Loading...
1/*
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>
26#include <linux/mmzone.h>
27#include <linux/screen_info.h>
28#include <linux/ioport.h>
29#include <linux/acpi.h>
30#include <linux/sfi.h>
31#include <linux/apm_bios.h>
32#include <linux/initrd.h>
33#include <linux/bootmem.h>
34#include <linux/memblock.h>
35#include <linux/seq_file.h>
36#include <linux/console.h>
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>
43#include <linux/iscsi_ibft.h>
44#include <linux/nodemask.h>
45#include <linux/kexec.h>
46#include <linux/dmi.h>
47#include <linux/pfn.h>
48#include <linux/pci.h>
49#include <asm/pci-direct.h>
50#include <linux/init_ohci1394_dma.h>
51#include <linux/kvm_para.h>
52#include <linux/dma-contiguous.h>
53
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>
59#include <linux/user.h>
60#include <linux/delay.h>
61
62#include <linux/kallsyms.h>
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>
70#include <linux/tboot.h>
71
72#include <video/edid.h>
73
74#include <asm/mtrr.h>
75#include <asm/apic.h>
76#include <asm/realmode.h>
77#include <asm/e820.h>
78#include <asm/mpspec.h>
79#include <asm/setup.h>
80#include <asm/efi.h>
81#include <asm/timer.h>
82#include <asm/i8259.h>
83#include <asm/sections.h>
84#include <asm/dmi.h>
85#include <asm/io_apic.h>
86#include <asm/ist.h>
87#include <asm/setup_arch.h>
88#include <asm/bios_ebda.h>
89#include <asm/cacheflush.h>
90#include <asm/processor.h>
91#include <asm/bugs.h>
92
93#include <asm/vsyscall.h>
94#include <asm/cpu.h>
95#include <asm/desc.h>
96#include <asm/dma.h>
97#include <asm/iommu.h>
98#include <asm/gart.h>
99#include <asm/mmu_context.h>
100#include <asm/proto.h>
101
102#include <asm/paravirt.h>
103#include <asm/hypervisor.h>
104#include <asm/olpc_ofw.h>
105
106#include <asm/percpu.h>
107#include <asm/topology.h>
108#include <asm/apicdef.h>
109#include <asm/amd_nb.h>
110#ifdef CONFIG_X86_64
111#include <asm/numa_64.h>
112#endif
113#include <asm/mce.h>
114#include <asm/alternative.h>
115#include <asm/prom.h>
116
117/*
118 * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
119 * The direct mapping extends to max_pfn_mapped, so that we can directly access
120 * apertures, ACPI and other tables without having to play with fixmaps.
121 */
122unsigned long max_low_pfn_mapped;
123unsigned long max_pfn_mapped;
124
125#ifdef CONFIG_DMI
126RESERVE_BRK(dmi_alloc, 65536);
127#endif
128
129
130static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
131unsigned long _brk_end = (unsigned long)__brk_base;
132
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
139int default_check_phys_apicid_present(int phys_apicid)
140{
141 return __default_check_phys_apicid_present(phys_apicid);
142}
143#endif
144
145#ifndef CONFIG_DEBUG_BOOT_PARAMS
146struct boot_params __initdata boot_params;
147#else
148struct boot_params boot_params;
149#endif
150
151/*
152 * Machine setup..
153 */
154static struct resource data_resource = {
155 .name = "Kernel data",
156 .start = 0,
157 .end = 0,
158 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
159};
160
161static struct resource code_resource = {
162 .name = "Kernel code",
163 .start = 0,
164 .end = 0,
165 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
166};
167
168static struct resource bss_resource = {
169 .name = "Kernel bss",
170 .start = 0,
171 .end = 0,
172 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
173};
174
175
176#ifdef CONFIG_X86_32
177/* cpu data as detected by the assembly code in head.S */
178struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
179/* common cpu data for all cpus */
180struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
181EXPORT_SYMBOL(boot_cpu_data);
182
183unsigned int def_to_bigsmp;
184
185/* for MCA, but anyone else can use it if they want */
186unsigned int machine_id;
187unsigned int machine_submodel_id;
188unsigned int BIOS_revision;
189
190struct apm_info apm_info;
191EXPORT_SYMBOL(apm_info);
192
193#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
194 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
195struct ist_info ist_info;
196EXPORT_SYMBOL(ist_info);
197#else
198struct ist_info ist_info;
199#endif
200
201#else
202struct cpuinfo_x86 boot_cpu_data __read_mostly = {
203 .x86_phys_bits = MAX_PHYSMEM_BITS,
204};
205EXPORT_SYMBOL(boot_cpu_data);
206#endif
207
208
209#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
210unsigned long mmu_cr4_features;
211#else
212unsigned long mmu_cr4_features = X86_CR4_PAE;
213#endif
214
215/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
216int bootloader_type, bootloader_version;
217
218/*
219 * Setup options
220 */
221struct screen_info screen_info;
222EXPORT_SYMBOL(screen_info);
223struct edid_info edid_info;
224EXPORT_SYMBOL_GPL(edid_info);
225
226extern int root_mountflags;
227
228unsigned long saved_video_mode;
229
230#define RAMDISK_IMAGE_START_MASK 0x07FF
231#define RAMDISK_PROMPT_FLAG 0x8000
232#define RAMDISK_LOAD_FLAG 0x4000
233
234static char __initdata command_line[COMMAND_LINE_SIZE];
235#ifdef CONFIG_CMDLINE_BOOL
236static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
237#endif
238
239#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
240struct edd edd;
241#ifdef CONFIG_EDD_MODULE
242EXPORT_SYMBOL(edd);
243#endif
244/**
245 * copy_edd() - Copy the BIOS EDD information
246 * from boot_params into a safe place.
247 *
248 */
249static inline void __init copy_edd(void)
250{
251 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
252 sizeof(edd.mbr_signature));
253 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
254 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
255 edd.edd_info_nr = boot_params.eddbuf_entries;
256}
257#else
258static inline void __init copy_edd(void)
259{
260}
261#endif
262
263void * __init extend_brk(size_t size, size_t align)
264{
265 size_t mask = align - 1;
266 void *ret;
267
268 BUG_ON(_brk_start == 0);
269 BUG_ON(align & mask);
270
271 _brk_end = (_brk_end + mask) & ~mask;
272 BUG_ON((char *)(_brk_end + size) > __brk_limit);
273
274 ret = (void *)_brk_end;
275 _brk_end += size;
276
277 memset(ret, 0, size);
278
279 return ret;
280}
281
282#ifdef CONFIG_X86_64
283static void __init init_gbpages(void)
284{
285 if (direct_gbpages && cpu_has_gbpages)
286 printk(KERN_INFO "Using GB pages for direct mapping\n");
287 else
288 direct_gbpages = 0;
289}
290#else
291static inline void init_gbpages(void)
292{
293}
294static void __init cleanup_highmap(void)
295{
296}
297#endif
298
299static void __init reserve_brk(void)
300{
301 if (_brk_end > _brk_start)
302 memblock_reserve(__pa(_brk_start),
303 __pa(_brk_end) - __pa(_brk_start));
304
305 /* Mark brk area as locked down and no longer taking any
306 new allocations */
307 _brk_start = 0;
308}
309
310#ifdef CONFIG_BLK_DEV_INITRD
311
312#define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
313static void __init relocate_initrd(void)
314{
315 /* Assume only end is not page aligned */
316 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
317 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
318 u64 area_size = PAGE_ALIGN(ramdisk_size);
319 u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
320 u64 ramdisk_here;
321 unsigned long slop, clen, mapaddr;
322 char *p, *q;
323
324 /* We need to move the initrd down into lowmem */
325 ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
326 PAGE_SIZE);
327
328 if (!ramdisk_here)
329 panic("Cannot find place for new RAMDISK of size %lld\n",
330 ramdisk_size);
331
332 /* Note: this includes all the lowmem currently occupied by
333 the initrd, we rely on that fact to keep the data intact. */
334 memblock_reserve(ramdisk_here, area_size);
335 initrd_start = ramdisk_here + PAGE_OFFSET;
336 initrd_end = initrd_start + ramdisk_size;
337 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
338 ramdisk_here, ramdisk_here + ramdisk_size - 1);
339
340 q = (char *)initrd_start;
341
342 /* Copy any lowmem portion of the initrd */
343 if (ramdisk_image < end_of_lowmem) {
344 clen = end_of_lowmem - ramdisk_image;
345 p = (char *)__va(ramdisk_image);
346 memcpy(q, p, clen);
