x86 boot: add code to add BIOS provided EFI memory entries to kernel
[deliverable/linux.git] / arch / x86 / kernel / setup_32.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>
30#include <linux/apm_bios.h>
31#include <linux/initrd.h>
32#include <linux/bootmem.h>
33#include <linux/seq_file.h>
34#include <linux/console.h>
35#include <linux/mca.h>
36#include <linux/root_dev.h>
37#include <linux/highmem.h>
38#include <linux/module.h>
39#include <linux/efi.h>
40#include <linux/init.h>
41#include <linux/edd.h>
138fe4e0 42#include <linux/iscsi_ibft.h>
1da177e4 43#include <linux/nodemask.h>
1bc3b91a 44#include <linux/kexec.h>
2030eae5 45#include <linux/crash_dump.h>
e9928674 46#include <linux/dmi.h>
22a9835c 47#include <linux/pfn.h>
376ff035 48#include <linux/pci.h>
f212ec4b 49#include <linux/init_ohci1394_dma.h>
790c73f6 50#include <linux/kvm_para.h>
1bc3b91a 51
1da177e4 52#include <video/edid.h>
1bc3b91a 53
093af8d7 54#include <asm/mtrr.h>
9635b47d 55#include <asm/apic.h>
1da177e4
LT
56#include <asm/e820.h>
57#include <asm/mpspec.h>
91023300 58#include <asm/mmzone.h>
1da177e4
LT
59#include <asm/setup.h>
60#include <asm/arch_hooks.h>
61#include <asm/sections.h>
62#include <asm/io_apic.h>
63#include <asm/ist.h>
64#include <asm/io.h>
7ce0bcfd 65#include <asm/vmi.h>
e75eac33 66#include <setup_arch.h>
ce3fe6b2 67#include <asm/bios_ebda.h>
00bf4098 68#include <asm/cacheflush.h>
2fde61fd 69#include <asm/processor.h>
1da177e4
LT
70
71/* This value is set up by the early boot code to point to the value
72 immediately after the boot time page tables. It contains a *physical*
73 address, and must not be in the .bss segment! */
74unsigned long init_pg_tables_end __initdata = ~0UL;
75
1da177e4
LT
76/*
77 * Machine setup..
78 */
c9cce83d
BW
79static struct resource data_resource = {
80 .name = "Kernel data",
81 .start = 0,
82 .end = 0,
83 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
84};
85
86static struct resource code_resource = {
87 .name = "Kernel code",
88 .start = 0,
89 .end = 0,
90 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
91};
92
93static struct resource bss_resource = {
94 .name = "Kernel bss",
95 .start = 0,
96 .end = 0,
97 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
98};
99
100static struct resource video_ram_resource = {
101 .name = "Video RAM area",
102 .start = 0xa0000,
103 .end = 0xbffff,
104 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
105};
106
107static struct resource standard_io_resources[] = { {
108 .name = "dma1",
109 .start = 0x0000,
110 .end = 0x001f,
111 .flags = IORESOURCE_BUSY | IORESOURCE_IO
112}, {
113 .name = "pic1",
114 .start = 0x0020,
115 .end = 0x0021,
116 .flags = IORESOURCE_BUSY | IORESOURCE_IO
117}, {
118 .name = "timer0",
119 .start = 0x0040,
120 .end = 0x0043,
121 .flags = IORESOURCE_BUSY | IORESOURCE_IO
122}, {
123 .name = "timer1",
124 .start = 0x0050,
125 .end = 0x0053,
126 .flags = IORESOURCE_BUSY | IORESOURCE_IO
127}, {
128 .name = "keyboard",
129 .start = 0x0060,
9096bd7a
HW
130 .end = 0x0060,
131 .flags = IORESOURCE_BUSY | IORESOURCE_IO
132}, {
133 .name = "keyboard",
134 .start = 0x0064,
135 .end = 0x0064,
c9cce83d
BW
136 .flags = IORESOURCE_BUSY | IORESOURCE_IO
137}, {
138 .name = "dma page reg",
139 .start = 0x0080,
140 .end = 0x008f,
141 .flags = IORESOURCE_BUSY | IORESOURCE_IO
142}, {
143 .name = "pic2",
144 .start = 0x00a0,
145 .end = 0x00a1,
