x86_64: Add vDSO for x86-64 with gettimeofday/clock_gettime/getcpu
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / arch / x86_64 / kernel / setup.c
CommitLineData
1da177e4
LT
1/*
2 * linux/arch/x86-64/kernel/setup.c
3 *
4 * Copyright (C) 1995 Linus Torvalds
5 *
6 * Nov 2001 Dave Jones <davej@suse.de>
7 * Forked from i386 setup code.
1da177e4
LT
8 */
9
10/*
11 * This file handles the architecture-dependent parts of initialization
12 */
13
14#include <linux/errno.h>
15#include <linux/sched.h>
16#include <linux/kernel.h>
17#include <linux/mm.h>
18#include <linux/stddef.h>
19#include <linux/unistd.h>
20#include <linux/ptrace.h>
21#include <linux/slab.h>
22#include <linux/user.h>
23#include <linux/a.out.h>
894673ee 24#include <linux/screen_info.h>
1da177e4
LT
25#include <linux/ioport.h>
26#include <linux/delay.h>
1da177e4
LT
27#include <linux/init.h>
28#include <linux/initrd.h>
29#include <linux/highmem.h>
30#include <linux/bootmem.h>
31#include <linux/module.h>
32#include <asm/processor.h>
33#include <linux/console.h>
34#include <linux/seq_file.h>
aac04b32 35#include <linux/crash_dump.h>
1da177e4
LT
36#include <linux/root_dev.h>
37#include <linux/pci.h>
38#include <linux/acpi.h>
39#include <linux/kallsyms.h>
40#include <linux/edd.h>
bbfceef4 41#include <linux/mmzone.h>
5f5609df 42#include <linux/kexec.h>
95235ca2 43#include <linux/cpufreq.h>
e9928674 44#include <linux/dmi.h>
17a941d8 45#include <linux/dma-mapping.h>
681558fd 46#include <linux/ctype.h>
bbfceef4 47
1da177e4
LT
48#include <asm/mtrr.h>
49#include <asm/uaccess.h>
50#include <asm/system.h>
51#include <asm/io.h>
52#include <asm/smp.h>
53#include <asm/msr.h>
54#include <asm/desc.h>
55#include <video/edid.h>
56#include <asm/e820.h>
57#include <asm/dma.h>
58#include <asm/mpspec.h>
59#include <asm/mmu_context.h>
60#include <asm/bootsetup.h>
61#include <asm/proto.h>
62#include <asm/setup.h>
63#include <asm/mach_apic.h>
64#include <asm/numa.h>
2bc0414e 65#include <asm/sections.h>
f2d3efed 66#include <asm/dmi.h>
1da177e4
LT
67
68/*
69 * Machine setup..
70 */
71
6c231b7b 72struct cpuinfo_x86 boot_cpu_data __read_mostly;
2ee60e17 73EXPORT_SYMBOL(boot_cpu_data);
1da177e4
LT
74
75unsigned long mmu_cr4_features;
76
1da177e4
LT
77/* Boot loader ID as an integer, for the benefit of proc_dointvec */
78int bootloader_type;
79
80unsigned long saved_video_mode;
81
f039b754
AK
82int force_mwait __cpuinitdata;
83
f2d3efed
AK
84/*
85 * Early DMI memory
86 */
87int dmi_alloc_index;
88char dmi_alloc_data[DMI_MAX_DATA];
89
1da177e4
LT
90/*
91 * Setup options
92 */
1da177e4 93struct screen_info screen_info;
2ee60e17 94EXPORT_SYMBOL(screen_info);
1da177e4
LT
95struct sys_desc_table_struct {
96 unsigned short length;
97 unsigned char table[0];
98};
99
100struct edid_info edid_info;
ba70710e 101EXPORT_SYMBOL_GPL(edid_info);
1da177e4
LT
102
103extern int root_mountflags;
1da177e4 104
adf48856 105char __initdata command_line[COMMAND_LINE_SIZE];
1da177e4
LT
106
107struct resource standard_io_resources[] = {
108 { .name = "dma1", .start = 0x00, .end = 0x1f,
109 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
110 { .name = "pic1", .start = 0x20, .end = 0x21,
111 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
112 { .name = "timer0", .start = 0x40, .end = 0x43,
113 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
114 { .name = "timer1", .start = 0x50, .end = 0x53,
115 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
116 { .name = "keyboard", .start = 0x60, .end = 0x6f,
117 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
118 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
119 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
120 { .name = "pic2", .start = 0xa0, .end = 0xa1,
121 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
122 { .name = "dma2", .start = 0xc0, .end = 0xdf,
123 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
124 { .name = "fpu", .start = 0xf0, .end = 0xff,
125 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
126};
127
1da177e4
LT
128#define IORESOURCE_RAM (IORESOURCE_BUSY | IORESOURCE_MEM)
129
130struct resource data_resource = {
131 .name = "Kernel data",
132 .start = 0,
133 .end = 0,
134 .flags = IORESOURCE_RAM,
135};
136struct resource code_resource = {
137 .name = "Kernel code",
138 .start = 0,
139 .end = 0,
140 .flags = IORESOURCE_RAM,
141};
142
2c8c0e6b
AK
143#ifdef CONFIG_PROC_VMCORE
144/* elfcorehdr= specifies the location of elf core header
145 * stored by the crashed kernel. This option will be passed
146 * by kexec loader to the capture kernel.
