]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - arch/x86_64/kernel/setup.c
[PATCH] swsusp: x86_64 mark special saveable/unsaveable pages
[mirror_ubuntu-hirsute-kernel.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.
8 *
9 * $Id$
10 */
11
12/*
13 * This file handles the architecture-dependent parts of initialization
14 */
15
16#include <linux/errno.h>
17#include <linux/sched.h>
18#include <linux/kernel.h>
19#include <linux/mm.h>
20#include <linux/stddef.h>
21#include <linux/unistd.h>
22#include <linux/ptrace.h>
23#include <linux/slab.h>
24#include <linux/user.h>
25#include <linux/a.out.h>
26#include <linux/tty.h>
27#include <linux/ioport.h>
28#include <linux/delay.h>
29#include <linux/config.h>
30#include <linux/init.h>
31#include <linux/initrd.h>
32#include <linux/highmem.h>
33#include <linux/bootmem.h>
34#include <linux/module.h>
35#include <asm/processor.h>
36#include <linux/console.h>
37#include <linux/seq_file.h>
aac04b32 38#include <linux/crash_dump.h>
1da177e4
LT
39#include <linux/root_dev.h>
40#include <linux/pci.h>
41#include <linux/acpi.h>
42#include <linux/kallsyms.h>
43#include <linux/edd.h>
bbfceef4 44#include <linux/mmzone.h>
5f5609df 45#include <linux/kexec.h>
95235ca2 46#include <linux/cpufreq.h>
e9928674 47#include <linux/dmi.h>
17a941d8 48#include <linux/dma-mapping.h>
681558fd 49#include <linux/ctype.h>
3e3318de 50#include <linux/suspend.h>
bbfceef4 51
1da177e4
LT
52#include <asm/mtrr.h>
53#include <asm/uaccess.h>
54#include <asm/system.h>
55#include <asm/io.h>
56#include <asm/smp.h>
57#include <asm/msr.h>
58#include <asm/desc.h>
59#include <video/edid.h>
60#include <asm/e820.h>
61#include <asm/dma.h>
62#include <asm/mpspec.h>
63#include <asm/mmu_context.h>
64#include <asm/bootsetup.h>
65#include <asm/proto.h>
66#include <asm/setup.h>
67#include <asm/mach_apic.h>
68#include <asm/numa.h>
17a941d8 69#include <asm/swiotlb.h>
2bc0414e 70#include <asm/sections.h>
17a941d8 71#include <asm/gart-mapping.h>
f2d3efed 72#include <asm/dmi.h>
1da177e4
LT
73
74/*
75 * Machine setup..
76 */
77
6c231b7b 78struct cpuinfo_x86 boot_cpu_data __read_mostly;
1da177e4
LT
79
80unsigned long mmu_cr4_features;
81
82int acpi_disabled;
83EXPORT_SYMBOL(acpi_disabled);
888ba6c6 84#ifdef CONFIG_ACPI
1da177e4
LT
85extern int __initdata acpi_ht;
86extern acpi_interrupt_flags acpi_sci_flags;
87int __initdata acpi_force = 0;
88#endif
89
90int acpi_numa __initdata;
91
1da177e4
LT
92/* Boot loader ID as an integer, for the benefit of proc_dointvec */
93int bootloader_type;
94
95unsigned long saved_video_mode;
96
f2d3efed
AK
97/*
98 * Early DMI memory
99 */
100int dmi_alloc_index;
101char dmi_alloc_data[DMI_MAX_DATA];
102
1da177e4
LT
103/*
104 * Setup options
105 */
1da177e4
LT
106struct screen_info screen_info;
107struct sys_desc_table_struct {
108 unsigned short length;
109 unsigned char table[0];
110};
111
112struct edid_info edid_info;
113struct e820map e820;
114
115extern int root_mountflags;
1da177e4
LT
116
117char command_line[COMMAND_LINE_SIZE];
118
119struct resource standard_io_resources[] = {
120 { .name = "dma1", .start = 0x00, .end = 0x1f,
121 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
122 { .name = "pic1", .start = 0x20, .end = 0x21,
123 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
124 { .name = "timer0", .start = 0x40, .end = 0x43,
125 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
126 { .name = "timer1", .start = 0x50, .end = 0x53,
127 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
128 { .name = "keyboard", .start = 0x60, .end = 0x6f,
129 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
130 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
131 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
132 { .name = "pic2", .start = 0xa0, .end = 0xa1,
133 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
134 { .name = "dma2", .start = 0xc0, .end = 0xdf,
135 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
136 { .name = "fpu", .start = 0xf0, .end = 0xff,
137 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
138};
139
140#define STANDARD_IO_RESOURCES \
141 (sizeof standard_io_resources / sizeof standard_io_resources[0])
142
143#define IORESOURCE_RAM (IORESOURCE_BUSY | IORESOURCE_MEM)
144
145struct resource data_resource = {
146 .name = "Kernel data",
147 .start = 0,
148 .end = 0,
149 .flags = IORESOURCE_RAM,
150};
151struct resource code_resource = {
152 .name = "Kernel code",
153 .start = 0,
154 .end = 0,
155 .flags = IORESOURCE_RAM,
156};
157
158#define IORESOURCE_ROM (IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM)
159
160static struct resource system_rom_resource = {
161 .name = "System ROM",
162 .start = 0xf0000,
163 .end = 0xfffff,
164 .flags = IORESOURCE_ROM,
165};
166
167static struct resource extension_rom_resource = {
168 .name = "Extension ROM",
169 .start = 0xe0000,
170 .end = 0xeffff,
171 .flags = IORESOURCE_ROM,
172};
173
174static struct resource adapter_rom_resources[] = {
175 { .name = "Adapter ROM", .start = 0xc8000, .end = 0,
176 .flags = IORESOURCE_ROM },
177 { .name = "Adapter ROM", .start = 0, .end = 0,
178 .flags = IORESOURCE_ROM },
179 { .name = "Adapter ROM", .start = 0, .end = 0,
180 .flags = IORESOURCE_ROM },
181 { .name = "Adapter ROM", .start = 0, .end = 0,
182 .flags = IORESOURCE_ROM },
183 { .name = "Adapter ROM", .start = 0, .end = 0,
184 .flags = IORESOURCE_ROM },
185 { .name = "Adapter ROM", .start = 0, .end = 0,
186 .flags = IORESOURCE_ROM }
187};
188
189#define ADAPTER_ROM_RESOURCES \
190 (sizeof adapter_rom_resources / sizeof adapter_rom_resources[0])
191
192static struct resource video_rom_resource = {
193 .name = "Video ROM",
194 .start = 0xc0000,
195 .end = 0xc7fff,
196 .flags = IORESOURCE_ROM,
197};
198
199static struct resource video_ram_resource = {
200 .name = "Video RAM area",
201 .start = 0xa0000,
202 .end = 0xbffff,
203 .flags = IORESOURCE_RAM,
204};
205
206#define romsignature(x) (*(unsigned short *)(x) == 0xaa55)
207
208static int __init romchecksum(unsigned char *rom, unsigned long length)
209{
210 unsigned char *p, sum = 0;
211
212 for (p = rom; p < rom + length; p++)
213 sum += *p;
214 return sum == 0;
215}
216
217static void __init probe_roms(void)
218{
219 unsigned long start, length, upper;
220 unsigned char *rom;
