x86, mm: Separate out calculate_table_space_size()
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / arch / x86 / kernel / setup.c
CommitLineData
1da177e4 1/*
1da177e4
LT
2 * Copyright (C) 1995 Linus Torvalds
3 *
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
5 *
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
8 *
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
11 *
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
14 *
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
17 *
18 */
19
20/*
21 * This file handles the architecture-dependent parts of initialization
22 */
23
24#include <linux/sched.h>
25#include <linux/mm.h>
05b79bdc 26#include <linux/mmzone.h>
894673ee 27#include <linux/screen_info.h>
1da177e4
LT
28#include <linux/ioport.h>
29#include <linux/acpi.h>
efafc8b2 30#include <linux/sfi.h>
1da177e4
LT
31#include <linux/apm_bios.h>
32#include <linux/initrd.h>
33#include <linux/bootmem.h>
72d7c3b3 34#include <linux/memblock.h>
1da177e4
LT
35#include <linux/seq_file.h>
36#include <linux/console.h>
1da177e4
LT
37#include <linux/root_dev.h>
38#include <linux/highmem.h>
39#include <linux/module.h>
40#include <linux/efi.h>
41#include <linux/init.h>
42#include <linux/edd.h>
138fe4e0 43#include <linux/iscsi_ibft.h>
1da177e4 44#include <linux/nodemask.h>
1bc3b91a 45#include <linux/kexec.h>
e9928674 46#include <linux/dmi.h>
22a9835c 47#include <linux/pfn.h>
376ff035 48#include <linux/pci.h>
46d671b5 49#include <asm/pci-direct.h>
f212ec4b 50#include <linux/init_ohci1394_dma.h>
790c73f6 51#include <linux/kvm_para.h>
0a2b9a6e 52#include <linux/dma-contiguous.h>
1bc3b91a 53
46d671b5
YL
54#include <linux/errno.h>
55#include <linux/kernel.h>
56#include <linux/stddef.h>
57#include <linux/unistd.h>
58#include <linux/ptrace.h>
46d671b5
YL
59#include <linux/user.h>
60#include <linux/delay.h>
46d671b5
YL
61
62#include <linux/kallsyms.h>
46d671b5
YL
63#include <linux/cpufreq.h>
64#include <linux/dma-mapping.h>
65#include <linux/ctype.h>
66#include <linux/uaccess.h>
67
68#include <linux/percpu.h>
69#include <linux/crash_dump.h>
69575d38 70#include <linux/tboot.h>
b3c869d3 71#include <linux/jiffies.h>
46d671b5 72
1da177e4 73#include <video/edid.h>
1bc3b91a 74
093af8d7 75#include <asm/mtrr.h>
9635b47d 76#include <asm/apic.h>
084ee1c6 77#include <asm/realmode.h>
1da177e4
LT
78#include <asm/e820.h>
79#include <asm/mpspec.h>
80#include <asm/setup.h>
55f26239 81#include <asm/efi.h>
8e6dafd6
IM
82#include <asm/timer.h>
83#include <asm/i8259.h>
1da177e4 84#include <asm/sections.h>
1c6e5503 85#include <asm/dmi.h>
1da177e4
LT
86#include <asm/io_apic.h>
87#include <asm/ist.h>
1164dd00 88#include <asm/setup_arch.h>
ce3fe6b2 89#include <asm/bios_ebda.h>
00bf4098 90#include <asm/cacheflush.h>
2fde61fd 91#include <asm/processor.h>
cc9f7a0c 92#include <asm/bugs.h>
1da177e4 93
46d671b5 94#include <asm/vsyscall.h>
6e5385d4 95#include <asm/cpu.h>
46d671b5
YL
96#include <asm/desc.h>
97#include <asm/dma.h>
46a7fa27 98#include <asm/iommu.h>
1d9b16d1 99#include <asm/gart.h>
46d671b5
YL
100#include <asm/mmu_context.h>
101#include <asm/proto.h>
102
46d671b5 103#include <asm/paravirt.h>
88b094fb 104#include <asm/hypervisor.h>
fd699c76 105#include <asm/olpc_ofw.h>
46d671b5
YL
106
107#include <asm/percpu.h>
46d671b5
YL
108#include <asm/topology.h>
109#include <asm/apicdef.h>
23ac4ae8 110#include <asm/amd_nb.h>
55f26239
YL
111#ifdef CONFIG_X86_64
112#include <asm/numa_64.h>
113#endif
a2202aa2 114#include <asm/mce.h>
f49aa448 115#include <asm/alternative.h>
da6b737b 116#include <asm/prom.h>
46d671b5 117
2b72394e
PE
118/*
119 * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
120 * The direct mapping extends to max_pfn_mapped, so that we can directly access
121 * apertures, ACPI and other tables without having to play with fixmaps.