347 q += clen;
348 ramdisk_image += clen;
349 ramdisk_size -= clen;
350 }
351
352 /* Copy the highmem portion of the initrd */
353 while (ramdisk_size) {
354 slop = ramdisk_image & ~PAGE_MASK;
355 clen = ramdisk_size;
356 if (clen > MAX_MAP_CHUNK-slop)
357 clen = MAX_MAP_CHUNK-slop;
358 mapaddr = ramdisk_image & PAGE_MASK;
359 p = early_memremap(mapaddr, clen+slop);
360 memcpy(q, p+slop, clen);
361 early_iounmap(p, clen+slop);
362 q += clen;
363 ramdisk_image += clen;
364 ramdisk_size -= clen;
365 }
366 /* high pages is not converted by early_res_to_bootmem */
367 ramdisk_image = boot_params.hdr.ramdisk_image;
368 ramdisk_size = boot_params.hdr.ramdisk_size;
369 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
370 " [mem %#010llx-%#010llx]\n",
371 ramdisk_image, ramdisk_image + ramdisk_size - 1,
372 ramdisk_here, ramdisk_here + ramdisk_size - 1);
373}
374
375static void __init reserve_initrd(void)
376{
377 /* Assume only end is not page aligned */
378 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
379 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
380 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
381 u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
382
383 if (!boot_params.hdr.type_of_loader ||
384 !ramdisk_image || !ramdisk_size)
385 return; /* No initrd provided by bootloader */
386
387 initrd_start = 0;
388
389 if (ramdisk_size >= (end_of_lowmem>>1)) {
390 panic("initrd too large to handle, "
391 "disabling initrd (%lld needed, %lld available)\n",
392 ramdisk_size, end_of_lowmem>>1);
393 }
394
395 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
396 ramdisk_end - 1);
397
398
399 if (ramdisk_end <= end_of_lowmem) {
400 /* All in lowmem, easy case */
401 /*
402 * don't need to reserve again, already reserved early
403 * in i386_start_kernel
404 */
405 initrd_start = ramdisk_image + PAGE_OFFSET;
406 initrd_end = initrd_start + ramdisk_size;
407 return;
408 }
409
410 relocate_initrd();
411
412 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
413}
414#else
415static void __init reserve_initrd(void)
416{
417}
418#endif /* CONFIG_BLK_DEV_INITRD */
419
420static void __init parse_setup_data(void)
421{
422 struct setup_data *data;
423 u64 pa_data;
424
425 if (boot_params.hdr.version < 0x0209)
426 return;
427 pa_data = boot_params.hdr.setup_data;
428 while (pa_data) {
429 u32 data_len, map_len;
430
431 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
432 (u64)sizeof(struct setup_data));
433 data = early_memremap(pa_data, map_len);
434 data_len = data->len + sizeof(struct setup_data);
435 if (data_len > map_len) {
436 early_iounmap(data, map_len);
437 data = early_memremap(pa_data, data_len);
438 map_len = data_len;
439 }
440
441 switch (data->type) {
442 case SETUP_E820_EXT:
443 parse_e820_ext(data);
444 break;
445 case SETUP_DTB:
446 add_dtb(pa_data);
447 break;
448 default:
449 break;
450 }
451 pa_data = data->next;
452 early_iounmap(data, map_len);
453 }
454}
455
456static void __init e820_reserve_setup_data(void)
457{
458 struct setup_data *data;
459 u64 pa_data;
460 int found = 0;
461
462 if (boot_params.hdr.version < 0x0209)
463 return;
464 pa_data = boot_params.hdr.setup_data;
465 while (pa_data) {
466 data = early_memremap(pa_data, sizeof(*data));
467 e820_update_range(pa_data, sizeof(*data)+data->len,
468 E820_RAM, E820_RESERVED_KERN);
469 found = 1;
470 pa_data = data->next;
471 early_iounmap(data, sizeof(*data));
472 }
473 if (!found)
474 return;
475
476 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
477 memcpy(&e820_saved, &e820, sizeof(struct e820map));
478 printk(KERN_INFO "extended physical RAM map:\n");
479 e820_print_map("reserve setup_data");
480}
481
482static void __init memblock_x86_reserve_range_setup_data(void)
483{
484 struct setup_data *data;
485 u64 pa_data;
486
487 if (boot_params.hdr.version < 0x0209)
488 return;
489 pa_data = boot_params.hdr.setup_data;
490 while (pa_data) {
491 data = early_memremap(pa_data, sizeof(*data));
492 memblock_reserve(pa_data, sizeof(*data) + data->len);
493 pa_data = data->next;
494 early_iounmap(data, sizeof(*data));
495 }
496}
497
498/*
499 * --------- Crashkernel reservation ------------------------------
500 */
501
502#ifdef CONFIG_KEXEC
503
504/*
505 * Keep the crash kernel below this limit. On 32 bits earlier kernels
506 * would limit the kernel to the low 512 MiB due to mapping restrictions.
507 * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
508 * limit once kexec-tools are fixed.
509 */
510#ifdef CONFIG_X86_32
511# define CRASH_KERNEL_ADDR_MAX (512 << 20)
512#else
513# define CRASH_KERNEL_ADDR_MAX (896 << 20)
514#endif
515
516static void __init reserve_crashkernel(void)
517{
518 unsigned long long total_mem;
519 unsigned long long crash_size, crash_base;
520 int ret;
521
522 total_mem = memblock_phys_mem_size();
523
524 ret = parse_crashkernel(boot_command_line, total_mem,
525 &crash_size, &crash_base);
526 if (ret != 0 || crash_size <= 0)
527 return;
528
529 /* 0 means: find the address automatically */
530 if (crash_base <= 0) {
531 const unsigned long long alignment = 16<<20; /* 16M */
532
533 /*
534 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
535 */
536 crash_base = memblock_find_in_range(alignment,
537 CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
538
539 if (!crash_base) {
540 pr_info("crashkernel reservation failed - No suitable area found.\n");
541 return;
542 }
543 } else {
544 unsigned long long start;
545
546 start = memblock_find_in_range(crash_base,
547 crash_base + crash_size, crash_size, 1<<20);
548 if (start != crash_base) {
549 pr_info("crashkernel reservation failed - memory is in use.\n");
550 return;
551 }
552 }
553 memblock_reserve(crash_base, crash_size);
554
555 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
556 "for crashkernel (System RAM: %ldMB)\n",
557 (unsigned long)(crash_size >> 20),
558 (unsigned long)(crash_base >> 20),
559 (unsigned long)(total_mem >> 20));
560
561 crashk_res.start = crash_base;
562 crashk_res.end = crash_base + crash_size - 1;
563 insert_resource(&iomem_resource, &crashk_res);
564}
565#else
566static void __init reserve_crashkernel(void)
567{
568}
569#endif
570
571static struct resource standard_io_resources[] = {
572 { .name = "dma1", .start = 0x00, .end = 0x1f,
573 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
574 { .name = "pic1", .start = 0x20, .end = 0x21,
575 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
576 { .name = "timer0", .start = 0x40, .end = 0x43,
577 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
578 { .name = "timer1", .start = 0x50, .end = 0x53,
579 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
580 { .name = "keyboard", .start = 0x60, .end = 0x60,
581 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
582 { .name = "keyboard", .start = 0x64, .end = 0x64,
583 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
584 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
585 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
586 { .name = "pic2", .start = 0xa0, .end = 0xa1,
587 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
588 { .name = "dma2", .start = 0xc0, .end = 0xdf,
589 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
590 { .name = "fpu", .start = 0xf0, .end = 0xff,
591 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
592};
593
594void __init reserve_standard_io_resources(void)
595{
596 int i;
597
598 /* request I/O space for devices used on all i[345]86 PCs */
599 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
600 request_resource(&ioport_resource, &standard_io_resources[i]);
601
602}
603
604static __init void reserve_ibft_region(void)
605{
606 unsigned long addr, size = 0;
607
608 addr = find_ibft_region(&size);
609
610 if (size)
611 memblock_reserve(addr, size);
612}
613
614static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
615
616static void __init trim_bios_range(void)
617{
618 /*
619 * A special case is the first 4Kb of memory;
620 * This is a BIOS owned area, not kernel ram, but generally
621 * not listed as such in the E820 table.