221 int i;
222
223 /* video rom */
224 upper = adapter_rom_resources[0].start;
225 for (start = video_rom_resource.start; start < upper; start += 2048) {
226 rom = isa_bus_to_virt(start);
227 if (!romsignature(rom))
228 continue;
229
230 video_rom_resource.start = start;
231
232 /* 0 < length <= 0x7f * 512, historically */
233 length = rom[2] * 512;
234
235 /* if checksum okay, trust length byte */
236 if (length && romchecksum(rom, length))
237 video_rom_resource.end = start + length - 1;
238
239 request_resource(&iomem_resource, &video_rom_resource);
240 break;
241 }
242
243 start = (video_rom_resource.end + 1 + 2047) & ~2047UL;
244 if (start < upper)
245 start = upper;
246
247 /* system rom */
248 request_resource(&iomem_resource, &system_rom_resource);
249 upper = system_rom_resource.start;
250
251 /* check for extension rom (ignore length byte!) */
252 rom = isa_bus_to_virt(extension_rom_resource.start);
253 if (romsignature(rom)) {
254 length = extension_rom_resource.end - extension_rom_resource.start + 1;
255 if (romchecksum(rom, length)) {
256 request_resource(&iomem_resource, &extension_rom_resource);
257 upper = extension_rom_resource.start;
258 }
259 }
260
261 /* check for adapter roms on 2k boundaries */
262 for (i = 0; i < ADAPTER_ROM_RESOURCES && start < upper; start += 2048) {
263 rom = isa_bus_to_virt(start);
264 if (!romsignature(rom))
265 continue;
266
267 /* 0 < length <= 0x7f * 512, historically */
268 length = rom[2] * 512;
269
270 /* but accept any length that fits if checksum okay */
271 if (!length || start + length > upper || !romchecksum(rom, length))
272 continue;
273
274 adapter_rom_resources[i].start = start;
275 adapter_rom_resources[i].end = start + length - 1;
276 request_resource(&iomem_resource, &adapter_rom_resources[i]);
277
278 start = adapter_rom_resources[i++].end & ~2047UL;
279 }
280}
281
681558fd
AK
282/* Check for full argument with no trailing characters */
283static int fullarg(char *p, char *arg)
284{
285 int l = strlen(arg);
286 return !memcmp(p, arg, l) && (p[l] == 0 || isspace(p[l]));
287}
288
1da177e4
LT
289static __init void parse_cmdline_early (char ** cmdline_p)
290{
291 char c = ' ', *to = command_line, *from = COMMAND_LINE;
292 int len = 0;
69cda7b1 293 int userdef = 0;
1da177e4 294
1da177e4
LT
295 for (;;) {
296 if (c != ' ')
297 goto next_char;
298
299#ifdef CONFIG_SMP
300 /*
301 * If the BIOS enumerates physical processors before logical,
302 * maxcpus=N at enumeration-time can be used to disable HT.
303 */
304 else if (!memcmp(from, "maxcpus=", 8)) {
305 extern unsigned int maxcpus;
306
307 maxcpus = simple_strtoul(from + 8, NULL, 0);
308 }
309#endif
888ba6c6 310#ifdef CONFIG_ACPI
1da177e4 311 /* "acpi=off" disables both ACPI table parsing and interpreter init */
681558fd 312 if (fullarg(from,"acpi=off"))
1da177e4
LT
313 disable_acpi();
314
681558fd 315 if (fullarg(from, "acpi=force")) {
1da177e4
LT
316 /* add later when we do DMI horrors: */
317 acpi_force = 1;
318 acpi_disabled = 0;
319 }
320
321 /* acpi=ht just means: do ACPI MADT parsing
322 at bootup, but don't enable the full ACPI interpreter */
681558fd 323 if (fullarg(from, "acpi=ht")) {
1da177e4
LT
324 if (!acpi_force)
325 disable_acpi();
326 acpi_ht = 1;
327 }
681558fd 328 else if (fullarg(from, "pci=noacpi"))
1da177e4 329 acpi_disable_pci();
681558fd 330 else if (fullarg(from, "acpi=noirq"))
1da177e4
LT
331 acpi_noirq_set();
332
681558fd 333 else if (fullarg(from, "acpi_sci=edge"))
1da177e4 334 acpi_sci_flags.trigger = 1;
681558fd 335 else if (fullarg(from, "acpi_sci=level"))
1da177e4 336 acpi_sci_flags.trigger = 3;
681558fd 337 else if (fullarg(from, "acpi_sci=high"))
1da177e4 338 acpi_sci_flags.polarity = 1;
681558fd 339 else if (fullarg(from, "acpi_sci=low"))
1da177e4
LT
340 acpi_sci_flags.polarity = 3;
341
342 /* acpi=strict disables out-of-spec workarounds */
681558fd 343 else if (fullarg(from, "acpi=strict")) {
1da177e4
LT
344 acpi_strict = 1;
345 }
22999244 346#ifdef CONFIG_X86_IO_APIC
681558fd 347 else if (fullarg(from, "acpi_skip_timer_override"))
22999244
AK
348 acpi_skip_timer_override = 1;
349#endif
1da177e4
LT
350#endif
351
681558fd 352 if (fullarg(from, "disable_timer_pin_1"))
66759a01 353 disable_timer_pin_1 = 1;
681558fd 354 if (fullarg(from, "enable_timer_pin_1"))
66759a01
CE
355 disable_timer_pin_1 = -1;
356
d1530d82
AK
357 if (fullarg(from, "nolapic") || fullarg(from, "disableapic")) {
358 clear_bit(X86_FEATURE_APIC, boot_cpu_data.x86_capability);
1da177e4 359 disable_apic = 1;
d1530d82 360 }
1da177e4 361
681558fd 362 if (fullarg(from, "noapic"))
1da177e4
LT
363 skip_ioapic_setup = 1;
364
681558fd 365 if (fullarg(from,"apic")) {
1da177e4
LT
366 skip_ioapic_setup = 0;
367 ioapic_force = 1;
368 }
369
370 if (!memcmp(from, "mem=", 4))
371 parse_memopt(from+4, &from);
372
69cda7b1 373 if (!memcmp(from, "memmap=", 7)) {
374 /* exactmap option is for used defined memory */
375 if (!memcmp(from+7, "exactmap", 8)) {
376#ifdef CONFIG_CRASH_DUMP
377 /* If we are doing a crash dump, we
378 * still need to know the real mem
379 * size before original memory map is
380 * reset.
381 */
382 saved_max_pfn = e820_end_of_ram();
383#endif
384 from += 8+7;
385 end_pfn_map = 0;
386 e820.nr_map = 0;
387 userdef = 1;
388 }
389 else {
390 parse_memmapopt(from+7, &from);
391 userdef = 1;
392 }
393 }
394
2b97690f 395#ifdef CONFIG_NUMA
1da177e4
LT
396 if (!memcmp(from, "numa=", 5))
397 numa_setup(from+5);
398#endif
399
1da177e4
LT
400 if (!memcmp(from,"iommu=",6)) {
401 iommu_setup(from+6);
402 }
1da177e4 403
681558fd 404 if (fullarg(from,"oops=panic"))
1da177e4
LT
405 panic_on_oops = 1;
406
407 if (!memcmp(from, "noexec=", 7))
408 nonx_setup(from + 7);
409
5f5609df
EB
410#ifdef CONFIG_KEXEC
411 /* crashkernel=size@addr specifies the location to reserve for
412 * a crash kernel. By reserving this memory we guarantee
413 * that linux never set's it up as a DMA target.
414 * Useful for holding code to do something appropriate
415 * after a kernel panic.
416 */
417 else if (!memcmp(from, "crashkernel=", 12)) {
418 unsigned long size, base;
419 size = memparse(from+12, &from);
420 if (*from == '@') {
421 base = memparse(from+1, &from);
422 /* FIXME: Do I want a sanity check
423 * to validate the memory range?
