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/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>
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>
92 #include <asm/kasan.h>
94 #include <asm/vsyscall.h>
98 #include <asm/iommu.h>
100 #include <asm/mmu_context.h>
101 #include <asm/proto.h>
103 #include <asm/paravirt.h>
104 #include <asm/hypervisor.h>
105 #include <asm/olpc_ofw.h>
107 #include <asm/percpu.h>
108 #include <asm/topology.h>
109 #include <asm/apicdef.h>
110 #include <asm/amd_nb.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>
119 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
120 * max_pfn_mapped: highest direct mapped pfn over 4GB
122 * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
123 * represented by pfn_mapped
125 unsigned long max_low_pfn_mapped;
126 unsigned long max_pfn_mapped;
129 RESERVE_BRK(dmi_alloc, 65536);
133 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
134 unsigned long _brk_end = (unsigned long)__brk_base;
137 int default_cpu_present_to_apicid(int mps_cpu)
139 return __default_cpu_present_to_apicid(mps_cpu);
142 int default_check_phys_apicid_present(int phys_apicid)
144 return __default_check_phys_apicid_present(phys_apicid);
148 struct boot_params boot_params;
153 static struct resource data_resource = {
154 .name = "Kernel data",
157 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
160 static struct resource code_resource = {
161 .name = "Kernel code",
164 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
167 static struct resource bss_resource = {
168 .name = "Kernel bss",
171 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
176 /* cpu data as detected by the assembly code in head.S */
177 struct cpuinfo_x86 new_cpu_data = {
180 /* common cpu data for all cpus */
181 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
184 EXPORT_SYMBOL(boot_cpu_data);
186 unsigned int def_to_bigsmp;
188 /* for MCA, but anyone else can use it if they want */
189 unsigned int machine_id;
190 unsigned int machine_submodel_id;
191 unsigned int BIOS_revision;
193 struct apm_info apm_info;
194 EXPORT_SYMBOL(apm_info);
196 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
197 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
198 struct ist_info ist_info;
199 EXPORT_SYMBOL(ist_info);
201 struct ist_info ist_info;
205 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
206 .x86_phys_bits = MAX_PHYSMEM_BITS,
208 EXPORT_SYMBOL(boot_cpu_data);
212 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
213 __visible unsigned long mmu_cr4_features __ro_after_init;
215 __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
218 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
219 int bootloader_type, bootloader_version;
224 struct screen_info screen_info;
225 EXPORT_SYMBOL(screen_info);
226 struct edid_info edid_info;
227 EXPORT_SYMBOL_GPL(edid_info);
229 extern int root_mountflags;
231 unsigned long saved_video_mode;
233 #define RAMDISK_IMAGE_START_MASK 0x07FF
234 #define RAMDISK_PROMPT_FLAG 0x8000
235 #define RAMDISK_LOAD_FLAG 0x4000
237 static char __initdata command_line[COMMAND_LINE_SIZE];
238 #ifdef CONFIG_CMDLINE_BOOL
239 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
242 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
244 #ifdef CONFIG_EDD_MODULE
248 * copy_edd() - Copy the BIOS EDD information
249 * from boot_params into a safe place.
