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/memblock.h>
34 #include <linux/seq_file.h>
35 #include <linux/console.h>
36 #include <linux/root_dev.h>
37 #include <linux/highmem.h>
38 #include <linux/export.h>
39 #include <linux/efi.h>
40 #include <linux/init.h>
41 #include <linux/edd.h>
42 #include <linux/iscsi_ibft.h>
43 #include <linux/nodemask.h>
44 #include <linux/kexec.h>
45 #include <linux/dmi.h>
46 #include <linux/pfn.h>
47 #include <linux/pci.h>
48 #include <asm/pci-direct.h>
49 #include <linux/init_ohci1394_dma.h>
50 #include <linux/kvm_para.h>
51 #include <linux/dma-contiguous.h>
53 #include <uapi/linux/mount.h>
55 #include <linux/errno.h>
56 #include <linux/kernel.h>
57 #include <linux/stddef.h>
58 #include <linux/unistd.h>
59 #include <linux/ptrace.h>
60 #include <linux/user.h>
61 #include <linux/delay.h>
63 #include <linux/kallsyms.h>
64 #include <linux/cpufreq.h>
65 #include <linux/dma-mapping.h>
66 #include <linux/ctype.h>
67 #include <linux/uaccess.h>
69 #include <linux/percpu.h>
70 #include <linux/crash_dump.h>
71 #include <linux/tboot.h>
72 #include <linux/jiffies.h>
73 #include <linux/mem_encrypt.h>
74 #include <linux/sizes.h>
76 #include <linux/usb/xhci-dbgp.h>
77 #include <video/edid.h>
81 #include <asm/realmode.h>
82 #include <asm/e820/api.h>
83 #include <asm/mpspec.h>
84 #include <asm/setup.h>
86 #include <asm/timer.h>
87 #include <asm/i8259.h>
88 #include <asm/sections.h>
89 #include <asm/io_apic.h>
91 #include <asm/setup_arch.h>
92 #include <asm/bios_ebda.h>
93 #include <asm/cacheflush.h>
94 #include <asm/processor.h>
96 #include <asm/kasan.h>
98 #include <asm/vsyscall.h>
100 #include <asm/desc.h>
102 #include <asm/iommu.h>
103 #include <asm/gart.h>
104 #include <asm/mmu_context.h>
105 #include <asm/proto.h>
107 #include <asm/paravirt.h>
108 #include <asm/hypervisor.h>
109 #include <asm/olpc_ofw.h>
111 #include <asm/percpu.h>
112 #include <asm/topology.h>
113 #include <asm/apicdef.h>
114 #include <asm/amd_nb.h>
116 #include <asm/alternative.h>
117 #include <asm/prom.h>
118 #include <asm/microcode.h>
119 #include <asm/kaslr.h>
120 #include <asm/unwind.h>
123 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
124 * max_pfn_mapped: highest direct mapped pfn over 4GB
126 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
127 * represented by pfn_mapped
129 unsigned long max_low_pfn_mapped;
130 unsigned long max_pfn_mapped;
133 RESERVE_BRK(dmi_alloc, 65536);
137 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
138 unsigned long _brk_end = (unsigned long)__brk_base;
140 struct boot_params boot_params;
145 static struct resource data_resource = {
146 .name = "Kernel data",
149 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
152 static struct resource code_resource = {
153 .name = "Kernel code",
156 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
159 static struct resource bss_resource = {
160 .name = "Kernel bss",
163 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
168 /* cpu data as detected by the assembly code in head_32.S */
169 struct cpuinfo_x86 new_cpu_data;
171 /* common cpu data for all cpus */
172 struct cpuinfo_x86 boot_cpu_data __read_mostly;
173 EXPORT_SYMBOL(boot_cpu_data);
175 unsigned int def_to_bigsmp;
177 /* for MCA, but anyone else can use it if they want */
178 unsigned int machine_id;
179 unsigned int machine_submodel_id;
180 unsigned int BIOS_revision;
182 struct apm_info apm_info;
183 EXPORT_SYMBOL(apm_info);
185 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
186 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
187 struct ist_info ist_info;
188 EXPORT_SYMBOL(ist_info);
190 struct ist_info ist_info;
194 struct cpuinfo_x86 boot_cpu_data __read_mostly;
195 EXPORT_SYMBOL(boot_cpu_data);
199 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
200 __visible unsigned long mmu_cr4_features __ro_after_init;
202 __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
205 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
206 int bootloader_type, bootloader_version;
211 struct screen_info screen_info;
212 EXPORT_SYMBOL(screen_info);
213 struct edid_info edid_info;
214 EXPORT_SYMBOL_GPL(edid_info);
216 extern int root_mountflags;
218 unsigned long saved_video_mode;
220 #define RAMDISK_IMAGE_START_MASK 0x07FF
221 #define RAMDISK_PROMPT_FLAG 0x8000
222 #define RAMDISK_LOAD_FLAG 0x4000
224 static char __initdata command_line[COMMAND_LINE_SIZE];
225 #ifdef CONFIG_CMDLINE_BOOL
226 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
229 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
231 #ifdef CONFIG_EDD_MODULE
235 * copy_edd() - Copy the BIOS EDD information
236 * from boot_params into a safe place.
