to_str(*type & 0xff));
468out_put_cpu: 465out_put_cpu:
469 put_cpu(); 466 put_cpu();
470} 467}
diff --git a/arch/x86/kernel/cpu/mtrr/if.c b/arch/x86/kernel/cpu/mtrr/if.c
index 4c4214690dd1..fb73a52913a4 100644
--- a/arch/x86/kernel/cpu/mtrr/if.c
+++ b/arch/x86/kernel/cpu/mtrr/if.c
@@ -377,10 +377,6 @@ static const struct file_operations mtrr_fops = {
377 .release = mtrr_close, 377 .release = mtrr_close,
378}; 378};
379 379
380
381static struct proc_dir_entry *proc_root_mtrr;
382
383
384static int mtrr_seq_show(struct seq_file *seq, void *offset) 380static int mtrr_seq_show(struct seq_file *seq, void *offset)
385{ 381{
386 char factor; 382 char factor;
@@ -423,11 +419,7 @@ static int __init mtrr_if_init(void)
423 (!cpu_has(c, X86_FEATURE_CENTAUR_MCR))) 419 (!cpu_has(c, X86_FEATURE_CENTAUR_MCR)))
424 return -ENODEV; 420 return -ENODEV;
425 421
426 proc_root_mtrr = 422 proc_create("mtrr", S_IWUSR | S_IRUGO, NULL, &mtrr_fops);
427 proc_create("mtrr", S_IWUSR | S_IRUGO, NULL, &mtrr_fops);
428
429 if (proc_root_mtrr)
430 proc_root_mtrr->owner = THIS_MODULE;
431 return 0; 423 return 0;
432} 424}
433 425
diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c
index d67e0e48bc2d..f93047fed791 100644
--- a/arch/x86/kernel/cpu/proc.c
+++ b/arch/x86/kernel/cpu/proc.c
@@ -14,7 +14,7 @@ static void show_cpuinfo_core(struct seq_file *m, struct cpuinfo_x86 *c,
14 if (c->x86_max_cores * smp_num_siblings > 1) { 14 if (c->x86_max_cores * smp_num_siblings > 1) {
15 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id); 15 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id);
16 seq_printf(m, "siblings\t: %d\n", 16 seq_printf(m, "siblings\t: %d\n",
17 cpus_weight(per_cpu(cpu_core_map, cpu))); 17 cpumask_weight(cpu_sibling_mask(cpu)));
18 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id); 18 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id);
19 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores); 19 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
20 seq_printf(m, "apicid\t\t: %d\n", c->apicid); 20 seq_printf(m, "apicid\t\t: %d\n", c->apicid);
@@ -143,9 +143,9 @@ static int show_cpuinfo(struct seq_file *m, void *v)
143static void *c_start(struct seq_file *m, loff_t *pos) 143static void *c_start(struct seq_file *m, loff_t *pos)
144{ 144{
145 if (*pos == 0) /* just in case, cpu 0 is not the first */ 145 if (*pos == 0) /* just in case, cpu 0 is not the first */
146 *pos = first_cpu(cpu_online_map); 146 *pos = cpumask_first(cpu_online_mask);
147 else 147 else
148 *pos = next_cpu_nr(*pos - 1, cpu_online_map); 148 *pos = cpumask_next(*pos - 1, cpu_online_mask);
149 if ((*pos) < nr_cpu_ids) 149 if ((*pos) < nr_cpu_ids)
150 return &cpu_data(*pos); 150 return &cpu_data(*pos);
151 return NULL; 151 return NULL;
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index 87d103ded1c3..95ea5fa7d444 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c
@@ -10,10 +10,12 @@
10#include <linux/kdebug.h> 10#include <linux/kdebug.h>
11#include <linux/module.h> 11#include <linux/module.h>
12#include <linux/ptrace.h> 12#include <linux/ptrace.h>
13#include <linux/ftrace.h>
13#include <linux/kexec.h> 14#include <linux/kexec.h>
14#include <linux/bug.h> 15#include <linux/bug.h>
15#include <linux/nmi.h> 16#include <linux/nmi.h>
16#include <linux/sysfs.h> 17#include <linux/sysfs.h>
18#include <linux/ftrace.h>
17 19
18#include <asm/stacktrace.h> 20#include <asm/stacktrace.h>
19 21
@@ -195,6 +197,11 @@ unsigned __kprobes long oops_begin(void)
195 int cpu; 197 int cpu;
196 unsigned long flags; 198 unsigned long flags;
197 199
200 /* notify the hw-branch tracer so it may disable tracing and
201 add the last trace to the trace buffer -
202 the earlier this happens, the more useful the trace. */
203 trace_hw_branch_oops();
204
198 oops_enter(); 205 oops_enter();
199 206
200 /* racy, but better than risking deadlock. */ 207 /* racy, but better than risking deadlock. */
diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c
index fb638d9ce6d2..ef2c3563357d 100644
--- a/arch/x86/kernel/e820.c
+++ b/arch/x86/kernel/e820.c
@@ -233,7 +233,7 @@ void __init e820_print_map(char *who)
233 */ 233 */
234 234
235int __init sanitize_e820_map(struct e820entry *biosmap, int max_nr_map, 235int __init sanitize_e820_map(struct e820entry *biosmap, int max_nr_map,
236 int *pnr_map) 236 u32 *pnr_map)
237{ 237{
238 struct change_member { 238 struct change_member {
239 struct e820entry *pbios; /* pointer to original bios entry */ 239 struct e820entry *pbios; /* pointer to original bios entry */
@@ -552,7 +552,7 @@ u64 __init e820_remove_range(u64 start, u64 size, unsigned old_type,
552 552
553void __init update_e820(void) 553void __init update_e820(void)
554{ 554{
555 int nr_map; 555 u32 nr_map;
556 556
557 nr_map = e820.nr_map; 557 nr_map = e820.nr_map;
558 if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr_map)) 558 if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr_map))
@@ -563,7 +563,7 @@ void __init update_e820(void)
563} 563}
564static void __init update_e820_saved(void) 564static void __init update_e820_saved(void)
565{ 565{
566 int nr_map; 566 u32 nr_map;
567 567
568 nr_map = e820_saved.nr_map; 568 nr_map = e820_saved.nr_map;
569 if (sanitize_e820_map(e820_saved.map, ARRAY_SIZE(e820_saved.map), &nr_map)) 569 if (sanitize_e820_map(e820_saved.map, ARRAY_SIZE(e820_saved.map), &nr_map))
@@ -1303,7 +1303,7 @@ early_param("memmap", parse_memmap_opt);
1303void __init finish_e820_parsing(void) 1303void __init finish_e820_parsing(void)
1304{ 1304{
1305 if (userdef) { 1305 if (userdef) {
1306 int nr = e820.nr_map; 1306 u32 nr = e820.nr_map;
1307 1307
1308 if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr) < 0) 1308 if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr) < 0)
1309 early_panic("Invalid user supplied memory map"); 1309 early_panic("Invalid user supplied memory map");
@@ -1386,7 +1386,7 @@ void __init e820_reserve_resources_late(void)
1386char *__init default_machine_specific_memory_setup(void) 1386char *__init default_machine_specific_memory_setup(void)
1387{ 1387{
1388 char *who = "BIOS-e820"; 1388 char *who = "BIOS-e820";
1389 int new_nr; 1389 u32 new_nr;
1390 /* 1390 /*
1391 * Try to copy the BIOS-supplied E820-map. 1391 * Try to copy the BIOS-supplied E820-map.
1392 * 1392 *
diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index 231bdd3c5b1c..18dfa30795c9 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -18,14 +18,28 @@
18#include <linux/init.h> 18#include <linux/init.h>
19#include <linux/list.h> 19#include <linux/list.h>
20 20
21#include <trace/syscall.h>
22
23#include <asm/cacheflush.h>
21#include <asm/ftrace.h> 24#include <asm/ftrace.h>
22#include <linux/ftrace.h>
23#include <asm/nops.h> 25#include <asm/nops.h>
24#include <asm/nmi.h> 26#include <asm/nmi.h>
25 27
26 28
27#ifdef CONFIG_DYNAMIC_FTRACE 29#ifdef CONFIG_DYNAMIC_FTRACE
28 30
31int ftrace_arch_code_modify_prepare(void)
32{
33 set_kernel_text_rw();
34 return 0;
35}
36
37int ftrace_arch_code_modify_post_process(void)
38{
39 set_kernel_text_ro();
40 return 0;
41}
42
29union ftrace_code_union { 43union ftrace_code_union {
30 char code[MCOUNT_INSN_SIZE]; 44 char code[MCOUNT_INSN_SIZE];
31 struct { 45 struct {
@@ -66,11 +80,11 @@ static unsigned char *ftrace_call_replace(unsigned long ip, unsigned long addr)
66 * 80 *
67 * 1) Put the instruction pointer into the IP buffer 81 * 1) Put the instruction pointer into the IP buffer
68 * and the new code into the "code" buffer. 82 * and the new code into the "code" buffer.
69 * 2) Set a flag that says we are modifying code 83 * 2) Wait for any running NMIs to finish and set a flag that says
70 * 3) Wait for any running NMIs to finish. 84 * we are modifying code, it is done in an atomic operation.
71 * 4) Write the code 85 * 3) Write the code
72 * 5) clear the flag. 86 * 4) clear the flag.
73 * 6) Wait for any running NMIs to finish. 87 * 5) Wait for any running NMIs to finish.
74 * 88 *
75 * If an NMI is executed, the first thing it does is to call 89 * If an NMI is executed, the first thing it does is to call
76 * "ftrace_nmi_enter". This will check if the flag is set to write 90 * "ftrace_nmi_enter". This will check if the flag is set to write
@@ -82,9 +96,9 @@ static unsigned char *ftrace_call_replace(unsigned long ip, unsigned long addr)
82 * are the same as what exists. 96 * are the same as what exists.
83 */ 97 */
84 98
85static atomic_t in_nmi = ATOMIC_INIT(0); 99#define MOD_CODE_WRITE_FLAG (1 << 31) /* set when NMI should do the write */
100static atomic_t nmi_running = ATOMIC_INIT(0);
86static int mod_code_status; /* holds return value of text write */ 101static int mod_code_status; /* holds return value of text write */
87static int mod_code_write; /* set when NMI should do the write */
88static void *mod_code_ip; /* holds the IP to write to */ 102static void *mod_code_ip; /* holds the IP to write to */
89static void *mod_code_newcode; /* holds the text to write to the IP */ 103static void *mod_code_newcode; /* holds the text to write to the IP */
90 104
@@ -101,6 +115,20 @@ int ftrace_arch_read_dyn_info(char *buf, int size)
101 return r; 115 return r;
102} 116}
103 117
118static void clear_mod_flag(void)
119{
120 int old = atomic_read(&nmi_running);
121
122 for (;;) {
123 int new = old & ~MOD_CODE_WRITE_FLAG;
124
125 if (old == new)
126 break;
127
128 old = atomic_cmpxchg(&nmi_running, old, new);
129 }
130}
131
104static void ftrace_mod_code(void) 132static void ftrace_mod_code(void)
105{ 133{
106 /* 134 /*
@@ -111,37 +139,52 @@ static void ftrace_mod_code(void)
111 */ 139 */
112 mod_code_status = probe_kernel_write(mod_code_ip, mod_code_newcode, 140 mod_code_status = probe_kernel_write(mod_code_ip, mod_code_newcode,
113 MCOUNT_INSN_SIZE); 141 MCOUNT_INSN_SIZE);
142
143 /* if we fail, then kill any new writers */
144 if (mod_code_status)
145 clear_mod_flag();
114} 146}
115 147
116void ftrace_nmi_enter(void) 148void ftrace_nmi_enter(void)
117{ 149{
118 atomic_inc(&in_nmi); 150 if (atomic_inc_return(&nmi_running) & MOD_CODE_WRITE_FLAG) {
119 /* Must have in_nmi seen before reading write flag */ 151 smp_rmb();
120 smp_mb();
121 if (mod_code_write) {
122 ftrace_mod_code(); 152 ftrace_mod_code();
123 atomic_inc(&nmi_update_count); 153 atomic_inc(&nmi_update_count);
124 } 154 }
155 /* Must have previous changes seen before executions */
156 smp_mb();
125} 157}
126 158
127void ftrace_nmi_exit(void) 159void ftrace_nmi_exit(void)
128{ 160{
129 /* Finish all executions before clearing in_nmi */ 161 /* Finish all executions before clearing nmi_running */
130 smp_wmb(); 162 smp_mb();
131 atomic_dec(&in_nmi); 163 atomic_dec(&nmi_running);
164}
165
166static void wait_for_nmi_and_set_mod_flag(void)
167{
168 if (!atomic_cmpxchg(&nmi_running, 0, MOD_CODE_WRITE_FLAG))
169 return;
170
171 do {
172 cpu_relax();
173 } while (atomic_cmpxchg(&nmi_running, 0, MOD_CODE_WRITE_FLAG));
174
175 nmi_wait_count++;
132} 176}
133 177
134static void wait_for_nmi(void) 178static void wait_for_nmi(void)
135{ 179{
136 int waited = 0; 180 if (!atomic_read(&nmi_running))
181 return;
137 182
138 while (atomic_read(&in_nmi)) { 183 do {
139 waited = 1;
140 cpu_relax(); 184 cpu_relax();
141 } 185 } while (atomic_read(&nmi_running));
142 186
143 if (waited) 187 nmi_wait_count++;
144 nmi_wait_count++;
145} 188}
146 189
147static int 190static int
@@ -151,14 +194,9 @@ do_ftrace_mod_code(unsigned long ip, void *new_code)
151 mod_code_newcode = new_code; 194 mod_code_newcode = new_code;
152 195
153 /* The buffers need to be visible before we let NMIs write them */ 196 /* The buffers need to be visible before we let NMIs write them */
154 smp_wmb();
155
156 mod_code_write = 1;
157
158 /* Make sure write bit is visible before we wait on NMIs */
159 smp_mb(); 197 smp_mb();
160 198
161 wait_for_nmi(); 199 wait_for_nmi_and_set_mod_flag();
162 200
163 /* Make sure all running NMIs have finished before we write the code */ 201 /* Make sure all running NMIs have finished before we write the code */
164 smp_mb(); 202 smp_mb();
@@ -166,13 +204,9 @@ do_ftrace_mod_code(unsigned long ip, void *new_code)
166 ftrace_mod_code(); 204 ftrace_mod_code();
167 205
168 /* Make sure the write happens before clearing the bit */ 206 /* Make sure the write happens before clearing the bit */
169 smp_wmb();
170
171 mod_code_write = 0;
172
173 /* make sure NMIs see the cleared bit */
174 smp_mb(); 207 smp_mb();
175 208
209 clear_mod_flag();
176 wait_for_nmi(); 210 wait_for_nmi();
177 211
178 return mod_code_status; 212 return mod_code_status;
@@ -368,100 +402,8 @@ int ftrace_disable_ftrace_graph_caller(void)
368 return ftrace_mod_jmp(ip, old_offset, new_offset); 402 return ftrace_mod_jmp(ip, old_offset, new_offset);
369} 403}
370 404
371#else /* CONFIG_DYNAMIC_FTRACE */
372
373/*
374 * These functions are picked from those used on
375 * this page for dynamic ftrace. They have been
376 * simplified to ignore all traces in NMI context.
377 */
378static atomic_t in_nmi;
379
380void ftrace_nmi_enter(void)
381{
382 atomic_inc(&in_nmi);
383}
384
385void ftrace_nmi_exit(void)
386{
387 atomic_dec(&in_nmi);
388}
389
390#endif /* !CONFIG_DYNAMIC_FTRACE */ 405#endif /* !CONFIG_DYNAMIC_FTRACE */
391 406
392/* Add a function return address to the trace stack on thread info.*/
393static int push_return_trace(unsigned long ret, unsigned long long time,
394 unsigned long func, int *depth)
395{
396 int index;
397
398 if (!current->ret_stack)
399 return -EBUSY;
400
401 /* The return trace stack is full */
402 if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) {
403 atomic_inc(&current->trace_overrun);
404 return -EBUSY;
405 }
406
407 index = ++current->curr_ret_stack;
408 barrier();
409 current->ret_stack[index].ret = ret;
410 current->ret_stack[index].func = func;
411 current->ret_stack[index].calltime = time;
412 *depth = index;
413
414 return 0;
415}
416
417/* Retrieve a function return address to the trace stack on thread info.*/
418static void pop_return_trace(struct ftrace_graph_ret *trace, unsigned long *ret)
419{
420 int index;
421
422 index = current->curr_ret_stack;
423
424 if (unlikely(index < 0)) {
425 ftrace_graph_stop();
426 WARN_ON(1);
427 /* Might as well panic, otherwise we have no where to go */
428 *ret = (unsigned long)panic;
429 return;
430 }
431
432 *ret = current->ret_stack[index].ret;
433 trace->func = current->ret_stack[index].func;
434 trace->calltime = current->ret_stack[index].calltime;
435 trace->overrun = atomic_read(&current->trace_overrun);
436 trace->depth = index;
437 barrier();
438 current->curr_ret_stack--;
439
440}
441
442/*
443 * Send the trace to the ring-buffer.
444 * @return the original return address.
445 */
446unsigned long ftrace_return_to_handler(void)
447{
448 struct ftrace_graph_ret trace;
449 unsigned long ret;
450
451 pop_return_trace(&trace, &ret);
452 trace.rettime = cpu_clock(raw_smp_processor_id());
453 ftrace_graph_return(&trace);
454
455 if (unlikely(!ret)) {
456 ftrace_graph_stop();
457 WARN_ON(1);
458 /* Might as well panic. What else to do? */
459 ret = (unsigned long)panic;
460 }
461
462 return ret;
463}
464
465/* 407/*
466 * Hook the return address and push it in the stack of return addrs 408 * Hook the return address and push it in the stack of return addrs
467 * in current thread info. 409 * in current thread info.
