x86, realmode: Separate real_mode reserve and setup
[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>
a2202aa2 111#include <asm/mce.h>
f49aa448 112#include <asm/alternative.h>
da6b737b 113#include <asm/prom.h>
46d671b5 114
2b72394e 115/*
66520ebc
JS
116 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
117 * max_pfn_mapped: highest direct mapped pfn over 4GB
118 *
119 * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
120 * represented by pfn_mapped
2b72394e
PE
121 */
122unsigned long max_low_pfn_mapped;
123unsigned long max_pfn_mapped;
124
e808bae2 125#ifdef CONFIG_DMI
796216a5 126RESERVE_BRK(dmi_alloc, 65536);
e808bae2 127#endif
796216a5 128
c0b5842a 129
93dbda7c
JF
130static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
131unsigned long _brk_end = (unsigned long)__brk_base;
132
c0b5842a
IM
133#ifdef CONFIG_X86_64
134int default_cpu_present_to_apicid(int mps_cpu)
135{
136 return __default_cpu_present_to_apicid(mps_cpu);
137}
138
e11dadab 139int default_check_phys_apicid_present(int phys_apicid)
c0b5842a 140{
e11dadab 141 return __default_check_phys_apicid_present(phys_apicid);
c0b5842a
IM
142}
143#endif
144
217b8ce8 145struct boot_params boot_params;
217b8ce8 146
1da177e4
LT
147/*
148 * Machine setup..
149 */
c9cce83d
BW
150static struct resource data_resource = {
151 .name = "Kernel data",
152 .start = 0,
153 .end = 0,
154 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
155};
156
157static struct resource code_resource = {
158 .name = "Kernel code",
159 .start = 0,
160 .end = 0,
161 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
162};
163
164static struct resource bss_resource = {
165 .name = "Kernel bss",
166 .start = 0,
167 .end = 0,
168 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
169};
170
7dea23ec
YL
171
172#ifdef CONFIG_X86_32
1da177e4 173/* cpu data as detected by the assembly code in head.S */
7dea23ec 174struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
1da177e4 175/* common cpu data for all cpus */
7dea23ec 176struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
129f6946 177EXPORT_SYMBOL(boot_cpu_data);
1da177e4 178
0c254e38
AS
179unsigned int def_to_bigsmp;
180
1da177e4
LT
181/* for MCA, but anyone else can use it if they want */
182unsigned int machine_id;
183unsigned int machine_submodel_id;
184unsigned int BIOS_revision;
1da177e4 185
7dea23ec
YL
186struct apm_info apm_info;
187EXPORT_SYMBOL(apm_info);
188
189#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
190 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
191struct ist_info ist_info;
192EXPORT_SYMBOL(ist_info);
193#else
194struct ist_info ist_info;
195#endif
196
197#else
ed26dbe5
JF
198struct cpuinfo_x86 boot_cpu_data __read_mostly = {
199 .x86_phys_bits = MAX_PHYSMEM_BITS,
200};
7dea23ec
YL
201EXPORT_SYMBOL(boot_cpu_data);
202#endif
203
204
205#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
206unsigned long mmu_cr4_features;
207#else
208unsigned long mmu_cr4_features = X86_CR4_PAE;
209#endif
210
5031296c
PA
211/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
212int bootloader_type, bootloader_version;
1da177e4 213
1da177e4
LT
214/*
215 * Setup options
216 */
1da177e4 217struct screen_info screen_info;
129f6946 218EXPORT_SYMBOL(screen_info);
1da177e4 219struct edid_info edid_info;
5e518d76 220EXPORT_SYMBOL_GPL(edid_info);
1da177e4 221
1da177e4
LT
222extern int root_mountflags;
223
e44b7b75 224unsigned long saved_video_mode;
1da177e4 225
cf8fa920 226#define RAMDISK_IMAGE_START_MASK 0x07FF
1da177e4 227#define RAMDISK_PROMPT_FLAG 0x8000
cf8fa920 228#define RAMDISK_LOAD_FLAG 0x4000
1da177e4 229
4e498b66 230static char __initdata command_line[COMMAND_LINE_SIZE];
516cbf37
TB
231#ifdef CONFIG_CMDLINE_BOOL
232static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
233#endif
1da177e4 234
1da177e4
LT
235#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
236struct edd edd;
237#ifdef CONFIG_EDD_MODULE
238EXPORT_SYMBOL(edd);
239#endif
240/**
241 * copy_edd() - Copy the BIOS EDD information
242 * from boot_params into a safe place.