622 *
623 * This typically reserves additional memory (64KiB by default)
624 * since some BIOSes are known to corrupt low memory. See the
625 * Kconfig help text for X86_RESERVE_LOW.
626 */
627 e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
628 E820_RAM, E820_RESERVED);
629
630 /*
631 * special case: Some BIOSen report the PC BIOS
632 * area (640->1Mb) as ram even though it is not.
633 * take them out.
634 */
635 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
636 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
637}
638
639static int __init parse_reservelow(char *p)
640{
641 unsigned long long size;
642
643 if (!p)
644 return -EINVAL;
645
646 size = memparse(p, &p);
647
648 if (size < 4096)
649 size = 4096;
650
651 if (size > 640*1024)
652 size = 640*1024;
653
654 reserve_low = size;
655
656 return 0;
657}
658
659early_param("reservelow", parse_reservelow);
660
661/*
662 * Determine if we were loaded by an EFI loader. If so, then we have also been
663 * passed the efi memmap, systab, etc., so we should use these data structures
664 * for initialization. Note, the efi init code path is determined by the
665 * global efi_enabled. This allows the same kernel image to be used on existing
666 * systems (with a traditional BIOS) as well as on EFI systems.
667 */
668/*
669 * setup_arch - architecture-specific boot-time initializations
670 *
671 * Note: On x86_64, fixmaps are ready for use even before this is called.
672 */
673
674void __init setup_arch(char **cmdline_p)
675{
676#ifdef CONFIG_X86_32
677 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
678 visws_early_detect();
679
680 /*
681 * copy kernel address range established so far and switch
682 * to the proper swapper page table
683 */
684 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
685 initial_page_table + KERNEL_PGD_BOUNDARY,
686 KERNEL_PGD_PTRS);
687
688 load_cr3(swapper_pg_dir);
689 __flush_tlb_all();
690#else
691 printk(KERN_INFO "Command line: %s\n", boot_command_line);
692#endif
693
694 /*
695 * If we have OLPC OFW, we might end up relocating the fixmap due to
696 * reserve_top(), so do this before touching the ioremap area.
697 */
698 olpc_ofw_detect();
699
700 early_trap_init();
701 early_cpu_init();
702 early_ioremap_init();
703
704 setup_olpc_ofw_pgd();
705
706 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
707 screen_info = boot_params.screen_info;
708 edid_info = boot_params.edid_info;
709#ifdef CONFIG_X86_32
710 apm_info.bios = boot_params.apm_bios_info;
711 ist_info = boot_params.ist_info;
712 if (boot_params.sys_desc_table.length != 0) {
713 machine_id = boot_params.sys_desc_table.table[0];
714 machine_submodel_id = boot_params.sys_desc_table.table[1];
715 BIOS_revision = boot_params.sys_desc_table.table[2];
716 }
717#endif
718 saved_video_mode = boot_params.hdr.vid_mode;
719 bootloader_type = boot_params.hdr.type_of_loader;
720 if ((bootloader_type >> 4) == 0xe) {
721 bootloader_type &= 0xf;
722 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
723 }
724 bootloader_version = bootloader_type & 0xf;
725 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
726
727#ifdef CONFIG_BLK_DEV_RAM
728 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
729 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
730 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
731#endif
732#ifdef CONFIG_EFI
733 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
734 "EL32", 4)) {
735 efi_enabled = 1;
736 efi_64bit = false;
737 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
738 "EL64", 4)) {
739 efi_enabled = 1;
740 efi_64bit = true;
741 }
742 if (efi_enabled && efi_memblock_x86_reserve_range())
743 efi_enabled = 0;
744#endif
745
746 x86_init.oem.arch_setup();
747
748 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
749 setup_memory_map();
750 parse_setup_data();
751 /* update the e820_saved too */
752 e820_reserve_setup_data();
753
754 copy_edd();
755
756 if (!boot_params.hdr.root_flags)
757 root_mountflags &= ~MS_RDONLY;
758 init_mm.start_code = (unsigned long) _text;
759 init_mm.end_code = (unsigned long) _etext;
760 init_mm.end_data = (unsigned long) _edata;
761 init_mm.brk = _brk_end;
762
763 code_resource.start = virt_to_phys(_text);
764 code_resource.end = virt_to_phys(_etext)-1;
765 data_resource.start = virt_to_phys(_etext);
766 data_resource.end = virt_to_phys(_edata)-1;
767 bss_resource.start = virt_to_phys(&__bss_start);
768 bss_resource.end = virt_to_phys(&__bss_stop)-1;
769
770#ifdef CONFIG_CMDLINE_BOOL
771#ifdef CONFIG_CMDLINE_OVERRIDE
772 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
773#else
774 if (builtin_cmdline[0]) {
775 /* append boot loader cmdline to builtin */
776 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
777 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
778 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
779 }
780#endif
781#endif
782
783 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
784 *cmdline_p = command_line;
785
786 /*
787 * x86_configure_nx() is called before parse_early_param() to detect
788 * whether hardware doesn't support NX (so that the early EHCI debug
789 * console setup can safely call set_fixmap()). It may then be called
790 * again from within noexec_setup() during parsing early parameters
791 * to honor the respective command line option.
792 */
793 x86_configure_nx();
794
795 parse_early_param();
796
797 x86_report_nx();
798
799 /* after early param, so could get panic from serial */
800 memblock_x86_reserve_range_setup_data();
801
802 if (acpi_mps_check()) {
803#ifdef CONFIG_X86_LOCAL_APIC
804 disable_apic = 1;
805#endif
806 setup_clear_cpu_cap(X86_FEATURE_APIC);
807 }
808
809#ifdef CONFIG_PCI
810 if (pci_early_dump_regs)
811 early_dump_pci_devices();
812#endif
813
814 finish_e820_parsing();
815
816 if (efi_enabled)
817 efi_init();
818
819 dmi_scan_machine();
820
821 /*
822 * VMware detection requires dmi to be available, so this
823 * needs to be done after dmi_scan_machine, for the BP.
824 */
825 init_hypervisor_platform();
826
827 x86_init.resources.probe_roms();
828
829 /* after parse_early_param, so could debug it */
830 insert_resource(&iomem_resource, &code_resource);
831 insert_resource(&iomem_resource, &data_resource);
832 insert_resource(&iomem_resource, &bss_resource);
833
834 trim_bios_range();
835#ifdef CONFIG_X86_32
836 if (ppro_with_ram_bug()) {
837 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
838 E820_RESERVED);
839 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
840 printk(KERN_INFO "fixed physical RAM map:\n");
841 e820_print_map("bad_ppro");
842 }
843#else
844 early_gart_iommu_check();
845#endif
846
847 /*
848 * partially used pages are not usable - thus
849 * we are rounding upwards:
850 */
851 max_pfn = e820_end_of_ram_pfn();
852
853 /* update e820 for memory not covered by WB MTRRs */
854 mtrr_bp_init();
855 if (mtrr_trim_uncached_memory(max_pfn))
856 max_pfn = e820_end_of_ram_pfn();
857
858#ifdef CONFIG_X86_32
859 /* max_low_pfn get updated here */
860 find_low_pfn_range();
861#else
862 num_physpages = max_pfn;
863
864 check_x2apic();
865
866 /* How many end-of-memory variables you have, grandma! */
867 /* need this before calling reserve_initrd */
868 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
869 max_low_pfn = e820_end_of_low_ram_pfn();
870 else
871 max_low_pfn = max_pfn;
872
873 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
874#endif
875
876 /*
877 * Find and reserve possible boot-time SMP configuration:
878 */
879 find_smp_config();
880
881 reserve_ibft_region();
882
883 /*
884 * Need to conclude brk, before memblock_x86_fill()
885 * it could use memblock_find_in_range, could overlap with
886 * brk area.
