1 // SPDX-License-Identifier: GPL-2.0-only
3 * Copyright (C) 1995 Linus Torvalds
5 * This file contains the setup_arch() code, which handles the architecture-dependent
6 * parts of early kernel initialization.
8 #include <linux/console.h>
9 #include <linux/crash_dump.h>
10 #include <linux/dmi.h>
11 #include <linux/efi.h>
12 #include <linux/init_ohci1394_dma.h>
13 #include <linux/initrd.h>
14 #include <linux/iscsi_ibft.h>
15 #include <linux/memblock.h>
16 #include <linux/pci.h>
17 #include <linux/root_dev.h>
18 #include <linux/sfi.h>
19 #include <linux/hugetlb.h>
20 #include <linux/tboot.h>
21 #include <linux/usb/xhci-dbgp.h>
23 #include <uapi/linux/mount.h>
28 #include <asm/bios_ebda.h>
33 #include <asm/hypervisor.h>
34 #include <asm/io_apic.h>
35 #include <asm/kasan.h>
36 #include <asm/kaslr.h>
39 #include <asm/realmode.h>
40 #include <asm/olpc_ofw.h>
41 #include <asm/pci-direct.h>
43 #include <asm/proto.h>
44 #include <asm/unwind.h>
45 #include <asm/vsyscall.h>
46 #include <linux/vmalloc.h>
49 * max_low_pfn_mapped: highest directly mapped pfn < 4 GB
50 * max_pfn_mapped: highest directly mapped pfn > 4 GB
52 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
53 * represented by pfn_mapped[].
55 unsigned long max_low_pfn_mapped;
56 unsigned long max_pfn_mapped;
59 RESERVE_BRK(dmi_alloc, 65536);
64 * Range of the BSS area. The size of the BSS area is determined
65 * at link time, with RESERVE_BRK*() facility reserving additional
68 unsigned long _brk_start = (unsigned long)__brk_base;
69 unsigned long _brk_end = (unsigned long)__brk_base;
71 struct boot_params boot_params;
74 * These are the four main kernel memory regions, we put them into
75 * the resource tree so that kdump tools and other debugging tools
79 static struct resource rodata_resource = {
80 .name = "Kernel rodata",
83 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
86 static struct resource data_resource = {
87 .name = "Kernel data",
90 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
93 static struct resource code_resource = {
94 .name = "Kernel code",
97 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
100 static struct resource bss_resource = {
101 .name = "Kernel bss",
104 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
109 /* CPU data as detected by the assembly code in head_32.S */
110 struct cpuinfo_x86 new_cpu_data;
112 /* Common CPU data for all CPUs */
113 struct cpuinfo_x86 boot_cpu_data __read_mostly;
114 EXPORT_SYMBOL(boot_cpu_data);
116 unsigned int def_to_bigsmp;
118 /* For MCA, but anyone else can use it if they want */
119 unsigned int machine_id;
120 unsigned int machine_submodel_id;
121 unsigned int BIOS_revision;
123 struct apm_info apm_info;
124 EXPORT_SYMBOL(apm_info);
126 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
127 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
128 struct ist_info ist_info;
129 EXPORT_SYMBOL(ist_info);
131 struct ist_info ist_info;
135 struct cpuinfo_x86 boot_cpu_data __read_mostly;
136 EXPORT_SYMBOL(boot_cpu_data);
140 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
141 __visible unsigned long mmu_cr4_features __ro_after_init;
143 __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
146 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
147 int bootloader_type, bootloader_version;
152 struct screen_info screen_info;
153 EXPORT_SYMBOL(screen_info);
154 struct edid_info edid_info;
155 EXPORT_SYMBOL_GPL(edid_info);
157 extern int root_mountflags;
159 unsigned long saved_video_mode;
161 #define RAMDISK_IMAGE_START_MASK 0x07FF
162 #define RAMDISK_PROMPT_FLAG 0x8000
163 #define RAMDISK_LOAD_FLAG 0x4000
165 static char __initdata command_line[COMMAND_LINE_SIZE];
166 #ifdef CONFIG_CMDLINE_BOOL
167 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
170 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
172 #ifdef CONFIG_EDD_MODULE
176 * copy_edd() - Copy the BIOS EDD information
177 * from boot_params into a safe place.