147 */
148static int __init setup_elfcorehdr(char *arg)
681558fd 149{
2c8c0e6b
AK
150 char *end;
151 if (!arg)
152 return -EINVAL;
153 elfcorehdr_addr = memparse(arg, &end);
154 return end > arg ? 0 : -EINVAL;
681558fd 155}
2c8c0e6b 156early_param("elfcorehdr", setup_elfcorehdr);
e2c03888
AK
157#endif
158
2b97690f 159#ifndef CONFIG_NUMA
bbfceef4
MT
160static void __init
161contig_initmem_init(unsigned long start_pfn, unsigned long end_pfn)
1da177e4 162{
bbfceef4
MT
163 unsigned long bootmap_size, bootmap;
164
bbfceef4
MT
165 bootmap_size = bootmem_bootmap_pages(end_pfn)<<PAGE_SHIFT;
166 bootmap = find_e820_area(0, end_pfn<<PAGE_SHIFT, bootmap_size);
167 if (bootmap == -1L)
168 panic("Cannot find bootmem map of size %ld\n",bootmap_size);
169 bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, end_pfn);
5cb248ab
MG
170 e820_register_active_regions(0, start_pfn, end_pfn);
171 free_bootmem_with_active_regions(0, end_pfn);
bbfceef4 172 reserve_bootmem(bootmap, bootmap_size);
1da177e4
LT
173}
174#endif
175
1da177e4
LT
176#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
177struct edd edd;
178#ifdef CONFIG_EDD_MODULE
179EXPORT_SYMBOL(edd);
180#endif
181/**
182 * copy_edd() - Copy the BIOS EDD information
183 * from boot_params into a safe place.
184 *
185 */
186static inline void copy_edd(void)
187{
188 memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature));
189 memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info));
190 edd.mbr_signature_nr = EDD_MBR_SIG_NR;
191 edd.edd_info_nr = EDD_NR;
192}
193#else
194static inline void copy_edd(void)
195{
196}
197#endif
198
199#define EBDA_ADDR_POINTER 0x40E
ac71d12c
AK
200
201unsigned __initdata ebda_addr;
202unsigned __initdata ebda_size;
203
204static void discover_ebda(void)
1da177e4 205{
ac71d12c 206 /*
1da177e4
LT
207 * there is a real-mode segmented pointer pointing to the
208 * 4K EBDA area at 0x40E
209 */
bdb96a66 210 ebda_addr = *(unsigned short *)__va(EBDA_ADDR_POINTER);
ac71d12c
AK
211 ebda_addr <<= 4;
212
bdb96a66 213 ebda_size = *(unsigned short *)__va(ebda_addr);
ac71d12c
AK
214
215 /* Round EBDA up to pages */
216 if (ebda_size == 0)
217 ebda_size = 1;
218 ebda_size <<= 10;
219 ebda_size = round_up(ebda_size + (ebda_addr & ~PAGE_MASK), PAGE_SIZE);
220 if (ebda_size > 64*1024)
221 ebda_size = 64*1024;
1da177e4
LT
222}
223
224void __init setup_arch(char **cmdline_p)
225{
adf48856 226 printk(KERN_INFO "Command line: %s\n", boot_command_line);
43c85c9c 227
1da177e4 228 ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
1da177e4
LT
229 screen_info = SCREEN_INFO;
230 edid_info = EDID_INFO;
231 saved_video_mode = SAVED_VIDEO_MODE;
232 bootloader_type = LOADER_TYPE;
233
234#ifdef CONFIG_BLK_DEV_RAM
235 rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK;
236 rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0);
237 rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0);
238#endif
239 setup_memory_region();
240 copy_edd();
241
242 if (!MOUNT_ROOT_RDONLY)
243 root_mountflags &= ~MS_RDONLY;
244 init_mm.start_code = (unsigned long) &_text;
245 init_mm.end_code = (unsigned long) &_etext;
246 init_mm.end_data = (unsigned long) &_edata;
247 init_mm.brk = (unsigned long) &_end;
248
e3ebadd9
LT
249 code_resource.start = virt_to_phys(&_text);
250 code_resource.end = virt_to_phys(&_etext)-1;
251 data_resource.start = virt_to_phys(&_etext);
252 data_resource.end = virt_to_phys(&_edata)-1;
1da177e4 253
1da177e4
LT
254 early_identify_cpu(&boot_cpu_data);
255
adf48856 256 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
2c8c0e6b
AK
257 *cmdline_p = command_line;
258
259 parse_early_param();
260
261 finish_e820_parsing();
9ca33eb6 262
5cb248ab 263 e820_register_active_regions(0, 0, -1UL);
1da177e4
LT
264 /*
265 * partially used pages are not usable - thus
266 * we are rounding upwards:
267 */
268 end_pfn = e820_end_of_ram();
caff0710 269 num_physpages = end_pfn;
1da177e4
LT
270
271 check_efer();
272
ac71d12c
AK
273 discover_ebda();
274
1da177e4
LT
275 init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT));
276
f2d3efed
AK
277 dmi_scan_machine();
278
888ba6c6 279#ifdef CONFIG_ACPI
1da177e4
LT
280 /*
281 * Initialize the ACPI boot-time table parser (gets the RSDP and SDT).
282 * Call this early for SRAT node setup.
283 */
284 acpi_boot_table_init();
285#endif
286
caff0710
JB
287 /* How many end-of-memory variables you have, grandma! */
288 max_low_pfn = end_pfn;
289 max_pfn = end_pfn;
290 high_memory = (void *)__va(end_pfn * PAGE_SIZE - 1) + 1;
291
5cb248ab
MG
292 /* Remove active ranges so rediscovery with NUMA-awareness happens */
293 remove_all_active_ranges();
294
1da177e4
LT
295#ifdef CONFIG_ACPI_NUMA
296 /*
297 * Parse SRAT to discover nodes.
