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>
29 #include <asm/bios_ebda.h>
34 #include <asm/hypervisor.h>
35 #include <asm/io_apic.h>
36 #include <asm/kasan.h>
37 #include <asm/kaslr.h>
40 #include <asm/realmode.h>
41 #include <asm/olpc_ofw.h>
42 #include <asm/pci-direct.h>
44 #include <asm/proto.h>
45 #include <asm/unwind.h>
46 #include <asm/vsyscall.h>
47 #include <linux/vmalloc.h>
50 * max_low_pfn_mapped: highest directly mapped pfn < 4 GB
51 * max_pfn_mapped: highest directly mapped pfn > 4 GB
53 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
54 * represented by pfn_mapped[].
56 unsigned long max_low_pfn_mapped;
57 unsigned long max_pfn_mapped;
60 RESERVE_BRK(dmi_alloc, 65536);
65 * Range of the BSS area. The size of the BSS area is determined
66 * at link time, with RESERVE_BRK*() facility reserving additional
69 unsigned long _brk_start = (unsigned long)__brk_base;
70 unsigned long _brk_end = (unsigned long)__brk_base;
72 struct boot_params boot_params;
75 * These are the four main kernel memory regions, we put them into
76 * the resource tree so that kdump tools and other debugging tools
80 static struct resource rodata_resource = {
81 .name = "Kernel rodata",
84 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
87 static struct resource data_resource = {
88 .name = "Kernel data",
91 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
94 static struct resource code_resource = {
95 .name = "Kernel code",
98 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
101 static struct resource bss_resource = {
102 .name = "Kernel bss",
105 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
110 /* CPU data as detected by the assembly code in head_32.S */
111 struct cpuinfo_x86 new_cpu_data;
113 /* Common CPU data for all CPUs */
114 struct cpuinfo_x86 boot_cpu_data __read_mostly;
115 EXPORT_SYMBOL(boot_cpu_data);
117 unsigned int def_to_bigsmp;
119 /* For MCA, but anyone else can use it if they want */
120 unsigned int machine_id;
121 unsigned int machine_submodel_id;
122 unsigned int BIOS_revision;
124 struct apm_info apm_info;
125 EXPORT_SYMBOL(apm_info);
127 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
128 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
129 struct ist_info ist_info;
130 EXPORT_SYMBOL(ist_info);
132 struct ist_info ist_info;
136 struct cpuinfo_x86 boot_cpu_data __read_mostly;
137 EXPORT_SYMBOL(boot_cpu_data);
141 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
142 __visible unsigned long mmu_cr4_features __ro_after_init;
144 __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
147 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
148 int bootloader_type, bootloader_version;
153 struct screen_info screen_info;
154 EXPORT_SYMBOL(screen_info);
155 struct edid_info edid_info;
156 EXPORT_SYMBOL_GPL(edid_info);
158 extern int root_mountflags;
160 unsigned long saved_video_mode;
162 #define RAMDISK_IMAGE_START_MASK 0x07FF
163 #define RAMDISK_PROMPT_FLAG 0x8000
164 #define RAMDISK_LOAD_FLAG 0x4000
166 static char __initdata command_line[COMMAND_LINE_SIZE];
167 #ifdef CONFIG_CMDLINE_BOOL
168 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
171 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
173 #ifdef CONFIG_EDD_MODULE
177 * copy_edd() - Copy the BIOS EDD information
178 * from boot_params into a safe place.