146 .flags = IORESOURCE_BUSY | IORESOURCE_IO
147}, {
148 .name = "dma2",
149 .start = 0x00c0,
150 .end = 0x00df,
151 .flags = IORESOURCE_BUSY | IORESOURCE_IO
152}, {
153 .name = "fpu",
154 .start = 0x00f0,
155 .end = 0x00ff,
156 .flags = IORESOURCE_BUSY | IORESOURCE_IO
157} };
1da177e4
LT
158
159/* cpu data as detected by the assembly code in head.S */
4a5d107a 160struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
1da177e4 161/* common cpu data for all cpus */
c3d8c141 162struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
129f6946 163EXPORT_SYMBOL(boot_cpu_data);
1da177e4 164
0c254e38
AS
165unsigned int def_to_bigsmp;
166
551889a6 167#ifndef CONFIG_X86_PAE
1da177e4 168unsigned long mmu_cr4_features;
551889a6
IC
169#else
170unsigned long mmu_cr4_features = X86_CR4_PAE;
171#endif
1da177e4 172
1da177e4
LT
173/* for MCA, but anyone else can use it if they want */
174unsigned int machine_id;
175unsigned int machine_submodel_id;
176unsigned int BIOS_revision;
1da177e4 177
1da177e4
LT
178/* Boot loader ID as an integer, for the benefit of proc_dointvec */
179int bootloader_type;
180
181/* user-defined highmem size */
182static unsigned int highmem_pages = -1;
183
184/*
185 * Setup options
186 */
1da177e4 187struct screen_info screen_info;
129f6946 188EXPORT_SYMBOL(screen_info);
1da177e4 189struct apm_info apm_info;
129f6946 190EXPORT_SYMBOL(apm_info);
1da177e4 191struct edid_info edid_info;
5e518d76 192EXPORT_SYMBOL_GPL(edid_info);
1da177e4 193struct ist_info ist_info;
129f6946
AD
194#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
195 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
196EXPORT_SYMBOL(ist_info);
197#endif
1da177e4
LT
198
199extern void early_cpu_init(void);
1da177e4
LT
200extern int root_mountflags;
201
e44b7b75 202unsigned long saved_video_mode;
1da177e4 203
cf8fa920 204#define RAMDISK_IMAGE_START_MASK 0x07FF
1da177e4 205#define RAMDISK_PROMPT_FLAG 0x8000
cf8fa920 206#define RAMDISK_LOAD_FLAG 0x4000
1da177e4 207
4e498b66 208static char __initdata command_line[COMMAND_LINE_SIZE];
1da177e4 209
6d7d7433 210#ifndef CONFIG_DEBUG_BOOT_PARAMS
48c7ae67 211struct boot_params __initdata boot_params;
6d7d7433
HY
212#else
213struct boot_params boot_params;
214#endif
1da177e4 215
1da177e4
LT
216#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
217struct edd edd;
218#ifdef CONFIG_EDD_MODULE
219EXPORT_SYMBOL(edd);
220#endif
221/**
222 * copy_edd() - Copy the BIOS EDD information
223 * from boot_params into a safe place.
224 *
225 */
226static inline void copy_edd(void)
227{
30c82645
PA
228 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
229 sizeof(edd.mbr_signature));
230 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
231 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
232 edd.edd_info_nr = boot_params.eddbuf_entries;
1da177e4
LT
233}
234#else
235static inline void copy_edd(void)
236{
237}
238#endif
239
1a3f239d
RR
240#ifdef CONFIG_PROC_VMCORE
241/* elfcorehdr= specifies the location of elf core header
242 * stored by the crashed kernel.
243 */
244static int __init parse_elfcorehdr(char *arg)
245{
246 if (!arg)
247 return -EINVAL;
66759a01 248
1a3f239d
RR
249 elfcorehdr_addr = memparse(arg, &arg);
250 return 0;
251}
252early_param("elfcorehdr", parse_elfcorehdr);
253#endif /* CONFIG_PROC_VMCORE */
1da177e4 254
1a3f239d
RR
255/*
256 * highmem=size forces highmem to be exactly 'size' bytes.
