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_32.S */
177 struct cpuinfo_x86 new_cpu_data;
179 /* common cpu data for all cpus */
180 struct cpuinfo_x86 boot_cpu_data __read_mostly;
181 EXPORT_SYMBOL(boot_cpu_data);
183 unsigned int def_to_bigsmp;
185 /* for MCA, but anyone else can use it if they want */
186 unsigned int machine_id;
187 unsigned int machine_submodel_id;
188 unsigned int BIOS_revision;
190 struct apm_info apm_info;
191 EXPORT_SYMBOL(apm_info);
193 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
194 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
195 struct ist_info ist_info;
196 EXPORT_SYMBOL(ist_info);
198 struct ist_info ist_info;
202 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
203 .x86_phys_bits = MAX_PHYSMEM_BITS,
205 EXPORT_SYMBOL(boot_cpu_data);
209 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
210 __visible unsigned long mmu_cr4_features __ro_after_init;
212 __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
215 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
216 int bootloader_type, bootloader_version;
221 struct screen_info screen_info;
222 EXPORT_SYMBOL(screen_info);
223 struct edid_info edid_info;
224 EXPORT_SYMBOL_GPL(edid_info);
226 extern int root_mountflags;
228 unsigned long saved_video_mode;
230 #define RAMDISK_IMAGE_START_MASK 0x07FF
231 #define RAMDISK_PROMPT_FLAG 0x8000
232 #define RAMDISK_LOAD_FLAG 0x4000
234 static char __initdata command_line[COMMAND_LINE_SIZE];
235 #ifdef CONFIG_CMDLINE_BOOL
236 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
239 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
241 #ifdef CONFIG_EDD_MODULE
245 * copy_edd() - Copy the BIOS EDD information
246 * from boot_params into a safe place.
249 static inline void __init copy_edd(void)
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;
258 static inline void __init copy_edd(void)
263 void * __init extend_brk(size_t size, size_t align)
265 size_t mask = align - 1;
268 BUG_ON(_brk_start == 0);
269 BUG_ON(align & mask);
271 _brk_end = (_brk_end + mask) & ~mask;
272 BUG_ON((char *)(_brk_end + size) > __brk_limit);
274 ret = (void *)_brk_end;
277 memset(ret, 0, size);
283 static void __init cleanup_highmap(void)
288 static void __init reserve_brk(void)
290 if (_brk_end > _brk_start)
291 memblock_reserve(__pa_symbol(_brk_start),
292 _brk_end - _brk_start);
294 /* Mark brk area as locked down and no longer taking any
299 u64 relocated_ramdisk;
301 #ifdef CONFIG_BLK_DEV_INITRD
303 static u64 __init get_ramdisk_image(void)
305 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
307 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
309 return ramdisk_image;
311 static u64 __init get_ramdisk_size(void)
313 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
315 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
320 static void __init relocate_initrd(void)
322 /* Assume only end is not page aligned */
323 u64 ramdisk_image = get_ramdisk_image();
324 u64 ramdisk_size = get_ramdisk_size();
325 u64 area_size = PAGE_ALIGN(ramdisk_size);
327 /* We need to move the initrd down into directly mapped mem */
328 relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
329 area_size, PAGE_SIZE);
331 if (!relocated_ramdisk)
332 panic("Cannot find place for new RAMDISK of size %lld\n",
335 /* Note: this includes all the mem currently occupied by
336 the initrd, we rely on that fact to keep the data intact. */
337 memblock_reserve(relocated_ramdisk, area_size);
338 initrd_start = relocated_ramdisk + PAGE_OFFSET;
339 initrd_end = initrd_start + ramdisk_size;
340 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
341 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
343 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
345 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
346 " [mem %#010llx-%#010llx]\n",
347 ramdisk_image, ramdisk_image + ramdisk_size - 1,
348 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