@@ -469,14 +411,13 @@ unsigned long ftrace_return_to_handler(void)
469void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr) 411void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
470{ 412{
471 unsigned long old; 413 unsigned long old;
472 unsigned long long calltime;
473 int faulted; 414 int faulted;
474 struct ftrace_graph_ent trace; 415 struct ftrace_graph_ent trace;
475 unsigned long return_hooker = (unsigned long) 416 unsigned long return_hooker = (unsigned long)
476 &return_to_handler; 417 &return_to_handler;
477 418
478 /* Nmi's are currently unsupported */ 419 /* Nmi's are currently unsupported */
479 if (unlikely(atomic_read(&in_nmi))) 420 if (unlikely(in_nmi()))
480 return; 421 return;
481 422
482 if (unlikely(atomic_read(&current->tracing_graph_pause))) 423 if (unlikely(atomic_read(&current->tracing_graph_pause)))
@@ -512,17 +453,7 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
512 return; 453 return;
513 } 454 }
514 455
515 if (unlikely(!__kernel_text_address(old))) { 456 if (ftrace_push_return_trace(old, self_addr, &trace.depth) == -EBUSY) {
516 ftrace_graph_stop();
517 *parent = old;
518 WARN_ON(1);
519 return;
520 }
521
522 calltime = cpu_clock(raw_smp_processor_id());
523
524 if (push_return_trace(old, calltime,
525 self_addr, &trace.depth) == -EBUSY) {
526 *parent = old; 457 *parent = old;
527 return; 458 return;
528 } 459 }
@@ -536,3 +467,66 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
536 } 467 }
537} 468}
538#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ 469#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
470
471#ifdef CONFIG_FTRACE_SYSCALLS
472
473extern unsigned long __start_syscalls_metadata[];
474extern unsigned long __stop_syscalls_metadata[];
475extern unsigned long *sys_call_table;
476
477static struct syscall_metadata **syscalls_metadata;
478
479static struct syscall_metadata *find_syscall_meta(unsigned long *syscall)
480{
481 struct syscall_metadata *start;
482 struct syscall_metadata *stop;
483 char str[KSYM_SYMBOL_LEN];
484
485
486 start = (struct syscall_metadata *)__start_syscalls_metadata;
487 stop = (struct syscall_metadata *)__stop_syscalls_metadata;
488 kallsyms_lookup((unsigned long) syscall, NULL, NULL, NULL, str);
489
490 for ( ; start < stop; start++) {
491 if (start->name && !strcmp(start->name, str))
492 return start;
493 }
494 return NULL;
495}
496
497struct syscall_metadata *syscall_nr_to_meta(int nr)
498{
499 if (!syscalls_metadata || nr >= FTRACE_SYSCALL_MAX || nr < 0)
500 return NULL;
501
502 return syscalls_metadata[nr];
503}
504
505void arch_init_ftrace_syscalls(void)
506{
507 int i;
508 struct syscall_metadata *meta;
509 unsigned long **psys_syscall_table = &sys_call_table;
510 static atomic_t refs;
511
512 if (atomic_inc_return(&refs) != 1)
513 goto end;
514
515 syscalls_metadata = kzalloc(sizeof(*syscalls_metadata) *
516 FTRACE_SYSCALL_MAX, GFP_KERNEL);
517 if (!syscalls_metadata) {
518 WARN_ON(1);
519 return;
520 }
521
522 for (i = 0; i < FTRACE_SYSCALL_MAX; i++) {
523 meta = find_syscall_meta(psys_syscall_table[i]);
524 syscalls_metadata[i] = meta;
525 }
526 return;
527
528 /* Paranoid: avoid overflow */
529end:
530 atomic_dec(&refs);
531}
532#endif
diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c
index a00545fe5cdd..648b3a2a3a44 100644
--- a/arch/x86/kernel/hpet.c
+++ b/arch/x86/kernel/hpet.c
@@ -80,6 +80,7 @@ static inline void hpet_clear_mapping(void)
80 */ 80 */
81static int boot_hpet_disable; 81static int boot_hpet_disable;
82int hpet_force_user; 82int hpet_force_user;
83static int hpet_verbose;
83 84
84static int __init hpet_setup(char *str) 85static int __init hpet_setup(char *str)
85{ 86{
@@ -88,6 +89,8 @@ static int __init hpet_setup(char *str)
88 boot_hpet_disable = 1; 89 boot_hpet_disable = 1;
89 if (!strncmp("force", str, 5)) 90 if (!strncmp("force", str, 5))
90 hpet_force_user = 1; 91 hpet_force_user = 1;
92 if (!strncmp("verbose", str, 7))
93 hpet_verbose = 1;
91 } 94 }
92 return 1; 95 return 1;
93} 96}
@@ -119,6 +122,43 @@ int is_hpet_enabled(void)
119} 122}
120EXPORT_SYMBOL_GPL(is_hpet_enabled); 123EXPORT_SYMBOL_GPL(is_hpet_enabled);
121 124
125static void _hpet_print_config(const char *function, int line)
126{
127 u32 i, timers, l, h;
128 printk(KERN_INFO "hpet: %s(%d):\n", function, line);
129 l = hpet_readl(HPET_ID);
130 h = hpet_readl(HPET_PERIOD);
131 timers = ((l & HPET_ID_NUMBER) >> HPET_ID_NUMBER_SHIFT) + 1;
132 printk(KERN_INFO "hpet: ID: 0x%x, PERIOD: 0x%x\n", l, h);
133 l = hpet_readl(HPET_CFG);
134 h = hpet_readl(HPET_STATUS);
135 printk(KERN_INFO "hpet: CFG: 0x%x, STATUS: 0x%x\n", l, h);
136 l = hpet_readl(HPET_COUNTER);
137 h = hpet_readl(HPET_COUNTER+4);
138 printk(KERN_INFO "hpet: COUNTER_l: 0x%x, COUNTER_h: 0x%x\n", l, h);
139
140 for (i = 0; i < timers; i++) {
141 l = hpet_readl(HPET_Tn_CFG(i));
142 h = hpet_readl(HPET_Tn_CFG(i)+4);
143 printk(KERN_INFO "hpet: T%d: CFG_l: 0x%x, CFG_h: 0x%x\n",
144 i, l, h);
145 l = hpet_readl(HPET_Tn_CMP(i));
146 h = hpet_readl(HPET_Tn_CMP(i)+4);
147 printk(KERN_INFO "hpet: T%d: CMP_l: 0x%x, CMP_h: 0x%x\n",
148 i, l, h);
149 l = hpet_readl(HPET_Tn_ROUTE(i));
150 h = hpet_readl(HPET_Tn_ROUTE(i)+4);
151 printk(KERN_INFO "hpet: T%d ROUTE_l: 0x%x, ROUTE_h: 0x%x\n",
152 i, l, h);
153 }
154}
155
156#define hpet_print_config() \
157do { \
158 if (hpet_verbose) \
159 _hpet_print_config(__FUNCTION__, __LINE__); \
160} while (0)
161
122/* 162/*
123 * When the hpet driver (/dev/hpet) is enabled, we need to reserve 163 * When the hpet driver (/dev/hpet) is enabled, we need to reserve
124 * timer 0 and timer 1 in case of RTC emulation. 164 * timer 0 and timer 1 in case of RTC emulation.
@@ -191,27 +231,37 @@ static struct clock_event_device hpet_clockevent = {
191 .rating = 50, 231 .rating = 50,
192}; 232};
193 233
194static void hpet_start_counter(void) 234static void hpet_stop_counter(void)
195{ 235{
196 unsigned long cfg = hpet_readl(HPET_CFG); 236 unsigned long cfg = hpet_readl(HPET_CFG);
197
198 cfg &= ~HPET_CFG_ENABLE; 237 cfg &= ~HPET_CFG_ENABLE;
199 hpet_writel(cfg, HPET_CFG); 238 hpet_writel(cfg, HPET_CFG);
200 hpet_writel(0, HPET_COUNTER); 239 hpet_writel(0, HPET_COUNTER);
201 hpet_writel(0, HPET_COUNTER + 4); 240 hpet_writel(0, HPET_COUNTER + 4);
241}
242
243static void hpet_start_counter(void)
244{
245 unsigned long cfg = hpet_readl(HPET_CFG);
202 cfg |= HPET_CFG_ENABLE; 246 cfg |= HPET_CFG_ENABLE;
203 hpet_writel(cfg, HPET_CFG); 247 hpet_writel(cfg, HPET_CFG);
204} 248}
205 249
250static void hpet_restart_counter(void)
251{
252 hpet_stop_counter();
253 hpet_start_counter();
254}
255
206static void hpet_resume_device(void) 256static void hpet_resume_device(void)
207{ 257{
208 force_hpet_resume(); 258 force_hpet_resume();
209} 259}
210 260
211static void hpet_restart_counter(void) 261static void hpet_resume_counter(void)
212{ 262{
213 hpet_resume_device(); 263 hpet_resume_device();
214 hpet_start_counter(); 264 hpet_restart_counter();
215} 265}
216 266
217static void hpet_enable_legacy_int(void) 267static void hpet_enable_legacy_int(void)
@@ -259,29 +309,23 @@ static int hpet_setup_msi_irq(unsigned int irq);
259static void hpet_set_mode(enum clock_event_mode mode, 309static void hpet_set_mode(enum clock_event_mode mode,
260 struct clock_event_device *evt, int timer) 310 struct clock_event_device *evt, int timer)
261{ 311{
262 unsigned long cfg, cmp, now; 312 unsigned long cfg;
263 uint64_t delta; 313 uint64_t delta;
264 314
265 switch (mode) { 315 switch (mode) {
266 case CLOCK_EVT_MODE_PERIODIC: 316 case CLOCK_EVT_MODE_PERIODIC:
317 hpet_stop_counter();
267 delta = ((uint64_t)(NSEC_PER_SEC/HZ)) * evt->mult; 318 delta = ((uint64_t)(NSEC_PER_SEC/HZ)) * evt->mult;
268 delta >>= evt->shift; 319 delta >>= evt->shift;
269 now = hpet_readl(HPET_COUNTER);
270 cmp = now + (unsigned long) delta;
271 cfg = hpet_readl(HPET_Tn_CFG(timer)); 320 cfg = hpet_readl(HPET_Tn_CFG(timer));
272 /* Make sure we use edge triggered interrupts */ 321 /* Make sure we use edge triggered interrupts */
273 cfg &= ~HPET_TN_LEVEL; 322 cfg &= ~HPET_TN_LEVEL;
274 cfg |= HPET_TN_ENABLE | HPET_TN_PERIODIC | 323 cfg |= HPET_TN_ENABLE | HPET_TN_PERIODIC |
275 HPET_TN_SETVAL | HPET_TN_32BIT; 324 HPET_TN_SETVAL | HPET_TN_32BIT;
276 hpet_writel(cfg, HPET_Tn_CFG(timer)); 325 hpet_writel(cfg, HPET_Tn_CFG(timer));
277 /*
278 * The first write after writing TN_SETVAL to the
279 * config register sets the counter value, the second
280 * write sets the period.
281 */
282 hpet_writel(cmp, HPET_Tn_CMP(timer));
283 udelay(1);
284 hpet_writel((unsigned long) delta, HPET_Tn_CMP(timer)); 326 hpet_writel((unsigned long) delta, HPET_Tn_CMP(timer));
327 hpet_start_counter();
328 hpet_print_config();
285 break; 329 break;
286 330
287 case CLOCK_EVT_MODE_ONESHOT: 331 case CLOCK_EVT_MODE_ONESHOT:
@@ -308,6 +352,7 @@ static void hpet_set_mode(enum clock_event_mode mode,
308 irq_set_affinity(hdev->irq, cpumask_of(hdev->cpu)); 352 irq_set_affinity(hdev->irq, cpumask_of(hdev->cpu));
309 enable_irq(hdev->irq); 353 enable_irq(hdev->irq);
310 } 354 }
355 hpet_print_config();
311 break; 356 break;
312 } 357 }
313} 358}
@@ -526,6 +571,7 @@ static void hpet_msi_capability_lookup(unsigned int start_timer)
526 571
527 num_timers = ((id & HPET_ID_NUMBER) >> HPET_ID_NUMBER_SHIFT); 572 num_timers = ((id & HPET_ID_NUMBER) >> HPET_ID_NUMBER_SHIFT);
528 num_timers++; /* Value read out starts from 0 */ 573 num_timers++; /* Value read out starts from 0 */
574 hpet_print_config();
529 575
530 hpet_devs = kzalloc(sizeof(struct hpet_dev) * num_timers, GFP_KERNEL); 576 hpet_devs = kzalloc(sizeof(struct hpet_dev) * num_timers, GFP_KERNEL);
531 if (!hpet_devs) 577 if (!hpet_devs)
@@ -695,7 +741,7 @@ static struct clocksource clocksource_hpet = {
695 .mask = HPET_MASK, 741 .mask = HPET_MASK,
696 .shift = HPET_SHIFT, 742 .shift = HPET_SHIFT,
697 .flags = CLOCK_SOURCE_IS_CONTINUOUS, 743 .flags = CLOCK_SOURCE_IS_CONTINUOUS,
698 .resume = hpet_restart_counter, 744 .resume = hpet_resume_counter,
699#ifdef CONFIG_X86_64 745#ifdef CONFIG_X86_64
700 .vread = vread_hpet, 746 .vread = vread_hpet,
701#endif 747#endif
@@ -707,7 +753,7 @@ static int hpet_clocksource_register(void)
707 cycle_t t1; 753 cycle_t t1;
708 754
709 /* Start the counter */ 755 /* Start the counter */
710 hpet_start_counter(); 756 hpet_restart_counter();
711 757
712 /* Verify whether hpet counter works */ 758 /* Verify whether hpet counter works */
713 t1 = read_hpet(); 759 t1 = read_hpet();
@@ -793,6 +839,7 @@ int __init hpet_enable(void)
793 * information and the number of channels 839 * information and the number of channels
794 */ 840 */
795 id = hpet_readl(HPET_ID); 841 id = hpet_readl(HPET_ID);
842 hpet_print_config();
796 843
797#ifdef CONFIG_HPET_EMULATE_RTC 844#ifdef CONFIG_HPET_EMULATE_RTC
798 /* 845 /*
@@ -845,6 +892,7 @@ static __init int hpet_late_init(void)
845 return -ENODEV; 892 return -ENODEV;
846 893
847 hpet_reserve_platform_timers(hpet_readl(HPET_ID)); 894 hpet_reserve_platform_timers(hpet_readl(HPET_ID));
895 hpet_print_config();
848 896
849 for_each_online_cpu(cpu) { 897 for_each_online_cpu(cpu) {
850 hpet_cpuhp_notify(NULL, CPU_ONLINE, (void *)(long)cpu); 898 hpet_cpuhp_notify(NULL, CPU_ONLINE, (void *)(long)cpu);
diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
index b8ac3b6cf776..c3fe010d74c8 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -58,9 +58,14 @@ static int show_other_interrupts(struct seq_file *p, int prec)
58 for_each_online_cpu(j) 58 for_each_online_cpu(j)
59 seq_printf(p, "%10u ", irq_stats(j)->apic_timer_irqs); 59 seq_printf(p, "%10u ", irq_stats(j)->apic_timer_irqs);
60 seq_printf(p, " Local timer interrupts\n"); 60 seq_printf(p, " Local timer interrupts\n");
61
62 seq_printf(p, "%*s: ", prec, "SPU");
63 for_each_online_cpu(j)
64 seq_printf(p, "%10u ", irq_stats(j)->irq_spurious_count);
65 seq_printf(p, " Spurious interrupts\n");
61#endif 66#endif
62 if (generic_interrupt_extension) { 67 if (generic_interrupt_extension) {
63 seq_printf(p, "PLT: "); 68 seq_printf(p, "%*s: ", prec, "PLT");
64 for_each_online_cpu(j) 69 for_each_online_cpu(j)
65 seq_printf(p, "%10u ", irq_stats(j)->generic_irqs); 70 seq_printf(p, "%10u ", irq_stats(j)->generic_irqs);
66 seq_printf(p, " Platform interrupts\n"); 71 seq_printf(p, " Platform interrupts\n");
@@ -91,12 +96,6 @@ static int show_other_interrupts(struct seq_file *p, int prec)
91 seq_printf(p, " Threshold APIC interrupts\n"); 96 seq_printf(p, " Threshold APIC interrupts\n");
92# endif 97# endif
93#endif 98#endif
94#ifdef CONFIG_X86_LOCAL_APIC
95 seq_printf(p, "%*s: ", prec, "SPU");
96 for_each_online_cpu(j)
97 seq_printf(p, "%10u ", irq_stats(j)->irq_spurious_count);
98 seq_printf(p, " Spurious interrupts\n");
99#endif
100 seq_printf(p, "%*s: %10u\n", prec, "ERR", atomic_read(&irq_err_count)); 99 seq_printf(p, "%*s: %10u\n", prec, "ERR", atomic_read(&irq_err_count));
101#if defined(CONFIG_X86_IO_APIC) 100#if defined(CONFIG_X86_IO_APIC)
102 seq_printf(p, "%*s: %10u\n", prec, "MIS", atomic_read(&irq_mis_count)); 101 seq_printf(p, "%*s: %10u\n", prec, "MIS", atomic_read(&irq_mis_count));
@@ -133,23 +132,15 @@ int show_interrupts(struct seq_file *p, void *v)
133 return 0; 132 return 0;
134 133
135 spin_lock_irqsave(&desc->lock, flags); 134 spin_lock_irqsave(&desc->lock, flags);
136#ifndef CONFIG_SMP
137 any_count = kstat_irqs(i);
138#else
139 for_each_online_cpu(j) 135 for_each_online_cpu(j)
140 any_count |= kstat_irqs_cpu(i, j); 136 any_count |= kstat_irqs_cpu(i, j);
141#endif
142 action = desc->action; 137 action = desc->action;
143 if (!action && !any_count) 138 if (!action && !any_count)
144 goto out; 139 goto out;
145 140
146 seq_printf(p, "%*d: ", prec, i); 141 seq_printf(p, "%*d: ", prec, i);
147#ifndef CONFIG_SMP
148 seq_printf(p, "%10u ", kstat_irqs(i));
149#else
150 for_each_online_cpu(j) 142 for_each_online_cpu(j)
151 seq_printf(p, "%10u ", kstat_irqs_cpu(i, j)); 143 seq_printf(p, "%10u ", kstat_irqs_cpu(i, j));
152#endif
153 seq_printf(p, " %8s", desc->chip->name); 144 seq_printf(p, " %8s", desc->chip->name);
154 seq_printf(p, "-%-8s", desc->name); 145 seq_printf(p, "-%-8s", desc->name);
155 146
@@ -174,6 +165,7 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
174 165
175#ifdef CONFIG_X86_LOCAL_APIC 166#ifdef CONFIG_X86_LOCAL_APIC
176 sum += irq_stats(cpu)->apic_timer_irqs; 167 sum += irq_stats(cpu)->apic_timer_irqs;
168 sum += irq_stats(cpu)->irq_spurious_count;
177#endif 169#endif
178 if (generic_interrupt_extension) 170 if (generic_interrupt_extension)
179 sum += irq_stats(cpu)->generic_irqs; 171 sum += irq_stats(cpu)->generic_irqs;
@@ -188,9 +180,6 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
188 sum += irq_stats(cpu)->irq_threshold_count; 180 sum += irq_stats(cpu)->irq_threshold_count;
189#endif 181#endif
190#endif 182#endif
191#ifdef CONFIG_X86_LOCAL_APIC
192 sum += irq_stats(cpu)->irq_spurious_count;
193#endif
194 return sum; 183 return sum;
195} 184}
196 185
diff --git a/arch/x86/kernel/irqinit_32.c b/arch/x86/kernel/irqinit_32.c
index bc1326105448..368b0a8836f9 100644
--- a/arch/x86/kernel/irqinit_32.c
+++ b/arch/x86/kernel/irqinit_32.c
@@ -50,7 +50,6 @@ static irqreturn_t math_error_irq(int cpl, void *dev_id)
50 */ 50 */
51static struct irqaction fpu_irq = { 51static struct irqaction fpu_irq = {
52 .handler = math_error_irq, 52 .handler = math_error_irq,
53 .mask = CPU_MASK_NONE,
54 .name = "fpu", 53 .name = "fpu",
55}; 54};
56 55
@@ -83,7 +82,6 @@ void __init init_ISA_irqs(void)
83 */ 82 */
84static struct irqaction irq2 = { 83static struct irqaction irq2 = {
85 .handler = no_action, 84 .handler = no_action,
86 .mask = CPU_MASK_NONE,
87 .name = "cascade", 85 .name = "cascade",
88}; 86};
89 87
diff --git a/arch/x86/kernel/irqinit_64.c b/arch/x86/kernel/irqinit_64.c
index c7a49e0ffbfb..8cd10537fd46 100644
--- a/arch/x86/kernel/irqinit_64.c
+++ b/arch/x86/kernel/irqinit_64.c
@@ -45,7 +45,6 @@
45 45
46static struct irqaction irq2 = { 46static struct irqaction irq2 = {
47 .handler = no_action, 47 .handler = no_action,
48 .mask = CPU_MASK_NONE,
49 .name = "cascade", 48 .name = "cascade",
50}; 49};
51DEFINE_PER_CPU(vector_irq_t, vector_irq) = { 50DEFINE_PER_CPU(vector_irq_t, vector_irq) = {
diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c
index 55b94614e348..7b5169d2b000 100644
--- a/arch/x86/kernel/kprobes.c
+++ b/arch/x86/kernel/kprobes.c
@@ -638,13 +638,13 @@ static void __used __kprobes kretprobe_trampoline_holder(void)
638#else 638#else
639 " pushf\n" 639 " pushf\n"
640 /* 640 /*
641 * Skip cs, ip, orig_ax. 641 * Skip cs, ip, orig_ax and gs.