257 * This works even on boxes that have no highmem otherwise.
258 * This also works to reduce highmem size on bigger boxes.
259 */
260static int __init parse_highmem(char *arg)
261{
262 if (!arg)
263 return -EINVAL;
9635b47d 264
1a3f239d
RR
265 highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
266 return 0;
267}
268early_param("highmem", parse_highmem);
9635b47d 269
1a3f239d
RR
270/*
271 * vmalloc=size forces the vmalloc area to be exactly 'size'
272 * bytes. This can be used to increase (or decrease) the
273 * vmalloc area - the default is 128m.
274 */
275static int __init parse_vmalloc(char *arg)
276{
277 if (!arg)
278 return -EINVAL;
1bc3b91a 279
1a3f239d
RR
280 __VMALLOC_RESERVE = memparse(arg, &arg);
281 return 0;
1da177e4 282}
1a3f239d 283early_param("vmalloc", parse_vmalloc);
1da177e4 284
461a9aff
ZA
285/*
286 * reservetop=size reserves a hole at the top of the kernel address space which
287 * a hypervisor can load into later. Needed for dynamically loaded hypervisors,
288 * so relocating the fixmap can be done before paging initialization.
289 */
290static int __init parse_reservetop(char *arg)
291{
292 unsigned long address;
293
294 if (!arg)
295 return -EINVAL;
296
297 address = memparse(arg, &arg);
298 reserve_top_address(address);
299 return 0;
300}
301early_param("reservetop", parse_reservetop);
302
1da177e4
LT
303/*
304 * Determine low and high memory ranges:
305 */
306unsigned long __init find_max_low_pfn(void)
307{
308 unsigned long max_low_pfn;
309
310 max_low_pfn = max_pfn;
311 if (max_low_pfn > MAXMEM_PFN) {
312 if (highmem_pages == -1)
313 highmem_pages = max_pfn - MAXMEM_PFN;
314 if (highmem_pages + MAXMEM_PFN < max_pfn)
315 max_pfn = MAXMEM_PFN + highmem_pages;
316 if (highmem_pages + MAXMEM_PFN > max_pfn) {
317 printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
318 highmem_pages = 0;
319 }
320 max_low_pfn = MAXMEM_PFN;
321#ifndef CONFIG_HIGHMEM
322 /* Maximum memory usable is what is directly addressable */
323 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
324 MAXMEM>>20);
325 if (max_pfn > MAX_NONPAE_PFN)
c673f1a9 326 printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
1da177e4
LT
327 else
328 printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
329 max_pfn = MAXMEM_PFN;
330#else /* !CONFIG_HIGHMEM */
c673f1a9 331#ifndef CONFIG_HIGHMEM64G
1da177e4
LT
332 if (max_pfn > MAX_NONPAE_PFN) {
333 max_pfn = MAX_NONPAE_PFN;
334 printk(KERN_WARNING "Warning only 4GB will be used.\n");
c673f1a9 335 printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
1da177e4 336 }
c673f1a9 337#endif /* !CONFIG_HIGHMEM64G */
1da177e4
LT
338#endif /* !CONFIG_HIGHMEM */
339 } else {
340 if (highmem_pages == -1)
341 highmem_pages = 0;
342#ifdef CONFIG_HIGHMEM
343 if (highmem_pages >= max_pfn) {
344 printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
345 highmem_pages = 0;
346 }
347 if (highmem_pages) {
348 if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
349 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
350 highmem_pages = 0;
351 }
352 max_low_pfn -= highmem_pages;
353 }
354#else
355 if (highmem_pages)
356 printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
357#endif
358 }
359 return max_low_pfn;
360}
361
dedd04be
AH
362#define BIOS_LOWMEM_KILOBYTES 0x413
363
1da177e4 364/*
dedd04be
AH
365 * The BIOS places the EBDA/XBDA at the top of conventional
366 * memory, and usually decreases the reported amount of
367 * conventional memory (int 0x12) too. This also contains a
368 * workaround for Dell systems that neglect to reserve EBDA.