122 */
123unsigned long max_low_pfn_mapped;
124unsigned long max_pfn_mapped;
125
e808bae2 126#ifdef CONFIG_DMI
796216a5 127RESERVE_BRK(dmi_alloc, 65536);
e808bae2 128#endif
796216a5 129
c0b5842a 130
93dbda7c
JF
131static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
132unsigned long _brk_end = (unsigned long)__brk_base;
133
c0b5842a
IM
134#ifdef CONFIG_X86_64
135int default_cpu_present_to_apicid(int mps_cpu)
136{
137 return __default_cpu_present_to_apicid(mps_cpu);
138}
139
e11dadab 140int default_check_phys_apicid_present(int phys_apicid)
c0b5842a 141{
e11dadab 142 return __default_check_phys_apicid_present(phys_apicid);
c0b5842a
IM
143}
144#endif
145
217b8ce8
YL
146#ifndef CONFIG_DEBUG_BOOT_PARAMS
147struct boot_params __initdata boot_params;
148#else
149struct boot_params boot_params;
150#endif
151
1da177e4
LT
152/*
153 * Machine setup..
154 */
c9cce83d
BW
155static struct resource data_resource = {
156 .name = "Kernel data",
157 .start = 0,
158 .end = 0,
159 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
160};
161
162static struct resource code_resource = {
163 .name = "Kernel code",
164 .start = 0,
165 .end = 0,
166 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
167};
168
169static struct resource bss_resource = {
170 .name = "Kernel bss",
171 .start = 0,
172 .end = 0,
173 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
174};
175
7dea23ec
YL
176
177#ifdef CONFIG_X86_32
1da177e4 178/* cpu data as detected by the assembly code in head.S */
7dea23ec 179struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
1da177e4 180/* common cpu data for all cpus */
7dea23ec 181struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
129f6946 182EXPORT_SYMBOL(boot_cpu_data);
1da177e4 183
0c254e38
AS
184unsigned int def_to_bigsmp;
185
1da177e4
LT
186/* for MCA, but anyone else can use it if they want */
187unsigned int machine_id;
188unsigned int machine_submodel_id;
189unsigned int BIOS_revision;
1da177e4 190
7dea23ec
YL
191struct apm_info apm_info;
192EXPORT_SYMBOL(apm_info);
193
194#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
195 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
196struct ist_info ist_info;
197EXPORT_SYMBOL(ist_info);
198#else
199struct ist_info ist_info;
200#endif
201
202#else
ed26dbe5
JF
203struct cpuinfo_x86 boot_cpu_data __read_mostly = {
204 .x86_phys_bits = MAX_PHYSMEM_BITS,
205};
7dea23ec
YL
206EXPORT_SYMBOL(boot_cpu_data);
207#endif
208
209
210#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
211unsigned long mmu_cr4_features;
212#else
213unsigned long mmu_cr4_features = X86_CR4_PAE;
214#endif
215
5031296c
PA
216/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
217int bootloader_type, bootloader_version;
1da177e4 218
1da177e4
LT
219/*
220 * Setup options
221 */
1da177e4 222struct screen_info screen_info;
129f6946 223EXPORT_SYMBOL(screen_info);
1da177e4 224struct edid_info edid_info;
5e518d76 225EXPORT_SYMBOL_GPL(edid_info);
1da177e4 226
1da177e4
LT
227extern int root_mountflags;
228
e44b7b75 229unsigned long saved_video_mode;
1da177e4 230
cf8fa920 231#define RAMDISK_IMAGE_START_MASK 0x07FF
1da177e4 232#define RAMDISK_PROMPT_FLAG 0x8000
cf8fa920 233#define RAMDISK_LOAD_FLAG 0x4000
1da177e4 234
4e498b66 235static char __initdata command_line[COMMAND_LINE_SIZE];
516cbf37
TB
236#ifdef CONFIG_CMDLINE_BOOL
237static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
238#endif
1da177e4 239
1da177e4
LT
240#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
241struct edd edd;
242#ifdef CONFIG_EDD_MODULE
243EXPORT_SYMBOL(edd);
244#endif
245/**
246 * copy_edd() - Copy the BIOS EDD information
247 * from boot_params into a safe place.