351 static void __init early_reserve_initrd(void)
353 /* Assume only end is not page aligned */
354 u64 ramdisk_image = get_ramdisk_image();
355 u64 ramdisk_size = get_ramdisk_size();
356 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
358 if (!boot_params.hdr.type_of_loader ||
359 !ramdisk_image || !ramdisk_size)
360 return; /* No initrd provided by bootloader */
362 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
364 static void __init reserve_initrd(void)
366 /* Assume only end is not page aligned */
367 u64 ramdisk_image = get_ramdisk_image();
368 u64 ramdisk_size = get_ramdisk_size();
369 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
372 if (!boot_params.hdr.type_of_loader ||
373 !ramdisk_image || !ramdisk_size)
374 return; /* No initrd provided by bootloader */
378 mapped_size = memblock_mem_size(max_pfn_mapped);
379 if (ramdisk_size >= (mapped_size>>1))
380 panic("initrd too large to handle, "
381 "disabling initrd (%lld needed, %lld available)\n",
382 ramdisk_size, mapped_size>>1);
384 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
387 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
388 PFN_DOWN(ramdisk_end))) {
389 /* All are mapped, easy case */
390 initrd_start = ramdisk_image + PAGE_OFFSET;
391 initrd_end = initrd_start + ramdisk_size;
397 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
401 static void __init early_reserve_initrd(void)
404 static void __init reserve_initrd(void)
407 #endif /* CONFIG_BLK_DEV_INITRD */
409 static void __init parse_setup_data(void)
411 struct setup_data *data;
412 u64 pa_data, pa_next;
414 pa_data = boot_params.hdr.setup_data;
416 u32 data_len, data_type;
418 data = early_memremap(pa_data, sizeof(*data));
419 data_len = data->len + sizeof(struct setup_data);
420 data_type = data->type;
421 pa_next = data->next;
422 early_memunmap(data, sizeof(*data));
426 parse_e820_ext(pa_data, data_len);
432 parse_efi_setup(pa_data, data_len);
441 static void __init e820_reserve_setup_data(void)
443 struct setup_data *data;
446 pa_data = boot_params.hdr.setup_data;
451 data = early_memremap(pa_data, sizeof(*data));
452 e820_update_range(pa_data, sizeof(*data)+data->len,
453 E820_RAM, E820_RESERVED_KERN);
454 pa_data = data->next;
455 early_memunmap(data, sizeof(*data));
458 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
459 memcpy(e820_saved, e820, sizeof(struct e820map));
460 printk(KERN_INFO "extended physical RAM map:\n");
461 e820_print_map("reserve setup_data");
464 static void __init memblock_x86_reserve_range_setup_data(void)
466 struct setup_data *data;
469 pa_data = boot_params.hdr.setup_data;
471 data = early_memremap(pa_data, sizeof(*data));
472 memblock_reserve(pa_data, sizeof(*data) + data->len);
473 pa_data = data->next;
474 early_memunmap(data, sizeof(*data));
479 * --------- Crashkernel reservation ------------------------------
482 #ifdef CONFIG_KEXEC_CORE
484 /* 16M alignment for crash kernel regions */
485 #define CRASH_ALIGN (16 << 20)
488 * Keep the crash kernel below this limit. On 32 bits earlier kernels
489 * would limit the kernel to the low 512 MiB due to mapping restrictions.
490 * On 64bit, old kexec-tools need to under 896MiB.
493 # define CRASH_ADDR_LOW_MAX (512 << 20)
494 # define CRASH_ADDR_HIGH_MAX (512 << 20)
496 # define CRASH_ADDR_LOW_MAX (896UL << 20)
497 # define CRASH_ADDR_HIGH_MAX MAXMEM
500 static int __init reserve_crashkernel_low(void)
503 unsigned long long base, low_base = 0, low_size = 0;
504 unsigned long total_low_mem;
507 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
509 /* crashkernel=Y,low */
510 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
513 * two parts from lib/swiotlb.c:
514 * -swiotlb size: user-specified with swiotlb= or default.
516 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
517 * to 8M for other buffers that may need to stay low too. Also
518 * make sure we allocate enough extra low memory so that we
519 * don't run out of DMA buffers for 32-bit devices.