252 static inline void __init copy_edd(void)
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;
261 static inline void __init copy_edd(void)
266 void * __init extend_brk(size_t size, size_t align)
268 size_t mask = align - 1;
271 BUG_ON(_brk_start == 0);
272 BUG_ON(align & mask);
274 _brk_end = (_brk_end + mask) & ~mask;
275 BUG_ON((char *)(_brk_end + size) > __brk_limit);
277 ret = (void *)_brk_end;
280 memset(ret, 0, size);
286 static void __init cleanup_highmap(void)
291 static void __init reserve_brk(void)
293 if (_brk_end > _brk_start)
294 memblock_reserve(__pa_symbol(_brk_start),
295 _brk_end - _brk_start);
297 /* Mark brk area as locked down and no longer taking any
302 u64 relocated_ramdisk;
304 #ifdef CONFIG_BLK_DEV_INITRD
306 static u64 __init get_ramdisk_image(void)
308 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
310 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
312 return ramdisk_image;
314 static u64 __init get_ramdisk_size(void)
316 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
318 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
323 static void __init relocate_initrd(void)
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);
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);
334 if (!relocated_ramdisk)
335 panic("Cannot find place for new RAMDISK of size %lld\n",
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);
346 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
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);
354 static void __init early_reserve_initrd(void)
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);
361 if (!boot_params.hdr.type_of_loader ||
362 !ramdisk_image || !ramdisk_size)
363 return; /* No initrd provided by bootloader */
365 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
367 static void __init reserve_initrd(void)
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);
375 if (!boot_params.hdr.type_of_loader ||
376 !ramdisk_image || !ramdisk_size)
377 return; /* No initrd provided by bootloader */
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);
387 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
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;
400 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
404 static void __init early_reserve_initrd(void)
407 static void __init reserve_initrd(void)
410 #endif /* CONFIG_BLK_DEV_INITRD */
412 static void __init parse_setup_data(void)
414 struct setup_data *data;
415 u64 pa_data, pa_next;
417 pa_data = boot_params.hdr.setup_data;
419 u32 data_len, data_type;
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));
429 parse_e820_ext(pa_data, data_len);
435 parse_efi_setup(pa_data, data_len);
444 static void __init e820_reserve_setup_data(void)
446 struct setup_data *data;
449 pa_data = boot_params.hdr.setup_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));
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");
467 static void __init memblock_x86_reserve_range_setup_data(void)
469 struct setup_data *data;
472 pa_data = boot_params.hdr.setup_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));
482 * --------- Crashkernel reservation ------------------------------
485 #ifdef CONFIG_KEXEC_CORE
487 /* 16M alignment for crash kernel regions */
488 #define CRASH_ALIGN (16 << 20)
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.
496 # define CRASH_ADDR_LOW_MAX (512 << 20)
497 # define CRASH_ADDR_HIGH_MAX (512 << 20)
499 # define CRASH_ADDR_LOW_MAX (896UL << 20)
500 # define CRASH_ADDR_HIGH_MAX MAXMEM
503 static int __init reserve_crashkernel_low(void)
506 unsigned long long base, low_base = 0, low_size = 0;
507 unsigned long total_low_mem;
510 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
512 /* crashkernel=Y,low */
513 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
516 * two parts from lib/swiotlb.c:
517 * -swiotlb size: user-specified with swiotlb= or default.
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.
524 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
526 /* passed with crashkernel=0,low ? */
531 low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN);
533 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
534 (unsigned long)(low_size >> 20));
538 ret = memblock_reserve(low_base, low_size);
540 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
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));
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);
556 static void __init reserve_crashkernel(void)
558 unsigned long long crash_size, crash_base, total_mem;
562 total_mem = memblock_phys_mem_size();
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)
575 /* 0 means: find the address automatically */
576 if (crash_base <= 0) {
578 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
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);
585 pr_info("crashkernel reservation failed - No suitable area found.\n");
590 unsigned long long start;
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");
600 ret = memblock_reserve(crash_base, crash_size);
602 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
606 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
607 memblock_free(crash_base, crash_size);
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));
616 crashk_res.start = crash_base;
617 crashk_res.end = crash_base + crash_size - 1;
618 insert_resource(&iomem_resource, &crashk_res);
621 static void __init reserve_crashkernel(void)
626 static 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 }
649 void __init reserve_standard_io_resources(void)
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]);
659 static __init void reserve_ibft_region(void)
661 unsigned long addr, size = 0;
663 addr = find_ibft_region(&size);
666 memblock_reserve(addr, size);
669 static bool __init snb_gfx_workaround_needed(void)
674 static const __initconst u16 snb_ids[] = {
684 /* Assume no if something weird is going on with PCI */
685 if (!early_pci_allowed())
688 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
689 if (vendor != 0x8086)
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])
702 * Sandy Bridge graphics has trouble with certain ranges, exclude
703 * them from allocation.
705 static void __init trim_snb_memory(void)
707 static const __initconst unsigned long bad_pages[] = {
716 if (!snb_gfx_workaround_needed())
719 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
722 * Reserve all memory below the 1 MB mark that has not
723 * already been reserved.
725 memblock_reserve(0, 1<<20);
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",
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.
739 * If this gets used more widely it could use a real dispatch mechanism.