248 *
249 */
9eaa192d 250static inline void __init copy_edd(void)
1da177e4 251{
30c82645
PA
252 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
253 sizeof(edd.mbr_signature));
254 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
255 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
256 edd.edd_info_nr = boot_params.eddbuf_entries;
1da177e4
LT
257}
258#else
9eaa192d 259static inline void __init copy_edd(void)
1da177e4
LT
260{
261}
262#endif
263
5368a2be
PA
264void * __init extend_brk(size_t size, size_t align)
265{
266 size_t mask = align - 1;
267 void *ret;
268
269 BUG_ON(_brk_start == 0);
270 BUG_ON(align & mask);
271
272 _brk_end = (_brk_end + mask) & ~mask;
273 BUG_ON((char *)(_brk_end + size) > __brk_limit);
274
275 ret = (void *)_brk_end;
276 _brk_end += size;
277
278 memset(ret, 0, size);
279
280 return ret;
281}
282
854c879f
PE
283#ifdef CONFIG_X86_64
284static void __init init_gbpages(void)
285{
286 if (direct_gbpages && cpu_has_gbpages)
287 printk(KERN_INFO "Using GB pages for direct mapping\n");
288 else
289 direct_gbpages = 0;
290}
291#else
292static inline void init_gbpages(void)
293{
294}
e5f15b45 295static void __init cleanup_highmap(void)
f005fe12
YL
296{
297}
854c879f
PE
298#endif
299
5368a2be
PA
300static void __init reserve_brk(void)
301{
302 if (_brk_end > _brk_start)
24aa0788
TH
303 memblock_reserve(__pa(_brk_start),
304 __pa(_brk_end) - __pa(_brk_start));
5368a2be
PA
305
306 /* Mark brk area as locked down and no longer taking any
307 new allocations */
308 _brk_start = 0;
309}
310
cf8fa920
PA
311#ifdef CONFIG_BLK_DEV_INITRD
312
eb1379cb
YL
313#define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
314static void __init relocate_initrd(void)
cf8fa920 315{
c967da6a 316 /* Assume only end is not page aligned */
ba5b14cc
YL
317 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
318 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
c967da6a 319 u64 area_size = PAGE_ALIGN(ramdisk_size);
8c5dd8f4 320 u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
ba5b14cc 321 u64 ramdisk_here;
eb1379cb
YL
322 unsigned long slop, clen, mapaddr;
323 char *p, *q;
cf8fa920
PA
324
325 /* We need to move the initrd down into lowmem */
a9ce6bc1 326 ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
4e29684c 327 PAGE_SIZE);
cf8fa920 328
1f5026a7 329 if (!ramdisk_here)
3945e2c9
YL
330 panic("Cannot find place for new RAMDISK of size %lld\n",
331 ramdisk_size);
332
cf8fa920
PA
333 /* Note: this includes all the lowmem currently occupied by
334 the initrd, we rely on that fact to keep the data intact. */
24aa0788 335 memblock_reserve(ramdisk_here, area_size);
cf8fa920
PA
336 initrd_start = ramdisk_here + PAGE_OFFSET;
337 initrd_end = initrd_start + ramdisk_size;
365811d6
BH
338 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
339 ramdisk_here, ramdisk_here + ramdisk_size - 1);
cf8fa920 340
cf8fa920
PA
341 q = (char *)initrd_start;
342
343 /* Copy any lowmem portion of the initrd */
344 if (ramdisk_image < end_of_lowmem) {
345 clen = end_of_lowmem - ramdisk_image;
346 p = (char *)__va(ramdisk_image);
347 memcpy(q, p, clen);
348 q += clen;
349 ramdisk_image += clen;
350 ramdisk_size -= clen;
351 }
352
353 /* Copy the highmem portion of the initrd */
354 while (ramdisk_size) {
355 slop = ramdisk_image & ~PAGE_MASK;
356 clen = ramdisk_size;
357 if (clen > MAX_MAP_CHUNK-slop)
358 clen = MAX_MAP_CHUNK-slop;
359 mapaddr = ramdisk_image & PAGE_MASK;
88b4c146 360 p = early_memremap(mapaddr, clen+slop);
cf8fa920 361 memcpy(q, p+slop, clen);
beacfaac 362 early_iounmap(p, clen+slop);
cf8fa920
PA
363 q += clen;
364 ramdisk_image += clen;
365 ramdisk_size -= clen;
366 }
a4c81cf6 367 /* high pages is not converted by early_res_to_bootmem */
ba5b14cc
YL
368 ramdisk_image = boot_params.hdr.ramdisk_image;
369 ramdisk_size = boot_params.hdr.ramdisk_size;
365811d6
BH
370 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
371 " [mem %#010llx-%#010llx]\n",
ba5b14cc
YL
372 ramdisk_image, ramdisk_image + ramdisk_size - 1,
373 ramdisk_here, ramdisk_here + ramdisk_size - 1);
eb1379cb 374}
9a27f5c5 375
eb1379cb
YL