521 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
523 /* passed with crashkernel=0,low ? */
528 low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN);
530 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
531 (unsigned long)(low_size >> 20));
535 ret = memblock_reserve(low_base, low_size);
537 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
541 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
542 (unsigned long)(low_size >> 20),
543 (unsigned long)(low_base >> 20),
544 (unsigned long)(total_low_mem >> 20));
546 crashk_low_res.start = low_base;
547 crashk_low_res.end = low_base + low_size - 1;
548 insert_resource(&iomem_resource, &crashk_low_res);
553 static void __init reserve_crashkernel(void)
555 unsigned long long crash_size, crash_base, total_mem;
559 total_mem = memblock_phys_mem_size();
562 ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
563 if (ret != 0 || crash_size <= 0) {
564 /* crashkernel=X,high */
565 ret = parse_crashkernel_high(boot_command_line, total_mem,
566 &crash_size, &crash_base);
567 if (ret != 0 || crash_size <= 0)
572 /* 0 means: find the address automatically */
573 if (crash_base <= 0) {
575 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
576 * as old kexec-tools loads bzImage below that, unless
577 * "crashkernel=size[KMG],high" is specified.
579 crash_base = memblock_find_in_range(CRASH_ALIGN,
580 high ? CRASH_ADDR_HIGH_MAX
581 : CRASH_ADDR_LOW_MAX,
582 crash_size, CRASH_ALIGN);
584 pr_info("crashkernel reservation failed - No suitable area found.\n");
589 unsigned long long start;
591 start = memblock_find_in_range(crash_base,
592 crash_base + crash_size,
593 crash_size, 1 << 20);
594 if (start != crash_base) {
595 pr_info("crashkernel reservation failed - memory is in use.\n");
599 ret = memblock_reserve(crash_base, crash_size);
601 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
605 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
606 memblock_free(crash_base, crash_size);
610 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
611 (unsigned long)(crash_size >> 20),
612 (unsigned long)(crash_base >> 20),
613 (unsigned long)(total_mem >> 20));
615 crashk_res.start = crash_base;
616 crashk_res.end = crash_base + crash_size - 1;
617 insert_resource(&iomem_resource, &crashk_res);
620 static void __init reserve_crashkernel(void)
625 static struct resource standard_io_resources[] = {
626 { .name = "dma1", .start = 0x00, .end = 0x1f,
627 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
628 { .name = "pic1", .start = 0x20, .end = 0x21,
629 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
630 { .name = "timer0", .start = 0x40, .end = 0x43,
631 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
632 { .name = "timer1", .start = 0x50, .end = 0x53,
633 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
634 { .name = "keyboard", .start = 0x60, .end = 0x60,
635 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
636 { .name = "keyboard", .start = 0x64, .end = 0x64,
637 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
638 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
639 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
640 { .name = "pic2", .start = 0xa0, .end = 0xa1,
641 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
642 { .name = "dma2", .start = 0xc0, .end = 0xdf,
643 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
644 { .name = "fpu", .start = 0xf0, .end = 0xff,
645 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
648 void __init reserve_standard_io_resources(void)
652 /* request I/O space for devices used on all i[345]86 PCs */
653 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
654 request_resource(&ioport_resource, &standard_io_resources[i]);
658 static __init void reserve_ibft_region(void)
660 unsigned long addr, size = 0;
662 addr = find_ibft_region(&size);
665 memblock_reserve(addr, size);
668 static bool __init snb_gfx_workaround_needed(void)
673 static const __initconst u16 snb_ids[] = {
683 /* Assume no if something weird is going on with PCI */
684 if (!early_pci_allowed())
687 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
688 if (vendor != 0x8086)
691 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
692 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
693 if (devid == snb_ids[i])
701 * Sandy Bridge graphics has trouble with certain ranges, exclude
702 * them from allocation.
704 static void __init trim_snb_memory(void)
706 static const __initconst unsigned long bad_pages[] = {
715 if (!snb_gfx_workaround_needed())
718 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
721 * Reserve all memory below the 1 MB mark that has not
722 * already been reserved.
724 memblock_reserve(0, 1<<20);
726 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
727 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
728 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
734 * Here we put platform-specific memory range workarounds, i.e.