369 * The same workaround also avoids a problem with the AMD768MPX
370 * chipset: reserve a page before VGA to prevent PCI prefetch
371 * into it (errata #56). Usually the page is reserved anyways,
372 * unless you have no PS/2 mouse plugged in.
1da177e4
LT
373 */
374static void __init reserve_ebda_region(void)
375{
dedd04be
AH
376 unsigned int lowmem, ebda_addr;
377
378 /* To determine the position of the EBDA and the */
379 /* end of conventional memory, we need to look at */
380 /* the BIOS data area. In a paravirtual environment */
381 /* that area is absent. We'll just have to assume */
382 /* that the paravirt case can handle memory setup */
383 /* correctly, without our help. */
2fde61fd 384 if (paravirt_enabled())
dedd04be 385 return;
dedd04be
AH
386
387 /* end of low (conventional) memory */
388 lowmem = *(unsigned short *)__va(BIOS_LOWMEM_KILOBYTES);
389 lowmem <<= 10;
390
391 /* start of EBDA area */
356fa0c6 392 ebda_addr = get_bios_ebda();
dedd04be
AH
393
394 /* Fixup: bios puts an EBDA in the top 64K segment */
395 /* of conventional memory, but does not adjust lowmem. */
396 if ((lowmem - ebda_addr) <= 0x10000)
397 lowmem = ebda_addr;
398
399 /* Fixup: bios does not report an EBDA at all. */
400 /* Some old Dells seem to need 4k anyhow (bugzilla 2990) */
401 if ((ebda_addr == 0) && (lowmem >= 0x9f000))
402 lowmem = 0x9f000;
403
404 /* Paranoia: should never happen, but... */
405 if ((lowmem == 0) || (lowmem >= 0x100000))
406 lowmem = 0x9f000;
407
408 /* reserve all memory between lowmem and the 1MB mark */
409 reserve_bootmem(lowmem, 0x100000 - lowmem, BOOTMEM_DEFAULT);
1da177e4
LT
410}
411
05b79bdc 412#ifndef CONFIG_NEED_MULTIPLE_NODES
a4928cff 413static void __init setup_bootmem_allocator(void);
1da177e4
LT
414static unsigned long __init setup_memory(void)
415{
416 /*
417 * partially used pages are not usable - thus
418 * we are rounding upwards:
419 */
420 min_low_pfn = PFN_UP(init_pg_tables_end);
421
1da177e4
LT
422 max_low_pfn = find_max_low_pfn();
423
424#ifdef CONFIG_HIGHMEM
425 highstart_pfn = highend_pfn = max_pfn;
426 if (max_pfn > max_low_pfn) {
427 highstart_pfn = max_low_pfn;
428 }
429 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
430 pages_to_mb(highend_pfn - highstart_pfn));
ba9c231f
JB
431 num_physpages = highend_pfn;
432 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
433#else
434 num_physpages = max_low_pfn;
435 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
436#endif
437#ifdef CONFIG_FLATMEM
438 max_mapnr = num_physpages;
1da177e4
LT
439#endif
440 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
441 pages_to_mb(max_low_pfn));
442
443 setup_bootmem_allocator();
444
445 return max_low_pfn;
446}
447
a4928cff 448static void __init zone_sizes_init(void)
1da177e4 449{
6391af17
MG
450 unsigned long max_zone_pfns[MAX_NR_ZONES];
451 memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
452 max_zone_pfns[ZONE_DMA] =
453 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
454 max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
1da177e4 455#ifdef CONFIG_HIGHMEM
6391af17 456 max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
4cfee88a
MG
457 add_active_range(0, 0, highend_pfn);
458#else
4cfee88a 459 add_active_range(0, 0, max_low_pfn);
1da177e4 460#endif
4cfee88a
MG
461