180 static inline void __init copy_edd(void)
182 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
183 sizeof(edd.mbr_signature));
184 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
185 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
186 edd.edd_info_nr = boot_params.eddbuf_entries;
189 static inline void __init copy_edd(void)
194 void * __init extend_brk(size_t size, size_t align)
196 size_t mask = align - 1;
199 BUG_ON(_brk_start == 0);
200 BUG_ON(align & mask);
202 _brk_end = (_brk_end + mask) & ~mask;
203 BUG_ON((char *)(_brk_end + size) > __brk_limit);
205 ret = (void *)_brk_end;
208 memset(ret, 0, size);
214 static void __init cleanup_highmap(void)
219 static void __init reserve_brk(void)
221 if (_brk_end > _brk_start)
222 memblock_reserve(__pa_symbol(_brk_start),
223 _brk_end - _brk_start);
225 /* Mark brk area as locked down and no longer taking any
230 u64 relocated_ramdisk;
232 #ifdef CONFIG_BLK_DEV_INITRD
234 static u64 __init get_ramdisk_image(void)
236 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
238 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
240 if (ramdisk_image == 0)
241 ramdisk_image = phys_initrd_start;
243 return ramdisk_image;
245 static u64 __init get_ramdisk_size(void)
247 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
249 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
251 if (ramdisk_size == 0)
252 ramdisk_size = phys_initrd_size;
257 static void __init relocate_initrd(void)
259 /* Assume only end is not page aligned */
260 u64 ramdisk_image = get_ramdisk_image();
261 u64 ramdisk_size = get_ramdisk_size();
262 u64 area_size = PAGE_ALIGN(ramdisk_size);
264 /* We need to move the initrd down into directly mapped mem */
265 relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
266 area_size, PAGE_SIZE);
268 if (!relocated_ramdisk)
269 panic("Cannot find place for new RAMDISK of size %lld\n",
272 /* Note: this includes all the mem currently occupied by
273 the initrd, we rely on that fact to keep the data intact. */
274 memblock_reserve(relocated_ramdisk, area_size);
275 initrd_start = relocated_ramdisk + PAGE_OFFSET;
276 initrd_end = initrd_start + ramdisk_size;
277 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
278 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
280 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
282 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
283 " [mem %#010llx-%#010llx]\n",
284 ramdisk_image, ramdisk_image + ramdisk_size - 1,
285 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
288 static void __init early_reserve_initrd(void)
290 /* Assume only end is not page aligned */
291 u64 ramdisk_image = get_ramdisk_image();
292 u64 ramdisk_size = get_ramdisk_size();
293 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
295 if (!boot_params.hdr.type_of_loader ||
296 !ramdisk_image || !ramdisk_size)
297 return; /* No initrd provided by bootloader */
299 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
301 static void __init reserve_initrd(void)
303 /* Assume only end is not page aligned */
304 u64 ramdisk_image = get_ramdisk_image();
305 u64 ramdisk_size = get_ramdisk_size();
306 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
309 if (!boot_params.hdr.type_of_loader ||
310 !ramdisk_image || !ramdisk_size)
311 return; /* No initrd provided by bootloader */
315 mapped_size = memblock_mem_size(max_pfn_mapped);
316 if (ramdisk_size >= (mapped_size>>1))
317 panic("initrd too large to handle, "
318 "disabling initrd (%lld needed, %lld available)\n",
319 ramdisk_size, mapped_size>>1);
321 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
324 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
325 PFN_DOWN(ramdisk_end))) {
326 /* All are mapped, easy case */
327 initrd_start = ramdisk_image + PAGE_OFFSET;
328 initrd_end = initrd_start + ramdisk_size;
334 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
338 static void __init early_reserve_initrd(void)
341 static void __init reserve_initrd(void)
344 #endif /* CONFIG_BLK_DEV_INITRD */
346 static void __init parse_setup_data(void)
348 struct setup_data *data;
349 u64 pa_data, pa_next;
351 pa_data = boot_params.hdr.setup_data;
353 u32 data_len, data_type;
355 data = early_memremap(pa_data, sizeof(*data));
356 data_len = data->len + sizeof(struct setup_data);
357 data_type = data->type;
358 pa_next = data->next;
359 early_memunmap(data, sizeof(*data));
363 e820__memory_setup_extended(pa_data, data_len);
369 parse_efi_setup(pa_data, data_len);
378 static void __init memblock_x86_reserve_range_setup_data(void)
380 struct setup_data *data;
383 pa_data = boot_params.hdr.setup_data;
385 data = early_memremap(pa_data, sizeof(*data));
386 memblock_reserve(pa_data, sizeof(*data) + data->len);
388 if (data->type == SETUP_INDIRECT &&
389 ((struct setup_indirect *)data->data)->type != SETUP_INDIRECT)
390 memblock_reserve(((struct setup_indirect *)data->data)->addr,
391 ((struct setup_indirect *)data->data)->len);
393 pa_data = data->next;
394 early_memunmap(data, sizeof(*data));
399 * --------- Crashkernel reservation ------------------------------
402 #ifdef CONFIG_KEXEC_CORE
404 /* 16M alignment for crash kernel regions */
405 #define CRASH_ALIGN SZ_16M
408 * Keep the crash kernel below this limit.