887 */
888 reserve_brk();
889
890 cleanup_highmap();
891
892 memblock.current_limit = get_max_mapped();
893 memblock_x86_fill();
894
895 /*
896 * The EFI specification says that boot service code won't be called
897 * after ExitBootServices(). This is, in fact, a lie.
898 */
899 if (efi_enabled)
900 efi_reserve_boot_services();
901
902 /* preallocate 4k for mptable mpc */
903 early_reserve_e820_mpc_new();
904
905#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
906 setup_bios_corruption_check();
907#endif
908
909 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
910 (max_pfn_mapped<<PAGE_SHIFT) - 1);
911
912 setup_real_mode();
913
914 init_gbpages();
915
916 /* max_pfn_mapped is updated here */
917 max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
918 max_pfn_mapped = max_low_pfn_mapped;
919
920#ifdef CONFIG_X86_64
921 if (max_pfn > max_low_pfn) {
922 max_pfn_mapped = init_memory_mapping(1UL<<32,
923 max_pfn<<PAGE_SHIFT);
924 /* can we preseve max_low_pfn ?*/
925 max_low_pfn = max_pfn;
926 }
927#endif
928 memblock.current_limit = get_max_mapped();
929 dma_contiguous_reserve(0);
930
931 /*
932 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
933 */
934
935#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
936 if (init_ohci1394_dma_early)
937 init_ohci1394_dma_on_all_controllers();
938#endif
939 /* Allocate bigger log buffer */
940 setup_log_buf(1);
941
942 reserve_initrd();
943
944 reserve_crashkernel();
945
946 vsmp_init();
947
948 io_delay_init();
949
950 /*
951 * Parse the ACPI tables for possible boot-time SMP configuration.
952 */
953 acpi_boot_table_init();
954
955 early_acpi_boot_init();
956
957 initmem_init();
958 memblock_find_dma_reserve();
959
960#ifdef CONFIG_KVM_CLOCK
961 kvmclock_init();
962#endif
963
964 x86_init.paging.pagetable_setup_start(swapper_pg_dir);
965 paging_init();
966 x86_init.paging.pagetable_setup_done(swapper_pg_dir);
967
968 if (boot_cpu_data.cpuid_level >= 0) {
969 /* A CPU has %cr4 if and only if it has CPUID */
970 mmu_cr4_features = read_cr4();
971 if (trampoline_cr4_features)
972 *trampoline_cr4_features = mmu_cr4_features;
973 }
974
975#ifdef CONFIG_X86_32
976 /* sync back kernel address range */
977 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
978 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
979 KERNEL_PGD_PTRS);
980#endif
981
982 tboot_probe();
983
984#ifdef CONFIG_X86_64
985 map_vsyscall();
986#endif
987
988 generic_apic_probe();
989
990 early_quirks();
991
992 /*
993 * Read APIC and some other early information from ACPI tables.
994 */
995 acpi_boot_init();
996 sfi_init();
997 x86_dtb_init();
998
999 /*
1000 * get boot-time SMP configuration:
1001 */
1002 if (smp_found_config)
1003 get_smp_config();
1004
1005 prefill_possible_map();
1006
1007 init_cpu_to_node();
1008
1009 init_apic_mappings();
1010 if (x86_io_apic_ops.init)
1011 x86_io_apic_ops.init();
1012
1013 kvm_guest_init();
1014
1015 e820_reserve_resources();
1016 e820_mark_nosave_regions(max_low_pfn);
1017
1018 x86_init.resources.reserve_resources();
1019
1020 e820_setup_gap();
1021
1022#ifdef CONFIG_VT
1023#if defined(CONFIG_VGA_CONSOLE)
1024 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1025 conswitchp = &vga_con;
1026#elif defined(CONFIG_DUMMY_CONSOLE)
1027 conswitchp = &dummy_con;
1028#endif
1029#endif
1030 x86_init.oem.banner();
1031
1032 x86_init.timers.wallclock_init();
1033
1034 x86_platform.wallclock_init();
1035
1036 mcheck_init();
1037
1038 arch_init_ideal_nops();
1039}
1040
1041#ifdef CONFIG_X86_32
1042
1043static struct resource video_ram_resource = {
1044 .name = "Video RAM area",
1045 .start = 0xa0000,
1046 .end = 0xbffff,
1047 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1048};
1049
1050void __init i386_reserve_resources(void)
1051{
1052 request_resource(&iomem_resource, &video_ram_resource);
1053 reserve_standard_io_resources();
1054}
1055
1056#endif /* CONFIG_X86_32 */
1/*
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>
26#include <linux/mmzone.h>
27#include <linux/screen_info.h>
28#include <linux/ioport.h>
29#include <linux/acpi.h>
30#include <linux/sfi.h>
31#include <linux/apm_bios.h>
32#include <linux/initrd.h>
33#include <linux/bootmem.h>
34#include <linux/memblock.h>
35#include <linux/seq_file.h>
36#include <linux/console.h>
37#include <linux/root_dev.h>
38#include <linux/highmem.h>
39#include <linux/export.h>
40#include <linux/efi.h>
41#include <linux/init.h>
42#include <linux/edd.h>
43#include <linux/iscsi_ibft.h>
44#include <linux/nodemask.h>
45#include <linux/kexec.h>
46#include <linux/dmi.h>
47#include <linux/pfn.h>
48#include <linux/pci.h>
49#include <asm/pci-direct.h>
50#include <linux/init_ohci1394_dma.h>
51#include <linux/kvm_para.h>
52#include <linux/dma-contiguous.h>
53
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>
59#include <linux/user.h>
60#include <linux/delay.h>
61
62#include <linux/kallsyms.h>
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>
70#include <linux/tboot.h>
71#include <linux/jiffies.h>
72
73#include <video/edid.h>
74
75#include <asm/mtrr.h>
76#include <asm/apic.h>
77#include <asm/realmode.h>
78#include <asm/e820.h>
79#include <asm/mpspec.h>
80#include <asm/setup.h>
81#include <asm/efi.h>
82#include <asm/timer.h>
83#include <asm/i8259.h>
84#include <asm/sections.h>
85#include <asm/io_apic.h>
86#include <asm/ist.h>
87#include <asm/setup_arch.h>
88#include <asm/bios_ebda.h>
89#include <asm/cacheflush.h>
90#include <asm/processor.h>
91#include <asm/bugs.h>
92#include <asm/kasan.h>
93
94#include <asm/vsyscall.h>
95#include <asm/cpu.h>
96#include <asm/desc.h>
97#include <asm/dma.h>
98#include <asm/iommu.h>
99#include <asm/gart.h>
100#include <asm/mmu_context.h>
101#include <asm/proto.h>
102
103#include <asm/paravirt.h>
104#include <asm/hypervisor.h>
105#include <asm/olpc_ofw.h>
106
107#include <asm/percpu.h>
108#include <asm/topology.h>
109#include <asm/apicdef.h>
110#include <asm/amd_nb.h>
111#include <asm/mce.h>
112#include <asm/alternative.h>
113#include <asm/prom.h>
114#include <asm/microcode.h>
115#include <asm/mmu_context.h>
116#include <asm/kaslr.h>
117
118/*
119 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
120 * max_pfn_mapped: highest direct mapped pfn over 4GB
121 *
122 * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
123 * represented by pfn_mapped
124 */
125unsigned long max_low_pfn_mapped;
126unsigned long max_pfn_mapped;
127
128#ifdef CONFIG_DMI
129RESERVE_BRK(dmi_alloc, 65536);
130#endif
131
132
133static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
134unsigned long _brk_end = (unsigned long)__brk_base;
135
136#ifdef CONFIG_X86_64
137int default_cpu_present_to_apicid(int mps_cpu)
138{
139 return __default_cpu_present_to_apicid(mps_cpu);
140}
141
142int default_check_phys_apicid_present(int phys_apicid)
143{
144 return __default_check_phys_apicid_present(phys_apicid);
145}
146#endif
147
148struct boot_params boot_params;
149
150/*
151 * Machine setup..