181 static inline void __init copy_edd(void)
183 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
184 sizeof(edd.mbr_signature));
185 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
186 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
187 edd.edd_info_nr = boot_params.eddbuf_entries;
190 static inline void __init copy_edd(void)
195 void * __init extend_brk(size_t size, size_t align)
197 size_t mask = align - 1;
200 BUG_ON(_brk_start == 0);
201 BUG_ON(align & mask);
203 _brk_end = (_brk_end + mask) & ~mask;
204 BUG_ON((char *)(_brk_end + size) > __brk_limit);
206 ret = (void *)_brk_end;
209 memset(ret, 0, size);
215 static void __init cleanup_highmap(void)
220 static void __init reserve_brk(void)
222 if (_brk_end > _brk_start)
223 memblock_reserve(__pa_symbol(_brk_start),
224 _brk_end - _brk_start);
226 /* Mark brk area as locked down and no longer taking any
231 u64 relocated_ramdisk;
233 #ifdef CONFIG_BLK_DEV_INITRD
235 static u64 __init get_ramdisk_image(void)
237 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
239 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
241 if (ramdisk_image == 0)
242 ramdisk_image = phys_initrd_start;
244 return ramdisk_image;
246 static u64 __init get_ramdisk_size(void)
248 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
250 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
252 if (ramdisk_size == 0)
253 ramdisk_size = phys_initrd_size;
258 static void __init relocate_initrd(void)
260 /* Assume only end is not page aligned */
261 u64 ramdisk_image = get_ramdisk_image();
262 u64 ramdisk_size = get_ramdisk_size();
263 u64 area_size = PAGE_ALIGN(ramdisk_size);
265 /* We need to move the initrd down into directly mapped mem */
266 relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
267 area_size, PAGE_SIZE);
269 if (!relocated_ramdisk)
270 panic("Cannot find place for new RAMDISK of size %lld\n",
273 /* Note: this includes all the mem currently occupied by
274 the initrd, we rely on that fact to keep the data intact. */
275 memblock_reserve(relocated_ramdisk, area_size);
276 initrd_start = relocated_ramdisk + PAGE_OFFSET;
277 initrd_end = initrd_start + ramdisk_size;
278 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
279 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
281 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
283 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
284 " [mem %#010llx-%#010llx]\n",
285 ramdisk_image, ramdisk_image + ramdisk_size - 1,
286 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
289 static void __init early_reserve_initrd(void)
291 /* Assume only end is not page aligned */
292 u64 ramdisk_image = get_ramdisk_image();
293 u64 ramdisk_size = get_ramdisk_size();
294 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
296 if (!boot_params.hdr.type_of_loader ||
297 !ramdisk_image || !ramdisk_size)
298 return; /* No initrd provided by bootloader */
300 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
302 static void __init reserve_initrd(void)
304 /* Assume only end is not page aligned */
305 u64 ramdisk_image = get_ramdisk_image();
306 u64 ramdisk_size = get_ramdisk_size();
307 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
310 if (!boot_params.hdr.type_of_loader ||
311 !ramdisk_image || !ramdisk_size)
312 return; /* No initrd provided by bootloader */
316 mapped_size = memblock_mem_size(max_pfn_mapped);
317 if (ramdisk_size >= (mapped_size>>1))
318 panic("initrd too large to handle, "
319 "disabling initrd (%lld needed, %lld available)\n",
320 ramdisk_size, mapped_size>>1);
322 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
325 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
326 PFN_DOWN(ramdisk_end))) {
327 /* All are mapped, easy case */
328 initrd_start = ramdisk_image + PAGE_OFFSET;
329 initrd_end = initrd_start + ramdisk_size;
335 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
339 static void __init early_reserve_initrd(void)
342 static void __init reserve_initrd(void)
345 #endif /* CONFIG_BLK_DEV_INITRD */
347 static void __init parse_setup_data(void)
349 struct setup_data *data;
350 u64 pa_data, pa_next;
352 pa_data = boot_params.hdr.setup_data;
354 u32 data_len, data_type;
356 data = early_memremap(pa_data, sizeof(*data));
357 data_len = data->len + sizeof(struct setup_data);
358 data_type = data->type;
359 pa_next = data->next;
360 early_memunmap(data, sizeof(*data));
364 e820__memory_setup_extended(pa_data, data_len);
370 parse_efi_setup(pa_data, data_len);
379 static void __init memblock_x86_reserve_range_setup_data(void)
381 struct setup_data *data;
384 pa_data = boot_params.hdr.setup_data;
386 data = early_memremap(pa_data, sizeof(*data));
387 memblock_reserve(pa_data, sizeof(*data) + data->len);
389 if (data->type == SETUP_INDIRECT &&
390 ((struct setup_indirect *)data->data)->type != SETUP_INDIRECT)
391 memblock_reserve(((struct setup_indirect *)data->data)->addr,
392 ((struct setup_indirect *)data->data)->len);
394 pa_data = data->next;
395 early_memunmap(data, sizeof(*data));
400 * --------- Crashkernel reservation ------------------------------
403 #ifdef CONFIG_KEXEC_CORE
405 /* 16M alignment for crash kernel regions */
406 #define CRASH_ALIGN SZ_16M
409 * Keep the crash kernel below this limit.