410 * Earlier 32-bits kernels would limit the kernel to the low 512 MB range
411 * due to mapping restrictions.
413 * 64-bit kdump kernels need to be restricted to be under 64 TB, which is
414 * the upper limit of system RAM in 4-level paging mode. Since the kdump
415 * jump could be from 5-level paging to 4-level paging, the jump will fail if
416 * the kernel is put above 64 TB, and during the 1st kernel bootup there's
417 * no good way to detect the paging mode of the target kernel which will be
418 * loaded for dumping.
421 # define CRASH_ADDR_LOW_MAX SZ_512M
422 # define CRASH_ADDR_HIGH_MAX SZ_512M
424 # define CRASH_ADDR_LOW_MAX SZ_4G
425 # define CRASH_ADDR_HIGH_MAX SZ_64T
428 static int __init reserve_crashkernel_low(void)
431 unsigned long long base, low_base = 0, low_size = 0;
432 unsigned long total_low_mem;
435 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
437 /* crashkernel=Y,low */
438 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
441 * two parts from kernel/dma/swiotlb.c:
442 * -swiotlb size: user-specified with swiotlb= or default.
444 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
445 * to 8M for other buffers that may need to stay low too. Also
446 * make sure we allocate enough extra low memory so that we
447 * don't run out of DMA buffers for 32-bit devices.
449 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
451 /* passed with crashkernel=0,low ? */
456 low_base = memblock_find_in_range(0, 1ULL << 32, low_size, CRASH_ALIGN);
458 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
459 (unsigned long)(low_size >> 20));
463 ret = memblock_reserve(low_base, low_size);
465 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
469 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
470 (unsigned long)(low_size >> 20),
471 (unsigned long)(low_base >> 20),
472 (unsigned long)(total_low_mem >> 20));
474 crashk_low_res.start = low_base;
475 crashk_low_res.end = low_base + low_size - 1;
476 insert_resource(&iomem_resource, &crashk_low_res);
481 static void __init reserve_crashkernel(void)
483 unsigned long long crash_size, crash_base, total_mem;
487 total_mem = memblock_phys_mem_size();
490 ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
491 if (ret != 0 || crash_size <= 0) {
492 /* crashkernel=X,high */
493 ret = parse_crashkernel_high(boot_command_line, total_mem,
494 &crash_size, &crash_base);
495 if (ret != 0 || crash_size <= 0)
500 if (xen_pv_domain()) {
501 pr_info("Ignoring crashkernel for a Xen PV domain\n");
505 /* 0 means: find the address automatically */
508 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
509 * crashkernel=x,high reserves memory over 4G, also allocates
510 * 256M extra low memory for DMA buffers and swiotlb.
511 * But the extra memory is not required for all machines.
512 * So try low memory first and fall back to high memory
513 * unless "crashkernel=size[KMG],high" is specified.