741 static void __init trim_platform_memory_ranges(void)
746 static void __init trim_bios_range(void)
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.
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.
757 e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
760 * special case: Some BIOSen report the PC BIOS
761 * area (640->1Mb) as ram even though it is not.
764 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
766 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
769 /* called before trim_bios_range() to spare extra sanitize */
770 static void __init e820_add_kernel_range(void)
772 u64 start = __pa_symbol(_text);
773 u64 size = __pa_symbol(_end) - start;
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.
782 if (e820_all_mapped(start, start + size, E820_RAM))
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);
790 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
792 static int __init parse_reservelow(char *p)
794 unsigned long long size;
799 size = memparse(p, &p);
812 early_param("reservelow", parse_reservelow);
814 static void __init trim_low_memory_range(void)
816 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
820 * Dump out kernel offset information on panic.
823 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
825 if (kaslr_enabled()) {
826 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
832 pr_emerg("Kernel Offset: disabled\n");
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.
846 * setup_arch - architecture-specific boot-time initializations
848 * Note: On x86_64, fixmaps are ready for use even before this is called.
851 void __init setup_arch(char **cmdline_p)
853 memblock_reserve(__pa_symbol(_text),
854 (unsigned long)__bss_stop - (unsigned long)_text);
856 early_reserve_initrd();
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.
865 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
868 * copy kernel address range established so far and switch
869 * to the proper swapper page table
871 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
872 initial_page_table + KERNEL_PGD_BOUNDARY,
875 load_cr3(swapper_pg_dir);
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.
887 printk(KERN_INFO "Command line: %s\n", boot_command_line);
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.
898 early_ioremap_init();
900 setup_olpc_ofw_pgd();
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;
906 apm_info.bios = boot_params.apm_bios_info;
907 ist_info = boot_params.ist_info;
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;
915 bootloader_version = bootloader_type & 0xf;
916 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
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);
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);
933 if (efi_enabled(EFI_BOOT))
934 efi_memblock_x86_reserve_range();
937 x86_init.oem.arch_setup();
939 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
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;
952 mpx_mm_init(&init_mm);
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;
961 #ifdef CONFIG_CMDLINE_BOOL
962 #ifdef CONFIG_CMDLINE_OVERRIDE
963 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
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);
974 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
975 *cmdline_p = command_line;
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.
990 /* after early param, so could get panic from serial */
991 memblock_x86_reserve_range_setup_data();
993 if (acpi_mps_check()) {
994 #ifdef CONFIG_X86_LOCAL_APIC
997 setup_clear_cpu_cap(X86_FEATURE_APIC);
1001 if (pci_early_dump_regs)
1002 early_dump_pci_devices();
1005 /* update the e820_saved too */
1006 e820_reserve_setup_data();
1007 finish_e820_parsing();
1009 if (efi_enabled(EFI_BOOT))
1014 dmi_set_dump_stack_arch_desc();
1017 * VMware detection requires dmi to be available, so this
1018 * needs to be done after dmi_scan_machine, for the BP.
1020 init_hypervisor_platform();
1022 x86_init.resources.probe_roms();
1024 /* after parse_early_param, so could debug it */
1025 insert_resource(&iomem_resource, &code_resource);
1026 insert_resource(&iomem_resource, &data_resource);
1027 insert_resource(&iomem_resource, &bss_resource);
1029 e820_add_kernel_range();
1031 #ifdef CONFIG_X86_32
1032 if (ppro_with_ram_bug()) {
1033 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1035 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
1036 printk(KERN_INFO "fixed physical RAM map:\n");
1037 e820_print_map("bad_ppro");
1040 early_gart_iommu_check();
1044 * partially used pages are not usable - thus
1045 * we are rounding upwards:
1047 max_pfn = e820_end_of_ram_pfn();
1049 /* update e820 for memory not covered by WB MTRRs */
1051 if (mtrr_trim_uncached_memory(max_pfn))
1052 max_pfn = e820_end_of_ram_pfn();
1054 max_possible_pfn = max_pfn;
1057 * Define random base addresses for memory sections after max_pfn is
1058 * defined and before each memory section base is used.