239 static inline void __init copy_edd(void)
241 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
242 sizeof(edd.mbr_signature));
243 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
244 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
245 edd.edd_info_nr = boot_params.eddbuf_entries;
248 static inline void __init copy_edd(void)
253 void * __init extend_brk(size_t size, size_t align)
255 size_t mask = align - 1;
258 BUG_ON(_brk_start == 0);
259 BUG_ON(align & mask);
261 _brk_end = (_brk_end + mask) & ~mask;
262 BUG_ON((char *)(_brk_end + size) > __brk_limit);
264 ret = (void *)_brk_end;
267 memset(ret, 0, size);
273 static void __init cleanup_highmap(void)
278 static void __init reserve_brk(void)
280 if (_brk_end > _brk_start)
281 memblock_reserve(__pa_symbol(_brk_start),
282 _brk_end - _brk_start);
284 /* Mark brk area as locked down and no longer taking any
289 u64 relocated_ramdisk;
291 #ifdef CONFIG_BLK_DEV_INITRD
293 static u64 __init get_ramdisk_image(void)
295 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
297 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
299 return ramdisk_image;
301 static u64 __init get_ramdisk_size(void)
303 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
305 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
310 static void __init relocate_initrd(void)
312 /* Assume only end is not page aligned */
313 u64 ramdisk_image = get_ramdisk_image();
314 u64 ramdisk_size = get_ramdisk_size();
315 u64 area_size = PAGE_ALIGN(ramdisk_size);
317 /* We need to move the initrd down into directly mapped mem */
318 relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
319 area_size, PAGE_SIZE);
321 if (!relocated_ramdisk)
322 panic("Cannot find place for new RAMDISK of size %lld\n",
325 /* Note: this includes all the mem currently occupied by
326 the initrd, we rely on that fact to keep the data intact. */
327 memblock_reserve(relocated_ramdisk, area_size);
328 initrd_start = relocated_ramdisk + PAGE_OFFSET;
329 initrd_end = initrd_start + ramdisk_size;
330 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
331 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
333 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
335 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
336 " [mem %#010llx-%#010llx]\n",
337 ramdisk_image, ramdisk_image + ramdisk_size - 1,
338 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
341 static void __init early_reserve_initrd(void)
343 /* Assume only end is not page aligned */
344 u64 ramdisk_image = get_ramdisk_image();
345 u64 ramdisk_size = get_ramdisk_size();
346 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
348 if (!boot_params.hdr.type_of_loader ||
349 !ramdisk_image || !ramdisk_size)
350 return; /* No initrd provided by bootloader */
352 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
354 static void __init 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);
362 if (!boot_params.hdr.type_of_loader ||
363 !ramdisk_image || !ramdisk_size)
364 return; /* No initrd provided by bootloader */
368 mapped_size = memblock_mem_size(max_pfn_mapped);
369 if (ramdisk_size >= (mapped_size>>1))
370 panic("initrd too large to handle, "
371 "disabling initrd (%lld needed, %lld available)\n",
372 ramdisk_size, mapped_size>>1);
374 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
377 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
378 PFN_DOWN(ramdisk_end))) {
379 /* All are mapped, easy case */
380 initrd_start = ramdisk_image + PAGE_OFFSET;
381 initrd_end = initrd_start + ramdisk_size;
387 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