424 */
425 crashk_res.start = base;
426 crashk_res.end = base + size - 1;
427 }
428 }
429#endif
430
aac04b32
VG
431#ifdef CONFIG_PROC_VMCORE
432 /* elfcorehdr= specifies the location of elf core header
433 * stored by the crashed kernel. This option will be passed
434 * by kexec loader to the capture kernel.
435 */
436 else if(!memcmp(from, "elfcorehdr=", 11))
437 elfcorehdr_addr = memparse(from+11, &from);
438#endif
e2c03888 439
d5176123 440#ifdef CONFIG_HOTPLUG_CPU
e2c03888
AK
441 else if (!memcmp(from, "additional_cpus=", 16))
442 setup_additional_cpus(from+16);
443#endif
444
1da177e4
LT
445 next_char:
446 c = *(from++);
447 if (!c)
448 break;
449 if (COMMAND_LINE_SIZE <= ++len)
450 break;
451 *(to++) = c;
452 }
69cda7b1 453 if (userdef) {
454 printk(KERN_INFO "user-defined physical RAM map:\n");
455 e820_print_map("user");
456 }
1da177e4
LT
457 *to = '\0';
458 *cmdline_p = command_line;
459}
460
2b97690f 461#ifndef CONFIG_NUMA
bbfceef4
MT
462static void __init
463contig_initmem_init(unsigned long start_pfn, unsigned long end_pfn)
1da177e4 464{
bbfceef4
MT
465 unsigned long bootmap_size, bootmap;
466
bbfceef4
MT
467 bootmap_size = bootmem_bootmap_pages(end_pfn)<<PAGE_SHIFT;
468 bootmap = find_e820_area(0, end_pfn<<PAGE_SHIFT, bootmap_size);
469 if (bootmap == -1L)
470 panic("Cannot find bootmem map of size %ld\n",bootmap_size);
471 bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, end_pfn);
472 e820_bootmem_free(NODE_DATA(0), 0, end_pfn << PAGE_SHIFT);
473 reserve_bootmem(bootmap, bootmap_size);
1da177e4
LT
474}
475#endif
476
477/* Use inline assembly to define this because the nops are defined
478 as inline assembly strings in the include files and we cannot
479 get them easily into strings. */
480asm("\t.data\nk8nops: "
481 K8_NOP1 K8_NOP2 K8_NOP3 K8_NOP4 K8_NOP5 K8_NOP6
482 K8_NOP7 K8_NOP8);
483
484extern unsigned char k8nops[];
485static unsigned char *k8_nops[ASM_NOP_MAX+1] = {
486 NULL,
487 k8nops,
488 k8nops + 1,
489 k8nops + 1 + 2,
490 k8nops + 1 + 2 + 3,
491 k8nops + 1 + 2 + 3 + 4,
492 k8nops + 1 + 2 + 3 + 4 + 5,
493 k8nops + 1 + 2 + 3 + 4 + 5 + 6,
494 k8nops + 1 + 2 + 3 + 4 + 5 + 6 + 7,
495};
496
7f6c5b04
AK
497extern char __vsyscall_0;
498
1da177e4
LT
499/* Replace instructions with better alternatives for this CPU type.
500
501 This runs before SMP is initialized to avoid SMP problems with
502 self modifying code. This implies that assymetric systems where
503 APs have less capabilities than the boot processor are not handled.
504 In this case boot with "noreplacement". */
505void apply_alternatives(void *start, void *end)
506{
507 struct alt_instr *a;
508 int diff, i, k;
509 for (a = start; (void *)a < end; a++) {
7f6c5b04
AK
510 u8 *instr;
511
1da177e4
LT
512 if (!boot_cpu_has(a->cpuid))
513 continue;
514
515 BUG_ON(a->replacementlen > a->instrlen);
7f6c5b04
AK
516 instr = a->instr;
517 /* vsyscall code is not mapped yet. resolve it manually. */
518 if (instr >= (u8 *)VSYSCALL_START && instr < (u8*)VSYSCALL_END)
519 instr = __va(instr - (u8*)VSYSCALL_START + (u8*)__pa_symbol(&__vsyscall_0));
520 __inline_memcpy(instr, a->replacement, a->replacementlen);
1da177e4
LT
521 diff = a->instrlen - a->replacementlen;
522
523 /* Pad the rest with nops */
524 for (i = a->replacementlen; diff > 0; diff -= k, i += k) {
525 k = diff;
526 if (k > ASM_NOP_MAX)
527 k = ASM_NOP_MAX;
7f6c5b04 528 __inline_memcpy(instr + i, k8_nops[k], k);
1da177e4
LT
529 }
530 }
531}
532
533static int no_replacement __initdata = 0;
534
535void __init alternative_instructions(void)
536{
537 extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
538 if (no_replacement)
539 return;
540 apply_alternatives(__alt_instructions, __alt_instructions_end);
541}
542
543static int __init noreplacement_setup(char *s)
544{
545 no_replacement = 1;
9b41046c 546 return 1;
1da177e4
LT
547}
548
549__setup("noreplacement", noreplacement_setup);
550
551#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
552struct edd edd;
553#ifdef CONFIG_EDD_MODULE
554EXPORT_SYMBOL(edd);
555#endif
556/**
557 * copy_edd() - Copy the BIOS EDD information
558 * from boot_params into a safe place.
559 *
560 */
561static inline void copy_edd(void)
562{
563 memcpy(edd.mbr_signature, EDD_MBR_SIGNATURE, sizeof(edd.mbr_signature));
564 memcpy(edd.edd_info, EDD_BUF, sizeof(edd.edd_info));
565 edd.mbr_signature_nr = EDD_MBR_SIG_NR;
566 edd.edd_info_nr = EDD_NR;
567}
568#else
569static inline void copy_edd(void)
570{
571}
572#endif
573
574#define EBDA_ADDR_POINTER 0x40E
ac71d12c
AK
575
576unsigned __initdata ebda_addr;
577unsigned __initdata ebda_size;
578
579static void discover_ebda(void)
1da177e4 580{
ac71d12c 581 /*
1da177e4
LT
582 * there is a real-mode segmented pointer pointing to the
583 * 4K EBDA area at 0x40E
584 */
ac71d12c
AK
585 ebda_addr = *(unsigned short *)EBDA_ADDR_POINTER;
586 ebda_addr <<= 4;
587
588 ebda_size = *(unsigned short *)(unsigned long)ebda_addr;
589
590 /* Round EBDA up to pages */
591 if (ebda_size == 0)
592 ebda_size = 1;
593 ebda_size <<= 10;
594 ebda_size = round_up(ebda_size + (ebda_addr & ~PAGE_MASK), PAGE_SIZE);
595 if (ebda_size > 64*1024)
596 ebda_size = 64*1024;
1da177e4
LT
597}
598
3e3318de
SL
599#ifdef CONFIG_SOFTWARE_SUSPEND
600static void __init mark_nosave_page_range(unsigned long start, unsigned long end)
601{
602 struct page *page;
603 while (start <= end) {
604 page = pfn_to_page(start);
605 SetPageNosave(page);
606 start++;
607 }
608}
609
610static void __init e820_nosave_reserved_pages(void)
611{
612 int i;
613 unsigned long r_start = 0, r_end = 0;
614
615 /* Assume e820 map is sorted */
616 for (i = 0; i < e820.nr_map; i++) {
617 struct e820entry *ei = &e820.map[i];
618 unsigned long start, end;
619
620 start = round_down(ei->addr, PAGE_SIZE);
621 end = round_up(ei->addr + ei->size, PAGE_SIZE);
622 if (start >= end)
623 continue;
624 if (ei->type == E820_RESERVED)
625 continue;
626 r_end = start>>PAGE_SHIFT;
627 /* swsusp ignores invalid pfn, ignore these pages here */
628 if (r_end > end_pfn)
629 r_end = end_pfn;
630 if (r_end > r_start)
631 mark_nosave_page_range(r_start, r_end-1);
632 if (r_end >= end_pfn)
633 break;
634 r_start = end>>PAGE_SHIFT;
635 }
636}
637
638static void __init e820_save_acpi_pages(void)
639{
640 int i;
641
642 /* Assume e820 map is sorted */
643 for (i = 0; i < e820.nr_map; i++) {
644 struct e820entry *ei = &e820.map[i];
645 unsigned long start, end;
646
647 start = ei->addr, PAGE_SIZE;
648 end = ei->addr + ei->size;
649 if (start >= end)
650 continue;
651 if (ei->type != E820_ACPI && ei->type != E820_NVS)
652 continue;
653 /*
654 * If the region is below end_pfn, it will be
655 * saved/restored by swsusp follow 'RAM' type.