642 * trampoline_handler() will plug in these values 642 * trampoline_handler() will plug in these values
643 */ 643 */
644 " subl $12, %esp\n" 644 " subl $16, %esp\n"
645 " pushl %fs\n" 645 " pushl %fs\n"
646 " pushl %ds\n"
647 " pushl %es\n" 646 " pushl %es\n"
647 " pushl %ds\n"
648 " pushl %eax\n" 648 " pushl %eax\n"
649 " pushl %ebp\n" 649 " pushl %ebp\n"
650 " pushl %edi\n" 650 " pushl %edi\n"
@@ -655,10 +655,10 @@ static void __used __kprobes kretprobe_trampoline_holder(void)
655 " movl %esp, %eax\n" 655 " movl %esp, %eax\n"
656 " call trampoline_handler\n" 656 " call trampoline_handler\n"
657 /* Move flags to cs */ 657 /* Move flags to cs */
658 " movl 52(%esp), %edx\n" 658 " movl 56(%esp), %edx\n"
659 " movl %edx, 48(%esp)\n" 659 " movl %edx, 52(%esp)\n"
660 /* Replace saved flags with true return address. */ 660 /* Replace saved flags with true return address. */
661 " movl %eax, 52(%esp)\n" 661 " movl %eax, 56(%esp)\n"
662 " popl %ebx\n" 662 " popl %ebx\n"
663 " popl %ecx\n" 663 " popl %ecx\n"
664 " popl %edx\n" 664 " popl %edx\n"
@@ -666,8 +666,8 @@ static void __used __kprobes kretprobe_trampoline_holder(void)
666 " popl %edi\n" 666 " popl %edi\n"
667 " popl %ebp\n" 667 " popl %ebp\n"
668 " popl %eax\n" 668 " popl %eax\n"
669 /* Skip ip, orig_ax, es, ds, fs */ 669 /* Skip ds, es, fs, gs, orig_ax and ip */
670 " addl $20, %esp\n" 670 " addl $24, %esp\n"
671 " popf\n" 671 " popf\n"
672#endif 672#endif
673 " ret\n"); 673 " ret\n");
@@ -691,6 +691,7 @@ static __used __kprobes void *trampoline_handler(struct pt_regs *regs)
691 regs->cs = __KERNEL_CS; 691 regs->cs = __KERNEL_CS;
692#else 692#else
693 regs->cs = __KERNEL_CS | get_kernel_rpl(); 693 regs->cs = __KERNEL_CS | get_kernel_rpl();
694 regs->gs = 0;
694#endif 695#endif
695 regs->ip = trampoline_address; 696 regs->ip = trampoline_address;
696 regs->orig_ax = ~0UL; 697 regs->orig_ax = ~0UL;
diff --git a/arch/x86/kernel/mfgpt_32.c b/arch/x86/kernel/mfgpt_32.c
index 8815f3c7fec7..846510b78a09 100644
--- a/arch/x86/kernel/mfgpt_32.c
+++ b/arch/x86/kernel/mfgpt_32.c
@@ -348,7 +348,6 @@ static irqreturn_t mfgpt_tick(int irq, void *dev_id)
348static struct irqaction mfgptirq = { 348static struct irqaction mfgptirq = {
349 .handler = mfgpt_tick, 349 .handler = mfgpt_tick,
350 .flags = IRQF_DISABLED | IRQF_NOBALANCING, 350 .flags = IRQF_DISABLED | IRQF_NOBALANCING,
351 .mask = CPU_MASK_NONE,
352 .name = "mfgpt-timer" 351 .name = "mfgpt-timer"
353}; 352};
354 353
diff --git a/arch/x86/kernel/microcode_amd.c b/arch/x86/kernel/microcode_amd.c
index c25fdb382292..453b5795a5c6 100644
--- a/arch/x86/kernel/microcode_amd.c
+++ b/arch/x86/kernel/microcode_amd.c
@@ -12,31 +12,30 @@
12 * 12 *
13 * Licensed under the terms of the GNU General Public 13 * Licensed under the terms of the GNU General Public
14 * License version 2. See file COPYING for details. 14 * License version 2. See file COPYING for details.
15*/ 15 */
16 16#include <linux/platform_device.h>
17#include <linux/capability.h> 17#include <linux/capability.h>
18#include <linux/kernel.h>
19#include <linux/init.h>
20#include <linux/sched.h>
21#include <linux/cpumask.h>
22#include <linux/module.h>
23#include <linux/slab.h>
24#include <linux/vmalloc.h>
25#include <linux/miscdevice.h> 18#include <linux/miscdevice.h>
19#include <linux/firmware.h>
26#include <linux/spinlock.h> 20#include <linux/spinlock.h>
27#include <linux/mm.h> 21#include <linux/cpumask.h>
28#include <linux/fs.h> 22#include <linux/pci_ids.h>
23#include <linux/uaccess.h>
24#include <linux/vmalloc.h>
25#include <linux/kernel.h>
26#include <linux/module.h>
29#include <linux/mutex.h> 27#include <linux/mutex.h>
28#include <linux/sched.h>
29#include <linux/init.h>
30#include <linux/slab.h>
30#include <linux/cpu.h> 31#include <linux/cpu.h>
31#include <linux/firmware.h>
32#include <linux/platform_device.h>
33#include <linux/pci.h> 32#include <linux/pci.h>
34#include <linux/pci_ids.h> 33#include <linux/fs.h>
35#include <linux/uaccess.h> 34#include <linux/mm.h>
36 35
37#include <asm/msr.h>
38#include <asm/processor.h>
39#include <asm/microcode.h> 36#include <asm/microcode.h>
37#include <asm/processor.h>
38#include <asm/msr.h>
40 39
41MODULE_DESCRIPTION("AMD Microcode Update Driver"); 40MODULE_DESCRIPTION("AMD Microcode Update Driver");
42MODULE_AUTHOR("Peter Oruba"); 41MODULE_AUTHOR("Peter Oruba");
@@ -72,8 +71,8 @@ struct microcode_header_amd {
72} __attribute__((packed)); 71} __attribute__((packed));
73 72
74struct microcode_amd { 73struct microcode_amd {
75 struct microcode_header_amd hdr; 74 struct microcode_header_amd hdr;
76 unsigned int mpb[0]; 75 unsigned int mpb[0];
77}; 76};
78 77
79#define UCODE_MAX_SIZE 2048 78#define UCODE_MAX_SIZE 2048
@@ -184,8 +183,8 @@ static int get_ucode_data(void *to, const u8 *from, size_t n)
184 return 0; 183 return 0;
185} 184}
186 185
187static void *get_next_ucode(const u8 *buf, unsigned int size, 186static void *
188 unsigned int *mc_size) 187get_next_ucode(const u8 *buf, unsigned int size, unsigned int *mc_size)
189{ 188{
190 unsigned int total_size; 189 unsigned int total_size;
191 u8 section_hdr[UCODE_CONTAINER_SECTION_HDR]; 190 u8 section_hdr[UCODE_CONTAINER_SECTION_HDR];
@@ -223,7 +222,6 @@ static void *get_next_ucode(const u8 *buf, unsigned int size,
223 return mc; 222 return mc;
224} 223}
225 224
226
227static int install_equiv_cpu_table(const u8 *buf) 225static int install_equiv_cpu_table(const u8 *buf)
228{ 226{
229 u8 *container_hdr[UCODE_CONTAINER_HEADER_SIZE]; 227 u8 *container_hdr[UCODE_CONTAINER_HEADER_SIZE];
@@ -372,4 +370,3 @@ struct microcode_ops * __init init_amd_microcode(void)
372{ 370{
373 return &microcode_amd_ops; 371 return &microcode_amd_ops;
374} 372}
375
diff --git a/arch/x86/kernel/microcode_core.c b/arch/x86/kernel/microcode_core.c
index c9b721ba968c..98c470c069d1 100644
--- a/arch/x86/kernel/microcode_core.c
+++ b/arch/x86/kernel/microcode_core.c
@@ -70,41 +70,41 @@
70 * Fix sigmatch() macro to handle old CPUs with pf == 0. 70 * Fix sigmatch() macro to handle old CPUs with pf == 0.
71 * Thanks to Stuart Swales for pointing out this bug. 71 * Thanks to Stuart Swales for pointing out this bug.
72 */ 72 */
73#include <linux/platform_device.h>
73#include <linux/capability.h> 74#include <linux/capability.h>
74#include <linux/kernel.h> 75#include <linux/miscdevice.h>
75#include <linux/init.h> 76#include <linux/firmware.h>
76#include <linux/sched.h>
77#include <linux/smp_lock.h> 77#include <linux/smp_lock.h>
78#include <linux/spinlock.h>
78#include <linux/cpumask.h> 79#include <linux/cpumask.h>
79#include <linux/module.h> 80#include <linux/uaccess.h>
80#include <linux/slab.h>
81#include <linux/vmalloc.h> 81#include <linux/vmalloc.h>
82#include <linux/miscdevice.h> 82#include <linux/kernel.h>
83#include <linux/spinlock.h> 83#include <linux/module.h>
84#include <linux/mm.h>
85#include <linux/fs.h>
86#include <linux/mutex.h> 84#include <linux/mutex.h>
85#include <linux/sched.h>
86#include <linux/init.h>
87#include <linux/slab.h>
87#include <linux/cpu.h> 88#include <linux/cpu.h>
88#include <linux/firmware.h> 89#include <linux/fs.h>
89#include <linux/platform_device.h> 90#include <linux/mm.h>
90 91
91#include <asm/msr.h>
92#include <asm/uaccess.h>
93#include <asm/processor.h>
94#include <asm/microcode.h> 92#include <asm/microcode.h>
93#include <asm/processor.h>
94#include <asm/msr.h>
95 95
96MODULE_DESCRIPTION("Microcode Update Driver"); 96MODULE_DESCRIPTION("Microcode Update Driver");
97MODULE_AUTHOR("Tigran Aivazian <tigran@aivazian.fsnet.co.uk>"); 97MODULE_AUTHOR("Tigran Aivazian <tigran@aivazian.fsnet.co.uk>");
98MODULE_LICENSE("GPL"); 98MODULE_LICENSE("GPL");
99 99
100#define MICROCODE_VERSION "2.00" 100#define MICROCODE_VERSION "2.00"
101 101
102static struct microcode_ops *microcode_ops; 102static struct microcode_ops *microcode_ops;
103 103
104/* no concurrent ->write()s are allowed on /dev/cpu/microcode */ 104/* no concurrent ->write()s are allowed on /dev/cpu/microcode */
105static DEFINE_MUTEX(microcode_mutex); 105static DEFINE_MUTEX(microcode_mutex);
106 106
107struct ucode_cpu_info ucode_cpu_info[NR_CPUS]; 107struct ucode_cpu_info ucode_cpu_info[NR_CPUS];
108EXPORT_SYMBOL_GPL(ucode_cpu_info); 108EXPORT_SYMBOL_GPL(ucode_cpu_info);
109 109
110#ifdef CONFIG_MICROCODE_OLD_INTERFACE 110#ifdef CONFIG_MICROCODE_OLD_INTERFACE
@@ -198,18 +198,33 @@ static void microcode_dev_exit(void)
198 198
199MODULE_ALIAS_MISCDEV(MICROCODE_MINOR); 199MODULE_ALIAS_MISCDEV(MICROCODE_MINOR);
200#else 200#else
201#define microcode_dev_init() 0 201#define microcode_dev_init() 0
202#define microcode_dev_exit() do { } while (0) 202#define microcode_dev_exit() do { } while (0)
203#endif 203#endif
204 204
205/* fake device for request_firmware */ 205/* fake device for request_firmware */
206static struct platform_device *microcode_pdev; 206static struct platform_device *microcode_pdev;
207
208static long reload_for_cpu(void *unused)
209{
210 struct ucode_cpu_info *uci = ucode_cpu_info + smp_processor_id();
211 int err = 0;
212
213 mutex_lock(&microcode_mutex);
214 if (uci->valid) {
215 err = microcode_ops->request_microcode_fw(smp_processor_id(),
216 &microcode_pdev->dev);
217 if (!err)
218 microcode_ops->apply_microcode(smp_processor_id());
219 }
220 mutex_unlock(&microcode_mutex);
221 return err;
222}
207 223
208static ssize_t reload_store(struct sys_device *dev, 224static ssize_t reload_store(struct sys_device *dev,
209 struct sysdev_attribute *attr, 225 struct sysdev_attribute *attr,
210 const char *buf, size_t sz) 226 const char *buf, size_t sz)
211{ 227{
212 struct ucode_cpu_info *uci = ucode_cpu_info + dev->id;
213 char *end; 228 char *end;
214 unsigned long val = simple_strtoul(buf, &end, 0); 229 unsigned long val = simple_strtoul(buf, &end, 0);
215 int err = 0; 230 int err = 0;
@@ -218,21 +233,9 @@ static ssize_t reload_store(struct sys_device *dev,
218 if (end == buf) 233 if (end == buf)
219 return -EINVAL; 234 return -EINVAL;
220 if (val == 1) { 235 if (val == 1) {
221 cpumask_t old = current->cpus_allowed;
222
223 get_online_cpus(); 236 get_online_cpus();
224 if (cpu_online(cpu)) { 237 if (cpu_online(cpu))
225 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu)); 238 err = work_on_cpu(cpu, reload_for_cpu, NULL);
226 mutex_lock(&microcode_mutex);
227 if (uci->valid) {
228 err = microcode_ops->request_microcode_fw(cpu,
229 &microcode_pdev->dev);
230 if (!err)
231 microcode_ops->apply_microcode(cpu);
232 }
233 mutex_unlock(&microcode_mutex);
234 set_cpus_allowed_ptr(current, &old);
235 }
236 put_online_cpus(); 239 put_online_cpus();
237 } 240 }
238 if (err) 241 if (err)
@@ -268,8 +271,8 @@ static struct attribute *mc_default_attrs[] = {
268}; 271};
269 272
270static struct attribute_group mc_attr_group = { 273static struct attribute_group mc_attr_group = {
271 .attrs = mc_default_attrs, 274 .attrs = mc_default_attrs,
272 .name = "microcode", 275 .name = "microcode",
273}; 276};
274 277
275static void __microcode_fini_cpu(int cpu) 278static void __microcode_fini_cpu(int cpu)
@@ -328,9 +331,9 @@ static int microcode_resume_cpu(int cpu)
328 return 0; 331 return 0;
329} 332}
330 333
331static void microcode_update_cpu(int cpu) 334static long microcode_update_cpu(void *unused)
332{ 335{
333 struct ucode_cpu_info *uci = ucode_cpu_info + cpu; 336 struct ucode_cpu_info *uci = ucode_cpu_info + smp_processor_id();
334 int err = 0; 337 int err = 0;
335 338
336 /* 339 /*
@@ -338,30 +341,27 @@ static void microcode_update_cpu(int cpu)
338 * otherwise just request a firmware: 341 * otherwise just request a firmware:
339 */ 342 */
340 if (uci->valid) { 343 if (uci->valid) {
341 err = microcode_resume_cpu(cpu); 344 err = microcode_resume_cpu(smp_processor_id());
342 } else { 345 } else {
343 collect_cpu_info(cpu); 346 collect_cpu_info(smp_processor_id());
344 if (uci->valid && system_state == SYSTEM_RUNNING) 347 if (uci->valid && system_state == SYSTEM_RUNNING)
345 err = microcode_ops->request_microcode_fw(cpu, 348 err = microcode_ops->request_microcode_fw(
349 smp_processor_id(),
346 &microcode_pdev->dev); 350 &microcode_pdev->dev);
347 } 351 }
348 if (!err) 352 if (!err)
349 microcode_ops->apply_microcode(cpu); 353 microcode_ops->apply_microcode(smp_processor_id());
354 return err;
350} 355}
351 356
352static void microcode_init_cpu(int cpu) 357static int microcode_init_cpu(int cpu)
353{ 358{
354 cpumask_t old = current->cpus_allowed; 359 int err;
355
356 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
357 /* We should bind the task to the CPU */
358 BUG_ON(raw_smp_processor_id() != cpu);
359
360 mutex_lock(&microcode_mutex); 360 mutex_lock(&microcode_mutex);
361 microcode_update_cpu(cpu); 361 err = work_on_cpu(cpu, microcode_update_cpu, NULL);
362 mutex_unlock(&microcode_mutex); 362 mutex_unlock(&microcode_mutex);
363 363
364 set_cpus_allowed_ptr(current, &old); 364 return err;
365} 365}
366 366
367static int mc_sysdev_add(struct sys_device *sys_dev) 367static int mc_sysdev_add(struct sys_device *sys_dev)
@@ -379,8 +379,9 @@ static int mc_sysdev_add(struct sys_device *sys_dev)
379 if (err) 379 if (err)
380 return err; 380 return err;
381 381
382 microcode_init_cpu(cpu); 382 err = microcode_init_cpu(cpu);
383 return 0; 383
384 return err;
384} 385}
385 386
386static int mc_sysdev_remove(struct sys_device *sys_dev) 387static int mc_sysdev_remove(struct sys_device *sys_dev)
@@ -404,14 +405,14 @@ static int mc_sysdev_resume(struct sys_device *dev)
404 return 0; 405 return 0;
405 406
406 /* only CPU 0 will apply ucode here */ 407 /* only CPU 0 will apply ucode here */
407 microcode_update_cpu(0); 408 microcode_update_cpu(NULL);
408 return 0; 409 return 0;
409} 410}
410 411
411static struct sysdev_driver mc_sysdev_driver = { 412static struct sysdev_driver mc_sysdev_driver = {
412 .add = mc_sysdev_add, 413 .add = mc_sysdev_add,
413 .remove = mc_sysdev_remove, 414 .remove = mc_sysdev_remove,
414 .resume = mc_sysdev_resume, 415 .resume = mc_sysdev_resume,
415}; 416};
416 417
417static __cpuinit int 418static __cpuinit int
@@ -424,7 +425,9 @@ mc_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu)
424 switch (action) { 425 switch (action) {
425 case CPU_ONLINE: 426 case CPU_ONLINE:
426 case CPU_ONLINE_FROZEN: 427 case CPU_ONLINE_FROZEN:
427 microcode_init_cpu(cpu); 428 if (microcode_init_cpu(cpu))
429 printk(KERN_ERR "microcode: failed to init CPU%d\n",
430 cpu);
428 case CPU_DOWN_FAILED: 431 case CPU_DOWN_FAILED:
429 case CPU_DOWN_FAILED_FROZEN: 432 case CPU_DOWN_FAILED_FROZEN:
430 pr_debug("microcode: CPU%d added\n", cpu); 433 pr_debug("microcode: CPU%d added\n", cpu);
@@ -448,7 +451,7 @@ mc_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu)
448} 451}
449 452
450static struct notifier_block __refdata mc_cpu_notifier = { 453static struct notifier_block __refdata mc_cpu_notifier = {
451 .notifier_call = mc_cpu_callback, 454 .notifier_call = mc_cpu_callback,
452}; 455};
453 456
454static int __init microcode_init(void) 457static int __init microcode_init(void)
diff --git a/arch/x86/kernel/microcode_intel.c b/arch/x86/kernel/microcode_intel.c
index 5e9f4fc51385..149b9ec7c1ab 100644
--- a/arch/x86/kernel/microcode_intel.c
+++ b/arch/x86/kernel/microcode_intel.c
@@ -70,28 +70,28 @@
70 * Fix sigmatch() macro to handle old CPUs with pf == 0. 70 * Fix sigmatch() macro to handle old CPUs with pf == 0.