152 */
153static struct resource data_resource = {
154 .name = "Kernel data",
155 .start = 0,
156 .end = 0,
157 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
158};
159
160static struct resource code_resource = {
161 .name = "Kernel code",
162 .start = 0,
163 .end = 0,
164 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
165};
166
167static struct resource bss_resource = {
168 .name = "Kernel bss",
169 .start = 0,
170 .end = 0,
171 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
172};
173
174
175#ifdef CONFIG_X86_32
176/* cpu data as detected by the assembly code in head.S */
177struct cpuinfo_x86 new_cpu_data = {
178 .wp_works_ok = -1,
179};
180/* common cpu data for all cpus */
181struct cpuinfo_x86 boot_cpu_data __read_mostly = {
182 .wp_works_ok = -1,
183};
184EXPORT_SYMBOL(boot_cpu_data);
185
186unsigned int def_to_bigsmp;
187
188/* for MCA, but anyone else can use it if they want */
189unsigned int machine_id;
190unsigned int machine_submodel_id;
191unsigned int BIOS_revision;
192
193struct apm_info apm_info;
194EXPORT_SYMBOL(apm_info);
195
196#if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
197 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
198struct ist_info ist_info;
199EXPORT_SYMBOL(ist_info);
200#else
201struct ist_info ist_info;
202#endif
203
204#else
205struct cpuinfo_x86 boot_cpu_data __read_mostly = {
206 .x86_phys_bits = MAX_PHYSMEM_BITS,
207};
208EXPORT_SYMBOL(boot_cpu_data);
209#endif
210
211
212#if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
213__visible unsigned long mmu_cr4_features __ro_after_init;
214#else
215__visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
216#endif
217
218/* Boot loader ID and version as integers, for the benefit of proc_dointvec */
219int bootloader_type, bootloader_version;
220
221/*
222 * Setup options
223 */
224struct screen_info screen_info;
225EXPORT_SYMBOL(screen_info);
226struct edid_info edid_info;
227EXPORT_SYMBOL_GPL(edid_info);
228
229extern int root_mountflags;
230
231unsigned long saved_video_mode;
232
233#define RAMDISK_IMAGE_START_MASK 0x07FF
234#define RAMDISK_PROMPT_FLAG 0x8000
235#define RAMDISK_LOAD_FLAG 0x4000
236
237static char __initdata command_line[COMMAND_LINE_SIZE];
238#ifdef CONFIG_CMDLINE_BOOL
239static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
240#endif
241
242#if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
243struct edd edd;
244#ifdef CONFIG_EDD_MODULE
245EXPORT_SYMBOL(edd);
246#endif
247/**
248 * copy_edd() - Copy the BIOS EDD information
249 * from boot_params into a safe place.
250 *
251 */
252static inline void __init copy_edd(void)
253{
254 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
255 sizeof(edd.mbr_signature));
256 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
257 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
258 edd.edd_info_nr = boot_params.eddbuf_entries;
259}
260#else
261static inline void __init copy_edd(void)
262{
263}
264#endif
265
266void * __init extend_brk(size_t size, size_t align)
267{
268 size_t mask = align - 1;
269 void *ret;
270
271 BUG_ON(_brk_start == 0);
272 BUG_ON(align & mask);
273
274 _brk_end = (_brk_end + mask) & ~mask;
275 BUG_ON((char *)(_brk_end + size) > __brk_limit);
276
277 ret = (void *)_brk_end;
278 _brk_end += size;
279
280 memset(ret, 0, size);
281
282 return ret;
283}
284
285#ifdef CONFIG_X86_32
286static void __init cleanup_highmap(void)
287{
288}
289#endif
290
291static void __init reserve_brk(void)
292{
293 if (_brk_end > _brk_start)
294 memblock_reserve(__pa_symbol(_brk_start),
295 _brk_end - _brk_start);
296
297 /* Mark brk area as locked down and no longer taking any
298 new allocations */
299 _brk_start = 0;
300}
301
302u64 relocated_ramdisk;
303
304#ifdef CONFIG_BLK_DEV_INITRD
305
306static u64 __init get_ramdisk_image(void)
307{
308 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
309
310 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
311
312 return ramdisk_image;
313}
314static u64 __init get_ramdisk_size(void)
315{
316 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
317
318 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
319
320 return ramdisk_size;
321}
322
323static void __init relocate_initrd(void)
324{
325 /* Assume only end is not page aligned */
326 u64 ramdisk_image = get_ramdisk_image();
327 u64 ramdisk_size = get_ramdisk_size();
328 u64 area_size = PAGE_ALIGN(ramdisk_size);
329
330 /* We need to move the initrd down into directly mapped mem */
331 relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
332 area_size, PAGE_SIZE);
333
334 if (!relocated_ramdisk)
335 panic("Cannot find place for new RAMDISK of size %lld\n",
336 ramdisk_size);
337
338 /* Note: this includes all the mem currently occupied by
339 the initrd, we rely on that fact to keep the data intact. */
340 memblock_reserve(relocated_ramdisk, area_size);
341 initrd_start = relocated_ramdisk + PAGE_OFFSET;
342 initrd_end = initrd_start + ramdisk_size;
343 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
344 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
345
346 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
347
348 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
349 " [mem %#010llx-%#010llx]\n",
350 ramdisk_image, ramdisk_image + ramdisk_size - 1,
351 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
352}
353
354static void __init early_reserve_initrd(void)
355{
356 /* Assume only end is not page aligned */
357 u64 ramdisk_image = get_ramdisk_image();
358 u64 ramdisk_size = get_ramdisk_size();
359 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
360
361 if (!boot_params.hdr.type_of_loader ||
362 !ramdisk_image || !ramdisk_size)
363 return; /* No initrd provided by bootloader */
364
365 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
366}
367static void __init reserve_initrd(void)
368{
369 /* Assume only end is not page aligned */
370 u64 ramdisk_image = get_ramdisk_image();
371 u64 ramdisk_size = get_ramdisk_size();
372 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
373 u64 mapped_size;
374
375 if (!boot_params.hdr.type_of_loader ||
376 !ramdisk_image || !ramdisk_size)
377 return; /* No initrd provided by bootloader */
378
379 initrd_start = 0;
380
381 mapped_size = memblock_mem_size(max_pfn_mapped);
382 if (ramdisk_size >= (mapped_size>>1))
383 panic("initrd too large to handle, "
384 "disabling initrd (%lld needed, %lld available)\n",
385 ramdisk_size, mapped_size>>1);
386
387 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
388 ramdisk_end - 1);
389
390 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
391 PFN_DOWN(ramdisk_end))) {
392 /* All are mapped, easy case */
393 initrd_start = ramdisk_image + PAGE_OFFSET;
394 initrd_end = initrd_start + ramdisk_size;
395 return;
396 }
397
398 relocate_initrd();
399
400 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
401}
402
403#else
404static void __init early_reserve_initrd(void)
405{
406}
407static void __init reserve_initrd(void)
408{
409}
410#endif /* CONFIG_BLK_DEV_INITRD */
411
412static void __init parse_setup_data(void)
413{
414 struct setup_data *data;
415 u64 pa_data, pa_next;
416
417 pa_data = boot_params.hdr.setup_data;
418 while (pa_data) {
419 u32 data_len, data_type;
420
421 data = early_memremap(pa_data, sizeof(*data));
422 data_len = data->len + sizeof(struct setup_data);
423 data_type = data->type;
424 pa_next = data->next;
425 early_memunmap(data, sizeof(*data));
426
427 switch (data_type) {
428 case SETUP_E820_EXT:
429 parse_e820_ext(pa_data, data_len);
430 break;
431 case SETUP_DTB:
432 add_dtb(pa_data);
433 break;
434 case SETUP_EFI:
435 parse_efi_setup(pa_data, data_len);
436 break;
437 default:
438 break;
439 }
440 pa_data = pa_next;
441 }
442}
443
444static void __init e820_reserve_setup_data(void)
445{
446 struct setup_data *data;
447 u64 pa_data;
448
449 pa_data = boot_params.hdr.setup_data;
450 if (!pa_data)
451 return;
452
453 while (pa_data) {
454 data = early_memremap(pa_data, sizeof(*data));
455 e820_update_range(pa_data, sizeof(*data)+data->len,
456 E820_RAM, E820_RESERVED_KERN);
457 pa_data = data->next;
458 early_memunmap(data, sizeof(*data));
459 }
460
461 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
462 memcpy(e820_saved, e820, sizeof(struct e820map));
463 printk(KERN_INFO "extended physical RAM map:\n");
464 e820_print_map("reserve setup_data");
465}
466
467static void __init memblock_x86_reserve_range_setup_data(void)
468{
469 struct setup_data *data;
470 u64 pa_data;
471
472 pa_data = boot_params.hdr.setup_data;
473 while (pa_data) {
474 data = early_memremap(pa_data, sizeof(*data));
475 memblock_reserve(pa_data, sizeof(*data) + data->len);
476 pa_data = data->next;
477 early_memunmap(data, sizeof(*data));
478 }
479}
480
481/*
482 * --------- Crashkernel reservation ------------------------------
483 */
484
485#ifdef CONFIG_KEXEC_CORE
486
487/* 16M alignment for crash kernel regions */
488#define CRASH_ALIGN (16 << 20)
489
490/*
491 * Keep the crash kernel below this limit. On 32 bits earlier kernels
492 * would limit the kernel to the low 512 MiB due to mapping restrictions.