411 * Earlier 32-bits kernels would limit the kernel to the low 512 MB range
412 * due to mapping restrictions.
414 * 64-bit kdump kernels need to be restricted to be under 64 TB, which is
415 * the upper limit of system RAM in 4-level paging mode. Since the kdump
416 * jump could be from 5-level paging to 4-level paging, the jump will fail if
417 * the kernel is put above 64 TB, and during the 1st kernel bootup there's
418 * no good way to detect the paging mode of the target kernel which will be
419 * loaded for dumping.
422 # define CRASH_ADDR_LOW_MAX SZ_512M
423 # define CRASH_ADDR_HIGH_MAX SZ_512M
425 # define CRASH_ADDR_LOW_MAX SZ_4G
426 # define CRASH_ADDR_HIGH_MAX SZ_64T
429 static int __init reserve_crashkernel_low(void)
432 unsigned long long base, low_base = 0, low_size = 0;
433 unsigned long total_low_mem;
436 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
438 /* crashkernel=Y,low */
439 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
442 * two parts from kernel/dma/swiotlb.c:
443 * -swiotlb size: user-specified with swiotlb= or default.
445 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
446 * to 8M for other buffers that may need to stay low too. Also
447 * make sure we allocate enough extra low memory so that we
448 * don't run out of DMA buffers for 32-bit devices.
450 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
452 /* passed with crashkernel=0,low ? */
457 low_base = memblock_find_in_range(0, 1ULL << 32, low_size, CRASH_ALIGN);
459 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
460 (unsigned long)(low_size >> 20));
464 ret = memblock_reserve(low_base, low_size);
466 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
470 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
471 (unsigned long)(low_size >> 20),
472 (unsigned long)(low_base >> 20),
473 (unsigned long)(total_low_mem >> 20));
475 crashk_low_res.start = low_base;
476 crashk_low_res.end = low_base + low_size - 1;
477 insert_resource(&iomem_resource, &crashk_low_res);
482 static void __init reserve_crashkernel(void)
484 unsigned long long crash_size, crash_base, total_mem;
488 total_mem = memblock_phys_mem_size();
491 ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
492 if (ret != 0 || crash_size <= 0) {
493 /* crashkernel=X,high */
494 ret = parse_crashkernel_high(boot_command_line, total_mem,
495 &crash_size, &crash_base);
496 if (ret != 0 || crash_size <= 0)
501 if (xen_pv_domain()) {
502 pr_info("Ignoring crashkernel for a Xen PV domain\n");
506 /* 0 means: find the address automatically */
509 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
510 * crashkernel=x,high reserves memory over 4G, also allocates
511 * 256M extra low memory for DMA buffers and swiotlb.
512 * But the extra memory is not required for all machines.
513 * So try low memory first and fall back to high memory
514 * unless "crashkernel=size[KMG],high" is specified.