71 * Thanks to Stuart Swales for pointing out this bug. 71 * Thanks to Stuart Swales for pointing out this bug.
72 */ 72 */
73#include <linux/platform_device.h>
73#include <linux/capability.h> 74#include <linux/capability.h>
74#include <linux/kernel.h> 75#include <linux/miscdevice.h>
75#include <linux/init.h> 76#include <linux/firmware.h>
76#include <linux/sched.h>
77#include <linux/smp_lock.h> 77#include <linux/smp_lock.h>
78#include <linux/spinlock.h>
78#include <linux/cpumask.h> 79#include <linux/cpumask.h>
79#include <linux/module.h> 80#include <linux/uaccess.h>
80#include <linux/slab.h>
81#include <linux/vmalloc.h> 81#include <linux/vmalloc.h>
82#include <linux/miscdevice.h> 82#include <linux/kernel.h>
83#include <linux/spinlock.h> 83#include <linux/module.h>
84#include <linux/mm.h>
85#include <linux/fs.h>
86#include <linux/mutex.h> 84#include <linux/mutex.h>
85#include <linux/sched.h>
86#include <linux/init.h>
87#include <linux/slab.h>
87#include <linux/cpu.h> 88#include <linux/cpu.h>
88#include <linux/firmware.h> 89#include <linux/fs.h>
89#include <linux/platform_device.h> 90#include <linux/mm.h>
90#include <linux/uaccess.h>
91 91
92#include <asm/msr.h>
93#include <asm/processor.h>
94#include <asm/microcode.h> 92#include <asm/microcode.h>
93#include <asm/processor.h>
94#include <asm/msr.h>
95 95
96MODULE_DESCRIPTION("Microcode Update Driver"); 96MODULE_DESCRIPTION("Microcode Update Driver");
97MODULE_AUTHOR("Tigran Aivazian <tigran@aivazian.fsnet.co.uk>"); 97MODULE_AUTHOR("Tigran Aivazian <tigran@aivazian.fsnet.co.uk>");
@@ -129,12 +129,13 @@ struct extended_sigtable {
129 struct extended_signature sigs[0]; 129 struct extended_signature sigs[0];
130}; 130};
131 131
132#define DEFAULT_UCODE_DATASIZE (2000) 132#define DEFAULT_UCODE_DATASIZE (2000)
133#define MC_HEADER_SIZE (sizeof(struct microcode_header_intel)) 133#define MC_HEADER_SIZE (sizeof(struct microcode_header_intel))
134#define DEFAULT_UCODE_TOTALSIZE (DEFAULT_UCODE_DATASIZE + MC_HEADER_SIZE) 134#define DEFAULT_UCODE_TOTALSIZE (DEFAULT_UCODE_DATASIZE + MC_HEADER_SIZE)
135#define EXT_HEADER_SIZE (sizeof(struct extended_sigtable)) 135#define EXT_HEADER_SIZE (sizeof(struct extended_sigtable))
136#define EXT_SIGNATURE_SIZE (sizeof(struct extended_signature)) 136#define EXT_SIGNATURE_SIZE (sizeof(struct extended_signature))
137#define DWSIZE (sizeof(u32)) 137#define DWSIZE (sizeof(u32))
138
138#define get_totalsize(mc) \ 139#define get_totalsize(mc) \
139 (((struct microcode_intel *)mc)->hdr.totalsize ? \ 140 (((struct microcode_intel *)mc)->hdr.totalsize ? \
140 ((struct microcode_intel *)mc)->hdr.totalsize : \ 141 ((struct microcode_intel *)mc)->hdr.totalsize : \
@@ -197,30 +198,31 @@ static inline int update_match_cpu(struct cpu_signature *csig, int sig, int pf)
197} 198}
198 199
199static inline int 200static inline int
200update_match_revision(struct microcode_header_intel *mc_header, int rev) 201update_match_revision(struct microcode_header_intel *mc_header, int rev)
201{ 202{
202 return (mc_header->rev <= rev) ? 0 : 1; 203 return (mc_header->rev <= rev) ? 0 : 1;
203} 204}
204 205
205static int microcode_sanity_check(void *mc) 206static int microcode_sanity_check(void *mc)
206{ 207{
208 unsigned long total_size, data_size, ext_table_size;
207 struct microcode_header_intel *mc_header = mc; 209 struct microcode_header_intel *mc_header = mc;
208 struct extended_sigtable *ext_header = NULL; 210 struct extended_sigtable *ext_header = NULL;
209 struct extended_signature *ext_sig;
210 unsigned long total_size, data_size, ext_table_size;
211 int sum, orig_sum, ext_sigcount = 0, i; 211 int sum, orig_sum, ext_sigcount = 0, i;
212 struct extended_signature *ext_sig;
212 213
213 total_size = get_totalsize(mc_header); 214 total_size = get_totalsize(mc_header);
214 data_size = get_datasize(mc_header); 215 data_size = get_datasize(mc_header);
216
215 if (data_size + MC_HEADER_SIZE > total_size) { 217 if (data_size + MC_HEADER_SIZE > total_size) {
216 printk(KERN_ERR "microcode: error! " 218 printk(KERN_ERR "microcode: error! "
217 "Bad data size in microcode data file\n"); 219 "Bad data size in microcode data file\n");
218 return -EINVAL; 220 return -EINVAL;
219 } 221 }
220 222
221 if (mc_header->ldrver != 1 || mc_header->hdrver != 1) { 223 if (mc_header->ldrver != 1 || mc_header->hdrver != 1) {
222 printk(KERN_ERR "microcode: error! " 224 printk(KERN_ERR "microcode: error! "
223 "Unknown microcode update format\n"); 225 "Unknown microcode update format\n");
224 return -EINVAL; 226 return -EINVAL;
225 } 227 }
226 ext_table_size = total_size - (MC_HEADER_SIZE + data_size); 228 ext_table_size = total_size - (MC_HEADER_SIZE + data_size);
@@ -318,11 +320,15 @@ get_matching_microcode(struct cpu_signature *cpu_sig, void *mc, int rev)
318 320
319static void apply_microcode(int cpu) 321static void apply_microcode(int cpu)
320{ 322{
323 struct microcode_intel *mc_intel;
324 struct ucode_cpu_info *uci;
321 unsigned long flags; 325 unsigned long flags;
322 unsigned int val[2]; 326 unsigned int val[2];
323 int cpu_num = raw_smp_processor_id(); 327 int cpu_num;
324 struct ucode_cpu_info *uci = ucode_cpu_info + cpu; 328
325 struct microcode_intel *mc_intel = uci->mc; 329 cpu_num = raw_smp_processor_id();
330 uci = ucode_cpu_info + cpu;
331 mc_intel = uci->mc;
326 332
327 /* We should bind the task to the CPU */ 333 /* We should bind the task to the CPU */
328 BUG_ON(cpu_num != cpu); 334 BUG_ON(cpu_num != cpu);
@@ -348,15 +354,17 @@ static void apply_microcode(int cpu)
348 spin_unlock_irqrestore(&microcode_update_lock, flags); 354 spin_unlock_irqrestore(&microcode_update_lock, flags);
349 if (val[1] != mc_intel->hdr.rev) { 355 if (val[1] != mc_intel->hdr.rev) {
350 printk(KERN_ERR "microcode: CPU%d update from revision " 356 printk(KERN_ERR "microcode: CPU%d update from revision "
351 "0x%x to 0x%x failed\n", cpu_num, uci->cpu_sig.rev, val[1]); 357 "0x%x to 0x%x failed\n",
358 cpu_num, uci->cpu_sig.rev, val[1]);
352 return; 359 return;
353 } 360 }
354 printk(KERN_INFO "microcode: CPU%d updated from revision " 361 printk(KERN_INFO "microcode: CPU%d updated from revision "
355 "0x%x to 0x%x, date = %04x-%02x-%02x \n", 362 "0x%x to 0x%x, date = %04x-%02x-%02x \n",
356 cpu_num, uci->cpu_sig.rev, val[1], 363 cpu_num, uci->cpu_sig.rev, val[1],
357 mc_intel->hdr.date & 0xffff, 364 mc_intel->hdr.date & 0xffff,
358 mc_intel->hdr.date >> 24, 365 mc_intel->hdr.date >> 24,
359 (mc_intel->hdr.date >> 16) & 0xff); 366 (mc_intel->hdr.date >> 16) & 0xff);
367
360 uci->cpu_sig.rev = val[1]; 368 uci->cpu_sig.rev = val[1];
361} 369}
362 370
@@ -404,18 +412,23 @@ static int generic_load_microcode(int cpu, void *data, size_t size,
404 leftover -= mc_size; 412 leftover -= mc_size;
405 } 413 }
406 414
407 if (new_mc) { 415 if (!new_mc)
408 if (!leftover) { 416 goto out;
409 if (uci->mc) 417
410 vfree(uci->mc); 418 if (leftover) {
411 uci->mc = (struct microcode_intel *)new_mc; 419 vfree(new_mc);
412 pr_debug("microcode: CPU%d found a matching microcode update with" 420 goto out;
413 " version 0x%x (current=0x%x)\n",
414 cpu, new_rev, uci->cpu_sig.rev);
415 } else
416 vfree(new_mc);
417 } 421 }
418 422
423 if (uci->mc)
424 vfree(uci->mc);
425 uci->mc = (struct microcode_intel *)new_mc;
426
427 pr_debug("microcode: CPU%d found a matching microcode update with"
428 " version 0x%x (current=0x%x)\n",
429 cpu, new_rev, uci->cpu_sig.rev);
430
431 out:
419 return (int)leftover; 432 return (int)leftover;
420} 433}
421 434
diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c
index dce99dca6cf8..70fd7e414c15 100644
--- a/arch/x86/kernel/mpparse.c
+++ b/arch/x86/kernel/mpparse.c
@@ -679,7 +679,7 @@ void __init get_smp_config(void)
679 __get_smp_config(0); 679 __get_smp_config(0);
680} 680}
681 681
682static void smp_reserve_bootmem(struct mpf_intel *mpf) 682static void __init smp_reserve_bootmem(struct mpf_intel *mpf)
683{ 683{
684 unsigned long size = get_mpc_size(mpf->physptr); 684 unsigned long size = get_mpc_size(mpf->physptr);
685#ifdef CONFIG_X86_32 685#ifdef CONFIG_X86_32
@@ -838,7 +838,7 @@ static int __init get_MP_intsrc_index(struct mpc_intsrc *m)
838 838
839static struct mpc_intsrc __initdata *m_spare[SPARE_SLOT_NUM]; 839static struct mpc_intsrc __initdata *m_spare[SPARE_SLOT_NUM];
840 840
841static void check_irq_src(struct mpc_intsrc *m, int *nr_m_spare) 841static void __init check_irq_src(struct mpc_intsrc *m, int *nr_m_spare)
842{ 842{
843 int i; 843 int i;
844 844
@@ -866,7 +866,8 @@ static void check_irq_src(struct mpc_intsrc *m, int *nr_m_spare)
866 } 866 }
867} 867}
868#else /* CONFIG_X86_IO_APIC */ 868#else /* CONFIG_X86_IO_APIC */
869static inline void check_irq_src(struct mpc_intsrc *m, int *nr_m_spare) {} 869static
870inline void __init check_irq_src(struct mpc_intsrc *m, int *nr_m_spare) {}
870#endif /* CONFIG_X86_IO_APIC */ 871#endif /* CONFIG_X86_IO_APIC */
871 872
872static int check_slot(unsigned long mpc_new_phys, unsigned long mpc_new_length, 873static int check_slot(unsigned long mpc_new_phys, unsigned long mpc_new_length,
diff --git a/arch/x86/kernel/pci-calgary_64.c b/arch/x86/kernel/pci-calgary_64.c
index d28bbdc35e4e..755c21e906f3 100644
--- a/arch/x86/kernel/pci-calgary_64.c
+++ b/arch/x86/kernel/pci-calgary_64.c
@@ -380,8 +380,9 @@ static inline struct iommu_table *find_iommu_table(struct device *dev)
380 return tbl; 380 return tbl;
381} 381}
382 382
383static void calgary_unmap_sg(struct device *dev, 383static void calgary_unmap_sg(struct device *dev, struct scatterlist *sglist,
384 struct scatterlist *sglist, int nelems, int direction) 384 int nelems,enum dma_data_direction dir,
385 struct dma_attrs *attrs)
385{ 386{
386 struct iommu_table *tbl = find_iommu_table(dev); 387 struct iommu_table *tbl = find_iommu_table(dev);
387 struct scatterlist *s; 388 struct scatterlist *s;
@@ -404,7 +405,8 @@ static void calgary_unmap_sg(struct device *dev,
404} 405}
405 406
406static int calgary_map_sg(struct device *dev, struct scatterlist *sg, 407static int calgary_map_sg(struct device *dev, struct scatterlist *sg,
407 int nelems, int direction) 408 int nelems, enum dma_data_direction dir,
409 struct dma_attrs *attrs)
408{ 410{
409 struct iommu_table *tbl = find_iommu_table(dev); 411 struct iommu_table *tbl = find_iommu_table(dev);
410 struct scatterlist *s; 412 struct scatterlist *s;
@@ -429,15 +431,14 @@ static int calgary_map_sg(struct device *dev, struct scatterlist *sg,
429 s->dma_address = (entry << PAGE_SHIFT) | s->offset; 431 s->dma_address = (entry << PAGE_SHIFT) | s->offset;
430 432
431 /* insert into HW table */ 433 /* insert into HW table */
432 tce_build(tbl, entry, npages, vaddr & PAGE_MASK, 434 tce_build(tbl, entry, npages, vaddr & PAGE_MASK, dir);
433 direction);
434 435
435 s->dma_length = s->length; 436 s->dma_length = s->length;
436 } 437 }
437 438
438 return nelems; 439 return nelems;
439error: 440error:
440 calgary_unmap_sg(dev, sg, nelems, direction); 441 calgary_unmap_sg(dev, sg, nelems, dir, NULL);
441 for_each_sg(sg, s, nelems, i) { 442 for_each_sg(sg, s, nelems, i) {
442 sg->dma_address = bad_dma_address; 443 sg->dma_address = bad_dma_address;
443 sg->dma_length = 0; 444 sg->dma_length = 0;
@@ -445,10 +446,12 @@ error:
445 return 0; 446 return 0;
446} 447}
447 448
448static dma_addr_t calgary_map_single(struct device *dev, phys_addr_t paddr, 449static dma_addr_t calgary_map_page(struct device *dev, struct page *page,
449 size_t size, int direction) 450 unsigned long offset, size_t size,
451 enum dma_data_direction dir,
452 struct dma_attrs *attrs)
450{ 453{
451 void *vaddr = phys_to_virt(paddr); 454 void *vaddr = page_address(page) + offset;
452 unsigned long uaddr; 455 unsigned long uaddr;
453 unsigned int npages; 456 unsigned int npages;
454 struct iommu_table *tbl = find_iommu_table(dev); 457 struct iommu_table *tbl = find_iommu_table(dev);
@@ -456,17 +459,18 @@ static dma_addr_t calgary_map_single(struct device *dev, phys_addr_t paddr,
456 uaddr = (unsigned long)vaddr; 459 uaddr = (unsigned long)vaddr;
457 npages = iommu_num_pages(uaddr, size, PAGE_SIZE); 460 npages = iommu_num_pages(uaddr, size, PAGE_SIZE);
458 461
459 return iommu_alloc(dev, tbl, vaddr, npages, direction); 462 return iommu_alloc(dev, tbl, vaddr, npages, dir);
460} 463}
461 464
462static void calgary_unmap_single(struct device *dev, dma_addr_t dma_handle, 465static void calgary_unmap_page(struct device *dev, dma_addr_t dma_addr,
463 size_t size, int direction) 466 size_t size, enum dma_data_direction dir,
467 struct dma_attrs *attrs)
464{ 468{
465 struct iommu_table *tbl = find_iommu_table(dev); 469 struct iommu_table *tbl = find_iommu_table(dev);
466 unsigned int npages; 470 unsigned int npages;
467 471
468 npages = iommu_num_pages(dma_handle, size, PAGE_SIZE); 472 npages = iommu_num_pages(dma_addr, size, PAGE_SIZE);
469 iommu_free(tbl, dma_handle, npages); 473 iommu_free(tbl, dma_addr, npages);
470} 474}
471 475
472static void* calgary_alloc_coherent(struct device *dev, size_t size, 476static void* calgary_alloc_coherent(struct device *dev, size_t size,
@@ -515,13 +519,13 @@ static void calgary_free_coherent(struct device *dev, size_t size,
515 free_pages((unsigned long)vaddr, get_order(size)); 519 free_pages((unsigned long)vaddr, get_order(size));
516} 520}
517 521
518static struct dma_mapping_ops calgary_dma_ops = { 522static struct dma_map_ops calgary_dma_ops = {
519 .alloc_coherent = calgary_alloc_coherent, 523 .alloc_coherent = calgary_alloc_coherent,
520 .free_coherent = calgary_free_coherent, 524 .free_coherent = calgary_free_coherent,
521 .map_single = calgary_map_single,
522 .unmap_single = calgary_unmap_single,
523 .map_sg = calgary_map_sg, 525 .map_sg = calgary_map_sg,
524 .unmap_sg = calgary_unmap_sg, 526 .unmap_sg = calgary_unmap_sg,
527 .map_page = calgary_map_page,
528 .unmap_page = calgary_unmap_page,
525}; 529};
526 530
527static inline void __iomem * busno_to_bbar(unsigned char num) 531static inline void __iomem * busno_to_bbar(unsigned char num)
diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c
index b25428533141..745579bc8256 100644
--- a/arch/x86/kernel/pci-dma.c
+++ b/arch/x86/kernel/pci-dma.c
@@ -1,4 +1,5 @@
1#include <linux/dma-mapping.h> 1#include <linux/dma-mapping.h>
2#include <linux/dma-debug.h>
2#include <linux/dmar.h> 3#include <linux/dmar.h>
3#include <linux/bootmem.h> 4#include <linux/bootmem.h>
4#include <linux/pci.h> 5#include <linux/pci.h>
@@ -12,7 +13,7 @@
12 13
13static int forbid_dac __read_mostly; 14static int forbid_dac __read_mostly;
14 15
15struct dma_mapping_ops *dma_ops; 16struct dma_map_ops *dma_ops;
16EXPORT_SYMBOL(dma_ops); 17EXPORT_SYMBOL(dma_ops);
17 18
18static int iommu_sac_force __read_mostly; 19static int iommu_sac_force __read_mostly;
@@ -39,11 +40,14 @@ EXPORT_SYMBOL(bad_dma_address);
39 to older i386. */ 40 to older i386. */
40struct device x86_dma_fallback_dev = { 41struct device x86_dma_fallback_dev = {
41 .init_name = "fallback device", 42 .init_name = "fallback device",
42 .coherent_dma_mask = DMA_32BIT_MASK, 43 .coherent_dma_mask = DMA_BIT_MASK(32),
43 .dma_mask = &x86_dma_fallback_dev.coherent_dma_mask, 44 .dma_mask = &x86_dma_fallback_dev.coherent_dma_mask,
44}; 45};
45EXPORT_SYMBOL(x86_dma_fallback_dev); 46EXPORT_SYMBOL(x86_dma_fallback_dev);
46 47
48/* Number of entries preallocated for DMA-API debugging */
49#define PREALLOC_DMA_DEBUG_ENTRIES 32768
50
47int dma_set_mask(struct device *dev, u64 mask) 51int dma_set_mask(struct device *dev, u64 mask)
48{ 52{
49 if (!dev->dma_mask || !dma_supported(dev, mask)) 53 if (!dev->dma_mask || !dma_supported(dev, mask))
@@ -144,7 +148,7 @@ again:
144 if (!is_buffer_dma_capable(dma_mask, addr, size)) { 148 if (!is_buffer_dma_capable(dma_mask, addr, size)) {
145 __free_pages(page, get_order(size)); 149 __free_pages(page, get_order(size));
146 150
147 if (dma_mask < DMA_32BIT_MASK && !(flag & GFP_DMA)) { 151 if (dma_mask < DMA_BIT_MASK(32) && !(flag & GFP_DMA)) {
148 flag = (flag & ~GFP_DMA32) | GFP_DMA; 152 flag = (flag & ~GFP_DMA32) | GFP_DMA;
149 goto again; 153 goto again;
150 } 154 }
@@ -224,7 +228,7 @@ early_param("iommu", iommu_setup);
224 228
225int dma_supported(struct device *dev, u64 mask) 229int dma_supported(struct device *dev, u64 mask)
226{ 230{
227 struct dma_mapping_ops *ops = get_dma_ops(dev); 231 struct dma_map_ops *ops = get_dma_ops(dev);
228 232
229#ifdef CONFIG_PCI 233#ifdef CONFIG_PCI
230 if (mask > 0xffffffff && forbid_dac > 0) { 234 if (mask > 0xffffffff && forbid_dac > 0) {
@@ -239,7 +243,7 @@ int dma_supported(struct device *dev, u64 mask)
239 /* Copied from i386. Doesn't make much sense, because it will 243 /* Copied from i386. Doesn't make much sense, because it will
240 only work for pci_alloc_coherent. 244 only work for pci_alloc_coherent.