656 */
657 if (start < (end_pfn << PAGE_SHIFT))
658 start = end_pfn << PAGE_SHIFT;
659 if (end > start)
660 swsusp_add_arch_pages(start, end);
661 }
662}
663
664extern char __start_rodata, __end_rodata;
665/*
666 * BIOS reserved region/hole - no save/restore
667 * ACPI NVS - save/restore
668 * ACPI Data - this is a little tricky, the mem could be used by OS after OS
669 * reads tables from the region, but anyway save/restore the memory hasn't any
670 * side effect and Linux runtime module load/unload might use it.
671 * kernel rodata - no save/restore (kernel rodata isn't changed)
672 */
673static int __init mark_nosave_pages(void)
674{
675 unsigned long pfn_start, pfn_end;
676
677 /* BIOS reserved regions & holes */
678 e820_nosave_reserved_pages();
679
680 /* kernel rodata */
681 pfn_start = round_up(__pa_symbol(&__start_rodata), PAGE_SIZE) >> PAGE_SHIFT;
682 pfn_end = round_down(__pa_symbol(&__end_rodata), PAGE_SIZE) >> PAGE_SHIFT;
683 mark_nosave_page_range(pfn_start, pfn_end-1);
684
685 /* record ACPI Data/NVS as saveable */
686 e820_save_acpi_pages();
687
688 return 0;
689}
690core_initcall(mark_nosave_pages);
691#endif
692
1da177e4
LT
693void __init setup_arch(char **cmdline_p)
694{
1da177e4
LT
695 unsigned long kernel_end;
696
697 ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
1da177e4
LT
698 screen_info = SCREEN_INFO;
699 edid_info = EDID_INFO;
700 saved_video_mode = SAVED_VIDEO_MODE;
701 bootloader_type = LOADER_TYPE;
702
703#ifdef CONFIG_BLK_DEV_RAM
704 rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK;
705 rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0);
706 rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0);
707#endif
708 setup_memory_region();
709 copy_edd();
710
711 if (!MOUNT_ROOT_RDONLY)
712 root_mountflags &= ~MS_RDONLY;
713 init_mm.start_code = (unsigned long) &_text;
714 init_mm.end_code = (unsigned long) &_etext;
715 init_mm.end_data = (unsigned long) &_edata;
716 init_mm.brk = (unsigned long) &_end;
717
718 code_resource.start = virt_to_phys(&_text);
719 code_resource.end = virt_to_phys(&_etext)-1;
720 data_resource.start = virt_to_phys(&_etext);
721 data_resource.end = virt_to_phys(&_edata)-1;
722
723 parse_cmdline_early(cmdline_p);
724
725 early_identify_cpu(&boot_cpu_data);
726
727 /*
728 * partially used pages are not usable - thus
729 * we are rounding upwards:
730 */
731 end_pfn = e820_end_of_ram();
1f50249e 732 num_physpages = end_pfn; /* for pfn_valid */
1da177e4
LT
733
734 check_efer();
735
ac71d12c
AK
736 discover_ebda();
737
1da177e4
LT
738 init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT));
739
f2d3efed
AK
740 dmi_scan_machine();
741
f6c2e333
SS
742 zap_low_mappings(0);
743
888ba6c6 744#ifdef CONFIG_ACPI
1da177e4
LT
745 /*
746 * Initialize the ACPI boot-time table parser (gets the RSDP and SDT).
747 * Call this early for SRAT node setup.
748 */
749 acpi_boot_table_init();
750#endif
751
752#ifdef CONFIG_ACPI_NUMA
753 /*
754 * Parse SRAT to discover nodes.
755 */
756 acpi_numa_init();
757#endif
758
2b97690f 759#ifdef CONFIG_NUMA
1da177e4
LT
760 numa_initmem_init(0, end_pfn);
761#else
bbfceef4 762 contig_initmem_init(0, end_pfn);
1da177e4
LT
763#endif
764
765 /* Reserve direct mapping */
766 reserve_bootmem_generic(table_start << PAGE_SHIFT,
767 (table_end - table_start) << PAGE_SHIFT);
768
769 /* reserve kernel */
770 kernel_end = round_up(__pa_symbol(&_end),PAGE_SIZE);
771 reserve_bootmem_generic(HIGH_MEMORY, kernel_end - HIGH_MEMORY);
772
773 /*
774 * reserve physical page 0 - it's a special BIOS page on many boxes,
775 * enabling clean reboots, SMP operation, laptop functions.
776 */
777 reserve_bootmem_generic(0, PAGE_SIZE);
778
779 /* reserve ebda region */
ac71d12c
AK
780 if (ebda_addr)
781 reserve_bootmem_generic(ebda_addr, ebda_size);
1da177e4
LT
782
783#ifdef CONFIG_SMP
784 /*
785 * But first pinch a few for the stack/trampoline stuff
786 * FIXME: Don't need the extra page at 4K, but need to fix
787 * trampoline before removing it. (see the GDT stuff)
788 */
789 reserve_bootmem_generic(PAGE_SIZE, PAGE_SIZE);
790
791 /* Reserve SMP trampoline */
792 reserve_bootmem_generic(SMP_TRAMPOLINE_BASE, PAGE_SIZE);
793#endif
794
795#ifdef CONFIG_ACPI_SLEEP
796 /*
797 * Reserve low memory region for sleep support.
798 */
799 acpi_reserve_bootmem();
800#endif
801#ifdef CONFIG_X86_LOCAL_APIC
802 /*
803 * Find and reserve possible boot-time SMP configuration:
804 */
805 find_smp_config();
806#endif
807#ifdef CONFIG_BLK_DEV_INITRD
808 if (LOADER_TYPE && INITRD_START) {
809 if (INITRD_START + INITRD_SIZE <= (end_pfn << PAGE_SHIFT)) {
810 reserve_bootmem_generic(INITRD_START, INITRD_SIZE);
811 initrd_start =
812 INITRD_START ? INITRD_START + PAGE_OFFSET : 0;
813 initrd_end = initrd_start+INITRD_SIZE;
814 }
815 else {
816 printk(KERN_ERR "initrd extends beyond end of memory "
817 "(0x%08lx > 0x%08lx)\ndisabling initrd\n",
818 (unsigned long)(INITRD_START + INITRD_SIZE),
819 (unsigned long)(end_pfn << PAGE_SHIFT));
820 initrd_start = 0;
821 }
822 }
823#endif
5f5609df
EB
824#ifdef CONFIG_KEXEC
825 if (crashk_res.start != crashk_res.end) {
826 reserve_bootmem(crashk_res.start,
827 crashk_res.end - crashk_res.start + 1);
828 }
829#endif
0d317fb7 830
1da177e4
LT
831 paging_init();
832
833 check_ioapic();
834
51f62e18
AR
835 /*
836 * set this early, so we dont allocate cpu0
837 * if MADT list doesnt list BSP first
838 * mpparse.c/MP_processor_info() allocates logical cpu numbers.