735 * memory known to be corrupt or otherwise in need to be reserved on
736 * specific platforms.
738 * If this gets used more widely it could use a real dispatch mechanism.
740 static void __init trim_platform_memory_ranges(void)
745 static void __init trim_bios_range(void)
748 * A special case is the first 4Kb of memory;
749 * This is a BIOS owned area, not kernel ram, but generally
750 * not listed as such in the E820 table.
752 * This typically reserves additional memory (64KiB by default)
753 * since some BIOSes are known to corrupt low memory. See the
754 * Kconfig help text for X86_RESERVE_LOW.
756 e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
759 * special case: Some BIOSen report the PC BIOS
760 * area (640->1Mb) as ram even though it is not.
763 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
765 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
768 /* called before trim_bios_range() to spare extra sanitize */
769 static void __init e820_add_kernel_range(void)
771 u64 start = __pa_symbol(_text);
772 u64 size = __pa_symbol(_end) - start;
775 * Complain if .text .data and .bss are not marked as E820_RAM and
776 * attempt to fix it by adding the range. We may have a confused BIOS,
777 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
778 * exclude kernel range. If we really are running on top non-RAM,
779 * we will crash later anyways.
781 if (e820_all_mapped(start, start + size, E820_RAM))
784 pr_warn(".text .data .bss are not marked as E820_RAM!\n");
785 e820_remove_range(start, size, E820_RAM, 0);
786 e820_add_region(start, size, E820_RAM);
789 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
791 static int __init parse_reservelow(char *p)
793 unsigned long long size;
798 size = memparse(p, &p);
811 early_param("reservelow", parse_reservelow);
813 static void __init trim_low_memory_range(void)
815 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
819 * Dump out kernel offset information on panic.
822 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
824 if (kaslr_enabled()) {
825 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
831 pr_emerg("Kernel Offset: disabled\n");
838 * Determine if we were loaded by an EFI loader. If so, then we have also been
839 * passed the efi memmap, systab, etc., so we should use these data structures
840 * for initialization. Note, the efi init code path is determined by the
841 * global efi_enabled. This allows the same kernel image to be used on existing
842 * systems (with a traditional BIOS) as well as on EFI systems.
845 * setup_arch - architecture-specific boot-time initializations
847 * Note: On x86_64, fixmaps are ready for use even before this is called.
850 void __init setup_arch(char **cmdline_p)
852 memblock_reserve(__pa_symbol(_text),
853 (unsigned long)__bss_stop - (unsigned long)_text);
855 early_reserve_initrd();
858 * At this point everything still needed from the boot loader
859 * or BIOS or kernel text should be early reserved or marked not
860 * RAM in e820. All other memory is free game.
864 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
867 * copy kernel address range established so far and switch
868 * to the proper swapper page table
870 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
871 initial_page_table + KERNEL_PGD_BOUNDARY,
874 load_cr3(swapper_pg_dir);
876 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
877 * a cr3 based tlb flush, so the following __flush_tlb_all()
878 * will not flush anything because the cpu quirk which clears
879 * X86_FEATURE_PGE has not been invoked yet. Though due to the
880 * load_cr3() above the TLB has been flushed already. The
881 * quirk is invoked before subsequent calls to __flush_tlb_all()
882 * so proper operation is guaranteed.
886 printk(KERN_INFO "Command line: %s\n", boot_command_line);
890 * If we have OLPC OFW, we might end up relocating the fixmap due to
891 * reserve_top(), so do this before touching the ioremap area.
897 early_ioremap_init();
899 setup_olpc_ofw_pgd();
901 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
902 screen_info = boot_params.screen_info;
903 edid_info = boot_params.edid_info;
905 apm_info.bios = boot_params.apm_bios_info;
906 ist_info = boot_params.ist_info;
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,
924 EFI32_LOADER_SIGNATURE, 4)) {
925 set_bit(EFI_BOOT, &efi.flags);
926 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
927 EFI64_LOADER_SIGNATURE, 4)) {
928 set_bit(EFI_BOOT, &efi.flags);
929 set_bit(EFI_64BIT, &efi.flags);
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;
944 if (!boot_params.hdr.root_flags)
945 root_mountflags &= ~MS_RDONLY;
946 init_mm.start_code = (unsigned long) _text;
947 init_mm.end_code = (unsigned long) _etext;
948 init_mm.end_data = (unsigned long) _edata;
949 init_mm.brk = _brk_end;
951 mpx_mm_init(&init_mm);
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.