241 The caller just has to use GFP_DMA in this case. */ 245 The caller just has to use GFP_DMA in this case. */
242 if (mask < DMA_24BIT_MASK) 246 if (mask < DMA_BIT_MASK(24))
243 return 0; 247 return 0;
244 248
245 /* Tell the device to use SAC when IOMMU force is on. This 249 /* Tell the device to use SAC when IOMMU force is on. This
@@ -254,7 +258,7 @@ int dma_supported(struct device *dev, u64 mask)
254 SAC for these. Assume all masks <= 40 bits are of this 258 SAC for these. Assume all masks <= 40 bits are of this
255 type. Normally this doesn't make any difference, but gives 259 type. Normally this doesn't make any difference, but gives
256 more gentle handling of IOMMU overflow. */ 260 more gentle handling of IOMMU overflow. */
257 if (iommu_sac_force && (mask >= DMA_40BIT_MASK)) { 261 if (iommu_sac_force && (mask >= DMA_BIT_MASK(40))) {
258 dev_info(dev, "Force SAC with mask %Lx\n", mask); 262 dev_info(dev, "Force SAC with mask %Lx\n", mask);
259 return 0; 263 return 0;
260 } 264 }
@@ -265,6 +269,12 @@ EXPORT_SYMBOL(dma_supported);
265 269
266static int __init pci_iommu_init(void) 270static int __init pci_iommu_init(void)
267{ 271{
272 dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES);
273
274#ifdef CONFIG_PCI
275 dma_debug_add_bus(&pci_bus_type);
276#endif
277
268 calgary_iommu_init(); 278 calgary_iommu_init();
269 279
270 intel_iommu_init(); 280 intel_iommu_init();
@@ -290,8 +300,7 @@ fs_initcall(pci_iommu_init);
290static __devinit void via_no_dac(struct pci_dev *dev) 300static __devinit void via_no_dac(struct pci_dev *dev)
291{ 301{
292 if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI && forbid_dac == 0) { 302 if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI && forbid_dac == 0) {
293 printk(KERN_INFO 303 dev_info(&dev->dev, "disabling DAC on VIA PCI bridge\n");
294 "PCI: VIA PCI bridge detected. Disabling DAC.\n");
295 forbid_dac = 1; 304 forbid_dac = 1;
296 } 305 }
297} 306}
diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c
index d5768b1af080..b284b58c035c 100644
--- a/arch/x86/kernel/pci-gart_64.c
+++ b/arch/x86/kernel/pci-gart_64.c
@@ -255,10 +255,13 @@ static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem,
255} 255}
256 256
257/* Map a single area into the IOMMU */ 257/* Map a single area into the IOMMU */
258static dma_addr_t 258static dma_addr_t gart_map_page(struct device *dev, struct page *page,
259gart_map_single(struct device *dev, phys_addr_t paddr, size_t size, int dir) 259 unsigned long offset, size_t size,
260 enum dma_data_direction dir,
261 struct dma_attrs *attrs)
260{ 262{
261 unsigned long bus; 263 unsigned long bus;
264 phys_addr_t paddr = page_to_phys(page) + offset;
262 265
263 if (!dev) 266 if (!dev)
264 dev = &x86_dma_fallback_dev; 267 dev = &x86_dma_fallback_dev;
@@ -275,8 +278,9 @@ gart_map_single(struct device *dev, phys_addr_t paddr, size_t size, int dir)
275/* 278/*
276 * Free a DMA mapping. 279 * Free a DMA mapping.
277 */ 280 */
278static void gart_unmap_single(struct device *dev, dma_addr_t dma_addr, 281static void gart_unmap_page(struct device *dev, dma_addr_t dma_addr,
279 size_t size, int direction) 282 size_t size, enum dma_data_direction dir,
283 struct dma_attrs *attrs)
280{ 284{
281 unsigned long iommu_page; 285 unsigned long iommu_page;
282 int npages; 286 int npages;
@@ -298,8 +302,8 @@ static void gart_unmap_single(struct device *dev, dma_addr_t dma_addr,
298/* 302/*
299 * Wrapper for pci_unmap_single working with scatterlists. 303 * Wrapper for pci_unmap_single working with scatterlists.
300 */ 304 */
301static void 305static void gart_unmap_sg(struct device *dev, struct scatterlist *sg, int nents,
302gart_unmap_sg(struct device *dev, struct scatterlist *sg, int nents, int dir) 306 enum dma_data_direction dir, struct dma_attrs *attrs)
303{ 307{
304 struct scatterlist *s; 308 struct scatterlist *s;
305 int i; 309 int i;
@@ -307,7 +311,7 @@ gart_unmap_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
307 for_each_sg(sg, s, nents, i) { 311 for_each_sg(sg, s, nents, i) {
308 if (!s->dma_length || !s->length) 312 if (!s->dma_length || !s->length)
309 break; 313 break;
310 gart_unmap_single(dev, s->dma_address, s->dma_length, dir); 314 gart_unmap_page(dev, s->dma_address, s->dma_length, dir, NULL);
311 } 315 }
312} 316}
313 317
@@ -329,7 +333,7 @@ static int dma_map_sg_nonforce(struct device *dev, struct scatterlist *sg,
329 addr = dma_map_area(dev, addr, s->length, dir, 0); 333 addr = dma_map_area(dev, addr, s->length, dir, 0);
330 if (addr == bad_dma_address) { 334 if (addr == bad_dma_address) {
331 if (i > 0) 335 if (i > 0)
332 gart_unmap_sg(dev, sg, i, dir); 336 gart_unmap_sg(dev, sg, i, dir, NULL);
333 nents = 0; 337 nents = 0;
334 sg[0].dma_length = 0; 338 sg[0].dma_length = 0;
335 break; 339 break;
@@ -400,8 +404,8 @@ dma_map_cont(struct device *dev, struct scatterlist *start, int nelems,
400 * DMA map all entries in a scatterlist. 404 * DMA map all entries in a scatterlist.
401 * Merge chunks that have page aligned sizes into a continuous mapping. 405 * Merge chunks that have page aligned sizes into a continuous mapping.
402 */ 406 */
403static int 407static int gart_map_sg(struct device *dev, struct scatterlist *sg, int nents,
404gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir) 408 enum dma_data_direction dir, struct dma_attrs *attrs)
405{ 409{
406 struct scatterlist *s, *ps, *start_sg, *sgmap; 410 struct scatterlist *s, *ps, *start_sg, *sgmap;
407 int need = 0, nextneed, i, out, start; 411 int need = 0, nextneed, i, out, start;
@@ -468,7 +472,7 @@ gart_map_sg(struct device *dev, struct scatterlist *sg, int nents, int dir)
468 472
469error: 473error:
470 flush_gart(); 474 flush_gart();
471 gart_unmap_sg(dev, sg, out, dir); 475 gart_unmap_sg(dev, sg, out, dir, NULL);
472 476
473 /* When it was forced or merged try again in a dumb way */ 477 /* When it was forced or merged try again in a dumb way */
474 if (force_iommu || iommu_merge) { 478 if (force_iommu || iommu_merge) {
@@ -521,7 +525,7 @@ static void
521gart_free_coherent(struct device *dev, size_t size, void *vaddr, 525gart_free_coherent(struct device *dev, size_t size, void *vaddr,
522 dma_addr_t dma_addr) 526 dma_addr_t dma_addr)
523{ 527{
524 gart_unmap_single(dev, dma_addr, size, DMA_BIDIRECTIONAL); 528 gart_unmap_page(dev, dma_addr, size, DMA_BIDIRECTIONAL, NULL);
525 free_pages((unsigned long)vaddr, get_order(size)); 529 free_pages((unsigned long)vaddr, get_order(size));
526} 530}
527 531
@@ -707,11 +711,11 @@ static __init int init_k8_gatt(struct agp_kern_info *info)
707 return -1; 711 return -1;
708} 712}
709 713
710static struct dma_mapping_ops gart_dma_ops = { 714static struct dma_map_ops gart_dma_ops = {
711 .map_single = gart_map_single,
712 .unmap_single = gart_unmap_single,
713 .map_sg = gart_map_sg, 715 .map_sg = gart_map_sg,
714 .unmap_sg = gart_unmap_sg, 716 .unmap_sg = gart_unmap_sg,
717 .map_page = gart_map_page,
718 .unmap_page = gart_unmap_page,
715 .alloc_coherent = gart_alloc_coherent, 719 .alloc_coherent = gart_alloc_coherent,
716 .free_coherent = gart_free_coherent, 720 .free_coherent = gart_free_coherent,
717}; 721};
diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c
index 8b02a3936d42..71d412a09f30 100644
--- a/arch/x86/kernel/pci-nommu.c
+++ b/arch/x86/kernel/pci-nommu.c
@@ -15,7 +15,7 @@ static int
15check_addr(char *name, struct device *hwdev, dma_addr_t bus, size_t size) 15check_addr(char *name, struct device *hwdev, dma_addr_t bus, size_t size)
16{ 16{
17 if (hwdev && !is_buffer_dma_capable(*hwdev->dma_mask, bus, size)) { 17 if (hwdev && !is_buffer_dma_capable(*hwdev->dma_mask, bus, size)) {
18 if (*hwdev->dma_mask >= DMA_32BIT_MASK) 18 if (*hwdev->dma_mask >= DMA_BIT_MASK(32))
19 printk(KERN_ERR 19 printk(KERN_ERR
20 "nommu_%s: overflow %Lx+%zu of device mask %Lx\n", 20 "nommu_%s: overflow %Lx+%zu of device mask %Lx\n",
21 name, (long long)bus, size, 21 name, (long long)bus, size,
@@ -25,19 +25,19 @@ check_addr(char *name, struct device *hwdev, dma_addr_t bus, size_t size)
25 return 1; 25 return 1;
26} 26}
27 27
28static dma_addr_t 28static dma_addr_t nommu_map_page(struct device *dev, struct page *page,
29nommu_map_single(struct device *hwdev, phys_addr_t paddr, size_t size, 29 unsigned long offset, size_t size,
30 int direction) 30 enum dma_data_direction dir,
31 struct dma_attrs *attrs)
31{ 32{
32 dma_addr_t bus = paddr; 33 dma_addr_t bus = page_to_phys(page) + offset;
33 WARN_ON(size == 0); 34 WARN_ON(size == 0);
34 if (!check_addr("map_single", hwdev, bus, size)) 35 if (!check_addr("map_single", dev, bus, size))
35 return bad_dma_address; 36 return bad_dma_address;
36 flush_write_buffers(); 37 flush_write_buffers();
37 return bus; 38 return bus;
38} 39}
39 40
40
41/* Map a set of buffers described by scatterlist in streaming 41/* Map a set of buffers described by scatterlist in streaming
42 * mode for DMA. This is the scatter-gather version of the 42 * mode for DMA. This is the scatter-gather version of the
43 * above pci_map_single interface. Here the scatter gather list 43 * above pci_map_single interface. Here the scatter gather list
@@ -54,7 +54,8 @@ nommu_map_single(struct device *hwdev, phys_addr_t paddr, size_t size,
54 * the same here. 54 * the same here.