516 crash_base = memblock_find_in_range(CRASH_ALIGN,
518 crash_size, CRASH_ALIGN);
520 crash_base = memblock_find_in_range(CRASH_ALIGN,
522 crash_size, CRASH_ALIGN);
524 pr_info("crashkernel reservation failed - No suitable area found.\n");
528 unsigned long long start;
530 start = memblock_find_in_range(crash_base,
531 crash_base + crash_size,
532 crash_size, 1 << 20);
533 if (start != crash_base) {
534 pr_info("crashkernel reservation failed - memory is in use.\n");
538 ret = memblock_reserve(crash_base, crash_size);
540 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
544 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
545 memblock_free(crash_base, crash_size);
549 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
550 (unsigned long)(crash_size >> 20),
551 (unsigned long)(crash_base >> 20),
552 (unsigned long)(total_mem >> 20));
554 crashk_res.start = crash_base;
555 crashk_res.end = crash_base + crash_size - 1;
556 insert_resource(&iomem_resource, &crashk_res);
559 static void __init reserve_crashkernel(void)
564 static struct resource standard_io_resources[] = {
565 { .name = "dma1", .start = 0x00, .end = 0x1f,
566 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
567 { .name = "pic1", .start = 0x20, .end = 0x21,
568 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
569 { .name = "timer0", .start = 0x40, .end = 0x43,
570 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
571 { .name = "timer1", .start = 0x50, .end = 0x53,
572 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
573 { .name = "keyboard", .start = 0x60, .end = 0x60,
574 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
575 { .name = "keyboard", .start = 0x64, .end = 0x64,
576 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
577 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
578 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
579 { .name = "pic2", .start = 0xa0, .end = 0xa1,
580 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
581 { .name = "dma2", .start = 0xc0, .end = 0xdf,
582 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
583 { .name = "fpu", .start = 0xf0, .end = 0xff,
584 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
587 void __init reserve_standard_io_resources(void)
591 /* request I/O space for devices used on all i[345]86 PCs */
592 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
593 request_resource(&ioport_resource, &standard_io_resources[i]);
597 static __init void reserve_ibft_region(void)
599 unsigned long addr, size = 0;
601 addr = find_ibft_region(&size);
604 memblock_reserve(addr, size);
607 static bool __init snb_gfx_workaround_needed(void)
612 static const __initconst u16 snb_ids[] = {
622 /* Assume no if something weird is going on with PCI */
623 if (!early_pci_allowed())
626 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
627 if (vendor != 0x8086)
630 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
631 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
632 if (devid == snb_ids[i])
640 * Sandy Bridge graphics has trouble with certain ranges, exclude
641 * them from allocation.
643 static void __init trim_snb_memory(void)
645 static const __initconst unsigned long bad_pages[] = {
654 if (!snb_gfx_workaround_needed())
657 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
660 * Reserve all memory below the 1 MB mark that has not
661 * already been reserved.
663 memblock_reserve(0, 1<<20);
665 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
666 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
667 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
673 * Here we put platform-specific memory range workarounds, i.e.
674 * memory known to be corrupt or otherwise in need to be reserved on
675 * specific platforms.
677 * If this gets used more widely it could use a real dispatch mechanism.
679 static void __init trim_platform_memory_ranges(void)
684 static void __init trim_bios_range(void)
687 * A special case is the first 4Kb of memory;
688 * This is a BIOS owned area, not kernel ram, but generally
689 * not listed as such in the E820 table.
691 * This typically reserves additional memory (64KiB by default)
692 * since some BIOSes are known to corrupt low memory. See the
693 * Kconfig help text for X86_RESERVE_LOW.
695 e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED);
698 * special case: Some BIOSes report the PC BIOS
699 * area (640Kb -> 1Mb) as RAM even though it is not.
702 e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1);
704 e820__update_table(e820_table);
707 /* called before trim_bios_range() to spare extra sanitize */
708 static void __init e820_add_kernel_range(void)
710 u64 start = __pa_symbol(_text);
711 u64 size = __pa_symbol(_end) - start;
714 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
715 * attempt to fix it by adding the range. We may have a confused BIOS,
716 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
717 * exclude kernel range. If we really are running on top non-RAM,
718 * we will crash later anyways.