987 #ifdef CONFIG_MEMORY_HOTPLUG
989 * Memory used by the kernel cannot be hot-removed because Linux
990 * cannot migrate the kernel pages. When memory hotplug is
991 * enabled, we should prevent memblock from allocating memory
994 * ACPI SRAT records all hotpluggable memory ranges. But before
995 * SRAT is parsed, we don't know about it.
997 * The kernel image is loaded into memory at very early time. We
998 * cannot prevent this anyway. So on NUMA system, we set any
999 * node the kernel resides in as un-hotpluggable.
1001 * Since on modern servers, one node could have double-digit
1002 * gigabytes memory, we can assume the memory around the kernel
1003 * image is also un-hotpluggable. So before SRAT is parsed, just
1004 * allocate memory near the kernel image to try the best to keep
1005 * the kernel away from hotpluggable memory.
1007 if (movable_node_is_enabled())
1008 memblock_set_bottom_up(true);
1013 /* after early param, so could get panic from serial */
1014 memblock_x86_reserve_range_setup_data();
1016 if (acpi_mps_check()) {
1017 #ifdef CONFIG_X86_LOCAL_APIC
1020 setup_clear_cpu_cap(X86_FEATURE_APIC);
1024 if (pci_early_dump_regs)
1025 early_dump_pci_devices();
1028 /* update the e820_saved too */
1029 e820_reserve_setup_data();
1030 finish_e820_parsing();
1032 if (efi_enabled(EFI_BOOT))
1037 dmi_set_dump_stack_arch_desc();
1040 * VMware detection requires dmi to be available, so this
1041 * needs to be done after dmi_scan_machine, for the BP.
1043 init_hypervisor_platform();
1045 x86_init.resources.probe_roms();
1047 /* after parse_early_param, so could debug it */
1048 insert_resource(&iomem_resource, &code_resource);
1049 insert_resource(&iomem_resource, &data_resource);
1050 insert_resource(&iomem_resource, &bss_resource);
1052 e820_add_kernel_range();
1054 #ifdef CONFIG_X86_32
1055 if (ppro_with_ram_bug()) {
1056 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1058 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
1059 printk(KERN_INFO "fixed physical RAM map:\n");
1060 e820_print_map("bad_ppro");
1063 early_gart_iommu_check();
1067 * partially used pages are not usable - thus
1068 * we are rounding upwards:
1070 max_pfn = e820_end_of_ram_pfn();
1072 /* update e820 for memory not covered by WB MTRRs */
1074 if (mtrr_trim_uncached_memory(max_pfn))
1075 max_pfn = e820_end_of_ram_pfn();
1077 max_possible_pfn = max_pfn;
1080 * Define random base addresses for memory sections after max_pfn is
1081 * defined and before each memory section base is used.
1083 kernel_randomize_memory();
1085 #ifdef CONFIG_X86_32
1086 /* max_low_pfn get updated here */
1087 find_low_pfn_range();
1091 /* How many end-of-memory variables you have, grandma! */
1092 /* need this before calling reserve_initrd */
1093 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1094 max_low_pfn = e820_end_of_low_ram_pfn();
1096 max_low_pfn = max_pfn;
1098 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1102 * Find and reserve possible boot-time SMP configuration:
1106 reserve_ibft_region();
1108 early_alloc_pgt_buf();
1111 * Need to conclude brk, before memblock_x86_fill()
1112 * it could use memblock_find_in_range, could overlap with
1119 memblock_set_current_limit(ISA_END_ADDRESS);
1120 memblock_x86_fill();
1122 reserve_bios_regions();
1124 if (efi_enabled(EFI_MEMMAP)) {
1130 * The EFI specification says that boot service code won't be
1131 * called after ExitBootServices(). This is, in fact, a lie.
