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>
22 #include <linux/static_call.h>
24 #include <uapi/linux/mount.h>
30 #include <asm/bios_ebda.h>
35 #include <asm/hypervisor.h>
36 #include <asm/io_apic.h>
37 #include <asm/kasan.h>
38 #include <asm/kaslr.h>
41 #include <asm/realmode.h>
42 #include <asm/olpc_ofw.h>
43 #include <asm/pci-direct.h>
45 #include <asm/proto.h>
46 #include <asm/unwind.h>
47 #include <asm/vsyscall.h>
48 #include <linux/vmalloc.h>
51 * max_low_pfn_mapped: highest directly mapped pfn < 4 GB
52 * max_pfn_mapped: highest directly mapped pfn > 4 GB
54 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
55 * represented by pfn_mapped[].
57 unsigned long max_low_pfn_mapped
;
58 unsigned long max_pfn_mapped
;
61 RESERVE_BRK(dmi_alloc
, 65536);
66 * Range of the BSS area. The size of the BSS area is determined
67 * at link time, with RESERVE_BRK*() facility reserving additional
70 unsigned long _brk_start
= (unsigned long)__brk_base
;
71 unsigned long _brk_end
= (unsigned long)__brk_base
;
73 struct boot_params boot_params
;
76 * These are the four main kernel memory regions, we put them into
77 * the resource tree so that kdump tools and other debugging tools
81 static struct resource rodata_resource
= {
82 .name
= "Kernel rodata",
85 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
88 static struct resource data_resource
= {
89 .name
= "Kernel data",
92 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
95 static struct resource code_resource
= {
96 .name
= "Kernel code",
99 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
102 static struct resource bss_resource
= {
103 .name
= "Kernel bss",
106 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
111 /* CPU data as detected by the assembly code in head_32.S */
112 struct cpuinfo_x86 new_cpu_data
;
114 /* Common CPU data for all CPUs */
115 struct cpuinfo_x86 boot_cpu_data __read_mostly
;
116 EXPORT_SYMBOL(boot_cpu_data
);
118 unsigned int def_to_bigsmp
;
120 /* For MCA, but anyone else can use it if they want */
121 unsigned int machine_id
;
122 unsigned int machine_submodel_id
;
123 unsigned int BIOS_revision
;
125 struct apm_info apm_info
;
126 EXPORT_SYMBOL(apm_info
);
128 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
129 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
130 struct ist_info ist_info
;
131 EXPORT_SYMBOL(ist_info
);
133 struct ist_info ist_info
;
137 struct cpuinfo_x86 boot_cpu_data __read_mostly
;
138 EXPORT_SYMBOL(boot_cpu_data
);
142 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
143 __visible
unsigned long mmu_cr4_features __ro_after_init
;
145 __visible
unsigned long mmu_cr4_features __ro_after_init
= X86_CR4_PAE
;
148 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
149 int bootloader_type
, bootloader_version
;
154 struct screen_info screen_info
;
155 EXPORT_SYMBOL(screen_info
);
156 struct edid_info edid_info
;
157 EXPORT_SYMBOL_GPL(edid_info
);
159 extern int root_mountflags
;
161 unsigned long saved_video_mode
;
163 #define RAMDISK_IMAGE_START_MASK 0x07FF
164 #define RAMDISK_PROMPT_FLAG 0x8000
165 #define RAMDISK_LOAD_FLAG 0x4000
167 static char __initdata command_line
[COMMAND_LINE_SIZE
];
168 #ifdef CONFIG_CMDLINE_BOOL
169 static char __initdata builtin_cmdline
[COMMAND_LINE_SIZE
] = CONFIG_CMDLINE
;
172 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
174 #ifdef CONFIG_EDD_MODULE
178 * copy_edd() - Copy the BIOS EDD information
179 * from boot_params into a safe place.