462 free_area_init_nodes(max_zone_pfns);
1da177e4
LT
463}
464#else
05b79bdc 465extern unsigned long __init setup_memory(void);
1da177e4 466extern void zone_sizes_init(void);
05b79bdc 467#endif /* !CONFIG_NEED_MULTIPLE_NODES */
1da177e4 468
d62cc471
BW
469static inline unsigned long long get_total_mem(void)
470{
471 unsigned long long total;
472
473 total = max_low_pfn - min_low_pfn;
474#ifdef CONFIG_HIGHMEM
475 total += highend_pfn - highstart_pfn;
476#endif
477
478 return total << PAGE_SHIFT;
479}
480
481#ifdef CONFIG_KEXEC
482static void __init reserve_crashkernel(void)
483{
484 unsigned long long total_mem;
485 unsigned long long crash_size, crash_base;
486 int ret;
487
488 total_mem = get_total_mem();
489
490 ret = parse_crashkernel(boot_command_line, total_mem,
491 &crash_size, &crash_base);
492 if (ret == 0 && crash_size > 0) {
493 if (crash_base > 0) {
494 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
495 "for crashkernel (System RAM: %ldMB)\n",
496 (unsigned long)(crash_size >> 20),
497 (unsigned long)(crash_base >> 20),
498 (unsigned long)(total_mem >> 20));
499 crashk_res.start = crash_base;
500 crashk_res.end = crash_base + crash_size - 1;
72a7fe39
BW
501 reserve_bootmem(crash_base, crash_size,
502 BOOTMEM_DEFAULT);
d62cc471
BW
503 } else
504 printk(KERN_INFO "crashkernel reservation failed - "
505 "you have to specify a base address\n");
506 }
507}
508#else
509static inline void __init reserve_crashkernel(void)
510{}
511#endif
512
cf8fa920
PA
513#ifdef CONFIG_BLK_DEV_INITRD
514
515static bool do_relocate_initrd = false;
516
517static void __init reserve_initrd(void)
518{
519 unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
520 unsigned long ramdisk_size = boot_params.hdr.ramdisk_size;
521 unsigned long ramdisk_end = ramdisk_image + ramdisk_size;
522 unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT;
523 unsigned long ramdisk_here;
524
525 initrd_start = 0;
526
527 if (!boot_params.hdr.type_of_loader ||
528 !ramdisk_image || !ramdisk_size)
529 return; /* No initrd provided by bootloader */
530
531 if (ramdisk_end < ramdisk_image) {
532 printk(KERN_ERR "initrd wraps around end of memory, "
533 "disabling initrd\n");
534 return;
535 }
536 if (ramdisk_size >= end_of_lowmem/2) {
537 printk(KERN_ERR "initrd too large to handle, "
538 "disabling initrd\n");
539 return;
540 }
541 if (ramdisk_end <= end_of_lowmem) {
542 /* All in lowmem, easy case */
72a7fe39 543 reserve_bootmem(ramdisk_image, ramdisk_size, BOOTMEM_DEFAULT);
cf8fa920
PA
544 initrd_start = ramdisk_image + PAGE_OFFSET;
545 initrd_end = initrd_start+ramdisk_size;
546 return;
547 }
548
549 /* We need to move the initrd down into lowmem */
550 ramdisk_here = (end_of_lowmem - ramdisk_size) & PAGE_MASK;
551
552 /* Note: this includes all the lowmem currently occupied by
553 the initrd, we rely on that fact to keep the data intact. */
72a7fe39 554 reserve_bootmem(ramdisk_here, ramdisk_size, BOOTMEM_DEFAULT);
cf8fa920
PA
555 initrd_start = ramdisk_here + PAGE_OFFSET;
556 initrd_end = initrd_start + ramdisk_size;
557
558 do_relocate_initrd = true;
559}
560
561#define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
562
563static void __init relocate_initrd(void)
564{
565 unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