391 static void __init early_reserve_initrd(void)
394 static void __init reserve_initrd(void)
397 #endif /* CONFIG_BLK_DEV_INITRD */
399 static void __init parse_setup_data(void)
401 struct setup_data *data;
402 u64 pa_data, pa_next;
404 pa_data = boot_params.hdr.setup_data;
406 u32 data_len, data_type;
408 data = early_memremap(pa_data, sizeof(*data));
409 data_len = data->len + sizeof(struct setup_data);
410 data_type = data->type;
411 pa_next = data->next;
412 early_memunmap(data, sizeof(*data));
416 e820__memory_setup_extended(pa_data, data_len);
422 parse_efi_setup(pa_data, data_len);
431 static void __init memblock_x86_reserve_range_setup_data(void)
433 struct setup_data *data;
436 pa_data = boot_params.hdr.setup_data;
438 data = early_memremap(pa_data, sizeof(*data));
439 memblock_reserve(pa_data, sizeof(*data) + data->len);
440 pa_data = data->next;
441 early_memunmap(data, sizeof(*data));
446 * --------- Crashkernel reservation ------------------------------
449 #ifdef CONFIG_KEXEC_CORE
451 /* 16M alignment for crash kernel regions */
452 #define CRASH_ALIGN SZ_16M
455 * Keep the crash kernel below this limit. On 32 bits earlier kernels
456 * would limit the kernel to the low 512 MiB due to mapping restrictions.
459 # define CRASH_ADDR_LOW_MAX SZ_512M
460 # define CRASH_ADDR_HIGH_MAX SZ_512M
462 # define CRASH_ADDR_LOW_MAX SZ_4G
463 # define CRASH_ADDR_HIGH_MAX MAXMEM
466 static int __init reserve_crashkernel_low(void)
469 unsigned long long base, low_base = 0, low_size = 0;
470 unsigned long total_low_mem;
473 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
475 /* crashkernel=Y,low */
476 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
479 * two parts from lib/swiotlb.c:
480 * -swiotlb size: user-specified with swiotlb= or default.
482 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
483 * to 8M for other buffers that may need to stay low too. Also
484 * make sure we allocate enough extra low memory so that we
485 * don't run out of DMA buffers for 32-bit devices.
487 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
489 /* passed with crashkernel=0,low ? */
494 low_base = memblock_find_in_range(0, 1ULL << 32, low_size, CRASH_ALIGN);
496 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
497 (unsigned long)(low_size >> 20));
501 ret = memblock_reserve(low_base, low_size);
503 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
507 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
508 (unsigned long)(low_size >> 20),
509 (unsigned long)(low_base >> 20),
510 (unsigned long)(total_low_mem >> 20));
512 crashk_low_res.start = low_base;
513 crashk_low_res.end = low_base + low_size - 1;
514 insert_resource(&iomem_resource, &crashk_low_res);
519 static void __init reserve_crashkernel(void)
521 unsigned long long crash_size, crash_base, total_mem;
525 total_mem = memblock_phys_mem_size();
528 ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
529 if (ret != 0 || crash_size <= 0) {
530 /* crashkernel=X,high */
531 ret = parse_crashkernel_high(boot_command_line, total_mem,
532 &crash_size, &crash_base);
533 if (ret != 0 || crash_size <= 0)
538 if (xen_pv_domain()) {
539 pr_info("Ignoring crashkernel for a Xen PV domain\n");
543 /* 0 means: find the address automatically */
546 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
547 * crashkernel=x,high reserves memory over 4G, also allocates
548 * 256M extra low memory for DMA buffers and swiotlb.
549 * But the extra memory is not required for all machines.
550 * So try low memory first and fall back to high memory
551 * unless "crashkernel=size[KMG],high" is specified.