1060 kernel_randomize_memory();
1062 #ifdef CONFIG_X86_32
1063 /* max_low_pfn get updated here */
1064 find_low_pfn_range();
1068 /* How many end-of-memory variables you have, grandma! */
1069 /* need this before calling reserve_initrd */
1070 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1071 max_low_pfn = e820_end_of_low_ram_pfn();
1073 max_low_pfn = max_pfn;
1075 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1079 * Find and reserve possible boot-time SMP configuration:
1083 reserve_ibft_region();
1085 early_alloc_pgt_buf();
1088 * Need to conclude brk, before memblock_x86_fill()
1089 * it could use memblock_find_in_range, could overlap with
1096 memblock_set_current_limit(ISA_END_ADDRESS);
1097 memblock_x86_fill();
1099 reserve_bios_regions();
1101 if (efi_enabled(EFI_MEMMAP)) {
1107 * The EFI specification says that boot service code won't be
1108 * called after ExitBootServices(). This is, in fact, a lie.
1110 efi_reserve_boot_services();
1113 /* preallocate 4k for mptable mpc */
1114 early_reserve_e820_mpc_new();
1116 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1117 setup_bios_corruption_check();
1120 #ifdef CONFIG_X86_32
1121 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1122 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1125 reserve_real_mode();
1127 trim_platform_memory_ranges();
1128 trim_low_memory_range();
1132 early_trap_pf_init();
1135 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1136 * with the current CR4 value. This may not be necessary, but
1137 * auditing all the early-boot CR4 manipulation would be needed to
1140 mmu_cr4_features = __read_cr4();
1142 memblock_set_current_limit(get_max_mapped());
1145 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1148 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1149 if (init_ohci1394_dma_early)
1150 init_ohci1394_dma_on_all_controllers();
1152 /* Allocate bigger log buffer */
1157 acpi_table_upgrade();
1164 * Parse the ACPI tables for possible boot-time SMP configuration.
1166 acpi_boot_table_init();
1168 early_acpi_boot_init();
1171 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1174 * Reserve memory for crash kernel after SRAT is parsed so that it
1175 * won't consume hotpluggable memory.
1177 reserve_crashkernel();
1179 memblock_find_dma_reserve();
1181 #ifdef CONFIG_KVM_GUEST
1185 x86_init.paging.pagetable_init();
1189 #ifdef CONFIG_X86_32
1190 /* sync back kernel address range */
1191 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1192 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1196 * sync back low identity map too. It is used for example
1197 * in the 32-bit EFI stub.
1199 clone_pgd_range(initial_page_table,
1200 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1201 min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY));
1208 generic_apic_probe();
1213 * Read APIC and some other early information from ACPI tables.
1220 * get boot-time SMP configuration:
1224 prefill_possible_map();
1228 init_apic_mappings();
1229 io_apic_init_mappings();
1233 e820_reserve_resources();
1234 e820_mark_nosave_regions(max_low_pfn);
1236 x86_init.resources.reserve_resources();
1241 #if defined(CONFIG_VGA_CONSOLE)
1242 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1243 conswitchp = &vga_con;
1244 #elif defined(CONFIG_DUMMY_CONSOLE)
1245 conswitchp = &dummy_con;
1248 x86_init.oem.banner();
1250 x86_init.timers.wallclock_init();
1254 arch_init_ideal_nops();
1256 register_refined_jiffies(CLOCK_TICK_RATE);
1259 if (efi_enabled(EFI_BOOT))
1260 efi_apply_memmap_quirks();
1264 #ifdef CONFIG_X86_32
1266 static struct resource video_ram_resource = {
1267 .name = "Video RAM area",
1270 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1273 void __init i386_reserve_resources(void)
1275 request_resource(&iomem_resource, &video_ram_resource);
1276 reserve_standard_io_resources();
1279 #endif /* CONFIG_X86_32 */
1281 static struct notifier_block kernel_offset_notifier = {
1282 .notifier_call = dump_kernel_offset
1285 static int __init register_kernel_offset_dumper(void)
1287 atomic_notifier_chain_register(&panic_notifier_list,
1288 &kernel_offset_notifier);
1291 __initcall(register_kernel_offset_dumper);
1293 void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1295 if (!boot_cpu_has(X86_FEATURE_OSPKE))
1298 seq_printf(m, "ProtectionKey: %8u\n", vma_pkey(vma));