839 */
840 cpu_set(0, cpu_present_map);
888ba6c6 841#ifdef CONFIG_ACPI
1da177e4
LT
842 /*
843 * Read APIC and some other early information from ACPI tables.
844 */
845 acpi_boot_init();
846#endif
847
05b3cbd8
RT
848 init_cpu_to_node();
849
1da177e4
LT
850#ifdef CONFIG_X86_LOCAL_APIC
851 /*
852 * get boot-time SMP configuration:
853 */
854 if (smp_found_config)
855 get_smp_config();
856 init_apic_mappings();
857#endif
858
859 /*
860 * Request address space for all standard RAM and ROM resources
861 * and also for regions reported as reserved by the e820.
862 */
863 probe_roms();
864 e820_reserve_resources();
865
866 request_resource(&iomem_resource, &video_ram_resource);
867
868 {
869 unsigned i;
870 /* request I/O space for devices used on all i[345]86 PCs */
871 for (i = 0; i < STANDARD_IO_RESOURCES; i++)
872 request_resource(&ioport_resource, &standard_io_resources[i]);
873 }
874
a1e97782 875 e820_setup_gap();
1da177e4
LT
876
877#ifdef CONFIG_GART_IOMMU
5b7b644c 878 iommu_hole_init();
1da177e4
LT
879#endif
880
881#ifdef CONFIG_VT
882#if defined(CONFIG_VGA_CONSOLE)
883 conswitchp = &vga_con;
884#elif defined(CONFIG_DUMMY_CONSOLE)
885 conswitchp = &dummy_con;
886#endif
887#endif
888}
889
e6982c67 890static int __cpuinit get_model_name(struct cpuinfo_x86 *c)
1da177e4
LT
891{
892 unsigned int *v;
893
ebfcaa96 894 if (c->extended_cpuid_level < 0x80000004)
1da177e4
LT
895 return 0;
896
897 v = (unsigned int *) c->x86_model_id;
898 cpuid(0x80000002, &v[0], &v[1], &v[2], &v[3]);
899 cpuid(0x80000003, &v[4], &v[5], &v[6], &v[7]);
900 cpuid(0x80000004, &v[8], &v[9], &v[10], &v[11]);
901 c->x86_model_id[48] = 0;
902 return 1;
903}
904
905
e6982c67 906static void __cpuinit display_cacheinfo(struct cpuinfo_x86 *c)
1da177e4
LT
907{
908 unsigned int n, dummy, eax, ebx, ecx, edx;
909
ebfcaa96 910 n = c->extended_cpuid_level;
1da177e4
LT
911
912 if (n >= 0x80000005) {
913 cpuid(0x80000005, &dummy, &ebx, &ecx, &edx);
914 printk(KERN_INFO "CPU: L1 I Cache: %dK (%d bytes/line), D cache %dK (%d bytes/line)\n",
915 edx>>24, edx&0xFF, ecx>>24, ecx&0xFF);
916 c->x86_cache_size=(ecx>>24)+(edx>>24);
917 /* On K8 L1 TLB is inclusive, so don't count it */
918 c->x86_tlbsize = 0;
919 }
920
921 if (n >= 0x80000006) {
922 cpuid(0x80000006, &dummy, &ebx, &ecx, &edx);
923 ecx = cpuid_ecx(0x80000006);
924 c->x86_cache_size = ecx >> 16;
925 c->x86_tlbsize += ((ebx >> 16) & 0xfff) + (ebx & 0xfff);
926
927 printk(KERN_INFO "CPU: L2 Cache: %dK (%d bytes/line)\n",
928 c->x86_cache_size, ecx & 0xFF);
929 }
930
931 if (n >= 0x80000007)
932 cpuid(0x80000007, &dummy, &dummy, &dummy, &c->x86_power);
933 if (n >= 0x80000008) {
934 cpuid(0x80000008, &eax, &dummy, &dummy, &dummy);
935 c->x86_virt_bits = (eax >> 8) & 0xff;
936 c->x86_phys_bits = eax & 0xff;
937 }
938}
939
3f098c26
AK
940#ifdef CONFIG_NUMA
941static int nearby_node(int apicid)
942{
943 int i;
944 for (i = apicid - 1; i >= 0; i--) {
945 int node = apicid_to_node[i];
946 if (node != NUMA_NO_NODE && node_online(node))
947 return node;
948 }
949 for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) {
950 int node = apicid_to_node[i];
951 if (node != NUMA_NO_NODE && node_online(node))
952 return node;
953 }
954 return first_node(node_online_map); /* Shouldn't happen */
955}
956#endif
957
63518644
AK
958/*
959 * On a AMD dual core setup the lower bits of the APIC id distingush the cores.
960 * Assumes number of cores is a power of two.
961 */
962static void __init amd_detect_cmp(struct cpuinfo_x86 *c)
963{
964#ifdef CONFIG_SMP
2942283e 965 int cpu = smp_processor_id();
b41e2939 966 unsigned bits;
3f098c26
AK
967#ifdef CONFIG_NUMA
968 int node = 0;
60c1bc82 969 unsigned apicid = hard_smp_processor_id();
3f098c26 970#endif
b41e2939
AK
971
972 bits = 0;
94605eff 973 while ((1 << bits) < c->x86_max_cores)
b41e2939
AK
974 bits++;
975
976 /* Low order bits define the core id (index of core in socket) */
977 cpu_core_id[cpu] = phys_proc_id[cpu] & ((1 << bits)-1);
978 /* Convert the APIC ID into the socket ID */
60c1bc82 979 phys_proc_id[cpu] = phys_pkg_id(bits);
63518644
AK
980
981#ifdef CONFIG_NUMA
3f098c26
AK
982 node = phys_proc_id[cpu];
983 if (apicid_to_node[apicid] != NUMA_NO_NODE)
984 node = apicid_to_node[apicid];
985 if (!node_online(node)) {
986 /* Two possibilities here:
987 - The CPU is missing memory and no node was created.
988 In that case try picking one from a nearby CPU
989 - The APIC IDs differ from the HyperTransport node IDs
990 which the K8 northbridge parsing fills in.
991 Assume they are all increased by a constant offset,
992 but in the same order as the HT nodeids.
