2 * Copyright (C) 1995 Linus Torvalds
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
21 * This file handles the architecture-dependent parts of initialization
24 #include <linux/sched.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>
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>
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>
68 #include <linux/percpu.h>
69 #include <linux/crash_dump.h>
70 #include <linux/tboot.h>
71 #include <linux/jiffies.h>
73 #include <video/edid.h>
77 #include <asm/realmode.h>
79 #include <asm/mpspec.h>
80 #include <asm/setup.h>
82 #include <asm/timer.h>
83 #include <asm/i8259.h>
84 #include <asm/sections.h>
85 #include <asm/io_apic.h>
87 #include <asm/setup_arch.h>
88 #include <asm/bios_ebda.h>
89 #include <asm/cacheflush.h>
90 #include <asm/processor.h>
93 #include <asm/vsyscall.h>
97 #include <asm/iommu.h>
99 #include <asm/mmu_context.h>
100 #include <asm/proto.h>
102 #include <asm/paravirt.h>
103 #include <asm/hypervisor.h>
104 #include <asm/olpc_ofw.h>
106 #include <asm/percpu.h>
107 #include <asm/topology.h>
108 #include <asm/apicdef.h>
109 #include <asm/amd_nb.h>
111 #include <asm/alternative.h>
112 #include <asm/prom.h>
115 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
116 * max_pfn_mapped: highest direct mapped pfn over 4GB
118 * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
119 * represented by pfn_mapped
121 unsigned long max_low_pfn_mapped;
122 unsigned long max_pfn_mapped;
125 RESERVE_BRK(dmi_alloc, 65536);
129 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
130 unsigned long _brk_end = (unsigned long)__brk_base;
133 int default_cpu_present_to_apicid(int mps_cpu)
135 return __default_cpu_present_to_apicid(mps_cpu);
138 int default_check_phys_apicid_present(int phys_apicid)
140 return __default_check_phys_apicid_present(phys_apicid);
144 struct boot_params boot_params;
149 static struct resource data_resource = {
150 .name = "Kernel data",
153 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
156 static struct resource code_resource = {
157 .name = "Kernel code",
160 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
163 static struct resource bss_resource = {
164 .name = "Kernel bss",
167 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
172 /* cpu data as detected by the assembly code in head.S */
173 struct cpuinfo_x86 new_cpu_data = {
176 /* common cpu data for all cpus */
177 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
180 EXPORT_SYMBOL(boot_cpu_data);
182 unsigned int def_to_bigsmp;
184 /* for MCA, but anyone else can use it if they want */
185 unsigned int machine_id;
186 unsigned int machine_submodel_id;
187 unsigned int BIOS_revision;
189 struct apm_info apm_info;
190 EXPORT_SYMBOL(apm_info);
192 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
193 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
194 struct ist_info ist_info;
195 EXPORT_SYMBOL(ist_info);
197 struct ist_info ist_info;
201 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
202 .x86_phys_bits = MAX_PHYSMEM_BITS,
204 EXPORT_SYMBOL(boot_cpu_data);
208 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
209 __visible unsigned long mmu_cr4_features;
211 __visible unsigned long mmu_cr4_features = X86_CR4_PAE;
214 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
215 int bootloader_type, bootloader_version;
220 struct screen_info screen_info;
221 EXPORT_SYMBOL(screen_info);
222 struct edid_info edid_info;
223 EXPORT_SYMBOL_GPL(edid_info);
225 extern int root_mountflags;
227 unsigned long saved_video_mode;
229 #define RAMDISK_IMAGE_START_MASK 0x07FF
230 #define RAMDISK_PROMPT_FLAG 0x8000
231 #define RAMDISK_LOAD_FLAG 0x4000
233 static char __initdata command_line[COMMAND_LINE_SIZE];
234 #ifdef CONFIG_CMDLINE_BOOL
235 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
238 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
240 #ifdef CONFIG_EDD_MODULE
244 * copy_edd() - Copy the BIOS EDD information
245 * from boot_params into a safe place.