298 */
299 acpi_numa_init();
300#endif
301
2b97690f 302#ifdef CONFIG_NUMA
1da177e4
LT
303 numa_initmem_init(0, end_pfn);
304#else
bbfceef4 305 contig_initmem_init(0, end_pfn);
1da177e4
LT
306#endif
307
308 /* Reserve direct mapping */
309 reserve_bootmem_generic(table_start << PAGE_SHIFT,
310 (table_end - table_start) << PAGE_SHIFT);
311
312 /* reserve kernel */
ceee8822
AK
313 reserve_bootmem_generic(__pa_symbol(&_text),
314 __pa_symbol(&_end) - __pa_symbol(&_text));
1da177e4
LT
315
316 /*
317 * reserve physical page 0 - it's a special BIOS page on many boxes,
318 * enabling clean reboots, SMP operation, laptop functions.
319 */
320 reserve_bootmem_generic(0, PAGE_SIZE);
321
322 /* reserve ebda region */
ac71d12c
AK
323 if (ebda_addr)
324 reserve_bootmem_generic(ebda_addr, ebda_size);
076422d2
AS
325#ifdef CONFIG_NUMA
326 /* reserve nodemap region */
327 if (nodemap_addr)
328 reserve_bootmem_generic(nodemap_addr, nodemap_size);
329#endif
1da177e4
LT
330
331#ifdef CONFIG_SMP
1da177e4 332 /* Reserve SMP trampoline */
90b1c208 333 reserve_bootmem_generic(SMP_TRAMPOLINE_BASE, 2*PAGE_SIZE);
1da177e4
LT
334#endif
335
336#ifdef CONFIG_ACPI_SLEEP
337 /*
338 * Reserve low memory region for sleep support.
339 */
340 acpi_reserve_bootmem();
341#endif
1da177e4
LT
342 /*
343 * Find and reserve possible boot-time SMP configuration:
344 */
345 find_smp_config();
1da177e4
LT
346#ifdef CONFIG_BLK_DEV_INITRD
347 if (LOADER_TYPE && INITRD_START) {
348 if (INITRD_START + INITRD_SIZE <= (end_pfn << PAGE_SHIFT)) {
349 reserve_bootmem_generic(INITRD_START, INITRD_SIZE);
19e5d9c0 350 initrd_start = INITRD_START + PAGE_OFFSET;
1da177e4
LT
351 initrd_end = initrd_start+INITRD_SIZE;
352 }
353 else {
354 printk(KERN_ERR "initrd extends beyond end of memory "
355 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
356 (unsigned long)(INITRD_START + INITRD_SIZE),
357 (unsigned long)(end_pfn << PAGE_SHIFT));
358 initrd_start = 0;
359 }
360 }
361#endif
5f5609df
EB
362#ifdef CONFIG_KEXEC
363 if (crashk_res.start != crashk_res.end) {
00212fef 364 reserve_bootmem_generic(crashk_res.start,
5f5609df
EB
365 crashk_res.end - crashk_res.start + 1);
366 }
367#endif
0d317fb7 368
1da177e4
LT
369 paging_init();
370
f157cbb1 371#ifdef CONFIG_PCI
dfa4698c 372 early_quirks();
f157cbb1 373#endif
1da177e4 374
51f62e18
AR
375 /*
376 * set this early, so we dont allocate cpu0
377 * if MADT list doesnt list BSP first
378 * mpparse.c/MP_processor_info() allocates logical cpu numbers.
379 */
380 cpu_set(0, cpu_present_map);
888ba6c6 381#ifdef CONFIG_ACPI
1da177e4
LT
382 /*
383 * Read APIC and some other early information from ACPI tables.
384 */
385 acpi_boot_init();
386#endif
387
05b3cbd8
RT
388 init_cpu_to_node();
389
1da177e4
LT
390 /*
391 * get boot-time SMP configuration:
392 */
393 if (smp_found_config)
394 get_smp_config();
395 init_apic_mappings();
1da177e4
LT
396
397 /*
fc986db4
AK
398 * We trust e820 completely. No explicit ROM probing in memory.
399 */
1da177e4 400 e820_reserve_resources();
e8eff5ac 401 e820_mark_nosave_regions();
1da177e4 402
1da177e4
LT
403 {
404 unsigned i;
405 /* request I/O space for devices used on all i[345]86 PCs */
9d0ef4fd 406 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
1da177e4
LT
407 request_resource(&ioport_resource, &standard_io_resources[i]);
408 }
409
a1e97782 410 e820_setup_gap();
1da177e4 411
1da177e4
LT
412#ifdef CONFIG_VT
413#if defined(CONFIG_VGA_CONSOLE)
414 conswitchp = &vga_con;
415#elif defined(CONFIG_DUMMY_CONSOLE)
416 conswitchp = &dummy_con;
417#endif
418#endif
419}
420
e6982c67 421static int __cpuinit get_model_name(struct cpuinfo_x86 *c)
1da177e4
LT
422{
423 unsigned int *v;
424
ebfcaa96 425 if (c->extended_cpuid_level < 0x80000004)
1da177e4
LT
426 return 0;
427
428 v = (unsigned int *) c->x86_model_id;
429 cpuid(0x80000002, &v[0], &v[1], &v[2], &v[3]);
430 cpuid(0x80000003, &v[4], &v[5], &v[6], &v[7]);
431 cpuid(0x80000004, &v[8], &v[9], &v[10], &v[11]);
432 c->x86_model_id[48] = 0;
433 return 1;
434}
435
436
e6982c67 437static void __cpuinit display_cacheinfo(struct cpuinfo_x86 *c)
1da177e4
LT
438{
439 unsigned int n, dummy, eax, ebx, ecx, edx;
440
ebfcaa96 441 n = c->extended_cpuid_level;