554 crash_base = memblock_find_in_range(CRASH_ALIGN,
556 crash_size, CRASH_ALIGN);
558 crash_base = memblock_find_in_range(CRASH_ALIGN,
560 crash_size, CRASH_ALIGN);
562 pr_info("crashkernel reservation failed - No suitable area found.\n");
566 unsigned long long start;
568 start = memblock_find_in_range(crash_base,
569 crash_base + crash_size,
570 crash_size, 1 << 20);
571 if (start != crash_base) {
572 pr_info("crashkernel reservation failed - memory is in use.\n");
576 ret = memblock_reserve(crash_base, crash_size);
578 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
582 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
583 memblock_free(crash_base, crash_size);
587 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
588 (unsigned long)(crash_size >> 20),
589 (unsigned long)(crash_base >> 20),
590 (unsigned long)(total_mem >> 20));
592 crashk_res.start = crash_base;
593 crashk_res.end = crash_base + crash_size - 1;
594 insert_resource(&iomem_resource, &crashk_res);
597 static void __init reserve_crashkernel(void)
602 static struct resource standard_io_resources[] = {
603 { .name = "dma1", .start = 0x00, .end = 0x1f,
604 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
605 { .name = "pic1", .start = 0x20, .end = 0x21,
606 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
607 { .name = "timer0", .start = 0x40, .end = 0x43,
608 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
609 { .name = "timer1", .start = 0x50, .end = 0x53,
610 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
611 { .name = "keyboard", .start = 0x60, .end = 0x60,
612 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
613 { .name = "keyboard", .start = 0x64, .end = 0x64,
614 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
615 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
616 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
617 { .name = "pic2", .start = 0xa0, .end = 0xa1,
618 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
619 { .name = "dma2", .start = 0xc0, .end = 0xdf,
620 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
621 { .name = "fpu", .start = 0xf0, .end = 0xff,
622 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
625 void __init reserve_standard_io_resources(void)
629 /* request I/O space for devices used on all i[345]86 PCs */
630 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
631 request_resource(&ioport_resource, &standard_io_resources[i]);
635 static __init void reserve_ibft_region(void)
637 unsigned long addr, size = 0;
639 addr = find_ibft_region(&size);
642 memblock_reserve(addr, size);
645 static bool __init snb_gfx_workaround_needed(void)
650 static const __initconst u16 snb_ids[] = {
660 /* Assume no if something weird is going on with PCI */
661 if (!early_pci_allowed())
664 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
665 if (vendor != 0x8086)
668 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
669 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
670 if (devid == snb_ids[i])
678 * Sandy Bridge graphics has trouble with certain ranges, exclude
679 * them from allocation.
681 static void __init trim_snb_memory(void)
683 static const __initconst unsigned long bad_pages[] = {
692 if (!snb_gfx_workaround_needed())
695 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
698 * Reserve all memory below the 1 MB mark that has not
699 * already been reserved.
701 memblock_reserve(0, 1<<20);
703 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
704 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
705 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
711 * Here we put platform-specific memory range workarounds, i.e.
712 * memory known to be corrupt or otherwise in need to be reserved on
713 * specific platforms.
715 * If this gets used more widely it could use a real dispatch mechanism.
717 static void __init trim_platform_memory_ranges(void)
722 static void __init trim_bios_range(void)
725 * A special case is the first 4Kb of memory;
726 * This is a BIOS owned area, not kernel ram, but generally
727 * not listed as such in the E820 table.
729 * This typically reserves additional memory (64KiB by default)
730 * since some BIOSes are known to corrupt low memory. See the
731 * Kconfig help text for X86_RESERVE_LOW.
733 e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED);
736 * special case: Some BIOSen report the PC BIOS
737 * area (640->1Mb) as ram even though it is not.
740 e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1);
742 e820__update_table(e820_table);
745 /* called before trim_bios_range() to spare extra sanitize */
746 static void __init e820_add_kernel_range(void)
748 u64 start = __pa_symbol(_text);
749 u64 size = __pa_symbol(_end) - start;
752 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
753 * attempt to fix it by adding the range. We may have a confused BIOS,
754 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
755 * exclude kernel range. If we really are running on top non-RAM,
756 * we will crash later anyways.