248 static inline void __init copy_edd(void)
250 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
251 sizeof(edd.mbr_signature));
252 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
253 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
254 edd.edd_info_nr = boot_params.eddbuf_entries;
257 static inline void __init copy_edd(void)
262 void * __init extend_brk(size_t size, size_t align)
264 size_t mask = align - 1;
267 BUG_ON(_brk_start == 0);
268 BUG_ON(align & mask);
270 _brk_end = (_brk_end + mask) & ~mask;
271 BUG_ON((char *)(_brk_end + size) > __brk_limit);
273 ret = (void *)_brk_end;
276 memset(ret, 0, size);
282 static void __init cleanup_highmap(void)
287 static void __init reserve_brk(void)
289 if (_brk_end > _brk_start)
290 memblock_reserve(__pa_symbol(_brk_start),
291 _brk_end - _brk_start);
293 /* Mark brk area as locked down and no longer taking any
298 #ifdef CONFIG_BLK_DEV_INITRD
300 static u64 __init get_ramdisk_image(void)
302 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
304 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
306 return ramdisk_image;
308 static u64 __init get_ramdisk_size(void)
310 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
312 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
317 #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
318 static void __init relocate_initrd(void)
320 /* Assume only end is not page aligned */
321 u64 ramdisk_image = get_ramdisk_image();
322 u64 ramdisk_size = get_ramdisk_size();
323 u64 area_size = PAGE_ALIGN(ramdisk_size);
325 unsigned long slop, clen, mapaddr;
328 /* We need to move the initrd down into directly mapped mem */
329 ramdisk_here = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
330 area_size, PAGE_SIZE);
333 panic("Cannot find place for new RAMDISK of size %lld\n",
336 /* Note: this includes all the mem currently occupied by
337 the initrd, we rely on that fact to keep the data intact. */
338 memblock_reserve(ramdisk_here, area_size);
339 initrd_start = ramdisk_here + PAGE_OFFSET;
340 initrd_end = initrd_start + ramdisk_size;
341 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
342 ramdisk_here, ramdisk_here + ramdisk_size - 1);
344 q = (char *)initrd_start;
346 /* Copy the initrd */
347 while (ramdisk_size) {
348 slop = ramdisk_image & ~PAGE_MASK;
350 if (clen > MAX_MAP_CHUNK-slop)
351 clen = MAX_MAP_CHUNK-slop;
352 mapaddr = ramdisk_image & PAGE_MASK;
353 p = early_memremap(mapaddr, clen+slop);
354 memcpy(q, p+slop, clen);
355 early_iounmap(p, clen+slop);
357 ramdisk_image += clen;
358 ramdisk_size -= clen;
361 ramdisk_image = get_ramdisk_image();
362 ramdisk_size = get_ramdisk_size();
363 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
364 " [mem %#010llx-%#010llx]\n",
365 ramdisk_image, ramdisk_image + ramdisk_size - 1,
366 ramdisk_here, ramdisk_here + ramdisk_size - 1);
369 static void __init early_reserve_initrd(void)
371 /* Assume only end is not page aligned */
372 u64 ramdisk_image = get_ramdisk_image();
373 u64 ramdisk_size = get_ramdisk_size();
374 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
376 if (!boot_params.hdr.type_of_loader ||
377 !ramdisk_image || !ramdisk_size)
378 return; /* No initrd provided by bootloader */
380 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
382 static void __init reserve_initrd(void)
384 /* Assume only end is not page aligned */
385 u64 ramdisk_image = get_ramdisk_image();