517 crash_base = memblock_find_in_range(CRASH_ALIGN,
519 crash_size, CRASH_ALIGN);
521 crash_base = memblock_find_in_range(CRASH_ALIGN,
523 crash_size, CRASH_ALIGN);
525 pr_info("crashkernel reservation failed - No suitable area found.\n");
529 unsigned long long start;
531 start = memblock_find_in_range(crash_base,
532 crash_base + crash_size,
533 crash_size, 1 << 20);
534 if (start != crash_base) {
535 pr_info("crashkernel reservation failed - memory is in use.\n");
539 ret = memblock_reserve(crash_base, crash_size);
541 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
545 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
546 memblock_free(crash_base, crash_size);
550 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
551 (unsigned long)(crash_size >> 20),
552 (unsigned long)(crash_base >> 20),
553 (unsigned long)(total_mem >> 20));
555 crashk_res.start = crash_base;
556 crashk_res.end = crash_base + crash_size - 1;
557 insert_resource(&iomem_resource, &crashk_res);
560 static void __init reserve_crashkernel(void)
565 static struct resource standard_io_resources[] = {
566 { .name = "dma1", .start = 0x00, .end = 0x1f,
567 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
568 { .name = "pic1", .start = 0x20, .end = 0x21,
569 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
570 { .name = "timer0", .start = 0x40, .end = 0x43,
571 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
572 { .name = "timer1", .start = 0x50, .end = 0x53,
573 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
574 { .name = "keyboard", .start = 0x60, .end = 0x60,
575 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
576 { .name = "keyboard", .start = 0x64, .end = 0x64,
577 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
578 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
579 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
580 { .name = "pic2", .start = 0xa0, .end = 0xa1,
581 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
582 { .name = "dma2", .start = 0xc0, .end = 0xdf,
583 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
584 { .name = "fpu", .start = 0xf0, .end = 0xff,
585 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
588 void __init reserve_standard_io_resources(void)
592 /* request I/O space for devices used on all i[345]86 PCs */
593 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
594 request_resource(&ioport_resource, &standard_io_resources[i]);
598 static __init void reserve_ibft_region(void)
600 unsigned long addr, size = 0;
602 addr = find_ibft_region(&size);
605 memblock_reserve(addr, size);
608 static bool __init snb_gfx_workaround_needed(void)
613 static const __initconst u16 snb_ids[] = {
623 /* Assume no if something weird is going on with PCI */
624 if (!early_pci_allowed())
627 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
628 if (vendor != 0x8086)
631 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
632 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
633 if (devid == snb_ids[i])
641 * Sandy Bridge graphics has trouble with certain ranges, exclude
642 * them from allocation.
644 static void __init trim_snb_memory(void)
646 static const __initconst unsigned long bad_pages[] = {
655 if (!snb_gfx_workaround_needed())
658 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
661 * Reserve all memory below the 1 MB mark that has not
662 * already been reserved.
664 memblock_reserve(0, 1<<20);
666 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
667 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
668 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
674 * Here we put platform-specific memory range workarounds, i.e.
675 * memory known to be corrupt or otherwise in need to be reserved on
676 * specific platforms.
678 * If this gets used more widely it could use a real dispatch mechanism.
680 static void __init trim_platform_memory_ranges(void)
685 static void __init trim_bios_range(void)
688 * A special case is the first 4Kb of memory;
689 * This is a BIOS owned area, not kernel ram, but generally
690 * not listed as such in the E820 table.
692 * This typically reserves additional memory (64KiB by default)
693 * since some BIOSes are known to corrupt low memory. See the
694 * Kconfig help text for X86_RESERVE_LOW.
696 e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED);
699 * special case: Some BIOSes report the PC BIOS
700 * area (640Kb -> 1Mb) as RAM even though it is not.
703 e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1);
705 e820__update_table(e820_table);
708 /* called before trim_bios_range() to spare extra sanitize */
709 static void __init e820_add_kernel_range(void)
711 u64 start = __pa_symbol(_text);
712 u64 size = __pa_symbol(_end) - start;
715 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
716 * attempt to fix it by adding the range. We may have a confused BIOS,
717 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
718 * exclude kernel range. If we really are running on top non-RAM,
719 * we will crash later anyways.
721 if (e820__mapped_all(start, start + size, E820_TYPE_RAM))
724 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
725 e820__range_remove(start, size, E820_TYPE_RAM, 0);
726 e820__range_add(start, size, E820_TYPE_RAM);
729 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
731 static int __init parse_reservelow(char *p)
733 unsigned long long size;
738 size = memparse(p, &p);
751 early_param("reservelow", parse_reservelow);
753 static void __init trim_low_memory_range(void)
755 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
759 * Dump out kernel offset information on panic.