376static void __init reserve_initrd(void)
377{
c967da6a 378 /* Assume only end is not page aligned */
eb1379cb
YL
379 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
380 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
c967da6a 381 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
8c5dd8f4 382 u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
eb1379cb
YL
383
384 if (!boot_params.hdr.type_of_loader ||
385 !ramdisk_image || !ramdisk_size)
386 return; /* No initrd provided by bootloader */
387
388 initrd_start = 0;
389
390 if (ramdisk_size >= (end_of_lowmem>>1)) {
ab7b64e9
PJ
391 panic("initrd too large to handle, "
392 "disabling initrd (%lld needed, %lld available)\n",
393 ramdisk_size, end_of_lowmem>>1);
eb1379cb
YL
394 }
395
365811d6
BH
396 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
397 ramdisk_end - 1);
eb1379cb
YL
398
399
400 if (ramdisk_end <= end_of_lowmem) {
401 /* All in lowmem, easy case */
402 /*
403 * don't need to reserve again, already reserved early
404 * in i386_start_kernel
405 */
406 initrd_start = ramdisk_image + PAGE_OFFSET;
407 initrd_end = initrd_start + ramdisk_size;
408 return;
409 }
410
eb1379cb 411 relocate_initrd();
8c5dd8f4 412
24aa0788 413 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
cf8fa920 414}
225c37d7 415#else
eb1379cb 416static void __init reserve_initrd(void)
225c37d7
YL
417{
418}
cf8fa920
PA
419#endif /* CONFIG_BLK_DEV_INITRD */
420
29f784e3 421static void __init parse_setup_data(void)
257b0fde
YL
422{
423 struct setup_data *data;
424 u64 pa_data;
425
426 if (boot_params.hdr.version < 0x0209)
427 return;
428 pa_data = boot_params.hdr.setup_data;
429 while (pa_data) {
f1c2b357
SAS
430 u32 data_len, map_len;
431
432 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
433 (u64)sizeof(struct setup_data));
434 data = early_memremap(pa_data, map_len);
435 data_len = data->len + sizeof(struct setup_data);
436 if (data_len > map_len) {
437 early_iounmap(data, map_len);
438 data = early_memremap(pa_data, data_len);
439 map_len = data_len;
440 }
441
257b0fde
YL
442 switch (data->type) {
443 case SETUP_E820_EXT:
f1c2b357 444 parse_e820_ext(data);
257b0fde 445 break;
da6b737b
SAS
446 case SETUP_DTB:
447 add_dtb(pa_data);
257b0fde
YL
448 break;
449 default:
450 break;
451 }
257b0fde 452 pa_data = data->next;
f1c2b357 453 early_iounmap(data, map_len);
257b0fde
YL
454 }
455}
456
a0a0becd 457static void __init e820_reserve_setup_data(void)
28bb2237
YL
458{
459 struct setup_data *data;
460 u64 pa_data;
d9a81b44 461 int found = 0;
28bb2237
YL
462
463 if (boot_params.hdr.version < 0x0209)
464 return;
465 pa_data = boot_params.hdr.setup_data;
466 while (pa_data) {
88b4c146 467 data = early_memremap(pa_data, sizeof(*data));
28bb2237
YL
468 e820_update_range(pa_data, sizeof(*data)+data->len,
469 E820_RAM, E820_RESERVED_KERN);
d9a81b44 470 found = 1;
28bb2237
YL
471 pa_data = data->next;
472 early_iounmap(data, sizeof(*data));
473 }
d9a81b44
YL
474 if (!found)
475 return;
476
28bb2237 477 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
a0a0becd 478 memcpy(&e820_saved, &e820, sizeof(struct e820map));
28bb2237
YL
479 printk(KERN_INFO "extended physical RAM map:\n");
480 e820_print_map("reserve setup_data");
481}
482
a9ce6bc1 483static void __init memblock_x86_reserve_range_setup_data(void)
a0a0becd
YL
484{
485 struct setup_data *data;
486 u64 pa_data;
a0a0becd
YL
487
488 if (boot_params.hdr.version < 0x0209)
489 return;
490 pa_data = boot_params.hdr.setup_data;
491 while (pa_data) {
88b4c146 492 data = early_memremap(pa_data, sizeof(*data));
24aa0788 493 memblock_reserve(pa_data, sizeof(*data) + data->len);
a0a0becd
YL
494 pa_data = data->next;
495 early_iounmap(data, sizeof(*data));
496 }
497}
498
ccb4defa
YL
499/*
500 * --------- Crashkernel reservation ------------------------------
501 */
502
503#ifdef CONFIG_KEXEC
32105f7f 504
7f8595bf
PA
505/*
506 * Keep the crash kernel below this limit. On 32 bits earlier kernels
507 * would limit the kernel to the low 512 MiB due to mapping restrictions.
508 * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
509 * limit once kexec-tools are fixed.