386 u64 ramdisk_size = get_ramdisk_size();
387 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
390 if (!boot_params.hdr.type_of_loader ||
391 !ramdisk_image || !ramdisk_size)
392 return; /* No initrd provided by bootloader */
396 mapped_size = memblock_mem_size(max_pfn_mapped);
397 if (ramdisk_size >= (mapped_size>>1))
398 panic("initrd too large to handle, "
399 "disabling initrd (%lld needed, %lld available)\n",
400 ramdisk_size, mapped_size>>1);
402 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
405 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
406 PFN_DOWN(ramdisk_end))) {
407 /* All are mapped, easy case */
408 initrd_start = ramdisk_image + PAGE_OFFSET;
409 initrd_end = initrd_start + ramdisk_size;
415 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
418 static void __init early_reserve_initrd(void)
421 static void __init reserve_initrd(void)
424 #endif /* CONFIG_BLK_DEV_INITRD */
426 static void __init parse_setup_data(void)
428 struct setup_data *data;
429 u64 pa_data, pa_next;
431 pa_data = boot_params.hdr.setup_data;
433 u32 data_len, map_len, data_type;
435 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
436 (u64)sizeof(struct setup_data));
437 data = early_memremap(pa_data, map_len);
438 data_len = data->len + sizeof(struct setup_data);
439 data_type = data->type;
440 pa_next = data->next;
441 early_iounmap(data, map_len);
445 parse_e820_ext(pa_data, data_len);
457 static void __init e820_reserve_setup_data(void)
459 struct setup_data *data;
463 pa_data = boot_params.hdr.setup_data;
465 data = early_memremap(pa_data, sizeof(*data));
466 e820_update_range(pa_data, sizeof(*data)+data->len,
467 E820_RAM, E820_RESERVED_KERN);
469 pa_data = data->next;
470 early_iounmap(data, sizeof(*data));
475 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
476 memcpy(&e820_saved, &e820, sizeof(struct e820map));
477 printk(KERN_INFO "extended physical RAM map:\n");
478 e820_print_map("reserve setup_data");
481 static void __init memblock_x86_reserve_range_setup_data(void)
483 struct setup_data *data;
486 pa_data = boot_params.hdr.setup_data;
488 data = early_memremap(pa_data, sizeof(*data));
489 memblock_reserve(pa_data, sizeof(*data) + data->len);
490 pa_data = data->next;
491 early_iounmap(data, sizeof(*data));
496 * --------- Crashkernel reservation ------------------------------
502 * Keep the crash kernel below this limit. On 32 bits earlier kernels
503 * would limit the kernel to the low 512 MiB due to mapping restrictions.
504 * On 64bit, old kexec-tools need to under 896MiB.
507 # define CRASH_KERNEL_ADDR_LOW_MAX (512 << 20)
508 # define CRASH_KERNEL_ADDR_HIGH_MAX (512 << 20)
510 # define CRASH_KERNEL_ADDR_LOW_MAX (896UL<<20)
511 # define CRASH_KERNEL_ADDR_HIGH_MAX MAXMEM
514 static void __init reserve_crashkernel_low(void)
517 const unsigned long long alignment = 16<<20; /* 16M */
518 unsigned long long low_base = 0, low_size = 0;
519 unsigned long total_low_mem;
520 unsigned long long base;
521 bool auto_set = false;
524 total_low_mem = memblock_mem_size(1UL<<(32-PAGE_SHIFT));
525 /* crashkernel=Y,low */
526 ret = parse_crashkernel_low(boot_command_line, total_low_mem,
530 * two parts from lib/swiotlb.c:
531 * swiotlb size: user specified with swiotlb= or default.
532 * swiotlb overflow buffer: now is hardcoded to 32k.
533 * We round it to 8M for other buffers that
534 * may need to stay low too.