762 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
764 if (kaslr_enabled()) {
765 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
771 pr_emerg("Kernel Offset: disabled\n");
778 * Determine if we were loaded by an EFI loader. If so, then we have also been
779 * passed the efi memmap, systab, etc., so we should use these data structures
780 * for initialization. Note, the efi init code path is determined by the
781 * global efi_enabled. This allows the same kernel image to be used on existing
782 * systems (with a traditional BIOS) as well as on EFI systems.
785 * setup_arch - architecture-specific boot-time initializations
787 * Note: On x86_64, fixmaps are ready for use even before this is called.
790 void __init setup_arch(char **cmdline_p)
793 * Reserve the memory occupied by the kernel between _text and
794 * __end_of_kernel_reserve symbols. Any kernel sections after the
795 * __end_of_kernel_reserve symbol must be explicitly reserved with a
796 * separate memblock_reserve() or they will be discarded.
798 memblock_reserve(__pa_symbol(_text),
799 (unsigned long)__end_of_kernel_reserve - (unsigned long)_text);
802 * Make sure page 0 is always reserved because on systems with
803 * L1TF its contents can be leaked to user processes.
805 memblock_reserve(0, PAGE_SIZE);
807 early_reserve_initrd();
810 * At this point everything still needed from the boot loader
811 * or BIOS or kernel text should be early reserved or marked not
812 * RAM in e820. All other memory is free game.
816 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
819 * copy kernel address range established so far and switch
820 * to the proper swapper page table
822 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
823 initial_page_table + KERNEL_PGD_BOUNDARY,
826 load_cr3(swapper_pg_dir);
828 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
829 * a cr3 based tlb flush, so the following __flush_tlb_all()
830 * will not flush anything because the CPU quirk which clears
831 * X86_FEATURE_PGE has not been invoked yet. Though due to the
832 * load_cr3() above the TLB has been flushed already. The
833 * quirk is invoked before subsequent calls to __flush_tlb_all()
834 * so proper operation is guaranteed.
838 printk(KERN_INFO "Command line: %s\n", boot_command_line);
839 boot_cpu_data.x86_phys_bits = MAX_PHYSMEM_BITS;
843 * If we have OLPC OFW, we might end up relocating the fixmap due to
844 * reserve_top(), so do this before touching the ioremap area.
848 idt_setup_early_traps();
850 arch_init_ideal_nops();
852 early_ioremap_init();
854 setup_olpc_ofw_pgd();
856 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
857 screen_info = boot_params.screen_info;
858 edid_info = boot_params.edid_info;
860 apm_info.bios = boot_params.apm_bios_info;
861 ist_info = boot_params.ist_info;
863 saved_video_mode = boot_params.hdr.vid_mode;
864 bootloader_type = boot_params.hdr.type_of_loader;
865 if ((bootloader_type >> 4) == 0xe) {
866 bootloader_type &= 0xf;
867 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
869 bootloader_version = bootloader_type & 0xf;
870 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
872 #ifdef CONFIG_BLK_DEV_RAM
873 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
876 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
877 EFI32_LOADER_SIGNATURE, 4)) {
878 set_bit(EFI_BOOT, &efi.flags);
879 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
880 EFI64_LOADER_SIGNATURE, 4)) {
881 set_bit(EFI_BOOT, &efi.flags);
882 set_bit(EFI_64BIT, &efi.flags);
886 x86_init.oem.arch_setup();
888 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
889 e820__memory_setup();
894 if (!boot_params.hdr.root_flags)
895 root_mountflags &= ~MS_RDONLY;
896 init_mm.start_code = (unsigned long) _text;
897 init_mm.end_code = (unsigned long) _etext;
898 init_mm.end_data = (unsigned long) _edata;
899 init_mm.brk = _brk_end;
901 code_resource.start = __pa_symbol(_text);
902 code_resource.end = __pa_symbol(_etext)-1;
903 rodata_resource.start = __pa_symbol(__start_rodata);
904 rodata_resource.end = __pa_symbol(__end_rodata)-1;
905 data_resource.start = __pa_symbol(_sdata);
906 data_resource.end = __pa_symbol(_edata)-1;
907 bss_resource.start = __pa_symbol(__bss_start);
908 bss_resource.end = __pa_symbol(__bss_stop)-1;
910 #ifdef CONFIG_CMDLINE_BOOL
911 #ifdef CONFIG_CMDLINE_OVERRIDE
912 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
914 if (builtin_cmdline[0]) {
915 /* append boot loader cmdline to builtin */
916 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
917 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
918 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
923 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
924 *cmdline_p = command_line;
927 * x86_configure_nx() is called before parse_early_param() to detect
928 * whether hardware doesn't support NX (so that the early EHCI debug
929 * console setup can safely call set_fixmap()). It may then be called
930 * again from within noexec_setup() during parsing early parameters
931 * to honor the respective command line option.