1da177e4
LT
442
443 if (n >= 0x80000005) {
444 cpuid(0x80000005, &dummy, &ebx, &ecx, &edx);
445 printk(KERN_INFO "CPU: L1 I Cache: %dK (%d bytes/line), D cache %dK (%d bytes/line)\n",
446 edx>>24, edx&0xFF, ecx>>24, ecx&0xFF);
447 c->x86_cache_size=(ecx>>24)+(edx>>24);
448 /* On K8 L1 TLB is inclusive, so don't count it */
449 c->x86_tlbsize = 0;
450 }
451
452 if (n >= 0x80000006) {
453 cpuid(0x80000006, &dummy, &ebx, &ecx, &edx);
454 ecx = cpuid_ecx(0x80000006);
455 c->x86_cache_size = ecx >> 16;
456 c->x86_tlbsize += ((ebx >> 16) & 0xfff) + (ebx & 0xfff);
457
458 printk(KERN_INFO "CPU: L2 Cache: %dK (%d bytes/line)\n",
459 c->x86_cache_size, ecx & 0xFF);
460 }
461
462 if (n >= 0x80000007)
463 cpuid(0x80000007, &dummy, &dummy, &dummy, &c->x86_power);
464 if (n >= 0x80000008) {
465 cpuid(0x80000008, &eax, &dummy, &dummy, &dummy);
466 c->x86_virt_bits = (eax >> 8) & 0xff;
467 c->x86_phys_bits = eax & 0xff;
468 }
469}
470
3f098c26
AK
471#ifdef CONFIG_NUMA
472static int nearby_node(int apicid)
473{
474 int i;
475 for (i = apicid - 1; i >= 0; i--) {
476 int node = apicid_to_node[i];
477 if (node != NUMA_NO_NODE && node_online(node))
478 return node;
479 }
480 for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) {
481 int node = apicid_to_node[i];
482 if (node != NUMA_NO_NODE && node_online(node))
483 return node;
484 }
485 return first_node(node_online_map); /* Shouldn't happen */
486}
487#endif
488
63518644
AK
489/*
490 * On a AMD dual core setup the lower bits of the APIC id distingush the cores.
491 * Assumes number of cores is a power of two.
492 */
493static void __init amd_detect_cmp(struct cpuinfo_x86 *c)
494{
495#ifdef CONFIG_SMP
b41e2939 496 unsigned bits;
3f098c26 497#ifdef CONFIG_NUMA
f3fa8ebc 498 int cpu = smp_processor_id();
3f098c26 499 int node = 0;
60c1bc82 500 unsigned apicid = hard_smp_processor_id();
3f098c26 501#endif
faee9a5d 502 unsigned ecx = cpuid_ecx(0x80000008);
b41e2939 503
faee9a5d 504 c->x86_max_cores = (ecx & 0xff) + 1;
b41e2939 505
faee9a5d
AK
506 /* CPU telling us the core id bits shift? */
507 bits = (ecx >> 12) & 0xF;
508
509 /* Otherwise recompute */
510 if (bits == 0) {
511 while ((1 << bits) < c->x86_max_cores)
512 bits++;
513 }
b41e2939
AK
514
515 /* Low order bits define the core id (index of core in socket) */
f3fa8ebc 516 c->cpu_core_id = c->phys_proc_id & ((1 << bits)-1);
b41e2939 517 /* Convert the APIC ID into the socket ID */
f3fa8ebc 518 c->phys_proc_id = phys_pkg_id(bits);
63518644
AK
519
520#ifdef CONFIG_NUMA
f3fa8ebc 521 node = c->phys_proc_id;
3f098c26
AK
522 if (apicid_to_node[apicid] != NUMA_NO_NODE)
523 node = apicid_to_node[apicid];
524 if (!node_online(node)) {
525 /* Two possibilities here:
526 - The CPU is missing memory and no node was created.
527 In that case try picking one from a nearby CPU
528 - The APIC IDs differ from the HyperTransport node IDs
529 which the K8 northbridge parsing fills in.
530 Assume they are all increased by a constant offset,
531 but in the same order as the HT nodeids.
532 If that doesn't result in a usable node fall back to the
533 path for the previous case. */
f3fa8ebc 534 int ht_nodeid = apicid - (cpu_data[0].phys_proc_id << bits);
3f098c26
AK
535 if (ht_nodeid >= 0 &&
536 apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
537 node = apicid_to_node[ht_nodeid];
538 /* Pick a nearby node */
539 if (!node_online(node))
540 node = nearby_node(apicid);
541 }
69d81fcd 542 numa_set_node(cpu, node);
3f098c26 543
e42f9437 544 printk(KERN_INFO "CPU %d/%x -> Node %d\n", cpu, apicid, node);
63518644 545#endif
63518644
AK
546#endif
547}
1da177e4 548
ed77504b 549static void __cpuinit init_amd(struct cpuinfo_x86 *c)
1da177e4 550{
7bcd3f34 551 unsigned level;
1da177e4 552
bc5e8fdf
LT
553#ifdef CONFIG_SMP
554 unsigned long value;
555
7d318d77
AK
556 /*
557 * Disable TLB flush filter by setting HWCR.FFDIS on K8
558 * bit 6 of msr C001_0015
559 *
560 * Errata 63 for SH-B3 steppings
561 * Errata 122 for all steppings (F+ have it disabled by default)
562 */
563 if (c->x86 == 15) {
564 rdmsrl(MSR_K8_HWCR, value);
565 value |= 1 << 6;
566 wrmsrl(MSR_K8_HWCR, value);
567 }
bc5e8fdf
LT
568#endif
569
1da177e4
LT