182 static inline void __init
copy_edd(void)
184 memcpy(edd
.mbr_signature
, boot_params
.edd_mbr_sig_buffer
,
185 sizeof(edd
.mbr_signature
));
186 memcpy(edd
.edd_info
, boot_params
.eddbuf
, sizeof(edd
.edd_info
));
187 edd
.mbr_signature_nr
= boot_params
.edd_mbr_sig_buf_entries
;
188 edd
.edd_info_nr
= boot_params
.eddbuf_entries
;
191 static inline void __init
copy_edd(void)
196 void * __init
extend_brk(size_t size
, size_t align
)
198 size_t mask
= align
- 1;
201 BUG_ON(_brk_start
== 0);
202 BUG_ON(align
& mask
);
204 _brk_end
= (_brk_end
+ mask
) & ~mask
;
205 BUG_ON((char *)(_brk_end
+ size
) > __brk_limit
);
207 ret
= (void *)_brk_end
;
210 memset(ret
, 0, size
);
216 static void __init
cleanup_highmap(void)
221 static void __init
reserve_brk(void)
223 if (_brk_end
> _brk_start
)
224 memblock_reserve(__pa_symbol(_brk_start
),
225 _brk_end
- _brk_start
);
227 /* Mark brk area as locked down and no longer taking any
232 u64 relocated_ramdisk
;
234 #ifdef CONFIG_BLK_DEV_INITRD
236 static u64 __init
get_ramdisk_image(void)
238 u64 ramdisk_image
= boot_params
.hdr
.ramdisk_image
;
240 ramdisk_image
|= (u64
)boot_params
.ext_ramdisk_image
<< 32;
242 if (ramdisk_image
== 0)
243 ramdisk_image
= phys_initrd_start
;
245 return ramdisk_image
;
247 static u64 __init
get_ramdisk_size(void)
249 u64 ramdisk_size
= boot_params
.hdr
.ramdisk_size
;
251 ramdisk_size
|= (u64
)boot_params
.ext_ramdisk_size
<< 32;
253 if (ramdisk_size
== 0)
254 ramdisk_size
= phys_initrd_size
;
259 static void __init
relocate_initrd(void)
261 /* Assume only end is not page aligned */
262 u64 ramdisk_image
= get_ramdisk_image();
263 u64 ramdisk_size
= get_ramdisk_size();
264 u64 area_size
= PAGE_ALIGN(ramdisk_size
);
266 /* We need to move the initrd down into directly mapped mem */
267 relocated_ramdisk
= memblock_phys_alloc_range(area_size
, PAGE_SIZE
, 0,
268 PFN_PHYS(max_pfn_mapped
));
269 if (!relocated_ramdisk
)
270 panic("Cannot find place for new RAMDISK of size %lld\n",
273 initrd_start
= relocated_ramdisk
+ PAGE_OFFSET
;
274 initrd_end
= initrd_start
+ ramdisk_size
;
275 printk(KERN_INFO
"Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
276 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
278 copy_from_early_mem((void *)initrd_start
, ramdisk_image
, ramdisk_size
);
280 printk(KERN_INFO
"Move RAMDISK from [mem %#010llx-%#010llx] to"
281 " [mem %#010llx-%#010llx]\n",
282 ramdisk_image
, ramdisk_image
+ ramdisk_size
- 1,
283 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
286 static void __init
early_reserve_initrd(void)
288 /* Assume only end is not page aligned */
289 u64 ramdisk_image
= get_ramdisk_image();
290 u64 ramdisk_size
= get_ramdisk_size();
291 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
293 if (!boot_params
.hdr
.type_of_loader
||
294 !ramdisk_image
|| !ramdisk_size
)
295 return; /* No initrd provided by bootloader */
297 memblock_reserve(ramdisk_image
, ramdisk_end
- ramdisk_image
);
300 static void __init
reserve_initrd(void)
302 /* Assume only end is not page aligned */
303 u64 ramdisk_image
= get_ramdisk_image();
304 u64 ramdisk_size
= get_ramdisk_size();
305 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
307 if (!boot_params
.hdr
.type_of_loader