566 unsigned long ramdisk_size = boot_params.hdr.ramdisk_size;
567 unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT;
568 unsigned long ramdisk_here;
569 unsigned long slop, clen, mapaddr;
570 char *p, *q;
571
572 if (!do_relocate_initrd)
573 return;
574
575 ramdisk_here = initrd_start - PAGE_OFFSET;
576
577 q = (char *)initrd_start;
578
579 /* Copy any lowmem portion of the initrd */
580 if (ramdisk_image < end_of_lowmem) {
581 clen = end_of_lowmem - ramdisk_image;
582 p = (char *)__va(ramdisk_image);
583 memcpy(q, p, clen);
584 q += clen;
585 ramdisk_image += clen;
586 ramdisk_size -= clen;
587 }
588
589 /* Copy the highmem portion of the initrd */
590 while (ramdisk_size) {
591 slop = ramdisk_image & ~PAGE_MASK;
592 clen = ramdisk_size;
593 if (clen > MAX_MAP_CHUNK-slop)
594 clen = MAX_MAP_CHUNK-slop;
595 mapaddr = ramdisk_image & PAGE_MASK;
beacfaac 596 p = early_ioremap(mapaddr, clen+slop);
cf8fa920 597 memcpy(q, p+slop, clen);
beacfaac 598 early_iounmap(p, clen+slop);
cf8fa920
PA
599 q += clen;
600 ramdisk_image += clen;
601 ramdisk_size -= clen;
602 }
603}
604
605#endif /* CONFIG_BLK_DEV_INITRD */
606
1da177e4
LT
607void __init setup_bootmem_allocator(void)
608{
609 unsigned long bootmap_size;
610 /*
611 * Initialize the boot-time allocator (with low memory only):
612 */
613 bootmap_size = init_bootmem(min_low_pfn, max_low_pfn);
614
615 register_bootmem_low_pages(max_low_pfn);
616
617 /*
618 * Reserve the bootmem bitmap itself as well. We do this in two
619 * steps (first step was init_bootmem()) because this catches
620 * the (very unlikely) case of us accidentally initializing the
621 * bootmem allocator with an invalid RAM area.
622 */
8621b81c 623 reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) +
72a7fe39
BW
624 bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text),
625 BOOTMEM_DEFAULT);
1da177e4
LT
626
627 /*
628 * reserve physical page 0 - it's a special BIOS page on many boxes,
629 * enabling clean reboots, SMP operation, laptop functions.
630 */
72a7fe39 631 reserve_bootmem(0, PAGE_SIZE, BOOTMEM_DEFAULT);
1da177e4 632
dedd04be 633 /* reserve EBDA region */
1da177e4
LT
634 reserve_ebda_region();
635
1da177e4
LT
636#ifdef CONFIG_SMP
637 /*
638 * But first pinch a few for the stack/trampoline stuff
639 * FIXME: Don't need the extra page at 4K, but need to fix
640 * trampoline before removing it. (see the GDT stuff)
641 */
72a7fe39 642 reserve_bootmem(PAGE_SIZE, PAGE_SIZE, BOOTMEM_DEFAULT);
1da177e4 643#endif
673d5b43 644#ifdef CONFIG_ACPI_SLEEP
1da177e4
LT
645 /*
646 * Reserve low memory region for sleep support.
647 */
648 acpi_reserve_bootmem();
649#endif
650#ifdef CONFIG_X86_FIND_SMP_CONFIG
651 /*
652 * Find and reserve possible boot-time SMP configuration:
653 */
654 find_smp_config();
655#endif
1da177e4 656#ifdef CONFIG_BLK_DEV_INITRD
cf8fa920 657 reserve_initrd();
1da177e4 658#endif
cf8fa920 659 numa_kva_reserve();
d62cc471 660 reserve_crashkernel();
138fe4e0
KR
661
662 reserve_ibft_region();
1da177e4
LT
663}
664
665/*
666 * The node 0 pgdat is initialized before all of these because
667 * it's needed for bootmem. node>0 pgdats have their virtual
668 * space allocated before the pagetables are in place to access
669 * them, so they can't be cleared then.
670 *
671 * This should all compile down to nothing when NUMA is off.