510 */
511#ifdef CONFIG_X86_32
512# define CRASH_KERNEL_ADDR_MAX (512 << 20)
513#else
514# define CRASH_KERNEL_ADDR_MAX (896 << 20)
515#endif
516
29f784e3 517static void __init reserve_crashkernel(void)
ccb4defa
YL
518{
519 unsigned long long total_mem;
520 unsigned long long crash_size, crash_base;
521 int ret;
522
09c71bfd 523 total_mem = memblock_phys_mem_size();
ccb4defa
YL
524
525 ret = parse_crashkernel(boot_command_line, total_mem,
526 &crash_size, &crash_base);
32105f7f
BW
527 if (ret != 0 || crash_size <= 0)
528 return;
529
530 /* 0 means: find the address automatically */
531 if (crash_base <= 0) {
44280733
YL
532 const unsigned long long alignment = 16<<20; /* 16M */
533
9f4c1396 534 /*
7f8595bf 535 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
9f4c1396
YL
536 */
537 crash_base = memblock_find_in_range(alignment,
7f8595bf 538 CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
9f4c1396 539
1f5026a7 540 if (!crash_base) {
44280733 541 pr_info("crashkernel reservation failed - No suitable area found.\n");
ccb4defa
YL
542 return;
543 }
32105f7f 544 } else {
44280733
YL
545 unsigned long long start;
546
9f4c1396
YL
547 start = memblock_find_in_range(crash_base,
548 crash_base + crash_size, crash_size, 1<<20);
44280733
YL
549 if (start != crash_base) {
550 pr_info("crashkernel reservation failed - memory is in use.\n");
ccb4defa
YL
551 return;
552 }
32105f7f 553 }
24aa0788 554 memblock_reserve(crash_base, crash_size);
ccb4defa 555
32105f7f
BW
556 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
557 "for crashkernel (System RAM: %ldMB)\n",
558 (unsigned long)(crash_size >> 20),
559 (unsigned long)(crash_base >> 20),
560 (unsigned long)(total_mem >> 20));
ccb4defa 561
32105f7f
BW
562 crashk_res.start = crash_base;
563 crashk_res.end = crash_base + crash_size - 1;
564 insert_resource(&iomem_resource, &crashk_res);
ccb4defa
YL
565}
566#else
29f784e3 567static void __init reserve_crashkernel(void)
ccb4defa
YL
568{
569}
570#endif
571
bdba0e70
YL
572static struct resource standard_io_resources[] = {
573 { .name = "dma1", .start = 0x00, .end = 0x1f,
574 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
575 { .name = "pic1", .start = 0x20, .end = 0x21,
576 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
577 { .name = "timer0", .start = 0x40, .end = 0x43,
578 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
579 { .name = "timer1", .start = 0x50, .end = 0x53,
580 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
581 { .name = "keyboard", .start = 0x60, .end = 0x60,
582 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
583 { .name = "keyboard", .start = 0x64, .end = 0x64,
584 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
585 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
586 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
587 { .name = "pic2", .start = 0xa0, .end = 0xa1,
588 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
589 { .name = "dma2", .start = 0xc0, .end = 0xdf,
590 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
591 { .name = "fpu", .start = 0xf0, .end = 0xff,
592 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
593};
594
8fee697d 595void __init reserve_standard_io_resources(void)
bdba0e70
YL
596{
597 int i;
598
599 /* request I/O space for devices used on all i[345]86 PCs */
600 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
601 request_resource(&ioport_resource, &standard_io_resources[i]);
602
603}
604
042be38e
YL
605static __init void reserve_ibft_region(void)
606{
607 unsigned long addr, size = 0;
608
609 addr = find_ibft_region(&size);
610
611 if (size)
24aa0788 612 memblock_reserve(addr, size);
042be38e
YL
613}
614
9ea77bdb 615static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
5649b7c3 616
1b5576e6
YL
617static void __init trim_bios_range(void)
618{
619 /*
620 * A special case is the first 4Kb of memory;
621 * This is a BIOS owned area, not kernel ram, but generally
622 * not listed as such in the E820 table.
d0cd7425
PA
623 *
624 * This typically reserves additional memory (64KiB by default)
625 * since some BIOSes are known to corrupt low memory. See the
9ea77bdb 626 * Kconfig help text for X86_RESERVE_LOW.
1b5576e6 627 */
9ea77bdb 628 e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
d0cd7425
PA
629 E820_RAM, E820_RESERVED);
630
1b5576e6
YL
631 /*
632 * special case: Some BIOSen report the PC BIOS
633 * area (640->1Mb) as ram even though it is not.
634 * take them out.
635 */
636 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
637 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
638}
639
9ea77bdb
PA
640static int __init parse_reservelow(char *p)
641{
642 unsigned long long size;
643
644 if (!p)
645 return -EINVAL;
646
647 size = memparse(p, &p);
648
649 if (size < 4096)
650 size = 4096;
651
652 if (size > 640*1024)
653 size = 640*1024;
654
655 reserve_low = size;
656
657 return 0;
658}
659
660early_param("reservelow", parse_reservelow);
661
1da177e4
LT
662/*
663 * Determine if we were loaded by an EFI loader. If so, then we have also been
664 * passed the efi memmap, systab, etc., so we should use these data structures
665 * for initialization. Note, the efi init code path is determined by the
666 * global efi_enabled. This allows the same kernel image to be used on existing
667 * systems (with a traditional BIOS) as well as on EFI systems.