||
308 !ramdisk_image
|| !ramdisk_size
)
309 return; /* No initrd provided by bootloader */
313 printk(KERN_INFO
"RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image
,
316 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image
),
317 PFN_DOWN(ramdisk_end
))) {
318 /* All are mapped, easy case */
319 initrd_start
= ramdisk_image
+ PAGE_OFFSET
;
320 initrd_end
= initrd_start
+ ramdisk_size
;
326 memblock_free(ramdisk_image
, ramdisk_end
- ramdisk_image
);
330 static void __init
early_reserve_initrd(void)
333 static void __init
reserve_initrd(void)
336 #endif /* CONFIG_BLK_DEV_INITRD */
338 static void __init
parse_setup_data(void)
340 struct setup_data
*data
;
341 u64 pa_data
, pa_next
;
343 pa_data
= boot_params
.hdr
.setup_data
;
345 u32 data_len
, data_type
;
347 data
= early_memremap(pa_data
, sizeof(*data
));
348 data_len
= data
->len
+ sizeof(struct setup_data
);
349 data_type
= data
->type
;
350 pa_next
= data
->next
;
351 early_memunmap(data
, sizeof(*data
));
355 e820__memory_setup_extended(pa_data
, data_len
);
361 parse_efi_setup(pa_data
, data_len
);
370 static void __init
memblock_x86_reserve_range_setup_data(void)
372 struct setup_data
*data
;
375 pa_data
= boot_params
.hdr
.setup_data
;
377 data
= early_memremap(pa_data
, sizeof(*data
));
378 memblock_reserve(pa_data
, sizeof(*data
) + data
->len
);
380 if (data
->type
== SETUP_INDIRECT
&&
381 ((struct setup_indirect
*)data
->data
)->type
!= SETUP_INDIRECT
)
382 memblock_reserve(((struct setup_indirect
*)data
->data
)->addr
,
383 ((struct setup_indirect
*)data
->data
)->len
);
385 pa_data
= data
->next
;
386 early_memunmap(data
, sizeof(*data
));
391 * --------- Crashkernel reservation ------------------------------
394 #ifdef CONFIG_KEXEC_CORE
396 /* 16M alignment for crash kernel regions */
397 #define CRASH_ALIGN SZ_16M
400 * Keep the crash kernel below this limit.
402 * Earlier 32-bits kernels would limit the kernel to the low 512 MB range
403 * due to mapping restrictions.
405 * 64-bit kdump kernels need to be restricted to be under 64 TB, which is
406 * the upper limit of system RAM in 4-level paging mode. Since the kdump
407 * jump could be from 5-level paging to 4-level paging, the jump will fail if
408 * the kernel is put above 64 TB, and during the 1st kernel bootup there's
409 * no good way to detect the paging mode of the target kernel which will be
410 * loaded for dumping.
413 # define CRASH_ADDR_LOW_MAX SZ_512M
414 # define CRASH_ADDR_HIGH_MAX SZ_512M
416 # define CRASH_ADDR_LOW_MAX SZ_4G
417 # define CRASH_ADDR_HIGH_MAX SZ_64T
420 static int __init
reserve_crashkernel_low(void)
423 unsigned long long base
, low_base
= 0, low_size
= 0;
424 unsigned long total_low_mem
;
427 total_low_mem
= memblock_mem_size(1UL << (32 - PAGE_SHIFT
));
429 /* crashkernel=Y,low */
430 ret
= parse_crashkernel_low(boot_command_line
, total_low_mem
, &low_size
, &base
);
433 * two parts from kernel/dma/swiotlb.c:
434 * -swiotlb size: user-specified with swiotlb= or default.
436 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
437 * to 8M for other buffers that may need to stay low too. Also
438 * make sure we allocate enough extra low memory so that we
439 * don't run out of DMA buffers for 32-bit devices.