937 if (efi_enabled(EFI_BOOT))
938 efi_memblock_x86_reserve_range();
939 #ifdef CONFIG_MEMORY_HOTPLUG
941 * Memory used by the kernel cannot be hot-removed because Linux
942 * cannot migrate the kernel pages. When memory hotplug is
943 * enabled, we should prevent memblock from allocating memory
946 * ACPI SRAT records all hotpluggable memory ranges. But before
947 * SRAT is parsed, we don't know about it.
949 * The kernel image is loaded into memory at very early time. We
950 * cannot prevent this anyway. So on NUMA system, we set any
951 * node the kernel resides in as un-hotpluggable.
953 * Since on modern servers, one node could have double-digit
954 * gigabytes memory, we can assume the memory around the kernel
955 * image is also un-hotpluggable. So before SRAT is parsed, just
956 * allocate memory near the kernel image to try the best to keep
957 * the kernel away from hotpluggable memory.
959 if (movable_node_is_enabled())
960 memblock_set_bottom_up(true);
965 /* after early param, so could get panic from serial */
966 memblock_x86_reserve_range_setup_data();
968 if (acpi_mps_check()) {
969 #ifdef CONFIG_X86_LOCAL_APIC
972 setup_clear_cpu_cap(X86_FEATURE_APIC);
975 e820__reserve_setup_data();
976 e820__finish_early_params();
978 if (efi_enabled(EFI_BOOT))
984 * VMware detection requires dmi to be available, so this
985 * needs to be done after dmi_setup(), for the boot CPU.
987 init_hypervisor_platform();
990 x86_init.resources.probe_roms();
992 /* after parse_early_param, so could debug it */
993 insert_resource(&iomem_resource, &code_resource);
994 insert_resource(&iomem_resource, &rodata_resource);
995 insert_resource(&iomem_resource, &data_resource);
996 insert_resource(&iomem_resource, &bss_resource);
998 e820_add_kernel_range();
1000 #ifdef CONFIG_X86_32
1001 if (ppro_with_ram_bug()) {
1002 e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM,
1003 E820_TYPE_RESERVED);
1004 e820__update_table(e820_table);
1005 printk(KERN_INFO "fixed physical RAM map:\n");
1006 e820__print_table("bad_ppro");
1009 early_gart_iommu_check();
1013 * partially used pages are not usable - thus
1014 * we are rounding upwards:
1016 max_pfn = e820__end_of_ram_pfn();
1018 /* update e820 for memory not covered by WB MTRRs */
1020 if (mtrr_trim_uncached_memory(max_pfn))
1021 max_pfn = e820__end_of_ram_pfn();
1023 max_possible_pfn = max_pfn;
1026 * This call is required when the CPU does not support PAT. If
1027 * mtrr_bp_init() invoked it already via pat_init() the call has no
1033 * Define random base addresses for memory sections after max_pfn is
1034 * defined and before each memory section base is used.