536 low_size = swiotlb_size_or_default() + (8UL<<20);
539 /* passed with crashkernel=0,low ? */
544 low_base = memblock_find_in_range(low_size, (1ULL<<32),
545 low_size, alignment);
549 pr_info("crashkernel low reservation failed - No suitable area found.\n");
554 memblock_reserve(low_base, low_size);
555 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
556 (unsigned long)(low_size >> 20),
557 (unsigned long)(low_base >> 20),
558 (unsigned long)(total_low_mem >> 20));
559 crashk_low_res.start = low_base;
560 crashk_low_res.end = low_base + low_size - 1;
561 insert_resource(&iomem_resource, &crashk_low_res);
565 static void __init reserve_crashkernel(void)
567 const unsigned long long alignment = 16<<20; /* 16M */
568 unsigned long long total_mem;
569 unsigned long long crash_size, crash_base;
573 total_mem = memblock_phys_mem_size();
576 ret = parse_crashkernel(boot_command_line, total_mem,
577 &crash_size, &crash_base);
578 if (ret != 0 || crash_size <= 0) {
579 /* crashkernel=X,high */
580 ret = parse_crashkernel_high(boot_command_line, total_mem,
581 &crash_size, &crash_base);
582 if (ret != 0 || crash_size <= 0)
587 /* 0 means: find the address automatically */
588 if (crash_base <= 0) {
590 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
592 crash_base = memblock_find_in_range(alignment,
593 high ? CRASH_KERNEL_ADDR_HIGH_MAX :
594 CRASH_KERNEL_ADDR_LOW_MAX,
595 crash_size, alignment);
598 pr_info("crashkernel reservation failed - No suitable area found.\n");
603 unsigned long long start;
605 start = memblock_find_in_range(crash_base,
606 crash_base + crash_size, crash_size, 1<<20);
607 if (start != crash_base) {
608 pr_info("crashkernel reservation failed - memory is in use.\n");
612 memblock_reserve(crash_base, crash_size);
614 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
615 "for crashkernel (System RAM: %ldMB)\n",
616 (unsigned long)(crash_size >> 20),
617 (unsigned long)(crash_base >> 20),
618 (unsigned long)(total_mem >> 20));
620 crashk_res.start = crash_base;
621 crashk_res.end = crash_base + crash_size - 1;
622 insert_resource(&iomem_resource, &crashk_res);
624 if (crash_base >= (1ULL<<32))
625 reserve_crashkernel_low();
628 static void __init reserve_crashkernel(void)
633 static struct resource standard_io_resources[] = {
634 { .name = "dma1", .start = 0x00, .end = 0x1f,
635 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
636 { .name = "pic1", .start = 0x20, .end = 0x21,
637 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
638 { .name = "timer0", .start = 0x40, .end = 0x43,
639 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
640 { .name = "timer1", .start = 0x50, .end = 0x53,
641 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
642 { .name = "keyboard", .start = 0x60, .end = 0x60,
643 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
644 { .name = "keyboard", .start = 0x64, .end = 0x64,
645 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
646 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
647 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
648 { .name = "pic2", .start = 0xa0, .end = 0xa1,
649 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
650 { .name = "dma2", .start = 0xc0, .end = 0xdf,
651 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
652 { .name = "fpu", .start = 0xf0, .end = 0xff,
653 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
656 void __init reserve_standard_io_resources(void)
660 /* request I/O space for devices used on all i[345]86 PCs */
661 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
662 request_resource(&ioport_resource, &standard_io_resources[i]);
666 static __init void reserve_ibft_region(void)
668 unsigned long addr, size = 0;
670 addr = find_ibft_region(&size);
673 memblock_reserve(addr, size);
676 static bool __init snb_gfx_workaround_needed(void)
681 static const __initconst u16 snb_ids[] = {
691 /* Assume no if something weird is going on with PCI */
692 if (!early_pci_allowed())
695 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
696 if (vendor != 0x8086)
699 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
700 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
701 if (devid == snb_ids[i])
709 * Sandy Bridge graphics has trouble with certain ranges, exclude
710 * them from allocation.
712 static void __init trim_snb_memory(void)
714 static const __initconst unsigned long bad_pages[] = {
723 if (!snb_gfx_workaround_needed())
726 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
729 * Reserve all memory below the 1 MB mark that has not
730 * already been reserved.
732 memblock_reserve(0, 1<<20);
734 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
735 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
736 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
742 * Here we put platform-specific memory range workarounds, i.e.