668 */
76934ed4
YL
669/*
670 * setup_arch - architecture-specific boot-time initializations
671 *
672 * Note: On x86_64, fixmaps are ready for use even before this is called.
673 */
674
1da177e4
LT
675void __init setup_arch(char **cmdline_p)
676{
76934ed4 677#ifdef CONFIG_X86_32
1da177e4 678 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
3b33553b 679 visws_early_detect();
b40827fa
BP
680
681 /*
682 * copy kernel address range established so far and switch
683 * to the proper swapper page table
684 */
685 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
686 initial_page_table + KERNEL_PGD_BOUNDARY,
687 KERNEL_PGD_PTRS);
688
689 load_cr3(swapper_pg_dir);
690 __flush_tlb_all();
76934ed4
YL
691#else
692 printk(KERN_INFO "Command line: %s\n", boot_command_line);
693#endif
1da177e4 694
9863c90f
AK
695 /*
696 * If we have OLPC OFW, we might end up relocating the fixmap due to
697 * reserve_top(), so do this before touching the ioremap area.
698 */
fd699c76
AS
699 olpc_ofw_detect();
700
29c84391 701 early_trap_init();
9e882c92 702 early_cpu_init();
1a98fd14
JF
703 early_ioremap_init();
704
fd699c76
AS
705 setup_olpc_ofw_pgd();
706
30c82645
PA
707 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
708 screen_info = boot_params.screen_info;
709 edid_info = boot_params.edid_info;
76934ed4 710#ifdef CONFIG_X86_32
30c82645
PA
711 apm_info.bios = boot_params.apm_bios_info;
712 ist_info = boot_params.ist_info;
76934ed4 713 if (boot_params.sys_desc_table.length != 0) {
30c82645
PA
714 machine_id = boot_params.sys_desc_table.table[0];
715 machine_submodel_id = boot_params.sys_desc_table.table[1];
716 BIOS_revision = boot_params.sys_desc_table.table[2];
1da177e4 717 }
76934ed4
YL
718#endif
719 saved_video_mode = boot_params.hdr.vid_mode;
30c82645 720 bootloader_type = boot_params.hdr.type_of_loader;
5031296c
PA
721 if ((bootloader_type >> 4) == 0xe) {
722 bootloader_type &= 0xf;
723 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
724 }
725 bootloader_version = bootloader_type & 0xf;
726 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
1da177e4
LT
727
728#ifdef CONFIG_BLK_DEV_RAM
30c82645
PA
729 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
730 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
731 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
1da177e4 732#endif
7465252e
YL
733#ifdef CONFIG_EFI
734 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
1adbfa35 735 "EL32", 4)) {
7465252e 736 efi_enabled = 1;
1adbfa35
OJ
737 efi_64bit = false;
738 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
739 "EL64", 4)) {
740 efi_enabled = 1;
741 efi_64bit = true;
7465252e 742 }
1adbfa35
OJ
743 if (efi_enabled && efi_memblock_x86_reserve_range())
744 efi_enabled = 0;
7465252e
YL
745#endif
746
42bbdb43 747 x86_init.oem.arch_setup();
2215e69d 748
419afdf5 749 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
0dbfafa5 750 setup_memory_map();
28bb2237 751 parse_setup_data();
a0a0becd
YL
752 /* update the e820_saved too */
753 e820_reserve_setup_data();
28bb2237 754
1da177e4
LT
755 copy_edd();
756
30c82645 757 if (!boot_params.hdr.root_flags)
1da177e4
LT
758 root_mountflags &= ~MS_RDONLY;
759 init_mm.start_code = (unsigned long) _text;
760 init_mm.end_code = (unsigned long) _etext;
761 init_mm.end_data = (unsigned long) _edata;
93dbda7c 762 init_mm.brk = _brk_end;
1da177e4
LT
763
764 code_resource.start = virt_to_phys(_text);
765 code_resource.end = virt_to_phys(_etext)-1;
766 data_resource.start = virt_to_phys(_etext);
767 data_resource.end = virt_to_phys(_edata)-1;
00bf4098
BW
768 bss_resource.start = virt_to_phys(&__bss_start);
769 bss_resource.end = virt_to_phys(&__bss_stop)-1;
1da177e4 770
516cbf37
TB
771#ifdef CONFIG_CMDLINE_BOOL
772#ifdef CONFIG_CMDLINE_OVERRIDE
773 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
774#else
775 if (builtin_cmdline[0]) {
776 /* append boot loader cmdline to builtin */
777 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
778 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
779 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
780 }
781#endif
782#endif
783
eda6da92
YL
784 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
785 *cmdline_p = command_line;
786
eda6da92 787 /*
4b0f3b81
KC
788 * x86_configure_nx() is called before parse_early_param() to detect
789 * whether hardware doesn't support NX (so that the early EHCI debug
790 * console setup can safely call set_fixmap()). It may then be called
791 * again from within noexec_setup() during parsing early parameters
792 * to honor the respective command line option.