720 if (e820__mapped_all(start, start + size, E820_TYPE_RAM))
723 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
724 e820__range_remove(start, size, E820_TYPE_RAM, 0);
725 e820__range_add(start, size, E820_TYPE_RAM);
728 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
730 static int __init parse_reservelow(char *p)
732 unsigned long long size;
737 size = memparse(p, &p);
750 early_param("reservelow", parse_reservelow);
752 static void __init trim_low_memory_range(void)
754 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
758 * Dump out kernel offset information on panic.
761 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
763 if (kaslr_enabled()) {
764 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
770 pr_emerg("Kernel Offset: disabled\n");
777 * Determine if we were loaded by an EFI loader. If so, then we have also been
778 * passed the efi memmap, systab, etc., so we should use these data structures
779 * for initialization. Note, the efi init code path is determined by the
780 * global efi_enabled. This allows the same kernel image to be used on existing
781 * systems (with a traditional BIOS) as well as on EFI systems.
784 * setup_arch - architecture-specific boot-time initializations
786 * Note: On x86_64, fixmaps are ready for use even before this is called.
789 void __init setup_arch(char **cmdline_p)
792 * Reserve the memory occupied by the kernel between _text and
793 * __end_of_kernel_reserve symbols. Any kernel sections after the
794 * __end_of_kernel_reserve symbol must be explicitly reserved with a
795 * separate memblock_reserve() or they will be discarded.
797 memblock_reserve(__pa_symbol(_text),
798 (unsigned long)__end_of_kernel_reserve - (unsigned long)_text);
801 * Make sure page 0 is always reserved because on systems with
802 * L1TF its contents can be leaked to user processes.
804 memblock_reserve(0, PAGE_SIZE);
806 early_reserve_initrd();
809 * At this point everything still needed from the boot loader
810 * or BIOS or kernel text should be early reserved or marked not
811 * RAM in e820. All other memory is free game.
815 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
818 * copy kernel address range established so far and switch
819 * to the proper swapper page table
821 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
822 initial_page_table + KERNEL_PGD_BOUNDARY,
825 load_cr3(swapper_pg_dir);
827 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
828 * a cr3 based tlb flush, so the following __flush_tlb_all()
829 * will not flush anything because the CPU quirk which clears
830 * X86_FEATURE_PGE has not been invoked yet. Though due to the
831 * load_cr3() above the TLB has been flushed already. The
832 * quirk is invoked before subsequent calls to __flush_tlb_all()
833 * so proper operation is guaranteed.
837 printk(KERN_INFO "Command line: %s\n", boot_command_line);
838 boot_cpu_data.x86_phys_bits = MAX_PHYSMEM_BITS;
842 * If we have OLPC OFW, we might end up relocating the fixmap due to
843 * reserve_top(), so do this before touching the ioremap area.
847 idt_setup_early_traps();
849 arch_init_ideal_nops();
851 early_ioremap_init();
853 setup_olpc_ofw_pgd();
855 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
856 screen_info = boot_params.screen_info;
857 edid_info = boot_params.edid_info;
859 apm_info.bios = boot_params.apm_bios_info;
860 ist_info = boot_params.ist_info;
862 saved_video_mode = boot_params.hdr.vid_mode;
863 bootloader_type = boot_params.hdr.type_of_loader;
864 if ((bootloader_type >> 4) == 0xe) {
865 bootloader_type &= 0xf;
866 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
868 bootloader_version = bootloader_type & 0xf;
869 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
871 #ifdef CONFIG_BLK_DEV_RAM
872 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
875 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
876 EFI32_LOADER_SIGNATURE, 4)) {
877 set_bit(EFI_BOOT, &efi.flags);
878 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
879 EFI64_LOADER_SIGNATURE, 4)) {
880 set_bit(EFI_BOOT, &efi.flags);
881 set_bit(EFI_64BIT, &efi.flags);
885 x86_init.oem.arch_setup();
887 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
888 e820__memory_setup();
893 if (!boot_params.hdr.root_flags)
894 root_mountflags &= ~MS_RDONLY;
895 init_mm.start_code = (unsigned long) _text;
896 init_mm.end_code = (unsigned long) _etext;
897 init_mm.end_data = (unsigned long) _edata;
898 init_mm.brk = _brk_end;
900 code_resource.start = __pa_symbol(_text);
901 code_resource.end = __pa_symbol(_etext)-1;
902 rodata_resource.start = __pa_symbol(__start_rodata);
903 rodata_resource.end = __pa_symbol(__end_rodata)-1;
904 data_resource.start = __pa_symbol(_sdata);
905 data_resource.end = __pa_symbol(_edata)-1;
906 bss_resource.start = __pa_symbol(__bss_start);
907 bss_resource.end = __pa_symbol(__bss_stop)-1;
909 #ifdef CONFIG_CMDLINE_BOOL
910 #ifdef CONFIG_CMDLINE_OVERRIDE
911 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
913 if (builtin_cmdline[0]) {
914 /* append boot loader cmdline to builtin */
915 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
916 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
917 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
922 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
923 *cmdline_p = command_line;
926 * x86_configure_nx() is called before parse_early_param() to detect
927 * whether hardware doesn't support NX (so that the early EHCI debug
928 * console setup can safely call set_fixmap()). It may then be called
929 * again from within noexec_setup() during parsing early parameters
930 * to honor the respective command line option.