743 * memory known to be corrupt or otherwise in need to be reserved on
744 * specific platforms.
746 * If this gets used more widely it could use a real dispatch mechanism.
748 static void __init trim_platform_memory_ranges(void)
753 static void __init trim_bios_range(void)
756 * A special case is the first 4Kb of memory;
757 * This is a BIOS owned area, not kernel ram, but generally
758 * not listed as such in the E820 table.
760 * This typically reserves additional memory (64KiB by default)
761 * since some BIOSes are known to corrupt low memory. See the
762 * Kconfig help text for X86_RESERVE_LOW.
764 e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
767 * special case: Some BIOSen report the PC BIOS
768 * area (640->1Mb) as ram even though it is not.
771 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
773 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
776 /* called before trim_bios_range() to spare extra sanitize */
777 static void __init e820_add_kernel_range(void)
779 u64 start = __pa_symbol(_text);
780 u64 size = __pa_symbol(_end) - start;
783 * Complain if .text .data and .bss are not marked as E820_RAM and
784 * attempt to fix it by adding the range. We may have a confused BIOS,
785 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
786 * exclude kernel range. If we really are running on top non-RAM,
787 * we will crash later anyways.
789 if (e820_all_mapped(start, start + size, E820_RAM))
792 pr_warn(".text .data .bss are not marked as E820_RAM!\n");
793 e820_remove_range(start, size, E820_RAM, 0);
794 e820_add_region(start, size, E820_RAM);
797 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
799 static int __init parse_reservelow(char *p)
801 unsigned long long size;
806 size = memparse(p, &p);
819 early_param("reservelow", parse_reservelow);
821 static void __init trim_low_memory_range(void)
823 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
827 * Dump out kernel offset information on panic.
830 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
832 pr_emerg("Kernel Offset: 0x%lx from 0x%lx "
833 "(relocation range: 0x%lx-0x%lx)\n",
834 (unsigned long)&_text - __START_KERNEL, __START_KERNEL,
835 __START_KERNEL_map, MODULES_VADDR-1);
841 * Determine if we were loaded by an EFI loader. If so, then we have also been
842 * passed the efi memmap, systab, etc., so we should use these data structures
843 * for initialization. Note, the efi init code path is determined by the
844 * global efi_enabled. This allows the same kernel image to be used on existing
845 * systems (with a traditional BIOS) as well as on EFI systems.
848 * setup_arch - architecture-specific boot-time initializations
850 * Note: On x86_64, fixmaps are ready for use even before this is called.
853 void __init setup_arch(char **cmdline_p)
855 memblock_reserve(__pa_symbol(_text),
856 (unsigned long)__bss_stop - (unsigned long)_text);
858 early_reserve_initrd();
861 * At this point everything still needed from the boot loader
862 * or BIOS or kernel text should be early reserved or marked not
863 * RAM in e820. All other memory is free game.
867 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
868 visws_early_detect();
871 * copy kernel address range established so far and switch
872 * to the proper swapper page table
874 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
875 initial_page_table + KERNEL_PGD_BOUNDARY,
878 load_cr3(swapper_pg_dir);
881 printk(KERN_INFO "Command line: %s\n", boot_command_line);
885 * If we have OLPC OFW, we might end up relocating the fixmap due to
886 * reserve_top(), so do this before touching the ioremap area.