570 /* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
571 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */
572 clear_bit(0*32+31, &c->x86_capability);
573
7bcd3f34
AK
574 /* On C+ stepping K8 rep microcode works well for copy/memset */
575 level = cpuid_eax(1);
576 if (c->x86 == 15 && ((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58))
577 set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability);
5b74e3ab
AK
578 if (c->x86 == 0x10)
579 set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability);
7bcd3f34 580
18bd057b
AK
581 /* Enable workaround for FXSAVE leak */
582 if (c->x86 >= 6)
583 set_bit(X86_FEATURE_FXSAVE_LEAK, &c->x86_capability);
584
e42f9437
RS
585 level = get_model_name(c);
586 if (!level) {
1da177e4
LT
587 switch (c->x86) {
588 case 15:
589 /* Should distinguish Models here, but this is only
590 a fallback anyways. */
591 strcpy(c->x86_model_id, "Hammer");
592 break;
593 }
594 }
595 display_cacheinfo(c);
596
130951cc
AK
597 /* c->x86_power is 8000_0007 edx. Bit 8 is constant TSC */
598 if (c->x86_power & (1<<8))
599 set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
600
faee9a5d
AK
601 /* Multi core CPU? */
602 if (c->extended_cpuid_level >= 0x80000008)
63518644 603 amd_detect_cmp(c);
1da177e4 604
240cd6a8
AK
605 /* Fix cpuid4 emulation for more */
606 num_cache_leaves = 3;
2049336f 607
61677965
AK
608 /* RDTSC can be speculated around */
609 clear_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
f039b754
AK
610
611 /* Family 10 doesn't support C states in MWAIT so don't use it */
612 if (c->x86 == 0x10 && !force_mwait)
613 clear_bit(X86_FEATURE_MWAIT, &c->x86_capability);
1da177e4
LT
614}
615
e6982c67 616static void __cpuinit detect_ht(struct cpuinfo_x86 *c)
1da177e4
LT
617{
618#ifdef CONFIG_SMP
619 u32 eax, ebx, ecx, edx;
94605eff 620 int index_msb, core_bits;
94605eff
SS
621
622 cpuid(1, &eax, &ebx, &ecx, &edx);
623
94605eff 624
e42f9437 625 if (!cpu_has(c, X86_FEATURE_HT))
1da177e4 626 return;
e42f9437
RS
627 if (cpu_has(c, X86_FEATURE_CMP_LEGACY))
628 goto out;
1da177e4 629
1da177e4 630 smp_num_siblings = (ebx & 0xff0000) >> 16;
94605eff 631
1da177e4
LT
632 if (smp_num_siblings == 1) {
633 printk(KERN_INFO "CPU: Hyper-Threading is disabled\n");
94605eff
SS
634 } else if (smp_num_siblings > 1 ) {
635
1da177e4
LT
636 if (smp_num_siblings > NR_CPUS) {
637 printk(KERN_WARNING "CPU: Unsupported number of the siblings %d", smp_num_siblings);
638 smp_num_siblings = 1;
639 return;
640 }
94605eff
SS
641
642 index_msb = get_count_order(smp_num_siblings);
f3fa8ebc 643 c->phys_proc_id = phys_pkg_id(index_msb);
3dd9d514 644
94605eff 645 smp_num_siblings = smp_num_siblings / c->x86_max_cores;
3dd9d514 646
94605eff
SS
647 index_msb = get_count_order(smp_num_siblings) ;
648
649 core_bits = get_count_order(c->x86_max_cores);
3dd9d514 650
f3fa8ebc 651 c->cpu_core_id = phys_pkg_id(index_msb) &
94605eff 652 ((1 << core_bits) - 1);
1da177e4 653 }
e42f9437
RS
654out:
655 if ((c->x86_max_cores * smp_num_siblings) > 1) {
656 printk(KERN_INFO "CPU: Physical Processor ID: %d\n", c->phys_proc_id);
657 printk(KERN_INFO "CPU: Processor Core ID: %d\n", c->cpu_core_id);
658 }
659
1da177e4
LT
660#endif
661}
662
3dd9d514
AK
663/*
664 * find out the number of processor cores on the die
665 */
e6982c67 666static int __cpuinit intel_num_cpu_cores(struct cpuinfo_x86 *c)
3dd9d514 667{
2bbc419f 668 unsigned int eax, t;
3dd9d514
AK
669
670 if (c->cpuid_level < 4)
671 return 1;
672
2bbc419f 673 cpuid_count(4, 0, &eax, &t, &t, &t);
3dd9d514
AK
674
675 if (eax & 0x1f)
676 return ((eax >> 26) + 1);
677 else
678 return 1;
679}
680
df0cc26b
AK
681static void srat_detect_node(void)
682{
683#ifdef CONFIG_NUMA
ddea7be0 684 unsigned node;
df0cc26b 685 int cpu = smp_processor_id();
e42f9437 686 int apicid = hard_smp_processor_id();
df0cc26b
AK
687
688 /* Don't do the funky fallback heuristics the AMD version employs
689 for now. */
e42f9437 690 node = apicid_to_node[apicid];
df0cc26b 691 if (node == NUMA_NO_NODE)
0d015324 692 node = first_node(node_online_map);
69d81fcd 693 numa_set_node(cpu, node);
df0cc26b 694
c31fbb1a 695 printk(KERN_INFO "CPU %d/%x -> Node %d\n", cpu, apicid, node);
df0cc26b
AK
696#endif