936 if (efi_enabled(EFI_BOOT))
937 efi_memblock_x86_reserve_range();
938 #ifdef CONFIG_MEMORY_HOTPLUG
940 * Memory used by the kernel cannot be hot-removed because Linux
941 * cannot migrate the kernel pages. When memory hotplug is
942 * enabled, we should prevent memblock from allocating memory
945 * ACPI SRAT records all hotpluggable memory ranges. But before
946 * SRAT is parsed, we don't know about it.
948 * The kernel image is loaded into memory at very early time. We
949 * cannot prevent this anyway. So on NUMA system, we set any
950 * node the kernel resides in as un-hotpluggable.
952 * Since on modern servers, one node could have double-digit
953 * gigabytes memory, we can assume the memory around the kernel
954 * image is also un-hotpluggable. So before SRAT is parsed, just
955 * allocate memory near the kernel image to try the best to keep
956 * the kernel away from hotpluggable memory.
958 if (movable_node_is_enabled())
959 memblock_set_bottom_up(true);
964 /* after early param, so could get panic from serial */
965 memblock_x86_reserve_range_setup_data();
967 if (acpi_mps_check()) {
968 #ifdef CONFIG_X86_LOCAL_APIC
971 setup_clear_cpu_cap(X86_FEATURE_APIC);
974 e820__reserve_setup_data();
975 e820__finish_early_params();
977 if (efi_enabled(EFI_BOOT))
983 * VMware detection requires dmi to be available, so this
984 * needs to be done after dmi_setup(), for the boot CPU.
986 init_hypervisor_platform();
989 x86_init.resources.probe_roms();
991 /* after parse_early_param, so could debug it */
992 insert_resource(&iomem_resource, &code_resource);
993 insert_resource(&iomem_resource, &rodata_resource);
994 insert_resource(&iomem_resource, &data_resource);
995 insert_resource(&iomem_resource, &bss_resource);
997 e820_add_kernel_range();
1000 if (ppro_with_ram_bug()) {
1001 e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM,
1002 E820_TYPE_RESERVED);
1003 e820__update_table(e820_table);
1004 printk(KERN_INFO "fixed physical RAM map:\n");
1005 e820__print_table("bad_ppro");
1008 early_gart_iommu_check();
1012 * partially used pages are not usable - thus
1013 * we are rounding upwards:
1015 max_pfn = e820__end_of_ram_pfn();
1017 /* update e820 for memory not covered by WB MTRRs */
1019 if (mtrr_trim_uncached_memory(max_pfn))
1020 max_pfn = e820__end_of_ram_pfn();
1022 max_possible_pfn = max_pfn;
1025 * This call is required when the CPU does not support PAT. If
1026 * mtrr_bp_init() invoked it already via pat_init() the call has no
1032 * Define random base addresses for memory sections after max_pfn is
1033 * defined and before each memory section base is used.