493 * On 64bit, old kexec-tools need to under 896MiB.
494 */
495#ifdef CONFIG_X86_32
496# define CRASH_ADDR_LOW_MAX (512 << 20)
497# define CRASH_ADDR_HIGH_MAX (512 << 20)
498#else
499# define CRASH_ADDR_LOW_MAX (896UL << 20)
500# define CRASH_ADDR_HIGH_MAX MAXMEM
501#endif
502
503static int __init reserve_crashkernel_low(void)
504{
505#ifdef CONFIG_X86_64
506 unsigned long long base, low_base = 0, low_size = 0;
507 unsigned long total_low_mem;
508 int ret;
509
510 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
511
512 /* crashkernel=Y,low */
513 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
514 if (ret) {
515 /*
516 * two parts from lib/swiotlb.c:
517 * -swiotlb size: user-specified with swiotlb= or default.
518 *
519 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
520 * to 8M for other buffers that may need to stay low too. Also
521 * make sure we allocate enough extra low memory so that we
522 * don't run out of DMA buffers for 32-bit devices.
523 */
524 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
525 } else {
526 /* passed with crashkernel=0,low ? */
527 if (!low_size)
528 return 0;
529 }
530
531 low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN);
532 if (!low_base) {
533 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
534 (unsigned long)(low_size >> 20));
535 return -ENOMEM;
536 }
537
538 ret = memblock_reserve(low_base, low_size);
539 if (ret) {
540 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
541 return ret;
542 }
543
544 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
545 (unsigned long)(low_size >> 20),
546 (unsigned long)(low_base >> 20),
547 (unsigned long)(total_low_mem >> 20));
548
549 crashk_low_res.start = low_base;
550 crashk_low_res.end = low_base + low_size - 1;
551 insert_resource(&iomem_resource, &crashk_low_res);
552#endif
553 return 0;
554}
555
556static void __init reserve_crashkernel(void)
557{
558 unsigned long long crash_size, crash_base, total_mem;
559 bool high = false;
560 int ret;
561
562 total_mem = memblock_phys_mem_size();
563
564 /* crashkernel=XM */
565 ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
566 if (ret != 0 || crash_size <= 0) {
567 /* crashkernel=X,high */
568 ret = parse_crashkernel_high(boot_command_line, total_mem,
569 &crash_size, &crash_base);
570 if (ret != 0 || crash_size <= 0)
571 return;
572 high = true;
573 }
574
575 /* 0 means: find the address automatically */
576 if (crash_base <= 0) {
577 /*
578 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
579 */
580 crash_base = memblock_find_in_range(CRASH_ALIGN,
581 high ? CRASH_ADDR_HIGH_MAX
582 : CRASH_ADDR_LOW_MAX,
583 crash_size, CRASH_ALIGN);
584 if (!crash_base) {
585 pr_info("crashkernel reservation failed - No suitable area found.\n");
586 return;
587 }
588
589 } else {
590 unsigned long long start;
591
592 start = memblock_find_in_range(crash_base,
593 crash_base + crash_size,
594 crash_size, 1 << 20);
595 if (start != crash_base) {
596 pr_info("crashkernel reservation failed - memory is in use.\n");
597 return;
598 }
599 }
600 ret = memblock_reserve(crash_base, crash_size);
601 if (ret) {
602 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
603 return;
604 }
605
606 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
607 memblock_free(crash_base, crash_size);
608 return;
609 }
610
611 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
612 (unsigned long)(crash_size >> 20),
613 (unsigned long)(crash_base >> 20),
614 (unsigned long)(total_mem >> 20));
615
616 crashk_res.start = crash_base;
617 crashk_res.end = crash_base + crash_size - 1;
618 insert_resource(&iomem_resource, &crashk_res);
619}
620#else
621static void __init reserve_crashkernel(void)
622{
623}
624#endif
625
626static struct resource standard_io_resources[] = {
627 { .name = "dma1", .start = 0x00, .end = 0x1f,
628 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
629 { .name = "pic1", .start = 0x20, .end = 0x21,
630 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
631 { .name = "timer0", .start = 0x40, .end = 0x43,
632 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
633 { .name = "timer1", .start = 0x50, .end = 0x53,
634 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
635 { .name = "keyboard", .start = 0x60, .end = 0x60,
636 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
637 { .name = "keyboard", .start = 0x64, .end = 0x64,
638 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
639 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
640 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
641 { .name = "pic2", .start = 0xa0, .end = 0xa1,
642 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
643 { .name = "dma2", .start = 0xc0, .end = 0xdf,
644 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
645 { .name = "fpu", .start = 0xf0, .end = 0xff,
646 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
647};
648
649void __init reserve_standard_io_resources(void)
650{
651 int i;
652
653 /* request I/O space for devices used on all i[345]86 PCs */
654 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
655 request_resource(&ioport_resource, &standard_io_resources[i]);
656
657}
658
659static __init void reserve_ibft_region(void)
660{
661 unsigned long addr, size = 0;
662
663 addr = find_ibft_region(&size);
664
665 if (size)
666 memblock_reserve(addr, size);
667}
668
669static bool __init snb_gfx_workaround_needed(void)
670{
671#ifdef CONFIG_PCI
672 int i;
673 u16 vendor, devid;
674 static const __initconst u16 snb_ids[] = {
675 0x0102,
676 0x0112,
677 0x0122,
678 0x0106,
679 0x0116,
680 0x0126,
681 0x010a,
682 };
683
684 /* Assume no if something weird is going on with PCI */
685 if (!early_pci_allowed())
686 return false;
687
688 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
689 if (vendor != 0x8086)
690 return false;
691
692 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
693 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
694 if (devid == snb_ids[i])
695 return true;
696#endif
697
698 return false;
699}
700
701/*
702 * Sandy Bridge graphics has trouble with certain ranges, exclude
703 * them from allocation.
704 */
705static void __init trim_snb_memory(void)
706{
707 static const __initconst unsigned long bad_pages[] = {
708 0x20050000,
709 0x20110000,
710 0x20130000,
711 0x20138000,
712 0x40004000,
713 };
714 int i;
715
716 if (!snb_gfx_workaround_needed())
717 return;
718
719 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
720
721 /*
722 * Reserve all memory below the 1 MB mark that has not
723 * already been reserved.
724 */
725 memblock_reserve(0, 1<<20);
726
727 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
728 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
729 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
730 bad_pages[i]);
731 }
732}
733
734/*
735 * Here we put platform-specific memory range workarounds, i.e.
736 * memory known to be corrupt or otherwise in need to be reserved on
737 * specific platforms.
738 *
739 * If this gets used more widely it could use a real dispatch mechanism.
740 */
741static void __init trim_platform_memory_ranges(void)
742{
743 trim_snb_memory();
744}
745
746static void __init trim_bios_range(void)
747{
748 /*
749 * A special case is the first 4Kb of memory;
750 * This is a BIOS owned area, not kernel ram, but generally
751 * not listed as such in the E820 table.
752 *
753 * This typically reserves additional memory (64KiB by default)
754 * since some BIOSes are known to corrupt low memory. See the
755 * Kconfig help text for X86_RESERVE_LOW.
756 */
757 e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
758
759 /*
760 * special case: Some BIOSen report the PC BIOS
761 * area (640->1Mb) as ram even though it is not.
762 * take them out.