243 *
244 */
9eaa192d 245static inline void __init copy_edd(void)
1da177e4 246{
30c82645
PA
247 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
248 sizeof(edd.mbr_signature));
249 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
250 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
251 edd.edd_info_nr = boot_params.eddbuf_entries;
1da177e4
LT
252}
253#else
9eaa192d 254static inline void __init copy_edd(void)
1da177e4
LT
255{
256}
257#endif
258
5368a2be
PA
259void * __init extend_brk(size_t size, size_t align)
260{
261 size_t mask = align - 1;
262 void *ret;
263
264 BUG_ON(_brk_start == 0);
265 BUG_ON(align & mask);
266
267 _brk_end = (_brk_end + mask) & ~mask;
268 BUG_ON((char *)(_brk_end + size) > __brk_limit);
269
270 ret = (void *)_brk_end;
271 _brk_end += size;
272
273 memset(ret, 0, size);
274
275 return ret;
276}
277
148b2098 278#ifdef CONFIG_X86_32
e5f15b45 279static void __init cleanup_highmap(void)
f005fe12
YL
280{
281}
854c879f
PE
282#endif
283
5368a2be
PA
284static void __init reserve_brk(void)
285{
286 if (_brk_end > _brk_start)
24aa0788
TH
287 memblock_reserve(__pa(_brk_start),
288 __pa(_brk_end) - __pa(_brk_start));
5368a2be
PA
289
290 /* Mark brk area as locked down and no longer taking any
291 new allocations */
292 _brk_start = 0;
293}
294
cf8fa920
PA
295#ifdef CONFIG_BLK_DEV_INITRD
296
eb1379cb
YL
297#define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
298static void __init relocate_initrd(void)
cf8fa920 299{
c967da6a 300 /* Assume only end is not page aligned */
ba5b14cc
YL
301 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
302 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
c967da6a 303 u64 area_size = PAGE_ALIGN(ramdisk_size);
ba5b14cc 304 u64 ramdisk_here;
eb1379cb
YL
305 unsigned long slop, clen, mapaddr;
306 char *p, *q;
cf8fa920 307
e8c57d40 308 /* We need to move the initrd down into directly mapped mem */
74f27655 309 ramdisk_here = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
e8c57d40 310 area_size, PAGE_SIZE);
cf8fa920 311
1f5026a7 312 if (!ramdisk_here)
3945e2c9
YL
313 panic("Cannot find place for new RAMDISK of size %lld\n",
314 ramdisk_size);
315
e8c57d40 316 /* Note: this includes all the mem currently occupied by
cf8fa920 317 the initrd, we rely on that fact to keep the data intact. */
24aa0788 318 memblock_reserve(ramdisk_here, area_size);
cf8fa920
PA
319 initrd_start = ramdisk_here + PAGE_OFFSET;
320 initrd_end = initrd_start + ramdisk_size;
365811d6
BH
321 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
322 ramdisk_here, ramdisk_here + ramdisk_size - 1);
cf8fa920 323
cf8fa920
PA
324 q = (char *)initrd_start;
325
e8c57d40 326 /* Copy the initrd */
cf8fa920
PA
327 while (ramdisk_size) {
328 slop = ramdisk_image & ~PAGE_MASK;
329 clen = ramdisk_size;
330 if (clen > MAX_MAP_CHUNK-slop)
331 clen = MAX_MAP_CHUNK-slop;
332 mapaddr = ramdisk_image & PAGE_MASK;
88b4c146 333 p = early_memremap(mapaddr, clen+slop);
cf8fa920 334 memcpy(q, p+slop, clen);
beacfaac 335 early_iounmap(p, clen+slop);
cf8fa920
PA
336 q += clen;
337 ramdisk_image += clen;
338 ramdisk_size -= clen;
339 }
e8c57d40 340
ba5b14cc
YL
341 ramdisk_image = boot_params.hdr.ramdisk_image;
342 ramdisk_size = boot_params.hdr.ramdisk_size;
365811d6
BH
343 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
344 " [mem %#010llx-%#010llx]\n",
ba5b14cc
YL
345 ramdisk_image, ramdisk_image + ramdisk_size - 1,
346 ramdisk_here, ramdisk_here + ramdisk_size - 1);
eb1379cb 347}
9a27f5c5 348
e8c57d40
YL
349static u64 __init get_mem_size(unsigned long limit_pfn)
350{
351 int i;
352 u64 mapped_pages = 0;
353 unsigned long start_pfn, end_pfn;
354
355 for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, NULL) {
356 start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
357 end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
358 mapped_pages += end_pfn - start_pfn;
359 }
360
361 return mapped_pages << PAGE_SHIFT;
362}
eb1379cb
YL
363static void __init reserve_initrd(void)
364{
c967da6a 365 /* Assume only end is not page aligned */