892 early_ioremap_init();
894 setup_olpc_ofw_pgd();
896 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
897 screen_info = boot_params.screen_info;
898 edid_info = boot_params.edid_info;
900 apm_info.bios = boot_params.apm_bios_info;
901 ist_info = boot_params.ist_info;
902 if (boot_params.sys_desc_table.length != 0) {
903 machine_id = boot_params.sys_desc_table.table[0];
904 machine_submodel_id = boot_params.sys_desc_table.table[1];
905 BIOS_revision = boot_params.sys_desc_table.table[2];
908 saved_video_mode = boot_params.hdr.vid_mode;
909 bootloader_type = boot_params.hdr.type_of_loader;
910 if ((bootloader_type >> 4) == 0xe) {
911 bootloader_type &= 0xf;
912 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
914 bootloader_version = bootloader_type & 0xf;
915 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
917 #ifdef CONFIG_BLK_DEV_RAM
918 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
919 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
920 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
923 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
925 set_bit(EFI_BOOT, &x86_efi_facility);
926 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
928 set_bit(EFI_BOOT, &x86_efi_facility);
929 set_bit(EFI_64BIT, &x86_efi_facility);
932 if (efi_enabled(EFI_BOOT))
933 efi_memblock_x86_reserve_range();
936 x86_init.oem.arch_setup();
938 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
941 /* update the e820_saved too */
942 e820_reserve_setup_data();
946 if (!boot_params.hdr.root_flags)
947 root_mountflags &= ~MS_RDONLY;
948 init_mm.start_code = (unsigned long) _text;
949 init_mm.end_code = (unsigned long) _etext;
950 init_mm.end_data = (unsigned long) _edata;
951 init_mm.brk = _brk_end;
953 code_resource.start = __pa_symbol(_text);
954 code_resource.end = __pa_symbol(_etext)-1;
955 data_resource.start = __pa_symbol(_etext);
956 data_resource.end = __pa_symbol(_edata)-1;
957 bss_resource.start = __pa_symbol(__bss_start);
958 bss_resource.end = __pa_symbol(__bss_stop)-1;
960 #ifdef CONFIG_CMDLINE_BOOL
961 #ifdef CONFIG_CMDLINE_OVERRIDE
962 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
964 if (builtin_cmdline[0]) {
965 /* append boot loader cmdline to builtin */
966 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
967 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
968 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
973 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
974 *cmdline_p = command_line;
977 * x86_configure_nx() is called before parse_early_param() to detect
978 * whether hardware doesn't support NX (so that the early EHCI debug
979 * console setup can safely call set_fixmap()). It may then be called
980 * again from within noexec_setup() during parsing early parameters
981 * to honor the respective command line option.
989 /* after early param, so could get panic from serial */
990 memblock_x86_reserve_range_setup_data();
992 if (acpi_mps_check()) {
993 #ifdef CONFIG_X86_LOCAL_APIC
996 setup_clear_cpu_cap(X86_FEATURE_APIC);
1000 if (pci_early_dump_regs)
1001 early_dump_pci_devices();
1004 finish_e820_parsing();
1006 if (efi_enabled(EFI_BOOT))
1010 dmi_set_dump_stack_arch_desc();
1013 * VMware detection requires dmi to be available, so this
1014 * needs to be done after dmi_scan_machine, for the BP.
1016 init_hypervisor_platform();
1018 x86_init.resources.probe_roms();
1020 /* after parse_early_param, so could debug it */
1021 insert_resource(&iomem_resource, &code_resource);
1022 insert_resource(&iomem_resource, &data_resource);
1023 insert_resource(&iomem_resource, &bss_resource);
1025 e820_add_kernel_range();
1027 #ifdef CONFIG_X86_32
1028 if (ppro_with_ram_bug()) {
1029 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1031 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
1032 printk(KERN_INFO "fixed physical RAM map:\n");
1033 e820_print_map("bad_ppro");
1036 early_gart_iommu_check();
1040 * partially used pages are not usable - thus
1041 * we are rounding upwards:
1043 max_pfn = e820_end_of_ram_pfn();
1045 /* update e820 for memory not covered by WB MTRRs */
1047 if (mtrr_trim_uncached_memory(max_pfn))
1048 max_pfn = e820_end_of_ram_pfn();
1050 #ifdef CONFIG_X86_32
1051 /* max_low_pfn get updated here */
1052 find_low_pfn_range();
1056 /* How many end-of-memory variables you have, grandma! */
1057 /* need this before calling reserve_initrd */
1058 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1059 max_low_pfn = e820_end_of_low_ram_pfn();
1061 max_low_pfn = max_pfn;
1063 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1067 * Find and reserve possible boot-time SMP configuration:
1071 reserve_ibft_region();
1073 early_alloc_pgt_buf();
1076 * Need to conclude brk, before memblock_x86_fill()
1077 * it could use memblock_find_in_range, could overlap with
1084 memblock_set_current_limit(ISA_END_ADDRESS);
1085 memblock_x86_fill();
1088 * The EFI specification says that boot service code won't be called
1089 * after ExitBootServices(). This is, in fact, a lie.