672 */
10079ae3 673static void __init remapped_pgdat_init(void)
1da177e4
LT
674{
675 int nid;
676
677 for_each_online_node(nid) {
678 if (nid != 0)
679 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
680 }
681}
682
1da177e4
LT
683#ifdef CONFIG_MCA
684static void set_mca_bus(int x)
685{
686 MCA_bus = x;
687}
688#else
689static void set_mca_bus(int x) { }
690#endif
691
ccf82085
GOC
692#ifdef CONFIG_NUMA
693/*
694 * In the golden day, when everything among i386 and x86_64 will be
695 * integrated, this will not live here
696 */
697void *x86_cpu_to_node_map_early_ptr;
698int x86_cpu_to_node_map_init[NR_CPUS] = {
699 [0 ... NR_CPUS-1] = NUMA_NO_NODE
700};
701DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
702#endif
703
1da177e4
LT
704/*
705 * Determine if we were loaded by an EFI loader. If so, then we have also been
706 * passed the efi memmap, systab, etc., so we should use these data structures
707 * for initialization. Note, the efi init code path is determined by the
708 * global efi_enabled. This allows the same kernel image to be used on existing
709 * systems (with a traditional BIOS) as well as on EFI systems.
710 */
711void __init setup_arch(char **cmdline_p)
712{
713 unsigned long max_low_pfn;
714
715 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
716 pre_setup_arch_hook();
717 early_cpu_init();
beacfaac 718 early_ioremap_init();
1da177e4 719
1da177e4 720#ifdef CONFIG_EFI
e429795c
HY
721 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
722 "EL32", 4))
1da177e4
LT
723 efi_enabled = 1;
724#endif
725
30c82645
PA
726 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
727 screen_info = boot_params.screen_info;
728 edid_info = boot_params.edid_info;
729 apm_info.bios = boot_params.apm_bios_info;
730 ist_info = boot_params.ist_info;
e44b7b75 731 saved_video_mode = boot_params.hdr.vid_mode;
30c82645
PA
732 if( boot_params.sys_desc_table.length != 0 ) {
733 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
734 machine_id = boot_params.sys_desc_table.table[0];
735 machine_submodel_id = boot_params.sys_desc_table.table[1];
736 BIOS_revision = boot_params.sys_desc_table.table[2];
1da177e4 737 }
30c82645 738 bootloader_type = boot_params.hdr.type_of_loader;
1da177e4
LT
739
740#ifdef CONFIG_BLK_DEV_RAM
30c82645
PA
741 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
742 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
743 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
1da177e4
LT
744#endif
745 ARCH_SETUP
2215e69d 746
0dbfafa5 747 setup_memory_map();
2215e69d 748
1da177e4
LT
749 copy_edd();
750
30c82645 751 if (!boot_params.hdr.root_flags)
1da177e4
LT
752 root_mountflags &= ~MS_RDONLY;
753 init_mm.start_code = (unsigned long) _text;
754 init_mm.end_code = (unsigned long) _etext;
755 init_mm.end_data = (unsigned long) _edata;
756 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
757
758 code_resource.start = virt_to_phys(_text);
759 code_resource.end = virt_to_phys(_etext)-1;
760 data_resource.start = virt_to_phys(_etext);
761 data_resource.end = virt_to_phys(_edata)-1;
00bf4098
BW
762 bss_resource.start = virt_to_phys(&__bss_start);
763 bss_resource.end = virt_to_phys(&__bss_stop)-1;
1da177e4 764
1a3f239d 765 parse_early_param();
1da177e4 766
0dbfafa5 767 finish_e820_parsing();
1a3f239d 768
4e498b66 769 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
1a3f239d 770 *cmdline_p = command_line;
99b7de33 771
8b2cb7a8
HY
772 if (efi_enabled)
773 efi_init();
774
093af8d7 775 /* update e820 for memory not covered by WB MTRRs */
fa5c4639 776 propagate_e820_map();
093af8d7
YL
777 mtrr_bp_init();
778 if (mtrr_trim_uncached_memory(max_pfn))
fa5c4639 779 propagate_e820_map();
76c32418
YL
780
781 max_low_pfn = setup_memory();
093af8d7 782
790c73f6
GOC
783#ifdef CONFIG_KVM_CLOCK
784 kvmclock_init();
785#endif
786
7ce0bcfd
ZA
787#ifdef CONFIG_VMI
788 /*
789 * Must be after max_low_pfn is determined, and before kernel
790 * pagetables are setup.