993 If that doesn't result in a usable node fall back to the
994 path for the previous case. */
995 int ht_nodeid = apicid - (phys_proc_id[0] << bits);
996 if (ht_nodeid >= 0 &&
997 apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
998 node = apicid_to_node[ht_nodeid];
999 /* Pick a nearby node */
1000 if (!node_online(node))
1001 node = nearby_node(apicid);
1002 }
69d81fcd 1003 numa_set_node(cpu, node);
3f098c26 1004
77d910f5
AK
1005 printk(KERN_INFO "CPU %d/%x(%d) -> Node %d -> Core %d\n",
1006 cpu, apicid, c->x86_max_cores, node, cpu_core_id[cpu]);
63518644 1007#endif
63518644
AK
1008#endif
1009}
1da177e4
LT
1010
1011static int __init init_amd(struct cpuinfo_x86 *c)
1012{
1013 int r;
7bcd3f34 1014 unsigned level;
1da177e4 1015
bc5e8fdf
LT
1016#ifdef CONFIG_SMP
1017 unsigned long value;
1018
7d318d77
AK
1019 /*
1020 * Disable TLB flush filter by setting HWCR.FFDIS on K8
1021 * bit 6 of msr C001_0015
1022 *
1023 * Errata 63 for SH-B3 steppings
1024 * Errata 122 for all steppings (F+ have it disabled by default)
1025 */
1026 if (c->x86 == 15) {
1027 rdmsrl(MSR_K8_HWCR, value);
1028 value |= 1 << 6;
1029 wrmsrl(MSR_K8_HWCR, value);
1030 }
bc5e8fdf
LT
1031#endif
1032
1da177e4
LT
1033 /* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
1034 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */
1035 clear_bit(0*32+31, &c->x86_capability);
1036
7bcd3f34
AK
1037 /* On C+ stepping K8 rep microcode works well for copy/memset */
1038 level = cpuid_eax(1);
1039 if (c->x86 == 15 && ((level >= 0x0f48 && level < 0x0f50) || level >= 0x0f58))
1040 set_bit(X86_FEATURE_REP_GOOD, &c->x86_capability);
1041
18bd057b
AK
1042 /* Enable workaround for FXSAVE leak */
1043 if (c->x86 >= 6)
1044 set_bit(X86_FEATURE_FXSAVE_LEAK, &c->x86_capability);
1045
1da177e4
LT
1046 r = get_model_name(c);
1047 if (!r) {
1048 switch (c->x86) {
1049 case 15:
1050 /* Should distinguish Models here, but this is only
1051 a fallback anyways. */
1052 strcpy(c->x86_model_id, "Hammer");
1053 break;
1054 }
1055 }
1056 display_cacheinfo(c);
1057
130951cc
AK
1058 /* c->x86_power is 8000_0007 edx. Bit 8 is constant TSC */
1059 if (c->x86_power & (1<<8))
1060 set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
1061
ebfcaa96 1062 if (c->extended_cpuid_level >= 0x80000008) {
94605eff 1063 c->x86_max_cores = (cpuid_ecx(0x80000008) & 0xff) + 1;
1da177e4 1064
63518644 1065 amd_detect_cmp(c);
1da177e4
LT
1066 }
1067
1068 return r;
1069}
1070
e6982c67 1071static void __cpuinit detect_ht(struct cpuinfo_x86 *c)
1da177e4
LT
1072{
1073#ifdef CONFIG_SMP
1074 u32 eax, ebx, ecx, edx;
94605eff 1075 int index_msb, core_bits;
1da177e4 1076 int cpu = smp_processor_id();
94605eff
SS
1077
1078 cpuid(1, &eax, &ebx, &ecx, &edx);
1079
94605eff 1080
63518644 1081 if (!cpu_has(c, X86_FEATURE_HT) || cpu_has(c, X86_FEATURE_CMP_LEGACY))
1da177e4
LT
1082 return;
1083
1da177e4 1084 smp_num_siblings = (ebx & 0xff0000) >> 16;
94605eff 1085
1da177e4
LT
1086 if (smp_num_siblings == 1) {
1087 printk(KERN_INFO "CPU: Hyper-Threading is disabled\n");
94605eff
SS
1088 } else if (smp_num_siblings > 1 ) {
1089
1da177e4
LT
1090 if (smp_num_siblings > NR_CPUS) {
1091 printk(KERN_WARNING "CPU: Unsupported number of the siblings %d", smp_num_siblings);
1092 smp_num_siblings = 1;
1093 return;
1094 }
94605eff
SS
1095
1096 index_msb = get_count_order(smp_num_siblings);
1da177e4 1097 phys_proc_id[cpu] = phys_pkg_id(index_msb);
94605eff 1098
1da177e4
LT
1099 printk(KERN_INFO "CPU: Physical Processor ID: %d\n",
1100 phys_proc_id[cpu]);
3dd9d514 1101
94605eff 1102 smp_num_siblings = smp_num_siblings / c->x86_max_cores;
3dd9d514 1103
94605eff
SS
1104 index_msb = get_count_order(smp_num_siblings) ;
1105
1106 core_bits = get_count_order(c->x86_max_cores);
3dd9d514 1107
94605eff
SS
1108 cpu_core_id[cpu] = phys_pkg_id(index_msb) &
1109 ((1 << core_bits) - 1);
3dd9d514 1110
94605eff 1111 if (c->x86_max_cores > 1)
3dd9d514
AK
1112 printk(KERN_INFO "CPU: Processor Core ID: %d\n",
1113 cpu_core_id[cpu]);
1da177e4
LT
1114 }
1115#endif
1116}
1117
3dd9d514
AK
1118/*
1119 * find out the number of processor cores on the die
1120 */
e6982c67 1121static int __cpuinit intel_num_cpu_cores(struct cpuinfo_x86 *c)
3dd9d514
AK
1122{
1123 unsigned int eax;
1124
1125 if (c->cpuid_level < 4)
1126 return 1;
1127
1128 __asm__("cpuid"
1129 : "=a" (eax)
1130 : "0" (4), "c" (0)
1131 : "bx", "dx");
1132
1133 if (eax & 0x1f)
1134 return ((eax >> 26) + 1);
1135 else
1136 return 1;
1137}
1138
df0cc26b
AK
1139static void srat_detect_node(void)
1140{
1141#ifdef CONFIG_NUMA
ddea7be0 1142 unsigned node;
df0cc26b
AK
1143 int cpu = smp_processor_id();
1144
1145 /* Don't do the funky fallback heuristics the AMD version employs
1146 for now. */
ddea7be0 1147 node = apicid_to_node[hard_smp_processor_id()];
df0cc26b 1148 if (node == NUMA_NO_NODE)
0d015324 1149 node = first_node(node_online_map);
69d81fcd 1150 numa_set_node(cpu, node);
df0cc26b
AK
1151
1152 if (acpi_numa > 0)
1153 printk(KERN_INFO "CPU %d -> Node %d\n", cpu, node);
1154#endif
1155}
1156
e6982c67 1157static void __cpuinit init_intel(struct cpuinfo_x86 *c)
1da177e4
LT
1158{
1159 /* Cache sizes */
1160 unsigned n;
1161
1162 init_intel_cacheinfo(c);
ebfcaa96 1163 n = c->extended_cpuid_level;
1da177e4
LT
1164 if (n >= 0x80000008) {
1165 unsigned eax = cpuid_eax(0x80000008);
1166 c->x86_virt_bits = (eax >> 8) & 0xff;
1167 c->x86_phys_bits = eax & 0xff;
af9c142d
SL
1168 /* CPUID workaround for Intel 0F34 CPU */
1169 if (c->x86_vendor == X86_VENDOR_INTEL &&
1170 c->x86 == 0xF && c->x86_model == 0x3 &&
1171 c->x86_mask == 0x4)
1172 c->x86_phys_bits = 36;
1da177e4
LT
1173 }
1174
1175 if (c->x86 == 15)
1176 c->x86_cache_alignment = c->x86_clflush_size * 2;
39b3a791
AK
1177 if ((c->x86 == 0xf && c->x86_model >= 0x03) ||
1178 (c->x86 == 0x6 && c->x86_model >= 0x0e))
c29601e9 1179 set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
c818a181 1180 set_bit(X86_FEATURE_SYNC_RDTSC, &c->x86_capability);