55 */ 55 */
56static int nommu_map_sg(struct device *hwdev, struct scatterlist *sg, 56static int nommu_map_sg(struct device *hwdev, struct scatterlist *sg,
57 int nents, int direction) 57 int nents, enum dma_data_direction dir,
58 struct dma_attrs *attrs)
58{ 59{
59 struct scatterlist *s; 60 struct scatterlist *s;
60 int i; 61 int i;
@@ -78,11 +79,11 @@ static void nommu_free_coherent(struct device *dev, size_t size, void *vaddr,
78 free_pages((unsigned long)vaddr, get_order(size)); 79 free_pages((unsigned long)vaddr, get_order(size));
79} 80}
80 81
81struct dma_mapping_ops nommu_dma_ops = { 82struct dma_map_ops nommu_dma_ops = {
82 .alloc_coherent = dma_generic_alloc_coherent, 83 .alloc_coherent = dma_generic_alloc_coherent,
83 .free_coherent = nommu_free_coherent, 84 .free_coherent = nommu_free_coherent,
84 .map_single = nommu_map_single,
85 .map_sg = nommu_map_sg, 85 .map_sg = nommu_map_sg,
86 .map_page = nommu_map_page,
86 .is_phys = 1, 87 .is_phys = 1,
87}; 88};
88 89
diff --git a/arch/x86/kernel/pci-swiotlb_64.c b/arch/x86/kernel/pci-swiotlb.c
index d59c91747665..221a3853e268 100644
--- a/arch/x86/kernel/pci-swiotlb_64.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -33,18 +33,11 @@ phys_addr_t swiotlb_bus_to_phys(dma_addr_t baddr)
33 return baddr; 33 return baddr;
34} 34}
35 35
36int __weak swiotlb_arch_range_needs_mapping(void *ptr, size_t size) 36int __weak swiotlb_arch_range_needs_mapping(phys_addr_t paddr, size_t size)
37{ 37{
38 return 0; 38 return 0;
39} 39}
40 40
41static dma_addr_t
42swiotlb_map_single_phys(struct device *hwdev, phys_addr_t paddr, size_t size,
43 int direction)
44{
45 return swiotlb_map_single(hwdev, phys_to_virt(paddr), size, direction);
46}
47
48static void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size, 41static void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
49 dma_addr_t *dma_handle, gfp_t flags) 42 dma_addr_t *dma_handle, gfp_t flags)
50{ 43{
@@ -57,20 +50,20 @@ static void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
57 return swiotlb_alloc_coherent(hwdev, size, dma_handle, flags); 50 return swiotlb_alloc_coherent(hwdev, size, dma_handle, flags);
58} 51}
59 52
60struct dma_mapping_ops swiotlb_dma_ops = { 53static struct dma_map_ops swiotlb_dma_ops = {
61 .mapping_error = swiotlb_dma_mapping_error, 54 .mapping_error = swiotlb_dma_mapping_error,
62 .alloc_coherent = x86_swiotlb_alloc_coherent, 55 .alloc_coherent = x86_swiotlb_alloc_coherent,
63 .free_coherent = swiotlb_free_coherent, 56 .free_coherent = swiotlb_free_coherent,
64 .map_single = swiotlb_map_single_phys,
65 .unmap_single = swiotlb_unmap_single,
66 .sync_single_for_cpu = swiotlb_sync_single_for_cpu, 57 .sync_single_for_cpu = swiotlb_sync_single_for_cpu,
67 .sync_single_for_device = swiotlb_sync_single_for_device, 58 .sync_single_for_device = swiotlb_sync_single_for_device,
68 .sync_single_range_for_cpu = swiotlb_sync_single_range_for_cpu, 59 .sync_single_range_for_cpu = swiotlb_sync_single_range_for_cpu,
69 .sync_single_range_for_device = swiotlb_sync_single_range_for_device, 60 .sync_single_range_for_device = swiotlb_sync_single_range_for_device,
70 .sync_sg_for_cpu = swiotlb_sync_sg_for_cpu, 61 .sync_sg_for_cpu = swiotlb_sync_sg_for_cpu,
71 .sync_sg_for_device = swiotlb_sync_sg_for_device, 62 .sync_sg_for_device = swiotlb_sync_sg_for_device,
72 .map_sg = swiotlb_map_sg, 63 .map_sg = swiotlb_map_sg_attrs,
73 .unmap_sg = swiotlb_unmap_sg, 64 .unmap_sg = swiotlb_unmap_sg_attrs,
65 .map_page = swiotlb_map_page,
66 .unmap_page = swiotlb_unmap_page,
74 .dma_supported = NULL, 67 .dma_supported = NULL,
75}; 68};
76 69
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 156f87582c6c..ca989158e847 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -8,7 +8,7 @@
8#include <linux/module.h> 8#include <linux/module.h>
9#include <linux/pm.h> 9#include <linux/pm.h>
10#include <linux/clockchips.h> 10#include <linux/clockchips.h>
11#include <linux/ftrace.h> 11#include <trace/power.h>
12#include <asm/system.h> 12#include <asm/system.h>
13#include <asm/apic.h> 13#include <asm/apic.h>
14#include <asm/idle.h> 14#include <asm/idle.h>
@@ -22,6 +22,9 @@ EXPORT_SYMBOL(idle_nomwait);
22 22
23struct kmem_cache *task_xstate_cachep; 23struct kmem_cache *task_xstate_cachep;
24 24
25DEFINE_TRACE(power_start);
26DEFINE_TRACE(power_end);
27
25int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src) 28int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
26{ 29{
27 *dst = *src; 30 *dst = *src;
@@ -325,7 +328,7 @@ void stop_this_cpu(void *dummy)
325 /* 328 /*
326 * Remove this CPU: 329 * Remove this CPU:
327 */ 330 */
328 cpu_clear(smp_processor_id(), cpu_online_map); 331 set_cpu_online(smp_processor_id(), false);
329 disable_local_APIC(); 332 disable_local_APIC();
330 333
331 for (;;) { 334 for (;;) {
@@ -475,12 +478,13 @@ static int __cpuinit check_c1e_idle(const struct cpuinfo_x86 *c)
475 return 1; 478 return 1;
476} 479}
477 480
478static cpumask_t c1e_mask = CPU_MASK_NONE; 481static cpumask_var_t c1e_mask;
479static int c1e_detected; 482static int c1e_detected;
480 483
481void c1e_remove_cpu(int cpu) 484void c1e_remove_cpu(int cpu)
482{ 485{
483 cpu_clear(cpu, c1e_mask); 486 if (c1e_mask != NULL)
487 cpumask_clear_cpu(cpu, c1e_mask);
484} 488}
485 489
486/* 490/*
@@ -509,8 +513,8 @@ static void c1e_idle(void)
509 if (c1e_detected) { 513 if (c1e_detected) {
510 int cpu = smp_processor_id(); 514 int cpu = smp_processor_id();
511 515
512 if (!cpu_isset(cpu, c1e_mask)) { 516 if (!cpumask_test_cpu(cpu, c1e_mask)) {
513 cpu_set(cpu, c1e_mask); 517 cpumask_set_cpu(cpu, c1e_mask);
514 /* 518 /*
515 * Force broadcast so ACPI can not interfere. Needs 519 * Force broadcast so ACPI can not interfere. Needs
516 * to run with interrupts enabled as it uses 520 * to run with interrupts enabled as it uses
@@ -562,6 +566,15 @@ void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c)
562 pm_idle = default_idle; 566 pm_idle = default_idle;
563} 567}
564 568
569void __init init_c1e_mask(void)
570{
571 /* If we're using c1e_idle, we need to allocate c1e_mask. */
572 if (pm_idle == c1e_idle) {
573 alloc_cpumask_var(&c1e_mask, GFP_KERNEL);
574 cpumask_clear(c1e_mask);
575 }
576}
577
565static int __init idle_setup(char *str) 578static int __init idle_setup(char *str)
566{ 579{
567 if (!str) 580 if (!str)
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 14014d766cad..76f8f84043a2 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -245,7 +245,7 @@ void prepare_to_copy(struct task_struct *tsk)
245 unlazy_fpu(tsk); 245 unlazy_fpu(tsk);
246} 246}
247 247
248int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, 248int copy_thread(unsigned long clone_flags, unsigned long sp,
249 unsigned long unused, 249 unsigned long unused,
250 struct task_struct *p, struct pt_regs *regs) 250 struct task_struct *p, struct pt_regs *regs)
251{ 251{
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index abb7e6a7f0c6..b751a41392b1 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -278,7 +278,7 @@ void prepare_to_copy(struct task_struct *tsk)
278 unlazy_fpu(tsk); 278 unlazy_fpu(tsk);
279} 279}
280 280
281int copy_thread(int nr, unsigned long clone_flags, unsigned long sp, 281int copy_thread(unsigned long clone_flags, unsigned long sp,
282 unsigned long unused, 282 unsigned long unused,
283 struct task_struct *p, struct pt_regs *regs) 283 struct task_struct *p, struct pt_regs *regs)
284{ 284{
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 19378715f415..23b7c8f017e2 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -34,6 +34,8 @@
34#include <asm/proto.h> 34#include <asm/proto.h>
35#include <asm/ds.h> 35#include <asm/ds.h>
36 36
37#include <trace/syscall.h>
38
37#include "tls.h" 39#include "tls.h"
38 40
39enum x86_regset { 41enum x86_regset {
@@ -1415,6 +1417,9 @@ asmregparm long syscall_trace_enter(struct pt_regs *regs)
1415 tracehook_report_syscall_entry(regs)) 1417 tracehook_report_syscall_entry(regs))
1416 ret = -1L; 1418 ret = -1L;
1417 1419
1420 if (unlikely(test_thread_flag(TIF_SYSCALL_FTRACE)))
1421 ftrace_syscall_enter(regs);
1422
1418 if (unlikely(current->audit_context)) { 1423 if (unlikely(current->audit_context)) {
1419 if (IS_IA32) 1424 if (IS_IA32)
1420 audit_syscall_entry(AUDIT_ARCH_I386, 1425 audit_syscall_entry(AUDIT_ARCH_I386,
@@ -1438,6 +1443,9 @@ asmregparm void syscall_trace_leave(struct pt_regs *regs)
1438 if (unlikely(current->audit_context)) 1443 if (unlikely(current->audit_context))
1439 audit_syscall_exit(AUDITSC_RESULT(regs->ax), regs->ax); 1444 audit_syscall_exit(AUDITSC_RESULT(regs->ax), regs->ax);
1440 1445
1446 if (unlikely(test_thread_flag(TIF_SYSCALL_FTRACE)))
1447 ftrace_syscall_exit(regs);
1448
1441 if (test_thread_flag(TIF_SYSCALL_TRACE)) 1449 if (test_thread_flag(TIF_SYSCALL_TRACE))
1442 tracehook_report_syscall_exit(regs, 0); 1450 tracehook_report_syscall_exit(regs, 0);
1443 1451
@@ -1455,6 +1463,6 @@ asmregparm void syscall_trace_leave(struct pt_regs *regs)
1455 * system call instruction. 1463 * system call instruction.
1456 */ 1464 */
1457 if (test_thread_flag(TIF_SINGLESTEP) && 1465 if (test_thread_flag(TIF_SINGLESTEP) &&
1458 tracehook_consider_fatal_signal(current, SIGTRAP, SIG_DFL)) 1466 tracehook_consider_fatal_signal(current, SIGTRAP))
1459 send_sigtrap(current, regs, 0, TRAP_BRKPT); 1467 send_sigtrap(current, regs, 0, TRAP_BRKPT);
1460} 1468}
diff --git a/arch/x86/kernel/quirks.c b/arch/x86/kernel/quirks.c
index 6a5a2970f4c5..e95022e4f5d5 100644
--- a/arch/x86/kernel/quirks.c
+++ b/arch/x86/kernel/quirks.c
@@ -171,7 +171,8 @@ DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH8_4,
171 ich_force_enable_hpet); 171 ich_force_enable_hpet);
172DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH9_7, 172DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH9_7,
173 ich_force_enable_hpet); 173 ich_force_enable_hpet);
174 174DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x3a16, /* ICH10 */
175 ich_force_enable_hpet);
175 176
176static struct pci_dev *cached_dev; 177static struct pci_dev *cached_dev;
177 178
diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c
index 2aef36d8aca2..1340dad417f4 100644
--- a/arch/x86/kernel/reboot.c
+++ b/arch/x86/kernel/reboot.c
@@ -224,6 +224,14 @@ static struct dmi_system_id __initdata reboot_dmi_table[] = {
224 DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"), 224 DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"),
225 }, 225 },
226 }, 226 },
227 { /* Handle problems with rebooting on Dell DXP061 */
228 .callback = set_bios_reboot,
229 .ident = "Dell DXP061",
230 .matches = {
231 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
232 DMI_MATCH(DMI_PRODUCT_NAME, "Dell DXP061"),
233 },
234 },
227 { } 235 { }
228}; 236};
229 237
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index a0d26237d7cf..b4158439bf63 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -1049,7 +1049,6 @@ void __init x86_quirk_trap_init(void)
1049static struct irqaction irq0 = { 1049static struct irqaction irq0 = {
1050 .handler = timer_interrupt, 1050 .handler = timer_interrupt,
1051 .flags = IRQF_DISABLED | IRQF_NOBALANCING | IRQF_IRQPOLL | IRQF_TIMER, 1051 .flags = IRQF_DISABLED | IRQF_NOBALANCING | IRQF_IRQPOLL | IRQF_TIMER,
1052 .mask = CPU_MASK_NONE,
1053 .name = "timer" 1052 .name = "timer"
1054}; 1053};
1055 1054
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index 400331b50a53..3a97a4cf1872 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -153,7 +153,6 @@ static struct page * __init pcpur_get_page(unsigned int cpu, int pageno)
153static ssize_t __init setup_pcpu_remap(size_t static_size) 153static ssize_t __init setup_pcpu_remap(size_t static_size)
154{ 154{
155 static struct vm_struct vm; 155 static struct vm_struct vm;
156 pg_data_t *last;
157 size_t ptrs_size, dyn_size; 156 size_t ptrs_size, dyn_size;
158 unsigned int cpu; 157 unsigned int cpu;
159 ssize_t ret; 158 ssize_t ret;
@@ -162,22 +161,9 @@ static ssize_t __init setup_pcpu_remap(size_t static_size)
162 * If large page isn't supported, there's no benefit in doing 161 * If large page isn't supported, there's no benefit in doing
163 * this. Also, on non-NUMA, embedding is better. 162 * this. Also, on non-NUMA, embedding is better.
164 */ 163 */
165 if (!cpu_has_pse || pcpu_need_numa()) 164 if (!cpu_has_pse || !pcpu_need_numa())
166 return -EINVAL; 165 return -EINVAL;
167 166
168 last = NULL;
169 for_each_possible_cpu(cpu) {
170 int node = early_cpu_to_node(cpu);
171
172 if (node_online(node) && NODE_DATA(node) &&
173 last && last != NODE_DATA(node))
174 goto proceed;
175
176 last = NODE_DATA(node);
177 }
178 return -EINVAL;
179
180proceed:
181 /* 167 /*
182 * Currently supports only single page. Supporting multiple 168 * Currently supports only single page. Supporting multiple
183 * pages won't be too difficult if it ever becomes necessary. 169 * pages won't be too difficult if it ever becomes necessary.
diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c
index dfcc74ab0ab6..14425166b8e3 100644
--- a/arch/x86/kernel/signal.c
+++ b/arch/x86/kernel/signal.c
@@ -221,7 +221,7 @@ get_sigframe(struct k_sigaction *ka, struct pt_regs *regs, size_t frame_size,
221 if (!onsigstack) { 221 if (!onsigstack) {
222 /* This is the X/Open sanctioned signal stack switching. */ 222 /* This is the X/Open sanctioned signal stack switching. */
223 if (ka->sa.sa_flags & SA_ONSTACK) { 223 if (ka->sa.sa_flags & SA_ONSTACK) {
224 if (sas_ss_flags(sp) == 0) 224 if (current->sas_ss_size)
225 sp = current->sas_ss_sp + current->sas_ss_size; 225 sp = current->sas_ss_sp + current->sas_ss_size;
226 } else { 226 } else {
227#ifdef CONFIG_X86_32 227#ifdef CONFIG_X86_32
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index ef7d10170c30..58d24ef917d8 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -101,11 +101,11 @@ EXPORT_SYMBOL(smp_num_siblings);
101DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID; 101DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID;
102 102
103/* representing HT siblings of each logical CPU */ 103/* representing HT siblings of each logical CPU */
104DEFINE_PER_CPU(cpumask_t, cpu_sibling_map); 104DEFINE_PER_CPU(cpumask_var_t, cpu_sibling_map);
105EXPORT_PER_CPU_SYMBOL(cpu_sibling_map); 105EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
106 106
107/* representing HT and core siblings of each logical CPU */ 107/* representing HT and core siblings of each logical CPU */
108DEFINE_PER_CPU(cpumask_t, cpu_core_map); 108DEFINE_PER_CPU(cpumask_var_t, cpu_core_map);
109EXPORT_PER_CPU_SYMBOL(cpu_core_map); 109EXPORT_PER_CPU_SYMBOL(cpu_core_map);
110 110
111/* Per CPU bogomips and other parameters */ 111/* Per CPU bogomips and other parameters */
@@ -115,11 +115,6 @@ EXPORT_PER_CPU_SYMBOL(cpu_info);
115atomic_t init_deasserted; 115atomic_t init_deasserted;
116 116
117#if defined(CONFIG_NUMA) && defined(CONFIG_X86_32) 117#if defined(CONFIG_NUMA) && defined(CONFIG_X86_32)
118
119/* which logical CPUs are on which nodes */
120cpumask_t node_to_cpumask_map[MAX_NUMNODES] __read_mostly =
121 { [0 ... MAX_NUMNODES-1] = CPU_MASK_NONE };
122EXPORT_SYMBOL(node_to_cpumask_map);
123/* which node each logical CPU is on */ 118/* which node each logical CPU is on */
124int cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 }; 119int cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
125EXPORT_SYMBOL(cpu_to_node_map); 120EXPORT_SYMBOL(cpu_to_node_map);
@@ -128,7 +123,7 @@ EXPORT_SYMBOL(cpu_to_node_map);
128static void map_cpu_to_node(int cpu, int node) 123static void map_cpu_to_node(int cpu, int node)
129{ 124{
130 printk(KERN_INFO "Mapping cpu %d to node %d\n", cpu, node); 125 printk(KERN_INFO "Mapping cpu %d to node %d\n", cpu, node);
131 cpumask_set_cpu(cpu, &node_to_cpumask_map[node]); 126 cpumask_set_cpu(cpu, node_to_cpumask_map[node]);
132 cpu_to_node_map[cpu] = node; 127 cpu_to_node_map[cpu] = node;
133} 128}
134 129
@@ -139,7 +134,7 @@ static void unmap_cpu_to_node(int cpu)
139 134
140 printk(KERN_INFO "Unmapping cpu %d from all nodes\n", cpu); 135 printk(KERN_INFO "Unmapping cpu %d from all nodes\n", cpu);
141 for (node = 0; node < MAX_NUMNODES; node++) 136 for (node = 0; node < MAX_NUMNODES; node++)
142 cpumask_clear_cpu(cpu, &node_to_cpumask_map[node]); 137 cpumask_clear_cpu(cpu, node_to_cpumask_map[node]);
143 cpu_to_node_map[cpu] = 0; 138 cpu_to_node_map[cpu] = 0;
144} 139}
145#else /* !(CONFIG_NUMA && CONFIG_X86_32) */ 140#else /* !(CONFIG_NUMA && CONFIG_X86_32) */
@@ -301,7 +296,7 @@ notrace static void __cpuinit start_secondary(void *unused)
301 __flush_tlb_all(); 296 __flush_tlb_all();