1133 efi_reserve_boot_services();
1136 /* preallocate 4k for mptable mpc */
1137 early_reserve_e820_mpc_new();
1139 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1140 setup_bios_corruption_check();
1143 #ifdef CONFIG_X86_32
1144 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1145 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1148 reserve_real_mode();
1150 trim_platform_memory_ranges();
1151 trim_low_memory_range();
1155 early_trap_pf_init();
1158 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1159 * with the current CR4 value. This may not be necessary, but
1160 * auditing all the early-boot CR4 manipulation would be needed to
1163 mmu_cr4_features = __read_cr4();
1165 memblock_set_current_limit(get_max_mapped());
1168 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1171 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1172 if (init_ohci1394_dma_early)
1173 init_ohci1394_dma_on_all_controllers();
1175 /* Allocate bigger log buffer */
1178 if (efi_enabled(EFI_BOOT)) {
1179 switch (boot_params.secure_boot) {
1180 case efi_secureboot_mode_disabled:
1181 pr_info("Secure boot disabled\n");
1183 case efi_secureboot_mode_enabled:
1184 pr_info("Secure boot enabled\n");
1187 pr_info("Secure boot could not be determined\n");
1194 acpi_table_upgrade();
1201 * Parse the ACPI tables for possible boot-time SMP configuration.
1203 acpi_boot_table_init();
1205 early_acpi_boot_init();
1208 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1211 * Reserve memory for crash kernel after SRAT is parsed so that it
1212 * won't consume hotpluggable memory.
1214 reserve_crashkernel();
1216 memblock_find_dma_reserve();
1218 #ifdef CONFIG_KVM_GUEST
1222 x86_init.paging.pagetable_init();
1226 #ifdef CONFIG_X86_32
1227 /* sync back kernel address range */
1228 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1229 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1233 * sync back low identity map too. It is used for example
1234 * in the 32-bit EFI stub.
1236 clone_pgd_range(initial_page_table,
1237 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1238 min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY));
1245 generic_apic_probe();
1250 * Read APIC and some other early information from ACPI tables.
1257 * get boot-time SMP configuration:
1262 * Systems w/o ACPI and mptables might not have it mapped the local
1263 * APIC yet, but prefill_possible_map() might need to access it.
1265 init_apic_mappings();
1267 prefill_possible_map();
1271 io_apic_init_mappings();
1275 e820_reserve_resources();
1276 e820_mark_nosave_regions(max_low_pfn);
1278 x86_init.resources.reserve_resources();
1283 #if defined(CONFIG_VGA_CONSOLE)
1284 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1285 conswitchp = &vga_con;
1286 #elif defined(CONFIG_DUMMY_CONSOLE)
1287 conswitchp = &dummy_con;
1290 x86_init.oem.banner();
1292 x86_init.timers.wallclock_init();
1296 arch_init_ideal_nops();
1298 register_refined_jiffies(CLOCK_TICK_RATE);
1301 if (efi_enabled(EFI_BOOT))
1302 efi_apply_memmap_quirks();
1306 #ifdef CONFIG_X86_32
1308 static struct resource video_ram_resource = {
1309 .name = "Video RAM area",
1312 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1315 void __init i386_reserve_resources(void)
1317 request_resource(&iomem_resource, &video_ram_resource);
1318 reserve_standard_io_resources();
1321 #endif /* CONFIG_X86_32 */
1323 static struct notifier_block kernel_offset_notifier = {
1324 .notifier_call = dump_kernel_offset
1327 static int __init register_kernel_offset_dumper(void)
1329 atomic_notifier_chain_register(&panic_notifier_list,
1330 &kernel_offset_notifier);
1333 __initcall(register_kernel_offset_dumper);
1335 void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1337 if (!boot_cpu_has(X86_FEATURE_OSPKE))
1340 seq_printf(m, "ProtectionKey: %8u\n", vma_pkey(vma));