eb1379cb
YL
366 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
367 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
c967da6a 368 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
e8c57d40 369 u64 mapped_size;
eb1379cb
YL
370
371 if (!boot_params.hdr.type_of_loader ||
372 !ramdisk_image || !ramdisk_size)
373 return; /* No initrd provided by bootloader */
374
375 initrd_start = 0;
376
74f27655 377 mapped_size = get_mem_size(max_pfn_mapped);
e8c57d40 378 if (ramdisk_size >= (mapped_size>>1))
ab7b64e9
PJ
379 panic("initrd too large to handle, "
380 "disabling initrd (%lld needed, %lld available)\n",
e8c57d40 381 ramdisk_size, mapped_size>>1);
eb1379cb 382
365811d6
BH
383 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
384 ramdisk_end - 1);
eb1379cb 385
74f27655 386 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
e8c57d40
YL
387 PFN_DOWN(ramdisk_end))) {
388 /* All are mapped, easy case */
eb1379cb
YL
389 /*
390 * don't need to reserve again, already reserved early
391 * in i386_start_kernel
392 */
393 initrd_start = ramdisk_image + PAGE_OFFSET;
394 initrd_end = initrd_start + ramdisk_size;
395 return;
396 }
397
eb1379cb 398 relocate_initrd();
8c5dd8f4 399
24aa0788 400 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
cf8fa920 401}
225c37d7 402#else
eb1379cb 403static void __init reserve_initrd(void)
225c37d7
YL
404{
405}
cf8fa920
PA
406#endif /* CONFIG_BLK_DEV_INITRD */
407
29f784e3 408static void __init parse_setup_data(void)
257b0fde
YL
409{
410 struct setup_data *data;
411 u64 pa_data;
412
413 if (boot_params.hdr.version < 0x0209)
414 return;
415 pa_data = boot_params.hdr.setup_data;
416 while (pa_data) {
f1c2b357
SAS
417 u32 data_len, map_len;
418
419 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
420 (u64)sizeof(struct setup_data));
421 data = early_memremap(pa_data, map_len);
422 data_len = data->len + sizeof(struct setup_data);
423 if (data_len > map_len) {
424 early_iounmap(data, map_len);
425 data = early_memremap(pa_data, data_len);
426 map_len = data_len;
427 }
428
257b0fde
YL
429 switch (data->type) {
430 case SETUP_E820_EXT:
f1c2b357 431 parse_e820_ext(data);
257b0fde 432 break;
da6b737b
SAS
433 case SETUP_DTB:
434 add_dtb(pa_data);
257b0fde
YL
435 break;
436 default:
437 break;
438 }
257b0fde 439 pa_data = data->next;
f1c2b357 440 early_iounmap(data, map_len);
257b0fde
YL
441 }
442}
443
a0a0becd 444static void __init e820_reserve_setup_data(void)
28bb2237
YL
445{
446 struct setup_data *data;
447 u64 pa_data;
d9a81b44 448 int found = 0;
28bb2237
YL
449
450 if (boot_params.hdr.version < 0x0209)
451 return;
452 pa_data = boot_params.hdr.setup_data;
453 while (pa_data) {
88b4c146 454 data = early_memremap(pa_data, sizeof(*data));
28bb2237
YL
455 e820_update_range(pa_data, sizeof(*data)+data->len,
456 E820_RAM, E820_RESERVED_KERN);
d9a81b44 457 found = 1;
28bb2237
YL
458 pa_data = data->next;
459 early_iounmap(data, sizeof(*data));
460 }
d9a81b44
YL
461 if (!found)
462 return;
463
28bb2237 464 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
a0a0becd 465 memcpy(&e820_saved, &e820, sizeof(struct e820map));
28bb2237
YL
466 printk(KERN_INFO "extended physical RAM map:\n");
467 e820_print_map("reserve setup_data");
468}
469
a9ce6bc1 470static void __init memblock_x86_reserve_range_setup_data(void)
a0a0becd
YL
471{
472 struct setup_data *data;
473 u64 pa_data;
a0a0becd
YL
474
475 if (boot_params.hdr.version < 0x0209)
476 return;
477 pa_data = boot_params.hdr.setup_data;
478 while (pa_data) {
88b4c146 479 data = early_memremap(pa_data, sizeof(*data));
24aa0788 480 memblock_reserve(pa_data, sizeof(*data) + data->len);
a0a0becd
YL
481 pa_data = data->next;
482 early_iounmap(data, sizeof(*data));
483 }
484}
485
ccb4defa
YL
486/*
487 * --------- Crashkernel reservation ------------------------------
488 */
489
490#ifdef CONFIG_KEXEC
32105f7f 491
7f8595bf
PA
492/*
493 * Keep the crash kernel below this limit. On 32 bits earlier kernels
494 * would limit the kernel to the low 512 MiB due to mapping restrictions.