94605eff 1181 c->x86_max_cores = intel_num_cpu_cores(c);
df0cc26b
AK
1182
1183 srat_detect_node();
1da177e4
LT
1184}
1185
672289e9 1186static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c)
1da177e4
LT
1187{
1188 char *v = c->x86_vendor_id;
1189
1190 if (!strcmp(v, "AuthenticAMD"))
1191 c->x86_vendor = X86_VENDOR_AMD;
1192 else if (!strcmp(v, "GenuineIntel"))
1193 c->x86_vendor = X86_VENDOR_INTEL;
1194 else
1195 c->x86_vendor = X86_VENDOR_UNKNOWN;
1196}
1197
1198struct cpu_model_info {
1199 int vendor;
1200 int family;
1201 char *model_names[16];
1202};
1203
1204/* Do some early cpuid on the boot CPU to get some parameter that are
1205 needed before check_bugs. Everything advanced is in identify_cpu
1206 below. */
e6982c67 1207void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c)
1da177e4
LT
1208{
1209 u32 tfms;
1210
1211 c->loops_per_jiffy = loops_per_jiffy;
1212 c->x86_cache_size = -1;
1213 c->x86_vendor = X86_VENDOR_UNKNOWN;
1214 c->x86_model = c->x86_mask = 0; /* So far unknown... */
1215 c->x86_vendor_id[0] = '\0'; /* Unset */
1216 c->x86_model_id[0] = '\0'; /* Unset */
1217 c->x86_clflush_size = 64;
1218 c->x86_cache_alignment = c->x86_clflush_size;
94605eff 1219 c->x86_max_cores = 1;
ebfcaa96 1220 c->extended_cpuid_level = 0;
1da177e4
LT
1221 memset(&c->x86_capability, 0, sizeof c->x86_capability);
1222
1223 /* Get vendor name */
1224 cpuid(0x00000000, (unsigned int *)&c->cpuid_level,
1225 (unsigned int *)&c->x86_vendor_id[0],
1226 (unsigned int *)&c->x86_vendor_id[8],
1227 (unsigned int *)&c->x86_vendor_id[4]);
1228
1229 get_cpu_vendor(c);
1230
1231 /* Initialize the standard set of capabilities */
1232 /* Note that the vendor-specific code below might override */
1233
1234 /* Intel-defined flags: level 0x00000001 */
1235 if (c->cpuid_level >= 0x00000001) {
1236 __u32 misc;
1237 cpuid(0x00000001, &tfms, &misc, &c->x86_capability[4],
1238 &c->x86_capability[0]);
1239 c->x86 = (tfms >> 8) & 0xf;
1240 c->x86_model = (tfms >> 4) & 0xf;
1241 c->x86_mask = tfms & 0xf;
f5f786d0 1242 if (c->x86 == 0xf)
1da177e4 1243 c->x86 += (tfms >> 20) & 0xff;
f5f786d0 1244 if (c->x86 >= 0x6)
1da177e4 1245 c->x86_model += ((tfms >> 16) & 0xF) << 4;
1da177e4
LT
1246 if (c->x86_capability[0] & (1<<19))
1247 c->x86_clflush_size = ((misc >> 8) & 0xff) * 8;
1da177e4
LT
1248 } else {
1249 /* Have CPUID level 0 only - unheard of */
1250 c->x86 = 4;
1251 }
a158608b
AK
1252
1253#ifdef CONFIG_SMP
b41e2939 1254 phys_proc_id[smp_processor_id()] = (cpuid_ebx(1) >> 24) & 0xff;
a158608b 1255#endif
1da177e4
LT
1256}
1257
1258/*
1259 * This does the hard work of actually picking apart the CPU stuff...
1260 */
e6982c67 1261void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
1da177e4
LT
1262{
1263 int i;
1264 u32 xlvl;
1265
1266 early_identify_cpu(c);
1267
1268 /* AMD-defined flags: level 0x80000001 */
1269 xlvl = cpuid_eax(0x80000000);
ebfcaa96 1270 c->extended_cpuid_level = xlvl;
1da177e4
LT
1271 if ((xlvl & 0xffff0000) == 0x80000000) {
1272 if (xlvl >= 0x80000001) {
1273 c->x86_capability[1] = cpuid_edx(0x80000001);
5b7abc6f 1274 c->x86_capability[6] = cpuid_ecx(0x80000001);
1da177e4
LT
1275 }
1276 if (xlvl >= 0x80000004)
1277 get_model_name(c); /* Default name */
1278 }
1279
1280 /* Transmeta-defined flags: level 0x80860001 */
1281 xlvl = cpuid_eax(0x80860000);
1282 if ((xlvl & 0xffff0000) == 0x80860000) {
1283 /* Don't set x86_cpuid_level here for now to not confuse. */
1284 if (xlvl >= 0x80860001)
1285 c->x86_capability[2] = cpuid_edx(0x80860001);
1286 }
1287
1e9f28fa
SS
1288 c->apicid = phys_pkg_id(0);
1289
1da177e4
LT
1290 /*
1291 * Vendor-specific initialization. In this section we
1292 * canonicalize the feature flags, meaning if there are
1293 * features a certain CPU supports which CPUID doesn't
1294 * tell us, CPUID claiming incorrect flags, or other bugs,
1295 * we handle them here.
1296 *
1297 * At the end of this section, c->x86_capability better
1298 * indicate the features this CPU genuinely supports!
1299 */
1300 switch (c->x86_vendor) {
1301 case X86_VENDOR_AMD:
1302 init_amd(c);
1303 break;
1304
1305 case X86_VENDOR_INTEL:
1306 init_intel(c);
1307 break;
1308
1309 case X86_VENDOR_UNKNOWN:
1310 default:
1311 display_cacheinfo(c);
1312 break;
1313 }
1314
1315 select_idle_routine(c);
1316 detect_ht(c);
1da177e4
LT
1317
1318 /*
1319 * On SMP, boot_cpu_data holds the common feature set between
1320 * all CPUs; so make sure that we indicate which features are
1321 * common between the CPUs. The first time this routine gets
1322 * executed, c == &boot_cpu_data.
1323 */
1324 if (c != &boot_cpu_data) {
1325 /* AND the already accumulated flags with these */
1326 for (i = 0 ; i < NCAPINTS ; i++)
1327 boot_cpu_data.x86_capability[i] &= c->x86_capability[i];
1328 }
1329
1330#ifdef CONFIG_X86_MCE
1331 mcheck_init(c);
1332#endif
3b520b23
SL
1333 if (c == &boot_cpu_data)
1334 mtrr_bp_init();
1335 else
1336 mtrr_ap_init();
1da177e4 1337#ifdef CONFIG_NUMA
3019e8eb 1338 numa_add_cpu(smp_processor_id());
1da177e4
LT
1339#endif
1340}
1341
1342
e6982c67 1343void __cpuinit print_cpu_info(struct cpuinfo_x86 *c)
1da177e4
LT
1344{
1345 if (c->x86_model_id[0])
1346 printk("%s", c->x86_model_id);
1347
1348 if (c->x86_mask || c->cpuid_level >= 0)
1349 printk(" stepping %02x\n", c->x86_mask);
1350 else
1351 printk("\n");
1352}
1353
1354/*
1355 * Get CPU information for use by the procfs.
1356 */
1357
1358static int show_cpuinfo(struct seq_file *m, void *v)
1359{
1360 struct cpuinfo_x86 *c = v;
1361
1362 /*
1363 * These flag bits must match the definitions in <asm/cpufeature.h>.
1364 * NULL means this bit is undefined or reserved; either way it doesn't
1365 * have meaning as far as Linux is concerned. Note that it's important
1366 * to realize there is a difference between this table and CPUID -- if
1367 * applications want to get the raw CPUID data, they should access
1368 * /dev/cpu/<cpu_nr>/cpuid instead.