eda6da92 793 */
4763ed4d 794 x86_configure_nx();
eda6da92
YL
795
796 parse_early_param();
797
4b0f3b81 798 x86_report_nx();
0ad5bce7 799
28bb2237 800 /* after early param, so could get panic from serial */
a9ce6bc1 801 memblock_x86_reserve_range_setup_data();
28bb2237 802
76934ed4 803 if (acpi_mps_check()) {
3eb11edc 804#ifdef CONFIG_X86_LOCAL_APIC
76934ed4 805 disable_apic = 1;
3eb11edc 806#endif
988781dc 807 setup_clear_cpu_cap(X86_FEATURE_APIC);
3c999f14
YL
808 }
809
dc7c65db
LT
810#ifdef CONFIG_PCI
811 if (pci_early_dump_regs)
812 early_dump_pci_devices();
813#endif
814
0dbfafa5 815 finish_e820_parsing();
1a3f239d 816
ff0c0874
BM
817 if (efi_enabled)
818 efi_init();
819
2216d199
YL
820 dmi_scan_machine();
821
88b094fb
AK
822 /*
823 * VMware detection requires dmi to be available, so this
824 * needs to be done after dmi_scan_machine, for the BP.
825 */
2d826404 826 init_hypervisor_platform();
88b094fb 827
f7cf5a5b 828 x86_init.resources.probe_roms();
41c094fd
YL
829
830 /* after parse_early_param, so could debug it */
831 insert_resource(&iomem_resource, &code_resource);
832 insert_resource(&iomem_resource, &data_resource);
833 insert_resource(&iomem_resource, &bss_resource);
834
1b5576e6 835 trim_bios_range();
76934ed4 836#ifdef CONFIG_X86_32
cc9f7a0c
YL
837 if (ppro_with_ram_bug()) {
838 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
839 E820_RESERVED);
840 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
841 printk(KERN_INFO "fixed physical RAM map:\n");
842 e820_print_map("bad_ppro");
843 }
76934ed4
YL
844#else
845 early_gart_iommu_check();
846#endif
cc9f7a0c 847
7b2a0a6c
YL
848 /*
849 * partially used pages are not usable - thus
850 * we are rounding upwards:
851 */
f361a450 852 max_pfn = e820_end_of_ram_pfn();
7b2a0a6c 853
093af8d7
YL
854 /* update e820 for memory not covered by WB MTRRs */
855 mtrr_bp_init();
2dc807b3 856 if (mtrr_trim_uncached_memory(max_pfn))
f361a450 857 max_pfn = e820_end_of_ram_pfn();
76c32418 858
76934ed4 859#ifdef CONFIG_X86_32
4e29684c 860 /* max_low_pfn get updated here */
2ec65f8b 861 find_low_pfn_range();
76934ed4
YL
862#else
863 num_physpages = max_pfn;
864
06cd9a7d 865 check_x2apic();
76934ed4
YL
866
867 /* How many end-of-memory variables you have, grandma! */
868 /* need this before calling reserve_initrd */
f361a450
YL
869 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
870 max_low_pfn = e820_end_of_low_ram_pfn();
871 else
872 max_low_pfn = max_pfn;
873
76934ed4 874 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
5394f80f
JF
875#endif
876
893f38d1
YL
877 /*
878 * Find and reserve possible boot-time SMP configuration:
879 */
880 find_smp_config();
881
042be38e
YL
882 reserve_ibft_region();
883
72d7c3b3
YL
884 /*
885 * Need to conclude brk, before memblock_x86_fill()
886 * it could use memblock_find_in_range, could overlap with
887 * brk area.
888 */
889 reserve_brk();
890
e5f15b45
YL
891 cleanup_highmap();
892
72d7c3b3
YL
893 memblock.current_limit = get_max_mapped();
894 memblock_x86_fill();
895
916f676f
MG
896 /*
897 * The EFI specification says that boot service code won't be called
898 * after ExitBootServices(). This is, in fact, a lie.
899 */
900 if (efi_enabled)
901 efi_reserve_boot_services();
902
72d7c3b3
YL
903 /* preallocate 4k for mptable mpc */
904 early_reserve_e820_mpc_new();
905
906#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
907 setup_bios_corruption_check();
908#endif
909
365811d6
BH
910 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
911 (max_pfn_mapped<<PAGE_SHIFT) - 1);
72d7c3b3 912
084ee1c6 913 setup_real_mode();
893f38d1 914
854c879f
PE
915 init_gbpages();
916
22ddfcaa 917 init_mem_mapping();
f361a450 918
72d7c3b3 919 memblock.current_limit = get_max_mapped();
0a2b9a6e 920 dma_contiguous_reserve(0);
4e29684c 921
e7b37895
YL
922 /*
923 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
924 */
925
926#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
927 if (init_ohci1394_dma_early)
928 init_ohci1394_dma_on_all_controllers();
929#endif
162a7e75
MT
930 /* Allocate bigger log buffer */
931 setup_log_buf(1);
e7b37895 932
2ec65f8b
YL
933 reserve_initrd();
934
44280733
YL
935 reserve_crashkernel();
936
76934ed4 937 vsmp_init();
76934ed4 938
1c6e5503
YL
939 io_delay_init();
940
1c6e5503
YL
941 /*
942 * Parse the ACPI tables for possible boot-time SMP configuration.