763 */
764 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
765
766 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
767}
768
769/* called before trim_bios_range() to spare extra sanitize */
770static void __init e820_add_kernel_range(void)
771{
772 u64 start = __pa_symbol(_text);
773 u64 size = __pa_symbol(_end) - start;
774
775 /*
776 * Complain if .text .data and .bss are not marked as E820_RAM and
777 * attempt to fix it by adding the range. We may have a confused BIOS,
778 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
779 * exclude kernel range. If we really are running on top non-RAM,
780 * we will crash later anyways.
781 */
782 if (e820_all_mapped(start, start + size, E820_RAM))
783 return;
784
785 pr_warn(".text .data .bss are not marked as E820_RAM!\n");
786 e820_remove_range(start, size, E820_RAM, 0);
787 e820_add_region(start, size, E820_RAM);
788}
789
790static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
791
792static int __init parse_reservelow(char *p)
793{
794 unsigned long long size;
795
796 if (!p)
797 return -EINVAL;
798
799 size = memparse(p, &p);
800
801 if (size < 4096)
802 size = 4096;
803
804 if (size > 640*1024)
805 size = 640*1024;
806
807 reserve_low = size;
808
809 return 0;
810}
811
812early_param("reservelow", parse_reservelow);
813
814static void __init trim_low_memory_range(void)
815{
816 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
817}
818
819/*
820 * Dump out kernel offset information on panic.
821 */
822static int
823dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
824{
825 if (kaslr_enabled()) {
826 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
827 kaslr_offset(),
828 __START_KERNEL,
829 __START_KERNEL_map,
830 MODULES_VADDR-1);
831 } else {
832 pr_emerg("Kernel Offset: disabled\n");
833 }
834
835 return 0;
836}
837
838/*
839 * Determine if we were loaded by an EFI loader. If so, then we have also been
840 * passed the efi memmap, systab, etc., so we should use these data structures
841 * for initialization. Note, the efi init code path is determined by the
842 * global efi_enabled. This allows the same kernel image to be used on existing
843 * systems (with a traditional BIOS) as well as on EFI systems.
844 */
845/*
846 * setup_arch - architecture-specific boot-time initializations
847 *
848 * Note: On x86_64, fixmaps are ready for use even before this is called.
849 */
850
851void __init setup_arch(char **cmdline_p)
852{
853 memblock_reserve(__pa_symbol(_text),
854 (unsigned long)__bss_stop - (unsigned long)_text);
855
856 early_reserve_initrd();
857
858 /*
859 * At this point everything still needed from the boot loader
860 * or BIOS or kernel text should be early reserved or marked not
861 * RAM in e820. All other memory is free game.
862 */
863
864#ifdef CONFIG_X86_32
865 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
866
867 /*
868 * copy kernel address range established so far and switch
869 * to the proper swapper page table
870 */
871 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
872 initial_page_table + KERNEL_PGD_BOUNDARY,
873 KERNEL_PGD_PTRS);
874
875 load_cr3(swapper_pg_dir);
876 /*
877 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
878 * a cr3 based tlb flush, so the following __flush_tlb_all()
879 * will not flush anything because the cpu quirk which clears
880 * X86_FEATURE_PGE has not been invoked yet. Though due to the
881 * load_cr3() above the TLB has been flushed already. The
882 * quirk is invoked before subsequent calls to __flush_tlb_all()
883 * so proper operation is guaranteed.
884 */
885 __flush_tlb_all();
886#else
887 printk(KERN_INFO "Command line: %s\n", boot_command_line);
888#endif
889
890 /*
891 * If we have OLPC OFW, we might end up relocating the fixmap due to
892 * reserve_top(), so do this before touching the ioremap area.
893 */
894 olpc_ofw_detect();
895
896 early_trap_init();
897 early_cpu_init();
898 early_ioremap_init();
899
900 setup_olpc_ofw_pgd();
901
902 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
903 screen_info = boot_params.screen_info;
904 edid_info = boot_params.edid_info;
905#ifdef CONFIG_X86_32
906 apm_info.bios = boot_params.apm_bios_info;
907 ist_info = boot_params.ist_info;
908#endif
909 saved_video_mode = boot_params.hdr.vid_mode;
910 bootloader_type = boot_params.hdr.type_of_loader;
911 if ((bootloader_type >> 4) == 0xe) {
912 bootloader_type &= 0xf;
913 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
914 }
915 bootloader_version = bootloader_type & 0xf;
916 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
917
918#ifdef CONFIG_BLK_DEV_RAM
919 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
920 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
921 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
922#endif
923#ifdef CONFIG_EFI
924 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
925 EFI32_LOADER_SIGNATURE, 4)) {
926 set_bit(EFI_BOOT, &efi.flags);
927 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
928 EFI64_LOADER_SIGNATURE, 4)) {
929 set_bit(EFI_BOOT, &efi.flags);
930 set_bit(EFI_64BIT, &efi.flags);
931 }
932
933 if (efi_enabled(EFI_BOOT))
934 efi_memblock_x86_reserve_range();
935#endif
936
937 x86_init.oem.arch_setup();
938
939 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
940 setup_memory_map();
941 parse_setup_data();
942
943 copy_edd();
944
945 if (!boot_params.hdr.root_flags)
946 root_mountflags &= ~MS_RDONLY;
947 init_mm.start_code = (unsigned long) _text;
948 init_mm.end_code = (unsigned long) _etext;
949 init_mm.end_data = (unsigned long) _edata;
950 init_mm.brk = _brk_end;
951
952 mpx_mm_init(&init_mm);
953
954 code_resource.start = __pa_symbol(_text);
955 code_resource.end = __pa_symbol(_etext)-1;
956 data_resource.start = __pa_symbol(_etext);
957 data_resource.end = __pa_symbol(_edata)-1;
958 bss_resource.start = __pa_symbol(__bss_start);
959 bss_resource.end = __pa_symbol(__bss_stop)-1;
960
961#ifdef CONFIG_CMDLINE_BOOL
962#ifdef CONFIG_CMDLINE_OVERRIDE
963 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
964#else
965 if (builtin_cmdline[0]) {
966 /* append boot loader cmdline to builtin */
967 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
968 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
969 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
970 }
971#endif
972#endif
973
974 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
975 *cmdline_p = command_line;
976
977 /*
978 * x86_configure_nx() is called before parse_early_param() to detect
979 * whether hardware doesn't support NX (so that the early EHCI debug
980 * console setup can safely call set_fixmap()). It may then be called
981 * again from within noexec_setup() during parsing early parameters
982 * to honor the respective command line option.
983 */
984 x86_configure_nx();
985
986 parse_early_param();
987
988#ifdef CONFIG_MEMORY_HOTPLUG
989 /*
990 * Memory used by the kernel cannot be hot-removed because Linux
991 * cannot migrate the kernel pages. When memory hotplug is
992 * enabled, we should prevent memblock from allocating memory
993 * for the kernel.
994 *
995 * ACPI SRAT records all hotpluggable memory ranges. But before
996 * SRAT is parsed, we don't know about it.
997 *
998 * The kernel image is loaded into memory at very early time. We
999 * cannot prevent this anyway. So on NUMA system, we set any
1000 * node the kernel resides in as un-hotpluggable.
1001 *
1002 * Since on modern servers, one node could have double-digit
1003 * gigabytes memory, we can assume the memory around the kernel
1004 * image is also un-hotpluggable. So before SRAT is parsed, just
1005 * allocate memory near the kernel image to try the best to keep
1006 * the kernel away from hotpluggable memory.
1007 */
1008 if (movable_node_is_enabled())
1009 memblock_set_bottom_up(true);
1010#endif
1011
1012 x86_report_nx();
1013
1014 /* after early param, so could get panic from serial */
1015 memblock_x86_reserve_range_setup_data();
1016
1017 if (acpi_mps_check()) {
1018#ifdef CONFIG_X86_LOCAL_APIC
1019 disable_apic = 1;
1020#endif
1021 setup_clear_cpu_cap(X86_FEATURE_APIC);
1022 }
1023
1024#ifdef CONFIG_PCI
1025 if (pci_early_dump_regs)
1026 early_dump_pci_devices();
1027#endif
1028
1029 /* update the e820_saved too */
1030 e820_reserve_setup_data();
1031 finish_e820_parsing();
1032
1033 if (efi_enabled(EFI_BOOT))
1034 efi_init();
1035
1036 dmi_scan_machine();
1037 dmi_memdev_walk();
1038 dmi_set_dump_stack_arch_desc();
1039
1040 /*
1041 * VMware detection requires dmi to be available, so this
1042 * needs to be done after dmi_scan_machine, for the BP.