495 * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
496 * limit once kexec-tools are fixed.
497 */
498#ifdef CONFIG_X86_32
499# define CRASH_KERNEL_ADDR_MAX (512 << 20)
500#else
501# define CRASH_KERNEL_ADDR_MAX (896 << 20)
502#endif
503
29f784e3 504static void __init reserve_crashkernel(void)
ccb4defa
YL
505{
506 unsigned long long total_mem;
507 unsigned long long crash_size, crash_base;
508 int ret;
509
09c71bfd 510 total_mem = memblock_phys_mem_size();
ccb4defa
YL
511
512 ret = parse_crashkernel(boot_command_line, total_mem,
513 &crash_size, &crash_base);
32105f7f
BW
514 if (ret != 0 || crash_size <= 0)
515 return;
516
517 /* 0 means: find the address automatically */
518 if (crash_base <= 0) {
44280733
YL
519 const unsigned long long alignment = 16<<20; /* 16M */
520
9f4c1396 521 /*
7f8595bf 522 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
9f4c1396
YL
523 */
524 crash_base = memblock_find_in_range(alignment,
7f8595bf 525 CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
9f4c1396 526
1f5026a7 527 if (!crash_base) {
44280733 528 pr_info("crashkernel reservation failed - No suitable area found.\n");
ccb4defa
YL
529 return;
530 }
32105f7f 531 } else {
44280733
YL
532 unsigned long long start;
533
9f4c1396
YL
534 start = memblock_find_in_range(crash_base,
535 crash_base + crash_size, crash_size, 1<<20);
44280733
YL
536 if (start != crash_base) {
537 pr_info("crashkernel reservation failed - memory is in use.\n");
ccb4defa
YL
538 return;
539 }
32105f7f 540 }
24aa0788 541 memblock_reserve(crash_base, crash_size);
ccb4defa 542
32105f7f
BW
543 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
544 "for crashkernel (System RAM: %ldMB)\n",
545 (unsigned long)(crash_size >> 20),
546 (unsigned long)(crash_base >> 20),
547 (unsigned long)(total_mem >> 20));
ccb4defa 548
32105f7f
BW
549 crashk_res.start = crash_base;
550 crashk_res.end = crash_base + crash_size - 1;
551 insert_resource(&iomem_resource, &crashk_res);
ccb4defa
YL
552}
553#else
29f784e3 554static void __init reserve_crashkernel(void)
ccb4defa
YL
555{
556}
557#endif
558
bdba0e70
YL
559static struct resource standard_io_resources[] = {
560 { .name = "dma1", .start = 0x00, .end = 0x1f,
561 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
562 { .name = "pic1", .start = 0x20, .end = 0x21,
563 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
564 { .name = "timer0", .start = 0x40, .end = 0x43,
565 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
566 { .name = "timer1", .start = 0x50, .end = 0x53,
567 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
568 { .name = "keyboard", .start = 0x60, .end = 0x60,
569 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
570 { .name = "keyboard", .start = 0x64, .end = 0x64,
571 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
572 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
573 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
574 { .name = "pic2", .start = 0xa0, .end = 0xa1,
575 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
576 { .name = "dma2", .start = 0xc0, .end = 0xdf,
577 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
578 { .name = "fpu", .start = 0xf0, .end = 0xff,
579 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
580};
581
8fee697d 582void __init reserve_standard_io_resources(void)
bdba0e70
YL
583{
584 int i;
585
586 /* request I/O space for devices used on all i[345]86 PCs */
587 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
588 request_resource(&ioport_resource, &standard_io_resources[i]);
589
590}
591
042be38e
YL
592static __init void reserve_ibft_region(void)
593{
594 unsigned long addr, size = 0;
595
596 addr = find_ibft_region(&size);
597
598 if (size)
24aa0788 599 memblock_reserve(addr, size);
042be38e
YL
600}
601
9ea77bdb 602static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
5649b7c3 603
a9acc536
JB
604static bool __init snb_gfx_workaround_needed(void)
605{
e43b3cec 606#ifdef CONFIG_PCI
a9acc536
JB
607 int i;
608 u16 vendor, devid;
ab3cd867 609 static const __initconst u16 snb_ids[] = {
a9acc536
JB
610 0x0102,
611 0x0112,
612 0x0122,
613 0x0106,
614 0x0116,
615 0x0126,
616 0x010a,
617 };
618
619 /* Assume no if something weird is going on with PCI */
620 if (!early_pci_allowed())
621 return false;
622
623 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
624 if (vendor != 0x8086)
625 return false;
626
627 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
628 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
629 if (devid == snb_ids[i])
630 return true;
e43b3cec 631#endif
a9acc536
JB
632
633 return false;
634}
635
636/*
637 * Sandy Bridge graphics has trouble with certain ranges, exclude
638 * them from allocation.