302#endif 297#endif
303 298
304 /* This must be done before setting cpu_online_map */ 299 /* This must be done before setting cpu_online_mask */
305 set_cpu_sibling_map(raw_smp_processor_id()); 300 set_cpu_sibling_map(raw_smp_processor_id());
306 wmb(); 301 wmb();
307 302
@@ -334,6 +329,23 @@ notrace static void __cpuinit start_secondary(void *unused)
334 cpu_idle(); 329 cpu_idle();
335} 330}
336 331
332#ifdef CONFIG_CPUMASK_OFFSTACK
333/* In this case, llc_shared_map is a pointer to a cpumask. */
334static inline void copy_cpuinfo_x86(struct cpuinfo_x86 *dst,
335 const struct cpuinfo_x86 *src)
336{
337 struct cpumask *llc = dst->llc_shared_map;
338 *dst = *src;
339 dst->llc_shared_map = llc;
340}
341#else
342static inline void copy_cpuinfo_x86(struct cpuinfo_x86 *dst,
343 const struct cpuinfo_x86 *src)
344{
345 *dst = *src;
346}
347#endif /* CONFIG_CPUMASK_OFFSTACK */
348
337/* 349/*
338 * The bootstrap kernel entry code has set these up. Save them for 350 * The bootstrap kernel entry code has set these up. Save them for
339 * a given CPU 351 * a given CPU
@@ -343,7 +355,7 @@ void __cpuinit smp_store_cpu_info(int id)
343{ 355{
344 struct cpuinfo_x86 *c = &cpu_data(id); 356 struct cpuinfo_x86 *c = &cpu_data(id);
345 357
346 *c = boot_cpu_data; 358 copy_cpuinfo_x86(c, &boot_cpu_data);
347 c->cpu_index = id; 359 c->cpu_index = id;
348 if (id != 0) 360 if (id != 0)
349 identify_secondary_cpu(c); 361 identify_secondary_cpu(c);
@@ -367,15 +379,15 @@ void __cpuinit set_cpu_sibling_map(int cpu)
367 cpumask_set_cpu(cpu, cpu_sibling_mask(i)); 379 cpumask_set_cpu(cpu, cpu_sibling_mask(i));
368 cpumask_set_cpu(i, cpu_core_mask(cpu)); 380 cpumask_set_cpu(i, cpu_core_mask(cpu));
369 cpumask_set_cpu(cpu, cpu_core_mask(i)); 381 cpumask_set_cpu(cpu, cpu_core_mask(i));
370 cpumask_set_cpu(i, &c->llc_shared_map); 382 cpumask_set_cpu(i, c->llc_shared_map);
371 cpumask_set_cpu(cpu, &o->llc_shared_map); 383 cpumask_set_cpu(cpu, o->llc_shared_map);
372 } 384 }
373 } 385 }
374 } else { 386 } else {
375 cpumask_set_cpu(cpu, cpu_sibling_mask(cpu)); 387 cpumask_set_cpu(cpu, cpu_sibling_mask(cpu));
376 } 388 }
377 389
378 cpumask_set_cpu(cpu, &c->llc_shared_map); 390 cpumask_set_cpu(cpu, c->llc_shared_map);
379 391
380 if (current_cpu_data.x86_max_cores == 1) { 392 if (current_cpu_data.x86_max_cores == 1) {
381 cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu)); 393 cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu));
@@ -386,8 +398,8 @@ void __cpuinit set_cpu_sibling_map(int cpu)
386 for_each_cpu(i, cpu_sibling_setup_mask) { 398 for_each_cpu(i, cpu_sibling_setup_mask) {
387 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID && 399 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
388 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) { 400 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
389 cpumask_set_cpu(i, &c->llc_shared_map); 401 cpumask_set_cpu(i, c->llc_shared_map);
390 cpumask_set_cpu(cpu, &cpu_data(i).llc_shared_map); 402 cpumask_set_cpu(cpu, cpu_data(i).llc_shared_map);
391 } 403 }
392 if (c->phys_proc_id == cpu_data(i).phys_proc_id) { 404 if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
393 cpumask_set_cpu(i, cpu_core_mask(cpu)); 405 cpumask_set_cpu(i, cpu_core_mask(cpu));
@@ -425,12 +437,7 @@ const struct cpumask *cpu_coregroup_mask(int cpu)
425 if (sched_mc_power_savings || sched_smt_power_savings) 437 if (sched_mc_power_savings || sched_smt_power_savings)
426 return cpu_core_mask(cpu); 438 return cpu_core_mask(cpu);
427 else 439 else
428 return &c->llc_shared_map; 440 return c->llc_shared_map;
429}
430
431cpumask_t cpu_coregroup_map(int cpu)
432{
433 return *cpu_coregroup_mask(cpu);
434} 441}
435 442
436static void impress_friends(void) 443static void impress_friends(void)
@@ -897,9 +904,8 @@ int __cpuinit native_cpu_up(unsigned int cpu)
897 */ 904 */
898static __init void disable_smp(void) 905static __init void disable_smp(void)
899{ 906{
900 /* use the read/write pointers to the present and possible maps */ 907 init_cpu_present(cpumask_of(0));
901 cpumask_copy(&cpu_present_map, cpumask_of(0)); 908 init_cpu_possible(cpumask_of(0));
902 cpumask_copy(&cpu_possible_map, cpumask_of(0));
903 smpboot_clear_io_apic_irqs(); 909 smpboot_clear_io_apic_irqs();
904 910
905 if (smp_found_config) 911 if (smp_found_config)
@@ -1031,6 +1037,8 @@ static void __init smp_cpu_index_default(void)
1031 */ 1037 */
1032void __init native_smp_prepare_cpus(unsigned int max_cpus) 1038void __init native_smp_prepare_cpus(unsigned int max_cpus)
1033{ 1039{
1040 unsigned int i;
1041
1034 preempt_disable(); 1042 preempt_disable();
1035 smp_cpu_index_default(); 1043 smp_cpu_index_default();
1036 current_cpu_data = boot_cpu_data; 1044 current_cpu_data = boot_cpu_data;
@@ -1044,6 +1052,14 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
1044 boot_cpu_logical_apicid = logical_smp_processor_id(); 1052 boot_cpu_logical_apicid = logical_smp_processor_id();
1045#endif 1053#endif
1046 current_thread_info()->cpu = 0; /* needed? */ 1054 current_thread_info()->cpu = 0; /* needed? */
1055 for_each_possible_cpu(i) {
1056 alloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL);
1057 alloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL);
1058 alloc_cpumask_var(&cpu_data(i).llc_shared_map, GFP_KERNEL);
1059 cpumask_clear(per_cpu(cpu_core_map, i));
1060 cpumask_clear(per_cpu(cpu_sibling_map, i));
1061 cpumask_clear(cpu_data(i).llc_shared_map);
1062 }
1047 set_cpu_sibling_map(0); 1063 set_cpu_sibling_map(0);
1048 1064
1049 enable_IR_x2apic(); 1065 enable_IR_x2apic();
@@ -1132,11 +1148,11 @@ early_param("possible_cpus", _setup_possible_cpus);
1132 1148
1133 1149
1134/* 1150/*
1135 * cpu_possible_map should be static, it cannot change as cpu's 1151 * cpu_possible_mask should be static, it cannot change as cpu's
1136 * are onlined, or offlined. The reason is per-cpu data-structures 1152 * are onlined, or offlined. The reason is per-cpu data-structures
1137 * are allocated by some modules at init time, and dont expect to 1153 * are allocated by some modules at init time, and dont expect to
1138 * do this dynamically on cpu arrival/departure. 1154 * do this dynamically on cpu arrival/departure.
1139 * cpu_present_map on the other hand can change dynamically. 1155 * cpu_present_mask on the other hand can change dynamically.
1140 * In case when cpu_hotplug is not compiled, then we resort to current 1156 * In case when cpu_hotplug is not compiled, then we resort to current
1141 * behaviour, which is cpu_possible == cpu_present. 1157 * behaviour, which is cpu_possible == cpu_present.
1142 * - Ashok Raj 1158 * - Ashok Raj
diff --git a/arch/x86/kernel/syscall_table_32.S b/arch/x86/kernel/syscall_table_32.S
index 3bdb64829b82..ff5c8736b491 100644
--- a/arch/x86/kernel/syscall_table_32.S
+++ b/arch/x86/kernel/syscall_table_32.S
@@ -332,3 +332,5 @@ ENTRY(sys_call_table)
332 .long sys_dup3 /* 330 */ 332 .long sys_dup3 /* 330 */
333 .long sys_pipe2 333 .long sys_pipe2
334 .long sys_inotify_init1 334 .long sys_inotify_init1
335 .long sys_preadv
336 .long sys_pwritev
diff --git a/arch/x86/kernel/time_64.c b/arch/x86/kernel/time_64.c
index 241ec3923f61..5ba343e61844 100644
--- a/arch/x86/kernel/time_64.c
+++ b/arch/x86/kernel/time_64.c
@@ -116,7 +116,6 @@ unsigned long __init calibrate_cpu(void)
116static struct irqaction irq0 = { 116static struct irqaction irq0 = {
117 .handler = timer_interrupt, 117 .handler = timer_interrupt,
118 .flags = IRQF_DISABLED | IRQF_IRQPOLL | IRQF_NOBALANCING | IRQF_TIMER, 118 .flags = IRQF_DISABLED | IRQF_IRQPOLL | IRQF_NOBALANCING | IRQF_TIMER,
119 .mask = CPU_MASK_NONE,
120 .name = "timer" 119 .name = "timer"
121}; 120};
122 121
@@ -125,7 +124,6 @@ void __init hpet_time_init(void)
125 if (!hpet_enable()) 124 if (!hpet_enable())
126 setup_pit_timer(); 125 setup_pit_timer();
127 126
128 irq0.mask = cpumask_of_cpu(0);
129 setup_irq(0, &irq0); 127 setup_irq(0, &irq0);
130} 128}
131 129
diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c
index 79c073247284..ed0c33761e6d 100644
--- a/arch/x86/kernel/tlb_uv.c
+++ b/arch/x86/kernel/tlb_uv.c
@@ -25,6 +25,8 @@ static int uv_bau_retry_limit __read_mostly;
25 25
26/* position of pnode (which is nasid>>1): */ 26/* position of pnode (which is nasid>>1): */
27static int uv_nshift __read_mostly; 27static int uv_nshift __read_mostly;
28/* base pnode in this partition */
29static int uv_partition_base_pnode __read_mostly;
28 30
29static unsigned long uv_mmask __read_mostly; 31static unsigned long uv_mmask __read_mostly;
30 32
@@ -32,6 +34,34 @@ static DEFINE_PER_CPU(struct ptc_stats, ptcstats);
32static DEFINE_PER_CPU(struct bau_control, bau_control); 34static DEFINE_PER_CPU(struct bau_control, bau_control);
33 35
34/* 36/*
37 * Determine the first node on a blade.
38 */
39static int __init blade_to_first_node(int blade)
40{
41 int node, b;
42
43 for_each_online_node(node) {
44 b = uv_node_to_blade_id(node);
45 if (blade == b)
46 return node;
47 }
48 return -1; /* shouldn't happen */
49}
50
51/*
52 * Determine the apicid of the first cpu on a blade.
53 */
54static int __init blade_to_first_apicid(int blade)
55{
56 int cpu;
57
58 for_each_present_cpu(cpu)
59 if (blade == uv_cpu_to_blade_id(cpu))
60 return per_cpu(x86_cpu_to_apicid, cpu);
61 return -1;
62}
63
64/*
35 * Free a software acknowledge hardware resource by clearing its Pending 65 * Free a software acknowledge hardware resource by clearing its Pending
36 * bit. This will return a reply to the sender. 66 * bit. This will return a reply to the sender.
37 * If the message has timed out, a reply has already been sent by the 67 * If the message has timed out, a reply has already been sent by the
@@ -67,7 +97,7 @@ static void uv_bau_process_message(struct bau_payload_queue_entry *msg,
67 msp = __get_cpu_var(bau_control).msg_statuses + msg_slot; 97 msp = __get_cpu_var(bau_control).msg_statuses + msg_slot;
68 cpu = uv_blade_processor_id(); 98 cpu = uv_blade_processor_id();
69 msg->number_of_cpus = 99 msg->number_of_cpus =
70 uv_blade_nr_online_cpus(uv_node_to_blade_id(numa_node_id())); 100 uv_blade_nr_online_cpus(uv_node_to_blade_id(numa_node_id()));
71 this_cpu_mask = 1UL << cpu; 101 this_cpu_mask = 1UL << cpu;
72 if (msp->seen_by.bits & this_cpu_mask) 102 if (msp->seen_by.bits & this_cpu_mask)
73 return; 103 return;
@@ -215,14 +245,14 @@ static int uv_wait_completion(struct bau_desc *bau_desc,
215 * Returns @flush_mask if some remote flushing remains to be done. The 245 * Returns @flush_mask if some remote flushing remains to be done. The
216 * mask will have some bits still set. 246 * mask will have some bits still set.
217 */ 247 */
218const struct cpumask *uv_flush_send_and_wait(int cpu, int this_blade, 248const struct cpumask *uv_flush_send_and_wait(int cpu, int this_pnode,
219 struct bau_desc *bau_desc, 249 struct bau_desc *bau_desc,
220 struct cpumask *flush_mask) 250 struct cpumask *flush_mask)
221{ 251{
222 int completion_status = 0; 252 int completion_status = 0;
223 int right_shift; 253 int right_shift;
224 int tries = 0; 254 int tries = 0;
225 int blade; 255 int pnode;
226 int bit; 256 int bit;
227 unsigned long mmr_offset; 257 unsigned long mmr_offset;
228 unsigned long index; 258 unsigned long index;
@@ -265,8 +295,8 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_blade,
265 * use the IPI method of shootdown on them. 295 * use the IPI method of shootdown on them.
266 */ 296 */
267 for_each_cpu(bit, flush_mask) { 297 for_each_cpu(bit, flush_mask) {
268 blade = uv_cpu_to_blade_id(bit); 298 pnode = uv_cpu_to_pnode(bit);
269 if (blade == this_blade) 299 if (pnode == this_pnode)
270 continue; 300 continue;
271 cpumask_clear_cpu(bit, flush_mask); 301 cpumask_clear_cpu(bit, flush_mask);
272 } 302 }
@@ -275,6 +305,8 @@ const struct cpumask *uv_flush_send_and_wait(int cpu, int this_blade,
275 return NULL; 305 return NULL;
276} 306}
277 307
308static DEFINE_PER_CPU(cpumask_var_t, uv_flush_tlb_mask);
309
278/** 310/**
279 * uv_flush_tlb_others - globally purge translation cache of a virtual 311 * uv_flush_tlb_others - globally purge translation cache of a virtual
280 * address or all TLB's 312 * address or all TLB's
@@ -304,20 +336,19 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
304 struct mm_struct *mm, 336 struct mm_struct *mm,
305 unsigned long va, unsigned int cpu) 337 unsigned long va, unsigned int cpu)
306{ 338{
307 static DEFINE_PER_CPU(cpumask_t, flush_tlb_mask); 339 struct cpumask *flush_mask = __get_cpu_var(uv_flush_tlb_mask);
308 struct cpumask *flush_mask = &__get_cpu_var(flush_tlb_mask);
309 int i; 340 int i;
310 int bit; 341 int bit;
311 int blade; 342 int pnode;
312 int uv_cpu; 343 int uv_cpu;
313 int this_blade; 344 int this_pnode;
314 int locals = 0; 345 int locals = 0;
315 struct bau_desc *bau_desc; 346 struct bau_desc *bau_desc;
316 347
317 cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu)); 348 cpumask_andnot(flush_mask, cpumask, cpumask_of(cpu));
318 349
319 uv_cpu = uv_blade_processor_id(); 350 uv_cpu = uv_blade_processor_id();
320 this_blade = uv_numa_blade_id(); 351 this_pnode = uv_hub_info->pnode;
321 bau_desc = __get_cpu_var(bau_control).descriptor_base; 352 bau_desc = __get_cpu_var(bau_control).descriptor_base;
322 bau_desc += UV_ITEMS_PER_DESCRIPTOR * uv_cpu; 353 bau_desc += UV_ITEMS_PER_DESCRIPTOR * uv_cpu;
323 354
@@ -325,13 +356,14 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
325 356
326 i = 0; 357 i = 0;
327 for_each_cpu(bit, flush_mask) { 358 for_each_cpu(bit, flush_mask) {
328 blade = uv_cpu_to_blade_id(bit); 359 pnode = uv_cpu_to_pnode(bit);
329 BUG_ON(blade > (UV_DISTRIBUTION_SIZE - 1)); 360 BUG_ON(pnode > (UV_DISTRIBUTION_SIZE - 1));
330 if (blade == this_blade) { 361 if (pnode == this_pnode) {
331 locals++; 362 locals++;
332 continue; 363 continue;
333 } 364 }
334 bau_node_set(blade, &bau_desc->distribution); 365 bau_node_set(pnode - uv_partition_base_pnode,
366 &bau_desc->distribution);
335 i++; 367 i++;
336 } 368 }
337 if (i == 0) { 369 if (i == 0) {
@@ -349,7 +381,7 @@ const struct cpumask *uv_flush_tlb_others(const struct cpumask *cpumask,
349 bau_desc->payload.address = va; 381 bau_desc->payload.address = va;
350 bau_desc->payload.sending_cpu = cpu; 382 bau_desc->payload.sending_cpu = cpu;
351 383
352 return uv_flush_send_and_wait(uv_cpu, this_blade, bau_desc, flush_mask); 384 return uv_flush_send_and_wait(uv_cpu, this_pnode, bau_desc, flush_mask);
353} 385}
354 386
355/* 387/*
@@ -417,24 +449,58 @@ void uv_bau_message_interrupt(struct pt_regs *regs)
417 set_irq_regs(old_regs); 449 set_irq_regs(old_regs);
418} 450}
419 451
452/*
453 * uv_enable_timeouts
454 *
455 * Each target blade (i.e. blades that have cpu's) needs to have
456 * shootdown message timeouts enabled. The timeout does not cause
457 * an interrupt, but causes an error message to be returned to
458 * the sender.
459 */
420static void uv_enable_timeouts(void) 460static void uv_enable_timeouts(void)
421{ 461{
422 int i;
423 int blade; 462 int blade;
424 int last_blade; 463 int nblades;
425 int pnode; 464 int pnode;
426 int cur_cpu = 0; 465 unsigned long mmr_image;
427 unsigned long apicid;
428 466
429 last_blade = -1; 467 nblades = uv_num_possible_blades();
430 for_each_online_node(i) { 468
431 blade = uv_node_to_blade_id(i); 469 for (blade = 0; blade < nblades; blade++) {
432 if (blade == last_blade) 470 if (!uv_blade_nr_possible_cpus(blade))
433 continue; 471 continue;
434 last_blade = blade; 472
435 apicid = per_cpu(x86_cpu_to_apicid, cur_cpu);
436 pnode = uv_blade_to_pnode(blade); 473 pnode = uv_blade_to_pnode(blade);
437 cur_cpu += uv_blade_nr_possible_cpus(i); 474 mmr_image =
475 uv_read_global_mmr64(pnode, UVH_LB_BAU_MISC_CONTROL);
476 /*
477 * Set the timeout period and then lock it in, in three
478 * steps; captures and locks in the period.
479 *
480 * To program the period, the SOFT_ACK_MODE must be off.
481 */
482 mmr_image &= ~((unsigned long)1 <<
483 UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT);
484 uv_write_global_mmr64
485 (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
486 /*
487 * Set the 4-bit period.
488 */
489 mmr_image &= ~((unsigned long)0xf <<
490 UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT);
491 mmr_image |= (UV_INTD_SOFT_ACK_TIMEOUT_PERIOD <<
492 UV_INTD_SOFT_ACK_TIMEOUT_PERIOD_SHIFT);
493 uv_write_global_mmr64
494 (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
495 /*
496 * Subsequent reversals of the timebase bit (3) cause an
497 * immediate timeout of one or all INTD resources as
498 * indicated in bits 2:0 (7 causes all of them to timeout).