1043 */
1044 init_hypervisor_platform();
1045
1046 x86_init.resources.probe_roms();
1047
1048 /* after parse_early_param, so could debug it */
1049 insert_resource(&iomem_resource, &code_resource);
1050 insert_resource(&iomem_resource, &data_resource);
1051 insert_resource(&iomem_resource, &bss_resource);
1052
1053 e820_add_kernel_range();
1054 trim_bios_range();
1055#ifdef CONFIG_X86_32
1056 if (ppro_with_ram_bug()) {
1057 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1058 E820_RESERVED);
1059 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
1060 printk(KERN_INFO "fixed physical RAM map:\n");
1061 e820_print_map("bad_ppro");
1062 }
1063#else
1064 early_gart_iommu_check();
1065#endif
1066
1067 /*
1068 * partially used pages are not usable - thus
1069 * we are rounding upwards:
1070 */
1071 max_pfn = e820_end_of_ram_pfn();
1072
1073 /* update e820 for memory not covered by WB MTRRs */
1074 mtrr_bp_init();
1075 if (mtrr_trim_uncached_memory(max_pfn))
1076 max_pfn = e820_end_of_ram_pfn();
1077
1078 max_possible_pfn = max_pfn;
1079
1080 /*
1081 * Define random base addresses for memory sections after max_pfn is
1082 * defined and before each memory section base is used.
1083 */
1084 kernel_randomize_memory();
1085
1086#ifdef CONFIG_X86_32
1087 /* max_low_pfn get updated here */
1088 find_low_pfn_range();
1089#else
1090 check_x2apic();
1091
1092 /* How many end-of-memory variables you have, grandma! */
1093 /* need this before calling reserve_initrd */
1094 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1095 max_low_pfn = e820_end_of_low_ram_pfn();
1096 else
1097 max_low_pfn = max_pfn;
1098
1099 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1100#endif
1101
1102 /*
1103 * Find and reserve possible boot-time SMP configuration:
1104 */
1105 find_smp_config();
1106
1107 reserve_ibft_region();
1108
1109 early_alloc_pgt_buf();
1110
1111 /*
1112 * Need to conclude brk, before memblock_x86_fill()
1113 * it could use memblock_find_in_range, could overlap with
1114 * brk area.
1115 */
1116 reserve_brk();
1117
1118 cleanup_highmap();
1119
1120 memblock_set_current_limit(ISA_END_ADDRESS);
1121 memblock_x86_fill();
1122
1123 reserve_bios_regions();
1124
1125 if (efi_enabled(EFI_MEMMAP)) {
1126 efi_fake_memmap();
1127 efi_find_mirror();
1128 efi_esrt_init();
1129
1130 /*
1131 * The EFI specification says that boot service code won't be
1132 * called after ExitBootServices(). This is, in fact, a lie.
1133 */
1134 efi_reserve_boot_services();
1135 }
1136
1137 /* preallocate 4k for mptable mpc */
1138 early_reserve_e820_mpc_new();
1139
1140#ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1141 setup_bios_corruption_check();
1142#endif
1143
1144#ifdef CONFIG_X86_32
1145 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1146 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1147#endif
1148
1149 reserve_real_mode();
1150
1151 trim_platform_memory_ranges();
1152 trim_low_memory_range();
1153
1154 init_mem_mapping();
1155
1156 early_trap_pf_init();
1157
1158 /*
1159 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1160 * with the current CR4 value. This may not be necessary, but
1161 * auditing all the early-boot CR4 manipulation would be needed to
1162 * rule it out.
1163 */
1164 mmu_cr4_features = __read_cr4();
1165
1166 memblock_set_current_limit(get_max_mapped());
1167
1168 /*
1169 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1170 */
1171
1172#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1173 if (init_ohci1394_dma_early)
1174 init_ohci1394_dma_on_all_controllers();
1175#endif
1176 /* Allocate bigger log buffer */
1177 setup_log_buf(1);
1178
1179 reserve_initrd();
1180
1181 acpi_table_upgrade();
1182
1183 vsmp_init();
1184
1185 io_delay_init();
1186
1187 /*
1188 * Parse the ACPI tables for possible boot-time SMP configuration.
1189 */
1190 acpi_boot_table_init();
1191
1192 early_acpi_boot_init();
1193
1194 initmem_init();
1195 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1196
1197 /*
1198 * Reserve memory for crash kernel after SRAT is parsed so that it
1199 * won't consume hotpluggable memory.
1200 */
1201 reserve_crashkernel();
1202
1203 memblock_find_dma_reserve();
1204
1205#ifdef CONFIG_KVM_GUEST
1206 kvmclock_init();
1207#endif
1208
1209 x86_init.paging.pagetable_init();
1210
1211 kasan_init();
1212
1213#ifdef CONFIG_X86_32
1214 /* sync back kernel address range */
1215 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1216 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1217 KERNEL_PGD_PTRS);
1218
1219 /*
1220 * sync back low identity map too. It is used for example
1221 * in the 32-bit EFI stub.
1222 */
1223 clone_pgd_range(initial_page_table,
1224 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1225 min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY));
1226#endif
1227
1228 tboot_probe();
1229
1230 map_vsyscall();
1231
1232 generic_apic_probe();
1233
1234 early_quirks();
1235
1236 /*
1237 * Read APIC and some other early information from ACPI tables.
1238 */
1239 acpi_boot_init();
1240 sfi_init();
1241 x86_dtb_init();
1242
1243 /*
1244 * get boot-time SMP configuration:
1245 */
1246 get_smp_config();
1247
1248 /*
1249 * Systems w/o ACPI and mptables might not have it mapped the local
1250 * APIC yet, but prefill_possible_map() might need to access it.
1251 */
1252 init_apic_mappings();
1253
1254 prefill_possible_map();
1255
1256 init_cpu_to_node();
1257
1258 io_apic_init_mappings();
1259
1260 kvm_guest_init();
1261
1262 e820_reserve_resources();
1263 e820_mark_nosave_regions(max_low_pfn);
1264
1265 x86_init.resources.reserve_resources();
1266
1267 e820_setup_gap();
1268
1269#ifdef CONFIG_VT
1270#if defined(CONFIG_VGA_CONSOLE)
1271 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1272 conswitchp = &vga_con;
1273#elif defined(CONFIG_DUMMY_CONSOLE)
1274 conswitchp = &dummy_con;
1275#endif
1276#endif
1277 x86_init.oem.banner();
1278
1279 x86_init.timers.wallclock_init();
1280
1281 mcheck_init();
1282
1283 arch_init_ideal_nops();
1284
1285 register_refined_jiffies(CLOCK_TICK_RATE);
1286
1287#ifdef CONFIG_EFI
1288 if (efi_enabled(EFI_BOOT))
1289 efi_apply_memmap_quirks();
1290#endif
1291}
1292
1293#ifdef CONFIG_X86_32
1294
1295static struct resource video_ram_resource = {
1296 .name = "Video RAM area",
1297 .start = 0xa0000,
1298 .end = 0xbffff,
1299 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1300};
1301
1302void __init i386_reserve_resources(void)
1303{
1304 request_resource(&iomem_resource, &video_ram_resource);
1305 reserve_standard_io_resources();
1306}
1307
1308#endif /* CONFIG_X86_32 */
1309
1310static struct notifier_block kernel_offset_notifier = {
1311 .notifier_call = dump_kernel_offset
1312};
1313
1314static int __init register_kernel_offset_dumper(void)
1315{
1316 atomic_notifier_chain_register(&panic_notifier_list,
1317 &kernel_offset_notifier);
1318 return 0;
1319}
1320__initcall(register_kernel_offset_dumper);
1321
1322void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1323{
1324 if (!boot_cpu_has(X86_FEATURE_OSPKE))
1325 return;
1326
1327 seq_printf(m, "ProtectionKey: %8u\n", vma_pkey(vma));
1328}