639 */
640static void __init trim_snb_memory(void)
641{
ab3cd867 642 static const __initconst unsigned long bad_pages[] = {
a9acc536
JB
643 0x20050000,
644 0x20110000,
645 0x20130000,
646 0x20138000,
647 0x40004000,
648 };
649 int i;
650
651 if (!snb_gfx_workaround_needed())
652 return;
653
654 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
655
656 /*
657 * Reserve all memory below the 1 MB mark that has not
658 * already been reserved.
659 */
660 memblock_reserve(0, 1<<20);
661
662 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
663 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
664 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
665 bad_pages[i]);
666 }
667}
668
669/*
670 * Here we put platform-specific memory range workarounds, i.e.
671 * memory known to be corrupt or otherwise in need to be reserved on
672 * specific platforms.
673 *
674 * If this gets used more widely it could use a real dispatch mechanism.
675 */
676static void __init trim_platform_memory_ranges(void)
677{
678 trim_snb_memory();
679}
680
1b5576e6
YL
681static void __init trim_bios_range(void)
682{
683 /*
684 * A special case is the first 4Kb of memory;
685 * This is a BIOS owned area, not kernel ram, but generally
686 * not listed as such in the E820 table.
d0cd7425
PA
687 *
688 * This typically reserves additional memory (64KiB by default)
689 * since some BIOSes are known to corrupt low memory. See the
9ea77bdb 690 * Kconfig help text for X86_RESERVE_LOW.
1b5576e6 691 */
9ea77bdb 692 e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
d0cd7425
PA
693 E820_RAM, E820_RESERVED);
694
1b5576e6
YL
695 /*
696 * special case: Some BIOSen report the PC BIOS
697 * area (640->1Mb) as ram even though it is not.
698 * take them out.
699 */
700 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
a9acc536 701
1b5576e6
YL
702 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
703}
704
b422a309
YL
705/* called before trim_bios_range() to spare extra sanitize */
706static void __init e820_add_kernel_range(void)
707{
708 u64 start = __pa_symbol(_text);
709 u64 size = __pa_symbol(_end) - start;
710
711 /*
712 * Complain if .text .data and .bss are not marked as E820_RAM and
713 * attempt to fix it by adding the range. We may have a confused BIOS,
714 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
715 * exclude kernel range. If we really are running on top non-RAM,
716 * we will crash later anyways.
717 */
718 if (e820_all_mapped(start, start + size, E820_RAM))
719 return;
720
721 pr_warn(".text .data .bss are not marked as E820_RAM!\n");
722 e820_remove_range(start, size, E820_RAM, 0);
723 e820_add_region(start, size, E820_RAM);
724}
725
9ea77bdb
PA
726static int __init parse_reservelow(char *p)
727{
728 unsigned long long size;
729
730 if (!p)
731 return -EINVAL;
732
733 size = memparse(p, &p);
734
735 if (size < 4096)
736 size = 4096;
737
738 if (size > 640*1024)
739 size = 640*1024;
740
741 reserve_low = size;
742
743 return 0;
744}
745
746early_param("reservelow", parse_reservelow);
747
1da177e4
LT
748/*
749 * Determine if we were loaded by an EFI loader. If so, then we have also been
750 * passed the efi memmap, systab, etc., so we should use these data structures
751 * for initialization. Note, the efi init code path is determined by the
752 * global efi_enabled. This allows the same kernel image to be used on existing
753 * systems (with a traditional BIOS) as well as on EFI systems.
754 */
76934ed4
YL
755/*
756 * setup_arch - architecture-specific boot-time initializations
757 *
758 * Note: On x86_64, fixmaps are ready for use even before this is called.
759 */
760
1da177e4
LT
761void __init setup_arch(char **cmdline_p)
762{
76934ed4 763#ifdef CONFIG_X86_32
1da177e4 764 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
3b33553b 765 visws_early_detect();
b40827fa
BP
766
767 /*
768 * copy kernel address range established so far and switch
769 * to the proper swapper page table
770 */
771 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
772 initial_page_table + KERNEL_PGD_BOUNDARY,
773 KERNEL_PGD_PTRS);
774
775 load_cr3(swapper_pg_dir);
776 __flush_tlb_all();
76934ed4
YL
777#else
778 printk(KERN_INFO "Command line: %s\n", boot_command_line);
779#endif
1da177e4 780
9863c90f
AK
781 /*
782 * If we have OLPC OFW, we might end up relocating the fixmap due to
783 * reserve_top(), so do this before touching the ioremap area.