758 if (e820__mapped_all(start, start + size, E820_TYPE_RAM))
761 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
762 e820__range_remove(start, size, E820_TYPE_RAM, 0);
763 e820__range_add(start, size, E820_TYPE_RAM);
766 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
768 static int __init parse_reservelow(char *p)
770 unsigned long long size;
775 size = memparse(p, &p);
788 early_param("reservelow", parse_reservelow);
790 static void __init trim_low_memory_range(void)
792 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
796 * Dump out kernel offset information on panic.
799 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
801 if (kaslr_enabled()) {
802 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
808 pr_emerg("Kernel Offset: disabled\n");
815 * Determine if we were loaded by an EFI loader. If so, then we have also been
816 * passed the efi memmap, systab, etc., so we should use these data structures
817 * for initialization. Note, the efi init code path is determined by the
818 * global efi_enabled. This allows the same kernel image to be used on existing
819 * systems (with a traditional BIOS) as well as on EFI systems.
822 * setup_arch - architecture-specific boot-time initializations
824 * Note: On x86_64, fixmaps are ready for use even before this is called.
827 void __init setup_arch(char **cmdline_p)
829 memblock_reserve(__pa_symbol(_text),
830 (unsigned long)__bss_stop - (unsigned long)_text);
833 * Make sure page 0 is always reserved because on systems with
834 * L1TF its contents can be leaked to user processes.
836 memblock_reserve(0, PAGE_SIZE);
838 early_reserve_initrd();
841 * At this point everything still needed from the boot loader
842 * or BIOS or kernel text should be early reserved or marked not
843 * RAM in e820. All other memory is free game.
847 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
850 * copy kernel address range established so far and switch
851 * to the proper swapper page table
853 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
854 initial_page_table + KERNEL_PGD_BOUNDARY,
857 load_cr3(swapper_pg_dir);
859 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
860 * a cr3 based tlb flush, so the following __flush_tlb_all()
861 * will not flush anything because the cpu quirk which clears
862 * X86_FEATURE_PGE has not been invoked yet. Though due to the
863 * load_cr3() above the TLB has been flushed already. The
864 * quirk is invoked before subsequent calls to __flush_tlb_all()
865 * so proper operation is guaranteed.
869 printk(KERN_INFO "Command line: %s\n", boot_command_line);
870 boot_cpu_data.x86_phys_bits = MAX_PHYSMEM_BITS;
874 * If we have OLPC OFW, we might end up relocating the fixmap due to
875 * reserve_top(), so do this before touching the ioremap area.
879 idt_setup_early_traps();
881 arch_init_ideal_nops();
883 early_ioremap_init();
885 setup_olpc_ofw_pgd();
887 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
888 screen_info = boot_params.screen_info;
889 edid_info = boot_params.edid_info;
891 apm_info.bios = boot_params.apm_bios_info;
892 ist_info = boot_params.ist_info;
894 saved_video_mode = boot_params.hdr.vid_mode;
895 bootloader_type = boot_params.hdr.type_of_loader;
896 if ((bootloader_type >> 4) == 0xe) {
897 bootloader_type &= 0xf;
898 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
900 bootloader_version = bootloader_type & 0xf;
901 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
903 #ifdef CONFIG_BLK_DEV_RAM
904 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
905 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
906 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
909 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
910 EFI32_LOADER_SIGNATURE, 4)) {
911 set_bit(EFI_BOOT, &efi.flags);
912 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
913 EFI64_LOADER_SIGNATURE, 4)) {
914 set_bit(EFI_BOOT, &efi.flags);
915 set_bit(EFI_64BIT, &efi.flags);
919 x86_init.oem.arch_setup();
921 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
922 e820__memory_setup();
927 if (!boot_params.hdr.root_flags)
928 root_mountflags &= ~MS_RDONLY;
929 init_mm.start_code = (unsigned long) _text;
930 init_mm.end_code = (unsigned long) _etext;
931 init_mm.end_data = (unsigned long) _edata;
932 init_mm.brk = _brk_end;
934 mpx_mm_init(&init_mm);
936 code_resource.start = __pa_symbol(_text);
937 code_resource.end = __pa_symbol(_etext)-1;
938 data_resource.start = __pa_symbol(_etext);
939 data_resource.end = __pa_symbol(_edata)-1;
940 bss_resource.start = __pa_symbol(__bss_start);
941 bss_resource.end = __pa_symbol(__bss_stop)-1;
943 #ifdef CONFIG_CMDLINE_BOOL
944 #ifdef CONFIG_CMDLINE_OVERRIDE
945 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
947 if (builtin_cmdline[0]) {
948 /* append boot loader cmdline to builtin */
949 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
950 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
951 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
956 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
957 *cmdline_p = command_line;
960 * x86_configure_nx() is called before parse_early_param() to detect
961 * whether hardware doesn't support NX (so that the early EHCI debug
962 * console setup can safely call set_fixmap()). It may then be called
963 * again from within noexec_setup() during parsing early parameters
964 * to honor the respective command line option.