1035 kernel_randomize_memory();
1037 #ifdef CONFIG_X86_32
1038 /* max_low_pfn get updated here */
1039 find_low_pfn_range();
1043 /* How many end-of-memory variables you have, grandma! */
1044 /* need this before calling reserve_initrd */
1045 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1046 max_low_pfn = e820__end_of_low_ram_pfn();
1048 max_low_pfn = max_pfn;
1050 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1054 * Find and reserve possible boot-time SMP configuration:
1058 reserve_ibft_region();
1060 early_alloc_pgt_buf();
1063 * Need to conclude brk, before e820__memblock_setup()
1064 * it could use memblock_find_in_range, could overlap with
1071 memblock_set_current_limit(ISA_END_ADDRESS);
1072 e820__memblock_setup();
1074 reserve_bios_regions();
1081 * The EFI specification says that boot service code won't be
1082 * called after ExitBootServices(). This is, in fact, a lie.
1084 efi_reserve_boot_services();
1086 /* preallocate 4k for mptable mpc */
1087 e820__memblock_alloc_reserved_mpc_new();
1089 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1090 setup_bios_corruption_check();
1093 #ifdef CONFIG_X86_32
1094 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1095 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1098 reserve_real_mode();
1100 trim_platform_memory_ranges();
1101 trim_low_memory_range();
1105 idt_setup_early_pf();
1108 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1109 * with the current CR4 value. This may not be necessary, but
1110 * auditing all the early-boot CR4 manipulation would be needed to
1113 * Mask off features that don't work outside long mode (just
1116 mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE;
1118 memblock_set_current_limit(get_max_mapped());
1121 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1124 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1125 if (init_ohci1394_dma_early)
1126 init_ohci1394_dma_on_all_controllers();
1128 /* Allocate bigger log buffer */
1131 if (efi_enabled(EFI_BOOT)) {
1132 switch (boot_params.secure_boot) {
1133 case efi_secureboot_mode_disabled:
1134 pr_info("Secure boot disabled\n");
1136 case efi_secureboot_mode_enabled:
1137 pr_info("Secure boot enabled\n");
1140 pr_info("Secure boot could not be determined\n");
1147 acpi_table_upgrade();
1153 early_platform_quirks();
1156 * Parse the ACPI tables for possible boot-time SMP configuration.
1158 acpi_boot_table_init();
1160 early_acpi_boot_init();
1163 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1165 if (boot_cpu_has(X86_FEATURE_GBPAGES))
1166 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1169 * Reserve memory for crash kernel after SRAT is parsed so that it
1170 * won't consume hotpluggable memory.
1172 reserve_crashkernel();
1174 memblock_find_dma_reserve();
1176 if (!early_xdbc_setup_hardware())
1177 early_xdbc_register_console();
1179 x86_init.paging.pagetable_init();
1184 * Sync back kernel address range.
1186 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1189 sync_initial_page_table();
1195 generic_apic_probe();
1200 * Read APIC and some other early information from ACPI tables.
1207 * get boot-time SMP configuration:
1212 * Systems w/o ACPI and mptables might not have it mapped the local
1213 * APIC yet, but prefill_possible_map() might need to access it.
1215 init_apic_mappings();
1217 prefill_possible_map();
1221 io_apic_init_mappings();
1223 x86_init.hyper.guest_late_init();
1225 e820__reserve_resources();
1226 e820__register_nosave_regions(max_pfn);
1228 x86_init.resources.reserve_resources();
1230 e820__setup_pci_gap();
1233 #if defined(CONFIG_VGA_CONSOLE)
1234 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1235 conswitchp = &vga_con;
1238 x86_init.oem.banner();
1240 x86_init.timers.wallclock_init();
1244 register_refined_jiffies(CLOCK_TICK_RATE);
1247 if (efi_enabled(EFI_BOOT))
1248 efi_apply_memmap_quirks();
1254 #ifdef CONFIG_X86_32
1256 static struct resource video_ram_resource = {
1257 .name = "Video RAM area",
1260 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1263 void __init i386_reserve_resources(void)
1265 request_resource(&iomem_resource, &video_ram_resource);
1266 reserve_standard_io_resources();
1269 #endif /* CONFIG_X86_32 */
1271 static struct notifier_block kernel_offset_notifier = {
1272 .notifier_call = dump_kernel_offset
1275 static int __init register_kernel_offset_dumper(void)
1277 atomic_notifier_chain_register(&panic_notifier_list,
1278 &kernel_offset_notifier);
1281 __initcall(register_kernel_offset_dumper);