697}
698
e6982c67 699static void __cpuinit init_intel(struct cpuinfo_x86 *c)
1da177e4
LT
700{
701 /* Cache sizes */
702 unsigned n;
703
704 init_intel_cacheinfo(c);
0080e667
VP
705 if (c->cpuid_level > 9 ) {
706 unsigned eax = cpuid_eax(10);
707 /* Check for version and the number of counters */
708 if ((eax & 0xff) && (((eax>>8) & 0xff) > 1))
709 set_bit(X86_FEATURE_ARCH_PERFMON, &c->x86_capability);
710 }
711
36b2a8d5
SE
712 if (cpu_has_ds) {
713 unsigned int l1, l2;
714 rdmsr(MSR_IA32_MISC_ENABLE, l1, l2);
ee58fad5
SE
715 if (!(l1 & (1<<11)))
716 set_bit(X86_FEATURE_BTS, c->x86_capability);
36b2a8d5
SE
717 if (!(l1 & (1<<12)))
718 set_bit(X86_FEATURE_PEBS, c->x86_capability);
719 }
720
ebfcaa96 721 n = c->extended_cpuid_level;
1da177e4
LT
722 if (n >= 0x80000008) {
723 unsigned eax = cpuid_eax(0x80000008);
724 c->x86_virt_bits = (eax >> 8) & 0xff;
725 c->x86_phys_bits = eax & 0xff;
af9c142d
SL
726 /* CPUID workaround for Intel 0F34 CPU */
727 if (c->x86_vendor == X86_VENDOR_INTEL &&
728 c->x86 == 0xF && c->x86_model == 0x3 &&
729 c->x86_mask == 0x4)
730 c->x86_phys_bits = 36;
1da177e4
LT
731 }
732
733 if (c->x86 == 15)
734 c->x86_cache_alignment = c->x86_clflush_size * 2;
39b3a791
AK
735 if ((c->x86 == 0xf && c->x86_model >= 0x03) ||
736 (c->x86 == 0x6 && c->x86_model >= 0x0e))
c29601e9 737 set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
27fbe5b2
AK
738 if (c->x86 == 6)
739 set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability);
f3d73707
AV
740 if (c->x86 == 15)
741 set_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
742 else
743 clear_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
94605eff 744 c->x86_max_cores = intel_num_cpu_cores(c);
df0cc26b
AK
745
746 srat_detect_node();
1da177e4
LT
747}
748
672289e9 749static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c)
1da177e4
LT
750{
751 char *v = c->x86_vendor_id;
752
753 if (!strcmp(v, "AuthenticAMD"))
754 c->x86_vendor = X86_VENDOR_AMD;
755 else if (!strcmp(v, "GenuineIntel"))
756 c->x86_vendor = X86_VENDOR_INTEL;
757 else
758 c->x86_vendor = X86_VENDOR_UNKNOWN;
759}
760
761struct cpu_model_info {
762 int vendor;
763 int family;
764 char *model_names[16];
765};
766
767/* Do some early cpuid on the boot CPU to get some parameter that are
768 needed before check_bugs. Everything advanced is in identify_cpu
769 below. */
e6982c67 770void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c)
1da177e4
LT
771{
772 u32 tfms;
773
774 c->loops_per_jiffy = loops_per_jiffy;
775 c->x86_cache_size = -1;
776 c->x86_vendor = X86_VENDOR_UNKNOWN;
777 c->x86_model = c->x86_mask = 0; /* So far unknown... */
778 c->x86_vendor_id[0] = '\0'; /* Unset */
779 c->x86_model_id[0] = '\0'; /* Unset */
780 c->x86_clflush_size = 64;
781 c->x86_cache_alignment = c->x86_clflush_size;
94605eff 782 c->x86_max_cores = 1;
ebfcaa96 783 c->extended_cpuid_level = 0;
1da177e4
LT
784 memset(&c->x86_capability, 0, sizeof c->x86_capability);
785
786 /* Get vendor name */
787 cpuid(0x00000000, (unsigned int *)&c->cpuid_level,
788 (unsigned int *)&c->x86_vendor_id[0],
789 (unsigned int *)&c->x86_vendor_id[8],
790 (unsigned int *)&c->x86_vendor_id[4]);
791
792 get_cpu_vendor(c);
793
794 /* Initialize the standard set of capabilities */
795 /* Note that the vendor-specific code below might override */
796
797 /* Intel-defined flags: level 0x00000001 */
798 if (c->cpuid_level >= 0x00000001) {
799 __u32 misc;
800 cpuid(0x00000001, &tfms, &misc, &c->x86_capability[4],
801 &c->x86_capability[0]);
802 c->x86 = (tfms >> 8) & 0xf;
803 c->x86_model = (tfms >> 4) & 0xf;
804 c->x86_mask = tfms & 0xf;
f5f786d0 805 if (c->x86 == 0xf)
1da177e4 806 c->x86 += (tfms >> 20) & 0xff;
f5f786d0 807 if (c->x86 >= 0x6)
1da177e4 808 c->x86_model += ((tfms >> 16) & 0xF) << 4;
1da177e4
LT
809 if (c->x86_capability[0] & (1<<19))
810 c->x86_clflush_size = ((misc >> 8) & 0xff) * 8;
1da177e4
LT
811 } else {
812 /* Have CPUID level 0 only - unheard of */
813 c->x86 = 4;
814 }
a158608b
AK
815
816#ifdef CONFIG_SMP
f3fa8ebc 817 c->phys_proc_id = (cpuid_ebx(1) >> 24) & 0xff;
a158608b 818#endif
1da177e4
LT
819}
820
821/*
822 * This does the hard work of actually picking apart the CPU stuff...