1091 if (efi_enabled(EFI_MEMMAP))
1092 efi_reserve_boot_services();
1094 /* preallocate 4k for mptable mpc */
1095 early_reserve_e820_mpc_new();
1097 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1098 setup_bios_corruption_check();
1101 #ifdef CONFIG_X86_32
1102 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1103 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1106 reserve_real_mode();
1108 trim_platform_memory_ranges();
1109 trim_low_memory_range();
1113 early_trap_pf_init();
1117 memblock_set_current_limit(get_max_mapped());
1118 dma_contiguous_reserve(0);
1121 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1124 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1125 if (init_ohci1394_dma_early)
1126 init_ohci1394_dma_on_all_controllers();
1128 /* Allocate bigger log buffer */
1133 #if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
1134 acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
1137 reserve_crashkernel();
1144 * Parse the ACPI tables for possible boot-time SMP configuration.
1146 acpi_boot_table_init();
1148 early_acpi_boot_init();
1151 memblock_find_dma_reserve();
1153 #ifdef CONFIG_KVM_GUEST
1157 x86_init.paging.pagetable_init();
1159 if (boot_cpu_data.cpuid_level >= 0) {
1160 /* A CPU has %cr4 if and only if it has CPUID */
1161 mmu_cr4_features = read_cr4();
1162 if (trampoline_cr4_features)
1163 *trampoline_cr4_features = mmu_cr4_features;
1166 #ifdef CONFIG_X86_32
1167 /* sync back kernel address range */
1168 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1169 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1175 #ifdef CONFIG_X86_64
1179 generic_apic_probe();
1184 * Read APIC and some other early information from ACPI tables.
1191 * get boot-time SMP configuration:
1193 if (smp_found_config)
1196 prefill_possible_map();
1200 init_apic_mappings();
1201 if (x86_io_apic_ops.init)
1202 x86_io_apic_ops.init();
1206 e820_reserve_resources();
1207 e820_mark_nosave_regions(max_low_pfn);
1209 x86_init.resources.reserve_resources();
1214 #if defined(CONFIG_VGA_CONSOLE)
1215 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1216 conswitchp = &vga_con;
1217 #elif defined(CONFIG_DUMMY_CONSOLE)
1218 conswitchp = &dummy_con;
1221 x86_init.oem.banner();
1223 x86_init.timers.wallclock_init();
1227 arch_init_ideal_nops();
1229 register_refined_jiffies(CLOCK_TICK_RATE);
1232 /* Once setup is done above, unmap the EFI memory map on
1233 * mismatched firmware/kernel archtectures since there is no
1234 * support for runtime services.
1236 if (efi_enabled(EFI_BOOT) && !efi_is_native()) {
1237 pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
1243 #ifdef CONFIG_X86_32
1245 static struct resource video_ram_resource = {
1246 .name = "Video RAM area",
1249 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1252 void __init i386_reserve_resources(void)
1254 request_resource(&iomem_resource, &video_ram_resource);
1255 reserve_standard_io_resources();
1258 #endif /* CONFIG_X86_32 */
1260 static struct notifier_block kernel_offset_notifier = {
1261 .notifier_call = dump_kernel_offset
1264 static int __init register_kernel_offset_dumper(void)
1266 atomic_notifier_chain_register(&panic_notifier_list,
1267 &kernel_offset_notifier);
1270 __initcall(register_kernel_offset_dumper);