1036 kernel_randomize_memory();
1038 #ifdef CONFIG_X86_32
1039 /* max_low_pfn get updated here */
1040 find_low_pfn_range();
1044 /* How many end-of-memory variables you have, grandma! */
1045 /* need this before calling reserve_initrd */
1046 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1047 max_low_pfn = e820__end_of_low_ram_pfn();
1049 max_low_pfn = max_pfn;
1051 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1055 * Find and reserve possible boot-time SMP configuration:
1059 reserve_ibft_region();
1061 early_alloc_pgt_buf();
1064 * Need to conclude brk, before e820__memblock_setup()
1065 * it could use memblock_find_in_range, could overlap with
1072 memblock_set_current_limit(ISA_END_ADDRESS);
1073 e820__memblock_setup();
1075 reserve_bios_regions();
1082 * The EFI specification says that boot service code won't be
1083 * called after ExitBootServices(). This is, in fact, a lie.
1085 efi_reserve_boot_services();
1087 /* preallocate 4k for mptable mpc */
1088 e820__memblock_alloc_reserved_mpc_new();
1090 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1091 setup_bios_corruption_check();
1094 #ifdef CONFIG_X86_32
1095 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1096 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1099 reserve_real_mode();
1101 trim_platform_memory_ranges();
1102 trim_low_memory_range();
1106 idt_setup_early_pf();
1109 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1110 * with the current CR4 value. This may not be necessary, but
1111 * auditing all the early-boot CR4 manipulation would be needed to
1114 * Mask off features that don't work outside long mode (just
1117 mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE;
1119 memblock_set_current_limit(get_max_mapped());
1122 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1125 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1126 if (init_ohci1394_dma_early)
1127 init_ohci1394_dma_on_all_controllers();
1129 /* Allocate bigger log buffer */
1132 if (efi_enabled(EFI_BOOT)) {
1133 switch (boot_params.secure_boot) {
1134 case efi_secureboot_mode_disabled:
1135 pr_info("Secure boot disabled\n");
1137 case efi_secureboot_mode_enabled:
1138 pr_info("Secure boot enabled\n");
1141 pr_info("Secure boot could not be determined\n");
1148 acpi_table_upgrade();
1154 early_platform_quirks();
1157 * Parse the ACPI tables for possible boot-time SMP configuration.
1159 acpi_boot_table_init();
1161 early_acpi_boot_init();
1164 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1166 if (boot_cpu_has(X86_FEATURE_GBPAGES))
1167 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1170 * Reserve memory for crash kernel after SRAT is parsed so that it
1171 * won't consume hotpluggable memory.
1173 reserve_crashkernel();
1175 memblock_find_dma_reserve();
1177 if (!early_xdbc_setup_hardware())
1178 early_xdbc_register_console();
1180 x86_init.paging.pagetable_init();
1185 * Sync back kernel address range.
1187 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1190 sync_initial_page_table();
1196 generic_apic_probe();
1201 * Read APIC and some other early information from ACPI tables.
1208 * get boot-time SMP configuration:
1213 * Systems w/o ACPI and mptables might not have it mapped the local
1214 * APIC yet, but prefill_possible_map() might need to access it.
1216 init_apic_mappings();
1218 prefill_possible_map();
1222 io_apic_init_mappings();
1224 x86_init.hyper.guest_late_init();
1226 e820__reserve_resources();
1227 e820__register_nosave_regions(max_pfn);
1229 x86_init.resources.reserve_resources();
1231 e820__setup_pci_gap();
1234 #if defined(CONFIG_VGA_CONSOLE)
1235 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1236 conswitchp = &vga_con;
1239 x86_init.oem.banner();
1241 x86_init.timers.wallclock_init();
1245 register_refined_jiffies(CLOCK_TICK_RATE);
1248 if (efi_enabled(EFI_BOOT))
1249 efi_apply_memmap_quirks();
1255 #ifdef CONFIG_X86_32
1257 static struct resource video_ram_resource = {
1258 .name = "Video RAM area",
1261 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1264 void __init i386_reserve_resources(void)
1266 request_resource(&iomem_resource, &video_ram_resource);
1267 reserve_standard_io_resources();
1270 #endif /* CONFIG_X86_32 */
1272 static struct notifier_block kernel_offset_notifier = {
1273 .notifier_call = dump_kernel_offset
1276 static int __init register_kernel_offset_dumper(void)
1278 atomic_notifier_chain_register(&panic_notifier_list,
1279 &kernel_offset_notifier);
1282 __initcall(register_kernel_offset_dumper);