1369 */
1370 static char *x86_cap_flags[] = {
1371 /* Intel-defined */
1372 "fpu", "vme", "de", "pse", "tsc", "msr", "pae", "mce",
1373 "cx8", "apic", NULL, "sep", "mtrr", "pge", "mca", "cmov",
1374 "pat", "pse36", "pn", "clflush", NULL, "dts", "acpi", "mmx",
1375 "fxsr", "sse", "sse2", "ss", "ht", "tm", "ia64", NULL,
1376
1377 /* AMD-defined */
3c3b73b6 1378 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1da177e4
LT
1379 NULL, NULL, NULL, "syscall", NULL, NULL, NULL, NULL,
1380 NULL, NULL, NULL, NULL, "nx", NULL, "mmxext", NULL,
3f98bc49 1381 NULL, "fxsr_opt", "rdtscp", NULL, NULL, "lm", "3dnowext", "3dnow",
1da177e4
LT
1382
1383 /* Transmeta-defined */
1384 "recovery", "longrun", NULL, "lrti", NULL, NULL, NULL, NULL,
1385 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1386 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1387 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1388
1389 /* Other (Linux-defined) */
622dcaf9 1390 "cxmmx", NULL, "cyrix_arr", "centaur_mcr", NULL,
c29601e9 1391 "constant_tsc", NULL, NULL,
1da177e4
LT
1392 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1393 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1394 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1395
1396 /* Intel-defined (#2) */
9d95dd84 1397 "pni", NULL, NULL, "monitor", "ds_cpl", "vmx", "smx", "est",
1da177e4
LT
1398 "tm2", NULL, "cid", NULL, NULL, "cx16", "xtpr", NULL,
1399 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1400 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1401
5b7abc6f
PA
1402 /* VIA/Cyrix/Centaur-defined */
1403 NULL, NULL, "rng", "rng_en", NULL, NULL, "ace", "ace_en",
1404 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1405 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1406 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1407
1da177e4 1408 /* AMD-defined (#2) */
3f98bc49 1409 "lahf_lm", "cmp_legacy", "svm", NULL, "cr8_legacy", NULL, NULL, NULL,
1da177e4
LT
1410 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1411 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
5b7abc6f 1412 NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
1da177e4
LT
1413 };
1414 static char *x86_power_flags[] = {
1415 "ts", /* temperature sensor */
1416 "fid", /* frequency id control */
1417 "vid", /* voltage id control */
1418 "ttp", /* thermal trip */
1419 "tm",
3f98bc49
AK
1420 "stc",
1421 NULL,
39b3a791 1422 /* nothing */ /* constant_tsc - moved to flags */
1da177e4
LT
1423 };
1424
1425
1426#ifdef CONFIG_SMP
1427 if (!cpu_online(c-cpu_data))
1428 return 0;
1429#endif
1430
1431 seq_printf(m,"processor\t: %u\n"
1432 "vendor_id\t: %s\n"
1433 "cpu family\t: %d\n"
1434 "model\t\t: %d\n"
1435 "model name\t: %s\n",
1436 (unsigned)(c-cpu_data),
1437 c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown",
1438 c->x86,
1439 (int)c->x86_model,
1440 c->x86_model_id[0] ? c->x86_model_id : "unknown");
1441
1442 if (c->x86_mask || c->cpuid_level >= 0)
1443 seq_printf(m, "stepping\t: %d\n", c->x86_mask);
1444 else
1445 seq_printf(m, "stepping\t: unknown\n");
1446
1447 if (cpu_has(c,X86_FEATURE_TSC)) {
95235ca2
VP
1448 unsigned int freq = cpufreq_quick_get((unsigned)(c-cpu_data));
1449 if (!freq)
1450 freq = cpu_khz;
1da177e4 1451 seq_printf(m, "cpu MHz\t\t: %u.%03u\n",
95235ca2 1452 freq / 1000, (freq % 1000));
1da177e4
LT
1453 }
1454
1455 /* Cache size */
1456 if (c->x86_cache_size >= 0)
1457 seq_printf(m, "cache size\t: %d KB\n", c->x86_cache_size);
1458
1459#ifdef CONFIG_SMP
94605eff 1460 if (smp_num_siblings * c->x86_max_cores > 1) {
db468681
AK
1461 int cpu = c - cpu_data;
1462 seq_printf(m, "physical id\t: %d\n", phys_proc_id[cpu]);
94605eff 1463 seq_printf(m, "siblings\t: %d\n", cpus_weight(cpu_core_map[cpu]));
d31ddaa1 1464 seq_printf(m, "core id\t\t: %d\n", cpu_core_id[cpu]);
94605eff 1465 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
db468681 1466 }
1da177e4
LT
1467#endif
1468
1469 seq_printf(m,
1470 "fpu\t\t: yes\n"
1471 "fpu_exception\t: yes\n"
1472 "cpuid level\t: %d\n"
1473 "wp\t\t: yes\n"
1474 "flags\t\t:",
1475 c->cpuid_level);
1476
1477 {
1478 int i;
1479 for ( i = 0 ; i < 32*NCAPINTS ; i++ )
3d1712c9 1480 if (cpu_has(c, i) && x86_cap_flags[i] != NULL)
1da177e4
LT
1481 seq_printf(m, " %s", x86_cap_flags[i]);
1482 }
1483
1484 seq_printf(m, "\nbogomips\t: %lu.%02lu\n",
1485 c->loops_per_jiffy/(500000/HZ),
1486 (c->loops_per_jiffy/(5000/HZ)) % 100);
1487
1488 if (c->x86_tlbsize > 0)
1489 seq_printf(m, "TLB size\t: %d 4K pages\n", c->x86_tlbsize);
1490 seq_printf(m, "clflush size\t: %d\n", c->x86_clflush_size);
1491 seq_printf(m, "cache_alignment\t: %d\n", c->x86_cache_alignment);
1492
1493 seq_printf(m, "address sizes\t: %u bits physical, %u bits virtual\n",
1494 c->x86_phys_bits, c->x86_virt_bits);
1495
1496 seq_printf(m, "power management:");
1497 {
1498 unsigned i;
1499 for (i = 0; i < 32; i++)
1500 if (c->x86_power & (1 << i)) {
3f98bc49
AK
1501 if (i < ARRAY_SIZE(x86_power_flags) &&
1502 x86_power_flags[i])
1503 seq_printf(m, "%s%s",
1504 x86_power_flags[i][0]?" ":"",
1505 x86_power_flags[i]);
1da177e4
LT
1506 else
1507 seq_printf(m, " [%d]", i);
1508 }
1509 }
1da177e4 1510
d31ddaa1 1511 seq_printf(m, "\n\n");
1da177e4
LT
1512
1513 return 0;
1514}
1515
1516static void *c_start(struct seq_file *m, loff_t *pos)
1517{
1518 return *pos < NR_CPUS ? cpu_data + *pos : NULL;
1519}
1520
1521static void *c_next(struct seq_file *m, void *v, loff_t *pos)
1522{
1523 ++*pos;
1524 return c_start(m, pos);
1525}
1526
1527static void c_stop(struct seq_file *m, void *v)
1528{
1529}
1530
1531struct seq_operations cpuinfo_op = {
1532 .start =c_start,
1533 .next = c_next,
1534 .stop = c_stop,
1535 .show = show_cpuinfo,
1536};
e9928674 1537
160bd18e
MP
1538#ifdef CONFIG_INPUT_PCSPKR
1539#include <linux/platform_device.h>
1540static __init int add_pcspkr(void)
1541{
1542 struct platform_device *pd;
1543 int ret;
1544
1545 pd = platform_device_alloc("pcspkr", -1);
1546 if (!pd)
1547 return -ENOMEM;
1548
1549 ret = platform_device_add(pd);
1550 if (ret)
1551 platform_device_put(pd);
1552
1553 return ret;
1554}
1555device_initcall(add_pcspkr);
1556#endif