784 */
fd699c76
AS
785 olpc_ofw_detect();
786
29c84391 787 early_trap_init();
9e882c92 788 early_cpu_init();
1a98fd14
JF
789 early_ioremap_init();
790
fd699c76
AS
791 setup_olpc_ofw_pgd();
792
30c82645
PA
793 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
794 screen_info = boot_params.screen_info;
795 edid_info = boot_params.edid_info;
76934ed4 796#ifdef CONFIG_X86_32
30c82645
PA
797 apm_info.bios = boot_params.apm_bios_info;
798 ist_info = boot_params.ist_info;
76934ed4 799 if (boot_params.sys_desc_table.length != 0) {
30c82645
PA
800 machine_id = boot_params.sys_desc_table.table[0];
801 machine_submodel_id = boot_params.sys_desc_table.table[1];
802 BIOS_revision = boot_params.sys_desc_table.table[2];
1da177e4 803 }
76934ed4
YL
804#endif
805 saved_video_mode = boot_params.hdr.vid_mode;
30c82645 806 bootloader_type = boot_params.hdr.type_of_loader;
5031296c
PA
807 if ((bootloader_type >> 4) == 0xe) {
808 bootloader_type &= 0xf;
809 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
810 }
811 bootloader_version = bootloader_type & 0xf;
812 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
1da177e4
LT
813
814#ifdef CONFIG_BLK_DEV_RAM
30c82645
PA
815 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
816 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
817 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
1da177e4 818#endif
7465252e
YL
819#ifdef CONFIG_EFI
820 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
1adbfa35 821 "EL32", 4)) {
7465252e 822 efi_enabled = 1;
1adbfa35
OJ
823 efi_64bit = false;
824 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
825 "EL64", 4)) {
826 efi_enabled = 1;
827 efi_64bit = true;
7465252e 828 }
1adbfa35
OJ
829 if (efi_enabled && efi_memblock_x86_reserve_range())
830 efi_enabled = 0;
7465252e
YL
831#endif
832
42bbdb43 833 x86_init.oem.arch_setup();
2215e69d 834
419afdf5 835 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
0dbfafa5 836 setup_memory_map();
28bb2237 837 parse_setup_data();
a0a0becd
YL
838 /* update the e820_saved too */
839 e820_reserve_setup_data();
28bb2237 840
1da177e4
LT
841 copy_edd();
842
30c82645 843 if (!boot_params.hdr.root_flags)
1da177e4
LT
844 root_mountflags &= ~MS_RDONLY;
845 init_mm.start_code = (unsigned long) _text;
846 init_mm.end_code = (unsigned long) _etext;
847 init_mm.end_data = (unsigned long) _edata;
93dbda7c 848 init_mm.brk = _brk_end;
1da177e4
LT
849
850 code_resource.start = virt_to_phys(_text);
851 code_resource.end = virt_to_phys(_etext)-1;
852 data_resource.start = virt_to_phys(_etext);
853 data_resource.end = virt_to_phys(_edata)-1;
00bf4098
BW
854 bss_resource.start = virt_to_phys(&__bss_start);
855 bss_resource.end = virt_to_phys(&__bss_stop)-1;
1da177e4 856
516cbf37
TB
857#ifdef CONFIG_CMDLINE_BOOL
858#ifdef CONFIG_CMDLINE_OVERRIDE
859 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
860#else
861 if (builtin_cmdline[0]) {
862 /* append boot loader cmdline to builtin */
863 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
864 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
865 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
866 }
867#endif
868#endif
869
eda6da92
YL
870 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
871 *cmdline_p = command_line;
872
eda6da92 873 /*
4b0f3b81
KC
874 * x86_configure_nx() is called before parse_early_param() to detect
875 * whether hardware doesn't support NX (so that the early EHCI debug
876 * console setup can safely call set_fixmap()). It may then be called
877 * again from within noexec_setup() during parsing early parameters
878 * to honor the respective command line option.
eda6da92 879 */
4763ed4d 880 x86_configure_nx();
eda6da92
YL
881
882 parse_early_param();
883
4b0f3b81 884 x86_report_nx();
0ad5bce7 885
28bb2237 886 /* after early param, so could get panic from serial */
a9ce6bc1 887 memblock_x86_reserve_range_setup_data();
28bb2237 888
76934ed4 889 if (acpi_mps_check()) {
3eb11edc 890#ifdef CONFIG_X86_LOCAL_APIC
76934ed4 891 disable_apic = 1;
3eb11edc 892#endif
988781dc 893 setup_clear_cpu_cap(X86_FEATURE_APIC);
3c999f14
YL
894 }
895
dc7c65db
LT
896#ifdef CONFIG_PCI
897 if (pci_early_dump_regs)
898 early_dump_pci_devices();
899#endif
900
0dbfafa5 901 finish_e820_parsing();
1a3f239d 902
ff0c0874
BM
903 if (efi_enabled)
904 efi_init();
905
2216d199
YL
906 dmi_scan_machine();
907
88b094fb
AK
908 /*
909 * VMware detection requires dmi to be available, so this
910 * needs to be done after dmi_scan_machine, for the BP.