823 */
e6982c67 824void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
1da177e4
LT
825{
826 int i;
827 u32 xlvl;
828
829 early_identify_cpu(c);
830
831 /* AMD-defined flags: level 0x80000001 */
832 xlvl = cpuid_eax(0x80000000);
ebfcaa96 833 c->extended_cpuid_level = xlvl;
1da177e4
LT
834 if ((xlvl & 0xffff0000) == 0x80000000) {
835 if (xlvl >= 0x80000001) {
836 c->x86_capability[1] = cpuid_edx(0x80000001);
5b7abc6f 837 c->x86_capability[6] = cpuid_ecx(0x80000001);
1da177e4
LT
838 }
839 if (xlvl >= 0x80000004)
840 get_model_name(c); /* Default name */
841 }
842
843 /* Transmeta-defined flags: level 0x80860001 */
844 xlvl = cpuid_eax(0x80860000);
845 if ((xlvl & 0xffff0000) == 0x80860000) {
846 /* Don't set x86_cpuid_level here for now to not confuse. */
847 if (xlvl >= 0x80860001)
848 c->x86_capability[2] = cpuid_edx(0x80860001);
849 }
850
1d67953f
VP
851 init_scattered_cpuid_features(c);
852
1e9f28fa
SS
853 c->apicid = phys_pkg_id(0);
854
1da177e4
LT
855 /*
856 * Vendor-specific initialization. In this section we
857 * canonicalize the feature flags, meaning if there are
858 * features a certain CPU supports which CPUID doesn't
859 * tell us, CPUID claiming incorrect flags, or other bugs,
860 * we handle them here.
861 *
862 * At the end of this section, c->x86_capability better
863 * indicate the features this CPU genuinely supports!
864 */
865 switch (c->x86_vendor) {
866 case X86_VENDOR_AMD:
867 init_amd(c);
868 break;
869
870 case X86_VENDOR_INTEL:
871 init_intel(c);
872 break;
873
874 case X86_VENDOR_UNKNOWN:
875 default:
876 display_cacheinfo(c);
877 break;
878 }
879
880 select_idle_routine(c);
881 detect_ht(c);
1da177e4
LT
882
883 /*
884 * On SMP, boot_cpu_data holds the common feature set between
885 * all CPUs; so make sure that we indicate which features are
886 * common between the CPUs. The first time this routine gets
887 * executed, c == &boot_cpu_data.
888 */
889 if (c != &boot_cpu_data) {
890 /* AND the already accumulated flags with these */
891 for (i = 0 ; i < NCAPINTS ; i++)
892 boot_cpu_data.x86_capability[i] &= c->x86_capability[i];
893 }
894
895#ifdef CONFIG_X86_MCE
896 mcheck_init(c);
897#endif
8bd99481 898 if (c != &boot_cpu_data)
3b520b23 899 mtrr_ap_init();
1da177e4 900#ifdef CONFIG_NUMA
3019e8eb 901 numa_add_cpu(smp_processor_id());
1da177e4
LT
902#endif
903}
904
905
e6982c67 906void __cpuinit print_cpu_info(struct cpuinfo_x86 *c)
1da177e4
LT
907{
908 if (c->x86_model_id[0])
909 printk("%s", c->x86_model_id);
910
911 if (c->x86_mask || c->cpuid_level >= 0)
912 printk(" stepping %02x\n", c->x86_mask);
913 else
914 printk("\n");
915}
916
917/*
918 * Get CPU information for use by the procfs.
919 */
920
921static int show_cpuinfo(struct seq_file *m, void *v)
922{
923 struct cpuinfo_x86 *c = v;
924
925 /*
926 * These flag bits must match the definitions in <asm/cpufeature.h>.
927 * NULL means this bit is undefined or reserved; either way it doesn't
928 * have meaning as far as Linux is concerned. Note that it's important
929 * to realize there is a difference between this table and CPUID -- if
930 * applications want to get the raw CPUID data, they should access
931 * /dev/cpu/<cpu_nr>/cpuid instead.
932 */
933 static char *x86_cap_flags[] = {
934 /* Intel-defined */
935 "fpu", "vme", "de", "pse", "tsc", "msr", "pae", "mce",
936 "cx8", "apic", NULL, "sep", "mtrr", "pge", "mca", "cmov",
937 "pat", "pse36", "pn", "clflush", NULL, "dts", "acpi", "mmx",
ec481536 938 "fxsr", "sse", "sse2", "ss", "ht", "tm", "ia64", "pbe",
1da177e4
LT
939
940 /* AMD-defined */
3c3b73b6 941 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1da177e4
LT
942 NULL, NULL, NULL, "syscall", NULL, NULL, NULL, NULL,
943 NULL, NULL, NULL, NULL, "nx", NULL, "mmxext", NULL,
f790cd30
AK
944 NULL, "fxsr_opt", "pdpe1gb", "rdtscp", NULL, "lm",
945 "3dnowext", "3dnow",
1da177e4
LT
946
947 /* Transmeta-defined */
948 "recovery", "longrun", NULL, "lrti", NULL, NULL, NULL, NULL,
949 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
950 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
951 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
952
953 /* Other (Linux-defined) */
ec481536
PA
954 "cxmmx", "k6_mtrr", "cyrix_arr", "centaur_mcr",
955 NULL, NULL, NULL, NULL,
956 "constant_tsc", "up", NULL, "arch_perfmon",
957 "pebs", "bts", NULL, "sync_rdtsc",
958 "rep_good", NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1da177e4
LT
959 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
960
961 /* Intel-defined (#2) */
9d95dd84 962 "pni", NULL, NULL, "monitor", "ds_cpl", "vmx", "smx", "est",
dcf10307 963 "tm2", "ssse3", "cid", NULL, NULL, "cx16", "xtpr", NULL,