970 if (efi_enabled(EFI_BOOT))
971 efi_memblock_x86_reserve_range();
972 #ifdef CONFIG_MEMORY_HOTPLUG
974 * Memory used by the kernel cannot be hot-removed because Linux
975 * cannot migrate the kernel pages. When memory hotplug is
976 * enabled, we should prevent memblock from allocating memory
979 * ACPI SRAT records all hotpluggable memory ranges. But before
980 * SRAT is parsed, we don't know about it.
982 * The kernel image is loaded into memory at very early time. We
983 * cannot prevent this anyway. So on NUMA system, we set any
984 * node the kernel resides in as un-hotpluggable.
986 * Since on modern servers, one node could have double-digit
987 * gigabytes memory, we can assume the memory around the kernel
988 * image is also un-hotpluggable. So before SRAT is parsed, just
989 * allocate memory near the kernel image to try the best to keep
990 * the kernel away from hotpluggable memory.
992 if (movable_node_is_enabled())
993 memblock_set_bottom_up(true);
998 /* after early param, so could get panic from serial */
999 memblock_x86_reserve_range_setup_data();
1001 if (acpi_mps_check()) {
1002 #ifdef CONFIG_X86_LOCAL_APIC
1005 setup_clear_cpu_cap(X86_FEATURE_APIC);
1008 e820__reserve_setup_data();
1009 e820__finish_early_params();
1011 if (efi_enabled(EFI_BOOT))
1017 * VMware detection requires dmi to be available, so this
1018 * needs to be done after dmi_setup(), for the boot CPU.
1020 init_hypervisor_platform();
1023 x86_init.resources.probe_roms();
1025 /* after parse_early_param, so could debug it */
1026 insert_resource(&iomem_resource, &code_resource);
1027 insert_resource(&iomem_resource, &data_resource);
1028 insert_resource(&iomem_resource, &bss_resource);
1030 e820_add_kernel_range();
1032 #ifdef CONFIG_X86_32
1033 if (ppro_with_ram_bug()) {
1034 e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM,
1035 E820_TYPE_RESERVED);
1036 e820__update_table(e820_table);
1037 printk(KERN_INFO "fixed physical RAM map:\n");
1038 e820__print_table("bad_ppro");
1041 early_gart_iommu_check();
1045 * partially used pages are not usable - thus
1046 * we are rounding upwards:
1048 max_pfn = e820__end_of_ram_pfn();
1050 /* update e820 for memory not covered by WB MTRRs */
1052 if (mtrr_trim_uncached_memory(max_pfn))
1053 max_pfn = e820__end_of_ram_pfn();
1055 max_possible_pfn = max_pfn;
1058 * This call is required when the CPU does not support PAT. If
1059 * mtrr_bp_init() invoked it already via pat_init() the call has no
1065 * Define random base addresses for memory sections after max_pfn is
1066 * defined and before each memory section base is used.