943 */
944 acpi_boot_table_init();
1c6e5503 945
2e42060c
JS
946 early_acpi_boot_init();
947
d8fc3afc 948 initmem_init();
6f2a7536 949 memblock_find_dma_reserve();
91467bdf 950
90993cdd 951#ifdef CONFIG_KVM_GUEST
790c73f6
GOC
952 kvmclock_init();
953#endif
954
7737b215 955 x86_init.paging.pagetable_init();
f212ec4b 956
4da9484b
PA
957 if (boot_cpu_data.cpuid_level >= 0) {
958 /* A CPU has %cr4 if and only if it has CPUID */
959 mmu_cr4_features = read_cr4();
cda846f1
JS
960 if (trampoline_cr4_features)
961 *trampoline_cr4_features = mmu_cr4_features;
4da9484b
PA
962 }
963
b40827fa
BP
964#ifdef CONFIG_X86_32
965 /* sync back kernel address range */
966 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
967 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
968 KERNEL_PGD_PTRS);
969#endif
fd89a137 970
31625340
JC
971 tboot_probe();
972
76934ed4
YL
973#ifdef CONFIG_X86_64
974 map_vsyscall();
975#endif
976
1a3f239d 977 generic_apic_probe();
1da177e4 978
54ef3400 979 early_quirks();
d44647b0 980
295deae4
YL
981 /*
982 * Read APIC and some other early information from ACPI tables.
983 */
1da177e4 984 acpi_boot_init();
efafc8b2 985 sfi_init();
a906fdaa 986 x86_dtb_init();
04606618 987
295deae4
YL
988 /*
989 * get boot-time SMP configuration:
990 */
e0da3364
YL
991 if (smp_found_config)
992 get_smp_config();
76934ed4 993
329513a3 994 prefill_possible_map();
301e6190 995
5f4765f9 996 init_cpu_to_node();
5f4765f9 997
76934ed4 998 init_apic_mappings();
9438ef7f
IM
999 if (x86_io_apic_ops.init)
1000 x86_io_apic_ops.init();
9d6a4d08 1001
295deae4 1002 kvm_guest_init();
1da177e4 1003
41c094fd 1004 e820_reserve_resources();
bf62f398 1005 e820_mark_nosave_regions(max_low_pfn);
1da177e4 1006
8fee697d 1007 x86_init.resources.reserve_resources();
41c094fd
YL
1008
1009 e820_setup_gap();
1010
1da177e4
LT
1011#ifdef CONFIG_VT
1012#if defined(CONFIG_VGA_CONSOLE)
1013 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1014 conswitchp = &vga_con;
1015#elif defined(CONFIG_DUMMY_CONSOLE)
1016 conswitchp = &dummy_con;
1017#endif
1018#endif
6f30c1ac 1019 x86_init.oem.banner();
a2202aa2 1020
6b617e22
FT
1021 x86_init.timers.wallclock_init();
1022
a2202aa2 1023 mcheck_init();
f49aa448 1024
dc326fca 1025 arch_init_ideal_nops();
b3c869d3
JS
1026
1027 register_refined_jiffies(CLOCK_TICK_RATE);
5189c2a7
OJ
1028
1029#ifdef CONFIG_EFI
1030 /* Once setup is done above, disable efi_enabled on mismatched
1031 * firmware/kernel archtectures since there is no support for
1032 * runtime services.
1033 */
1034 if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) {
1035 pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
1036 efi_unmap_memmap();
1037 efi_enabled = 0;
1038 }
1039#endif
1da177e4 1040}
5649b7c3 1041
9be1b56a
IM
1042#ifdef CONFIG_X86_32
1043
8fee697d
TG
1044static struct resource video_ram_resource = {
1045 .name = "Video RAM area",
1046 .start = 0xa0000,
1047 .end = 0xbffff,
1048 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
9be1b56a
IM
1049};
1050
8fee697d 1051void __init i386_reserve_resources(void)
9be1b56a 1052{
8fee697d
TG
1053 request_resource(&iomem_resource, &video_ram_resource);
1054 reserve_standard_io_resources();
9be1b56a
IM
1055}
1056
9be1b56a 1057#endif /* CONFIG_X86_32 */