f790cd30 964 NULL, NULL, "dca", NULL, NULL, NULL, NULL, "popcnt",
1da177e4
LT
965 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
966
5b7abc6f
PA
967 /* VIA/Cyrix/Centaur-defined */
968 NULL, NULL, "rng", "rng_en", NULL, NULL, "ace", "ace_en",
ec481536 969 "ace2", "ace2_en", "phe", "phe_en", "pmm", "pmm_en", NULL, NULL,
5b7abc6f
PA
970 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
971 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
972
1da177e4 973 /* AMD-defined (#2) */
f790cd30
AK
974 "lahf_lm", "cmp_legacy", "svm", "extapic", "cr8_legacy",
975 "altmovcr8", "abm", "sse4a",
976 "misalignsse", "3dnowprefetch",
977 "osvw", "ibs", NULL, NULL, NULL, NULL,
1da177e4 978 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
5b7abc6f 979 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1d67953f
VP
980
981 /* Auxiliary (Linux-defined) */
982 "ida", NULL, NULL, NULL, NULL, NULL, NULL, NULL,
983 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
984 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
985 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1da177e4
LT
986 };
987 static char *x86_power_flags[] = {
988 "ts", /* temperature sensor */
989 "fid", /* frequency id control */
990 "vid", /* voltage id control */
991 "ttp", /* thermal trip */
992 "tm",
3f98bc49 993 "stc",
f790cd30
AK
994 "100mhzsteps",
995 "hwpstate",
d824395c
JR
996 "", /* tsc invariant mapped to constant_tsc */
997 /* nothing */
1da177e4
LT
998 };
999
1000
1001#ifdef CONFIG_SMP
1002 if (!cpu_online(c-cpu_data))
1003 return 0;
1004#endif
1005
1006 seq_printf(m,"processor\t: %u\n"
1007 "vendor_id\t: %s\n"
1008 "cpu family\t: %d\n"
1009 "model\t\t: %d\n"
1010 "model name\t: %s\n",
1011 (unsigned)(c-cpu_data),
1012 c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown",
1013 c->x86,
1014 (int)c->x86_model,
1015 c->x86_model_id[0] ? c->x86_model_id : "unknown");
1016
1017 if (c->x86_mask || c->cpuid_level >= 0)
1018 seq_printf(m, "stepping\t: %d\n", c->x86_mask);
1019 else
1020 seq_printf(m, "stepping\t: unknown\n");
1021
1022 if (cpu_has(c,X86_FEATURE_TSC)) {
95235ca2
VP
1023 unsigned int freq = cpufreq_quick_get((unsigned)(c-cpu_data));
1024 if (!freq)
1025 freq = cpu_khz;
1da177e4 1026 seq_printf(m, "cpu MHz\t\t: %u.%03u\n",
95235ca2 1027 freq / 1000, (freq % 1000));
1da177e4
LT
1028 }
1029
1030 /* Cache size */
1031 if (c->x86_cache_size >= 0)
1032 seq_printf(m, "cache size\t: %d KB\n", c->x86_cache_size);
1033
1034#ifdef CONFIG_SMP
94605eff 1035 if (smp_num_siblings * c->x86_max_cores > 1) {
db468681 1036 int cpu = c - cpu_data;
f3fa8ebc 1037 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id);
94605eff 1038 seq_printf(m, "siblings\t: %d\n", cpus_weight(cpu_core_map[cpu]));
f3fa8ebc 1039 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id);
94605eff 1040 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
db468681 1041 }
1da177e4
LT
1042#endif
1043
1044 seq_printf(m,
1045 "fpu\t\t: yes\n"
1046 "fpu_exception\t: yes\n"
1047 "cpuid level\t: %d\n"
1048 "wp\t\t: yes\n"
1049 "flags\t\t:",
1050 c->cpuid_level);
1051
1052 {
1053 int i;
1054 for ( i = 0 ; i < 32*NCAPINTS ; i++ )
3d1712c9 1055 if (cpu_has(c, i) && x86_cap_flags[i] != NULL)
1da177e4
LT
1056 seq_printf(m, " %s", x86_cap_flags[i]);
1057 }
1058
1059 seq_printf(m, "\nbogomips\t: %lu.%02lu\n",
1060 c->loops_per_jiffy/(500000/HZ),
1061 (c->loops_per_jiffy/(5000/HZ)) % 100);
1062
1063 if (c->x86_tlbsize > 0)
1064 seq_printf(m, "TLB size\t: %d 4K pages\n", c->x86_tlbsize);
1065 seq_printf(m, "clflush size\t: %d\n", c->x86_clflush_size);
1066 seq_printf(m, "cache_alignment\t: %d\n", c->x86_cache_alignment);
1067
1068 seq_printf(m, "address sizes\t: %u bits physical, %u bits virtual\n",
1069 c->x86_phys_bits, c->x86_virt_bits);
1070
1071 seq_printf(m, "power management:");
1072 {
1073 unsigned i;
1074 for (i = 0; i < 32; i++)
1075 if (c->x86_power & (1 << i)) {
3f98bc49
AK
1076 if (i < ARRAY_SIZE(x86_power_flags) &&
1077 x86_power_flags[i])
1078 seq_printf(m, "%s%s",
1079 x86_power_flags[i][0]?" ":"",
1080 x86_power_flags[i]);
1da177e4
LT
1081 else
1082 seq_printf(m, " [%d]", i);
1083 }
1084 }
1da177e4 1085
d31ddaa1 1086 seq_printf(m, "\n\n");
1da177e4
LT
1087
1088 return 0;
1089}
1090
1091static void *c_start(struct seq_file *m, loff_t *pos)
1092{
1093 return *pos < NR_CPUS ? cpu_data + *pos : NULL;
1094}
1095
1096static void *c_next(struct seq_file *m, void *v, loff_t *pos)
1097{
1098 ++*pos;
1099 return c_start(m, pos);
1100}
1101
1102static void c_stop(struct seq_file *m, void *v)
1103{
1104}
1105
1106struct seq_operations cpuinfo_op = {
1107 .start =c_start,
1108 .next = c_next,
1109 .stop = c_stop,
1110 .show = show_cpuinfo,
1111};