791 */
792 vmi_init();
793#endif
0cf1bfd2 794 kvm_guest_init();
7ce0bcfd 795
1da177e4
LT
796 /*
797 * NOTE: before this point _nobody_ is allowed to allocate
798 * any memory using the bootmem allocator. Although the
27b46d76 799 * allocator is now initialised only the first 8Mb of the kernel
1da177e4
LT
800 * virtual address space has been mapped. All allocations before
801 * paging_init() has completed must use the alloc_bootmem_low_pages()
802 * variant (which allocates DMA'able memory) and care must be taken
803 * not to exceed the 8Mb limit.
804 */
805
806#ifdef CONFIG_SMP
807 smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
808#endif
809 paging_init();
f212ec4b
BK
810
811 /*
812 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
813 */
814
815#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
816 if (init_ohci1394_dma_early)
817 init_ohci1394_dma_on_all_controllers();
818#endif
819
1da177e4 820 remapped_pgdat_init();
05b79bdc 821 sparse_init();
1da177e4
LT
822 zone_sizes_init();
823
824 /*
825 * NOTE: at this point the bootmem allocator is fully available.
826 */
827
cf8fa920
PA
828#ifdef CONFIG_BLK_DEV_INITRD
829 relocate_initrd();
830#endif
831
6996d3b6
JF
832 paravirt_post_allocator_init();
833
1da177e4
LT
834 dmi_scan_machine();
835
cf8fa920 836 io_delay_init();
b02aae9c 837
ccf82085
GOC
838#ifdef CONFIG_X86_SMP
839 /*
840 * setup to use the early static init tables during kernel startup
841 * X86_SMP will exclude sub-arches that don't deal well with it.
842 */
843 x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init;
844 x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init;
845#ifdef CONFIG_NUMA
846 x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init;
847#endif
848#endif
849
1da177e4 850#ifdef CONFIG_X86_GENERICARCH
1a3f239d 851 generic_apic_probe();
cf8fa920 852#endif
1da177e4 853
888ba6c6 854#ifdef CONFIG_ACPI
1da177e4
LT
855 /*
856 * Parse the ACPI tables for possible boot-time SMP configuration.
857 */
858 acpi_boot_table_init();
d44647b0
AC
859#endif
860
54ef3400 861 early_quirks();
d44647b0
AC
862
863#ifdef CONFIG_ACPI
1da177e4
LT
864 acpi_boot_init();
865
911a62d4
VP
866#if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
867 if (def_to_bigsmp)
868 printk(KERN_WARNING "More than 8 CPUs detected and "
869 "CONFIG_X86_PC cannot handle it.\nUse "
870 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
871#endif
872#endif
1da177e4
LT
873#ifdef CONFIG_X86_LOCAL_APIC
874 if (smp_found_config)
875 get_smp_config();
876#endif
877
b79cd8f1 878 e820_setup_gap();
1c10070a 879 e820_mark_nosave_regions();
1da177e4
LT
880
881#ifdef CONFIG_VT
882#if defined(CONFIG_VGA_CONSOLE)
883 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
884 conswitchp = &vga_con;
885#elif defined(CONFIG_DUMMY_CONSOLE)
886 conswitchp = &dummy_con;
887#endif
888#endif
889}
c9cce83d
BW
890
891/*
892 * Request address space for all standard resources
893 *
894 * This is called just before pcibios_init(), which is also a
895 * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
896 */
897static int __init request_standard_resources(void)
898{
899 int i;
900
901 printk(KERN_INFO "Setting up standard PCI resources\n");
2215e69d 902 init_iomem_resources(&code_resource, &data_resource, &bss_resource);
c9cce83d 903
c9cce83d
BW
904 request_resource(&iomem_resource, &video_ram_resource);
905
906 /* request I/O space for devices used on all i[345]86 PCs */
907 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
908 request_resource(&ioport_resource, &standard_io_resources[i]);
909 return 0;
910}
911
912subsys_initcall(request_standard_resources);
This page took 0.428416 seconds and 5 git commands to generate.