911 */
2d826404 912 init_hypervisor_platform();
88b094fb 913
f7cf5a5b 914 x86_init.resources.probe_roms();
41c094fd
YL
915
916 /* after parse_early_param, so could debug it */
917 insert_resource(&iomem_resource, &code_resource);
918 insert_resource(&iomem_resource, &data_resource);
919 insert_resource(&iomem_resource, &bss_resource);
920
b422a309 921 e820_add_kernel_range();
1b5576e6 922 trim_bios_range();
76934ed4 923#ifdef CONFIG_X86_32
cc9f7a0c
YL
924 if (ppro_with_ram_bug()) {
925 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
926 E820_RESERVED);
927 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
928 printk(KERN_INFO "fixed physical RAM map:\n");
929 e820_print_map("bad_ppro");
930 }
76934ed4
YL
931#else
932 early_gart_iommu_check();
933#endif
cc9f7a0c 934
7b2a0a6c
YL
935 /*
936 * partially used pages are not usable - thus
937 * we are rounding upwards:
938 */
f361a450 939 max_pfn = e820_end_of_ram_pfn();
7b2a0a6c 940
093af8d7
YL
941 /* update e820 for memory not covered by WB MTRRs */
942 mtrr_bp_init();
2dc807b3 943 if (mtrr_trim_uncached_memory(max_pfn))
f361a450 944 max_pfn = e820_end_of_ram_pfn();
76c32418 945
76934ed4 946#ifdef CONFIG_X86_32
4e29684c 947 /* max_low_pfn get updated here */
2ec65f8b 948 find_low_pfn_range();
76934ed4
YL
949#else
950 num_physpages = max_pfn;
951
06cd9a7d 952 check_x2apic();
76934ed4
YL
953
954 /* How many end-of-memory variables you have, grandma! */
955 /* need this before calling reserve_initrd */
f361a450
YL
956 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
957 max_low_pfn = e820_end_of_low_ram_pfn();
958 else
959 max_low_pfn = max_pfn;
960
76934ed4 961 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
5394f80f
JF
962#endif
963
893f38d1
YL
964 /*
965 * Find and reserve possible boot-time SMP configuration:
966 */
967 find_smp_config();
968
042be38e
YL
969 reserve_ibft_region();
970
8d57470d
YL
971 early_alloc_pgt_buf();
972
72d7c3b3
YL
973 /*
974 * Need to conclude brk, before memblock_x86_fill()
975 * it could use memblock_find_in_range, could overlap with
976 * brk area.
977 */
978 reserve_brk();
979
e5f15b45
YL
980 cleanup_highmap();
981
dd7dfad7 982 memblock.current_limit = ISA_END_ADDRESS;
72d7c3b3
YL
983 memblock_x86_fill();
984
916f676f
MG
985 /*
986 * The EFI specification says that boot service code won't be called
987 * after ExitBootServices(). This is, in fact, a lie.
988 */
989 if (efi_enabled)
990 efi_reserve_boot_services();
991
72d7c3b3
YL
992 /* preallocate 4k for mptable mpc */
993 early_reserve_e820_mpc_new();
994
995#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
996 setup_bios_corruption_check();
997#endif
998
365811d6
BH
999 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1000 (max_pfn_mapped<<PAGE_SHIFT) - 1);
72d7c3b3 1001
4f7b9226 1002 reserve_real_mode();
893f38d1 1003
a9acc536
JB
1004 trim_platform_memory_ranges();
1005
22ddfcaa 1006 init_mem_mapping();
f361a450 1007
4f7b9226
YL
1008 setup_real_mode();
1009
72d7c3b3 1010 memblock.current_limit = get_max_mapped();
0a2b9a6e 1011 dma_contiguous_reserve(0);
4e29684c 1012
e7b37895
YL
1013 /*
1014 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1015 */
1016
1017#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1018 if (init_ohci1394_dma_early)
1019 init_ohci1394_dma_on_all_controllers();
1020#endif
162a7e75
MT
1021 /* Allocate bigger log buffer */
1022 setup_log_buf(1);
e7b37895 1023
2ec65f8b
YL
1024 reserve_initrd();
1025
3dfd8235 1026#if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
53aac44c 1027 acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
3dfd8235 1028#endif
53aac44c 1029
44280733
YL
1030 reserve_crashkernel();
1031
76934ed4 1032 vsmp_init();
76934ed4 1033
1c6e5503
YL
1034 io_delay_init();
1035
1c6e5503
YL
1036 /*
1037 * Parse the ACPI tables for possible boot-time SMP configuration.