499 */
500 mmr_image |= ((unsigned long)1 <<
501 UV_ENABLE_INTD_SOFT_ACK_MODE_SHIFT);
502 uv_write_global_mmr64
503 (pnode, UVH_LB_BAU_MISC_CONTROL, mmr_image);
438 } 504 }
439} 505}
440 506
@@ -481,8 +547,7 @@ static int uv_ptc_seq_show(struct seq_file *file, void *data)
481 stat->requestee, stat->onetlb, stat->alltlb, 547 stat->requestee, stat->onetlb, stat->alltlb,
482 stat->s_retry, stat->d_retry, stat->ptc_i); 548 stat->s_retry, stat->d_retry, stat->ptc_i);
483 seq_printf(file, "%lx %ld %ld %ld %ld %ld %ld\n", 549 seq_printf(file, "%lx %ld %ld %ld %ld %ld %ld\n",
484 uv_read_global_mmr64(uv_blade_to_pnode 550 uv_read_global_mmr64(uv_cpu_to_pnode(cpu),
485 (uv_cpu_to_blade_id(cpu)),
486 UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE), 551 UVH_LB_BAU_INTD_SOFTWARE_ACKNOWLEDGE),
487 stat->sflush, stat->dflush, 552 stat->sflush, stat->dflush,
488 stat->retriesok, stat->nomsg, 553 stat->retriesok, stat->nomsg,
@@ -616,16 +681,18 @@ static struct bau_control * __init uv_table_bases_init(int blade, int node)
616 * finish the initialization of the per-blade control structures 681 * finish the initialization of the per-blade control structures
617 */ 682 */
618static void __init 683static void __init
619uv_table_bases_finish(int blade, int node, int cur_cpu, 684uv_table_bases_finish(int blade,
620 struct bau_control *bau_tablesp, 685 struct bau_control *bau_tablesp,
621 struct bau_desc *adp) 686 struct bau_desc *adp)
622{ 687{
623 struct bau_control *bcp; 688 struct bau_control *bcp;
624 int i; 689 int cpu;
625 690
626 for (i = cur_cpu; i < cur_cpu + uv_blade_nr_possible_cpus(blade); i++) { 691 for_each_present_cpu(cpu) {
627 bcp = (struct bau_control *)&per_cpu(bau_control, i); 692 if (blade != uv_cpu_to_blade_id(cpu))
693 continue;
628 694
695 bcp = (struct bau_control *)&per_cpu(bau_control, cpu);
629 bcp->bau_msg_head = bau_tablesp->va_queue_first; 696 bcp->bau_msg_head = bau_tablesp->va_queue_first;
630 bcp->va_queue_first = bau_tablesp->va_queue_first; 697 bcp->va_queue_first = bau_tablesp->va_queue_first;
631 bcp->va_queue_last = bau_tablesp->va_queue_last; 698 bcp->va_queue_last = bau_tablesp->va_queue_last;
@@ -648,11 +715,10 @@ uv_activation_descriptor_init(int node, int pnode)
648 struct bau_desc *adp; 715 struct bau_desc *adp;
649 struct bau_desc *ad2; 716 struct bau_desc *ad2;
650 717
651 adp = (struct bau_desc *) 718 adp = (struct bau_desc *)kmalloc_node(16384, GFP_KERNEL, node);
652 kmalloc_node(16384, GFP_KERNEL, node);
653 BUG_ON(!adp); 719 BUG_ON(!adp);
654 720
655 pa = __pa((unsigned long)adp); 721 pa = uv_gpa(adp); /* need the real nasid*/
656 n = pa >> uv_nshift; 722 n = pa >> uv_nshift;
657 m = pa & uv_mmask; 723 m = pa & uv_mmask;
658 724
@@ -666,8 +732,12 @@ uv_activation_descriptor_init(int node, int pnode)
666 for (i = 0, ad2 = adp; i < UV_ACTIVATION_DESCRIPTOR_SIZE; i++, ad2++) { 732 for (i = 0, ad2 = adp; i < UV_ACTIVATION_DESCRIPTOR_SIZE; i++, ad2++) {
667 memset(ad2, 0, sizeof(struct bau_desc)); 733 memset(ad2, 0, sizeof(struct bau_desc));
668 ad2->header.sw_ack_flag = 1; 734 ad2->header.sw_ack_flag = 1;
669 ad2->header.base_dest_nodeid = 735 /*
670 uv_blade_to_pnode(uv_cpu_to_blade_id(0)); 736 * base_dest_nodeid is the first node in the partition, so
737 * the bit map will indicate partition-relative node numbers.
738 * note that base_dest_nodeid is actually a nasid.
739 */
740 ad2->header.base_dest_nodeid = uv_partition_base_pnode << 1;
671 ad2->header.command = UV_NET_ENDPOINT_INTD; 741 ad2->header.command = UV_NET_ENDPOINT_INTD;
672 ad2->header.int_both = 1; 742 ad2->header.int_both = 1;
673 /* 743 /*
@@ -685,6 +755,8 @@ static struct bau_payload_queue_entry * __init
685uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp) 755uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp)
686{ 756{
687 struct bau_payload_queue_entry *pqp; 757 struct bau_payload_queue_entry *pqp;
758 unsigned long pa;
759 int pn;
688 char *cp; 760 char *cp;
689 761
690 pqp = (struct bau_payload_queue_entry *) kmalloc_node( 762 pqp = (struct bau_payload_queue_entry *) kmalloc_node(
@@ -695,10 +767,14 @@ uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp)
695 cp = (char *)pqp + 31; 767 cp = (char *)pqp + 31;
696 pqp = (struct bau_payload_queue_entry *)(((unsigned long)cp >> 5) << 5); 768 pqp = (struct bau_payload_queue_entry *)(((unsigned long)cp >> 5) << 5);
697 bau_tablesp->va_queue_first = pqp; 769 bau_tablesp->va_queue_first = pqp;
770 /*
771 * need the pnode of where the memory was really allocated
772 */
773 pa = uv_gpa(pqp);
774 pn = pa >> uv_nshift;
698 uv_write_global_mmr64(pnode, 775 uv_write_global_mmr64(pnode,
699 UVH_LB_BAU_INTD_PAYLOAD_QUEUE_FIRST, 776 UVH_LB_BAU_INTD_PAYLOAD_QUEUE_FIRST,
700 ((unsigned long)pnode << 777 ((unsigned long)pn << UV_PAYLOADQ_PNODE_SHIFT) |
701 UV_PAYLOADQ_PNODE_SHIFT) |
702 uv_physnodeaddr(pqp)); 778 uv_physnodeaddr(pqp));
703 uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_TAIL, 779 uv_write_global_mmr64(pnode, UVH_LB_BAU_INTD_PAYLOAD_QUEUE_TAIL,
704 uv_physnodeaddr(pqp)); 780 uv_physnodeaddr(pqp));
@@ -714,8 +790,9 @@ uv_payload_queue_init(int node, int pnode, struct bau_control *bau_tablesp)
714/* 790/*
715 * Initialization of each UV blade's structures 791 * Initialization of each UV blade's structures
716 */ 792 */
717static int __init uv_init_blade(int blade, int node, int cur_cpu) 793static int __init uv_init_blade(int blade)
718{ 794{
795 int node;
719 int pnode; 796 int pnode;
720 unsigned long pa; 797 unsigned long pa;
721 unsigned long apicid; 798 unsigned long apicid;
@@ -723,16 +800,17 @@ static int __init uv_init_blade(int blade, int node, int cur_cpu)
723 struct bau_payload_queue_entry *pqp; 800 struct bau_payload_queue_entry *pqp;
724 struct bau_control *bau_tablesp; 801 struct bau_control *bau_tablesp;
725 802
803 node = blade_to_first_node(blade);
726 bau_tablesp = uv_table_bases_init(blade, node); 804 bau_tablesp = uv_table_bases_init(blade, node);
727 pnode = uv_blade_to_pnode(blade); 805 pnode = uv_blade_to_pnode(blade);
728 adp = uv_activation_descriptor_init(node, pnode); 806 adp = uv_activation_descriptor_init(node, pnode);
729 pqp = uv_payload_queue_init(node, pnode, bau_tablesp); 807 pqp = uv_payload_queue_init(node, pnode, bau_tablesp);
730 uv_table_bases_finish(blade, node, cur_cpu, bau_tablesp, adp); 808 uv_table_bases_finish(blade, bau_tablesp, adp);
731 /* 809 /*
732 * the below initialization can't be in firmware because the 810 * the below initialization can't be in firmware because the
733 * messaging IRQ will be determined by the OS 811 * messaging IRQ will be determined by the OS
734 */ 812 */
735 apicid = per_cpu(x86_cpu_to_apicid, cur_cpu); 813 apicid = blade_to_first_apicid(blade);
736 pa = uv_read_global_mmr64(pnode, UVH_BAU_DATA_CONFIG); 814 pa = uv_read_global_mmr64(pnode, UVH_BAU_DATA_CONFIG);
737 if ((pa & 0xff) != UV_BAU_MESSAGE) { 815 if ((pa & 0xff) != UV_BAU_MESSAGE) {
738 uv_write_global_mmr64(pnode, UVH_BAU_DATA_CONFIG, 816 uv_write_global_mmr64(pnode, UVH_BAU_DATA_CONFIG,
@@ -747,40 +825,34 @@ static int __init uv_init_blade(int blade, int node, int cur_cpu)
747static int __init uv_bau_init(void) 825static int __init uv_bau_init(void)
748{ 826{
749 int blade; 827 int blade;
750 int node;
751 int nblades; 828 int nblades;
752 int last_blade;
753 int cur_cpu; 829 int cur_cpu;
754 830
755 if (!is_uv_system()) 831 if (!is_uv_system())
756 return 0; 832 return 0;
757 833
834 for_each_possible_cpu(cur_cpu)
835 alloc_cpumask_var_node(&per_cpu(uv_flush_tlb_mask, cur_cpu),
836 GFP_KERNEL, cpu_to_node(cur_cpu));
837
758 uv_bau_retry_limit = 1; 838 uv_bau_retry_limit = 1;
759 uv_nshift = uv_hub_info->n_val; 839 uv_nshift = uv_hub_info->n_val;
760 uv_mmask = (1UL << uv_hub_info->n_val) - 1; 840 uv_mmask = (1UL << uv_hub_info->n_val) - 1;
761 nblades = 0; 841 nblades = uv_num_possible_blades();
762 last_blade = -1; 842
763 cur_cpu = 0;
764 for_each_online_node(node) {
765 blade = uv_node_to_blade_id(node);
766 if (blade == last_blade)
767 continue;
768 last_blade = blade;
769 nblades++;
770 }
771 uv_bau_table_bases = (struct bau_control **) 843 uv_bau_table_bases = (struct bau_control **)
772 kmalloc(nblades * sizeof(struct bau_control *), GFP_KERNEL); 844 kmalloc(nblades * sizeof(struct bau_control *), GFP_KERNEL);
773 BUG_ON(!uv_bau_table_bases); 845 BUG_ON(!uv_bau_table_bases);
774 846
775 last_blade = -1; 847 uv_partition_base_pnode = 0x7fffffff;
776 for_each_online_node(node) { 848 for (blade = 0; blade < nblades; blade++)
777 blade = uv_node_to_blade_id(node); 849 if (uv_blade_nr_possible_cpus(blade) &&
778 if (blade == last_blade) 850 (uv_blade_to_pnode(blade) < uv_partition_base_pnode))
779 continue; 851 uv_partition_base_pnode = uv_blade_to_pnode(blade);
780 last_blade = blade; 852 for (blade = 0; blade < nblades; blade++)
781 uv_init_blade(blade, node, cur_cpu); 853 if (uv_blade_nr_possible_cpus(blade))
782 cur_cpu += uv_blade_nr_possible_cpus(blade); 854 uv_init_blade(blade);
783 } 855
784 alloc_intr_gate(UV_BAU_MESSAGE, uv_bau_message_intr1); 856 alloc_intr_gate(UV_BAU_MESSAGE, uv_bau_message_intr1);
785 uv_enable_timeouts(); 857 uv_enable_timeouts();
786 858
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 462b9ba67e92..7a567ebe6361 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -17,20 +17,21 @@
17#include <asm/delay.h> 17#include <asm/delay.h>
18#include <asm/hypervisor.h> 18#include <asm/hypervisor.h>
19 19
20unsigned int cpu_khz; /* TSC clocks / usec, not used here */ 20unsigned int __read_mostly cpu_khz; /* TSC clocks / usec, not used here */
21EXPORT_SYMBOL(cpu_khz); 21EXPORT_SYMBOL(cpu_khz);
22unsigned int tsc_khz; 22
23unsigned int __read_mostly tsc_khz;
23EXPORT_SYMBOL(tsc_khz); 24EXPORT_SYMBOL(tsc_khz);
24 25
25/* 26/*
26 * TSC can be unstable due to cpufreq or due to unsynced TSCs 27 * TSC can be unstable due to cpufreq or due to unsynced TSCs
27 */ 28 */
28static int tsc_unstable; 29static int __read_mostly tsc_unstable;
29 30
30/* native_sched_clock() is called before tsc_init(), so 31/* native_sched_clock() is called before tsc_init(), so
31 we must start with the TSC soft disabled to prevent 32 we must start with the TSC soft disabled to prevent
32 erroneous rdtsc usage on !cpu_has_tsc processors */ 33 erroneous rdtsc usage on !cpu_has_tsc processors */
33static int tsc_disabled = -1; 34static int __read_mostly tsc_disabled = -1;
34 35
35static int tsc_clocksource_reliable; 36static int tsc_clocksource_reliable;
36/* 37/*
@@ -543,8 +544,6 @@ unsigned long native_calibrate_tsc(void)
543 return tsc_pit_min; 544 return tsc_pit_min;
544} 545}
545 546
546#ifdef CONFIG_X86_32
547/* Only called from the Powernow K7 cpu freq driver */
548int recalibrate_cpu_khz(void) 547int recalibrate_cpu_khz(void)
549{ 548{
550#ifndef CONFIG_SMP 549#ifndef CONFIG_SMP
@@ -566,7 +565,6 @@ int recalibrate_cpu_khz(void)
566 565
567EXPORT_SYMBOL(recalibrate_cpu_khz); 566EXPORT_SYMBOL(recalibrate_cpu_khz);
568 567
569#endif /* CONFIG_X86_32 */
570 568
571/* Accelerators for sched_clock() 569/* Accelerators for sched_clock()
572 * convert from cycles(64bits) => nanoseconds (64bits) 570 * convert from cycles(64bits) => nanoseconds (64bits)
diff --git a/arch/x86/kernel/uv_sysfs.c b/arch/x86/kernel/uv_sysfs.c
index 67f9b9dbf800..36afb98675a4 100644
--- a/arch/x86/kernel/uv_sysfs.c
+++ b/arch/x86/kernel/uv_sysfs.c
@@ -21,6 +21,7 @@
21 21
22#include <linux/sysdev.h> 22#include <linux/sysdev.h>
23#include <asm/uv/bios.h> 23#include <asm/uv/bios.h>
24#include <asm/uv/uv.h>
24 25
25struct kobject *sgi_uv_kobj; 26struct kobject *sgi_uv_kobj;
26 27
@@ -47,6 +48,9 @@ static int __init sgi_uv_sysfs_init(void)
47{ 48{
48 unsigned long ret; 49 unsigned long ret;
49 50
51 if (!is_uv_system())
52 return -ENODEV;
53
50 if (!sgi_uv_kobj) 54 if (!sgi_uv_kobj)
51 sgi_uv_kobj = kobject_create_and_add("sgi_uv", firmware_kobj); 55 sgi_uv_kobj = kobject_create_and_add("sgi_uv", firmware_kobj);
52 if (!sgi_uv_kobj) { 56 if (!sgi_uv_kobj) {
diff --git a/arch/x86/kernel/vmiclock_32.c b/arch/x86/kernel/vmiclock_32.c
index 33a788d5879c..d303369a7bad 100644
--- a/arch/x86/kernel/vmiclock_32.c
+++ b/arch/x86/kernel/vmiclock_32.c
@@ -202,7 +202,6 @@ static struct irqaction vmi_clock_action = {
202 .name = "vmi-timer", 202 .name = "vmi-timer",
203 .handler = vmi_timer_interrupt, 203 .handler = vmi_timer_interrupt,
204 .flags = IRQF_DISABLED | IRQF_NOBALANCING | IRQF_TIMER, 204 .flags = IRQF_DISABLED | IRQF_NOBALANCING | IRQF_TIMER,
205 .mask = CPU_MASK_ALL,
206}; 205};
207 206
208static void __devinit vmi_time_init_clockevent(void) 207static void __devinit vmi_time_init_clockevent(void)
diff --git a/arch/x86/kernel/vsmp_64.c b/arch/x86/kernel/vsmp_64.c
index 74de562812cc..a1d804bcd483 100644
--- a/arch/x86/kernel/vsmp_64.c
+++ b/arch/x86/kernel/vsmp_64.c
@@ -22,7 +22,7 @@
22#include <asm/paravirt.h> 22#include <asm/paravirt.h>
23#include <asm/setup.h> 23#include <asm/setup.h>
24 24
25#ifdef CONFIG_PARAVIRT 25#if defined CONFIG_PCI && defined CONFIG_PARAVIRT
26/* 26/*
27 * Interrupt control on vSMPowered systems: 27 * Interrupt control on vSMPowered systems:
28 * ~AC is a shadow of IF. If IF is 'on' AC should be 'off' 28 * ~AC is a shadow of IF. If IF is 'on' AC should be 'off'
@@ -114,6 +114,7 @@ static void __init set_vsmp_pv_ops(void)
114} 114}
115#endif 115#endif
116 116
117#ifdef CONFIG_PCI
117static int is_vsmp = -1; 118static int is_vsmp = -1;
118 119
119static void __init detect_vsmp_box(void) 120static void __init detect_vsmp_box(void)
@@ -139,6 +140,15 @@ int is_vsmp_box(void)
139 } 140 }
140} 141}
141 142
143#else
144static void __init detect_vsmp_box(void)
145{
146}
147int is_vsmp_box(void)
148{
149 return 0;
150}
151#endif
142void __init vsmp_init(void) 152void __init vsmp_init(void)
143{ 153{
144 detect_vsmp_box(); 154 detect_vsmp_box();
diff --git a/arch/x86/kernel/xsave.c b/arch/x86/kernel/xsave.c
index 2b54fe002e94..0a5b04aa98f1 100644
--- a/arch/x86/kernel/xsave.c
+++ b/arch/x86/kernel/xsave.c
@@ -324,7 +324,7 @@ void __ref xsave_cntxt_init(void)
324 } 324 }
325 325
326 /* 326 /*
327 * for now OS knows only about FP/SSE 327 * Support only the state known to OS.
328 */ 328 */
329 pcntxt_mask = pcntxt_mask & XCNTXT_MASK; 329 pcntxt_mask = pcntxt_mask & XCNTXT_MASK;
330 xsave_init(); 330 xsave_init();
generated by cgit v1.2.2 (git 2.25.0) at 2025-08-12 02:36:27 -0400