1068 kernel_randomize_memory();
1070 #ifdef CONFIG_X86_32
1071 /* max_low_pfn get updated here */
1072 find_low_pfn_range();
1076 /* How many end-of-memory variables you have, grandma! */
1077 /* need this before calling reserve_initrd */
1078 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1079 max_low_pfn = e820__end_of_low_ram_pfn();
1081 max_low_pfn = max_pfn;
1083 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1087 * Find and reserve possible boot-time SMP configuration:
1091 reserve_ibft_region();
1093 early_alloc_pgt_buf();
1096 * Need to conclude brk, before e820__memblock_setup()
1097 * it could use memblock_find_in_range, could overlap with
1104 memblock_set_current_limit(ISA_END_ADDRESS);
1105 e820__memblock_setup();
1107 reserve_bios_regions();
1109 if (efi_enabled(EFI_MEMMAP)) {
1115 * The EFI specification says that boot service code won't be
1116 * called after ExitBootServices(). This is, in fact, a lie.
1118 efi_reserve_boot_services();
1121 /* preallocate 4k for mptable mpc */
1122 e820__memblock_alloc_reserved_mpc_new();
1124 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1125 setup_bios_corruption_check();
1128 #ifdef CONFIG_X86_32
1129 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1130 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1133 reserve_real_mode();
1135 trim_platform_memory_ranges();
1136 trim_low_memory_range();
1140 idt_setup_early_pf();
1143 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1144 * with the current CR4 value. This may not be necessary, but
1145 * auditing all the early-boot CR4 manipulation would be needed to
1148 * Mask off features that don't work outside long mode (just
1151 mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE;
1153 memblock_set_current_limit(get_max_mapped());
1156 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1159 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1160 if (init_ohci1394_dma_early)
1161 init_ohci1394_dma_on_all_controllers();
1163 /* Allocate bigger log buffer */
1166 if (efi_enabled(EFI_BOOT)) {
1167 switch (boot_params.secure_boot) {
1168 case efi_secureboot_mode_disabled:
1169 pr_info("Secure boot disabled\n");
1171 case efi_secureboot_mode_enabled:
1172 pr_info("Secure boot enabled\n");
1175 pr_info("Secure boot could not be determined\n");
1182 acpi_table_upgrade();
1188 early_platform_quirks();
1191 * Parse the ACPI tables for possible boot-time SMP configuration.
1193 acpi_boot_table_init();
1195 early_acpi_boot_init();
1198 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1201 * Reserve memory for crash kernel after SRAT is parsed so that it
1202 * won't consume hotpluggable memory.
1204 reserve_crashkernel();
1206 memblock_find_dma_reserve();
1208 if (!early_xdbc_setup_hardware())
1209 early_xdbc_register_console();
1211 x86_init.paging.pagetable_init();
1216 * Sync back kernel address range.
1218 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1221 sync_initial_page_table();
1227 generic_apic_probe();
1232 * Read APIC and some other early information from ACPI tables.
1239 * get boot-time SMP configuration:
1244 * Systems w/o ACPI and mptables might not have it mapped the local
1245 * APIC yet, but prefill_possible_map() might need to access it.
1247 init_apic_mappings();
1249 prefill_possible_map();
1253 io_apic_init_mappings();
1255 x86_init.hyper.guest_late_init();
1257 e820__reserve_resources();
1258 e820__register_nosave_regions(max_pfn);
1260 x86_init.resources.reserve_resources();
1262 e820__setup_pci_gap();
1265 #if defined(CONFIG_VGA_CONSOLE)
1266 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1267 conswitchp = &vga_con;
1268 #elif defined(CONFIG_DUMMY_CONSOLE)
1269 conswitchp = &dummy_con;
1272 x86_init.oem.banner();
1274 x86_init.timers.wallclock_init();
1278 register_refined_jiffies(CLOCK_TICK_RATE);
1281 if (efi_enabled(EFI_BOOT))
1282 efi_apply_memmap_quirks();
1288 #ifdef CONFIG_X86_32
1290 static struct resource video_ram_resource = {
1291 .name = "Video RAM area",
1294 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1297 void __init i386_reserve_resources(void)
1299 request_resource(&iomem_resource, &video_ram_resource);
1300 reserve_standard_io_resources();
1303 #endif /* CONFIG_X86_32 */
1305 static struct notifier_block kernel_offset_notifier = {
1306 .notifier_call = dump_kernel_offset
1309 static int __init register_kernel_offset_dumper(void)
1311 atomic_notifier_chain_register(&panic_notifier_list,
1312 &kernel_offset_notifier);
1315 __initcall(register_kernel_offset_dumper);