1038 */
1039 acpi_boot_table_init();
1c6e5503 1040
2e42060c
JS
1041 early_acpi_boot_init();
1042
d8fc3afc 1043 initmem_init();
6f2a7536 1044 memblock_find_dma_reserve();
91467bdf 1045
90993cdd 1046#ifdef CONFIG_KVM_GUEST
790c73f6
GOC
1047 kvmclock_init();
1048#endif
1049
7737b215 1050 x86_init.paging.pagetable_init();
f212ec4b 1051
4da9484b
PA
1052 if (boot_cpu_data.cpuid_level >= 0) {
1053 /* A CPU has %cr4 if and only if it has CPUID */
1054 mmu_cr4_features = read_cr4();
cda846f1
JS
1055 if (trampoline_cr4_features)
1056 *trampoline_cr4_features = mmu_cr4_features;
4da9484b
PA
1057 }
1058
b40827fa
BP
1059#ifdef CONFIG_X86_32
1060 /* sync back kernel address range */
1061 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1062 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1063 KERNEL_PGD_PTRS);
1064#endif
fd89a137 1065
31625340
JC
1066 tboot_probe();
1067
76934ed4
YL
1068#ifdef CONFIG_X86_64
1069 map_vsyscall();
1070#endif
1071
1a3f239d 1072 generic_apic_probe();
1da177e4 1073
54ef3400 1074 early_quirks();
d44647b0 1075
295deae4
YL
1076 /*
1077 * Read APIC and some other early information from ACPI tables.
1078 */
1da177e4 1079 acpi_boot_init();
efafc8b2 1080 sfi_init();
a906fdaa 1081 x86_dtb_init();
04606618 1082
295deae4
YL
1083 /*
1084 * get boot-time SMP configuration:
1085 */
e0da3364
YL
1086 if (smp_found_config)
1087 get_smp_config();
76934ed4 1088
329513a3 1089 prefill_possible_map();
301e6190 1090
5f4765f9 1091 init_cpu_to_node();
5f4765f9 1092
76934ed4 1093 init_apic_mappings();
9438ef7f
IM
1094 if (x86_io_apic_ops.init)
1095 x86_io_apic_ops.init();
9d6a4d08 1096
295deae4 1097 kvm_guest_init();
1da177e4 1098
41c094fd 1099 e820_reserve_resources();
bf62f398 1100 e820_mark_nosave_regions(max_low_pfn);
1da177e4 1101
8fee697d 1102 x86_init.resources.reserve_resources();
41c094fd
YL
1103
1104 e820_setup_gap();
1105
1da177e4
LT
1106#ifdef CONFIG_VT
1107#if defined(CONFIG_VGA_CONSOLE)
1108 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1109 conswitchp = &vga_con;
1110#elif defined(CONFIG_DUMMY_CONSOLE)
1111 conswitchp = &dummy_con;
1112#endif
1113#endif
6f30c1ac 1114 x86_init.oem.banner();
a2202aa2 1115
6b617e22
FT
1116 x86_init.timers.wallclock_init();
1117
a2202aa2 1118 mcheck_init();
f49aa448 1119
dc326fca 1120 arch_init_ideal_nops();
b3c869d3
JS
1121
1122 register_refined_jiffies(CLOCK_TICK_RATE);
5189c2a7
OJ
1123
1124#ifdef CONFIG_EFI
1125 /* Once setup is done above, disable efi_enabled on mismatched
1126 * firmware/kernel archtectures since there is no support for
1127 * runtime services.
1128 */
1129 if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) {
1130 pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
1131 efi_unmap_memmap();
1132 efi_enabled = 0;
1133 }
1134#endif
1da177e4 1135}
5649b7c3 1136
9be1b56a
IM
1137#ifdef CONFIG_X86_32
1138
8fee697d
TG
1139static struct resource video_ram_resource = {
1140 .name = "Video RAM area",
1141 .start = 0xa0000,
1142 .end = 0xbffff,
1143 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
9be1b56a
IM
1144};
1145
8fee697d 1146void __init i386_reserve_resources(void)
9be1b56a 1147{
8fee697d
TG
1148 request_resource(&iomem_resource, &video_ram_resource);
1149 reserve_standard_io_resources();
9be1b56a
IM
1150}
1151
9be1b56a 1152#endif /* CONFIG_X86_32 */