441 low_size
= max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
443 /* passed with crashkernel=0,low ? */
448 low_base
= memblock_find_in_range(0, 1ULL << 32, low_size
, CRASH_ALIGN
);
450 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
451 (unsigned long)(low_size
>> 20));
455 ret
= memblock_reserve(low_base
, low_size
);
457 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__
);
461 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
462 (unsigned long)(low_size
>> 20),
463 (unsigned long)(low_base
>> 20),
464 (unsigned long)(total_low_mem
>> 20));
466 crashk_low_res
.start
= low_base
;
467 crashk_low_res
.end
= low_base
+ low_size
- 1;
468 insert_resource(&iomem_resource
, &crashk_low_res
);
473 static void __init
reserve_crashkernel(void)
475 unsigned long long crash_size
, crash_base
, total_mem
;
479 total_mem
= memblock_phys_mem_size();
482 ret
= parse_crashkernel(boot_command_line
, total_mem
, &crash_size
, &crash_base
);
483 if (ret
!= 0 || crash_size
<= 0) {
484 /* crashkernel=X,high */
485 ret
= parse_crashkernel_high(boot_command_line
, total_mem
,
486 &crash_size
, &crash_base
);
487 if (ret
!= 0 || crash_size
<= 0)
492 if (xen_pv_domain()) {
493 pr_info("Ignoring crashkernel for a Xen PV domain\n");
497 /* 0 means: find the address automatically */
500 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
501 * crashkernel=x,high reserves memory over 4G, also allocates
502 * 256M extra low memory for DMA buffers and swiotlb.
503 * But the extra memory is not required for all machines.
504 * So try low memory first and fall back to high memory
505 * unless "crashkernel=size[KMG],high" is specified.
508 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
510 crash_size
, CRASH_ALIGN
);
512 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
514 crash_size
, CRASH_ALIGN
);
516 pr_info("crashkernel reservation failed - No suitable area found.\n");
520 unsigned long long start
;
522 start
= memblock_find_in_range(crash_base
,
523 crash_base
+ crash_size
,
524 crash_size
, 1 << 20);
525 if (start
!= crash_base
) {
526 pr_info("crashkernel reservation failed - memory is in use.\n");
530 ret
= memblock_reserve(crash_base
, crash_size
);
532 pr_err("%s: Error reserving crashkernel memblock.\n", __func__
);
536 if (crash_base
>= (1ULL << 32) && reserve_crashkernel_low()) {
537 memblock_free(crash_base
, crash_size
);
541 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
542 (unsigned long)(crash_size
>> 20),
543 (unsigned long)(crash_base
>> 20),
544 (unsigned long)(total_mem
>> 20));
546 crashk_res
.start
= crash_base
;
547 crashk_res
.end
= crash_base
+ crash_size
- 1;
548 insert_resource(&iomem_resource
, &crashk_res
);
551 static void __init
reserve_crashkernel(void)
556 static struct resource standard_io_resources
[] = {
557 { .name
= "dma1", .start
= 0x00, .end
= 0x1f,
558 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
559 { .name
= "pic1", .start
= 0x20, .end
= 0x21,
560 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
561 { .name
= "timer0", .start
= 0x40, .end
= 0x43,
562 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
563 { .name
= "timer1", .start
= 0x50, .end
= 0x53,
564 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
565 { .name
= "keyboard", .start
= 0x60, .end
= 0x60,
566 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
567 { .name
= "keyboard", .start
= 0x64, .end
= 0x64,
568 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
569 { .name
= "dma page reg", .start
= 0x80, .end
= 0x8f,
570 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
571 { .name
= "pic2", .start
= 0xa0, .end
= 0xa1,
572 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
573 { .name
= "dma2", .start
= 0xc0, .end
= 0xdf,
574 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
575 { .name
= "fpu", .start
= 0xf0, .end
= 0xff,
576 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
}
579 void __init
reserve_standard_io_resources(void)
583 /* request I/O space for devices used on all i[345]86 PCs */
584 for (i
= 0; i
< ARRAY_SIZE(standard_io_resources
); i
++)
585 request_resource(&ioport_resource
, &standard_io_resources
[i
]);
589 static __init
void reserve_ibft_region(void)
591 unsigned long addr
, size
= 0;
593 addr
= find_ibft_region(&size
);
596 memblock_reserve(addr
, size
);
599 static bool __init
snb_gfx_workaround_needed(void)
604 static const __initconst u16 snb_ids
[] = {
614 /* Assume no if something weird is going on with PCI */
615 if (!early_pci_allowed())
618 vendor
= read_pci_config_16(0, 2, 0, PCI_VENDOR_ID
);
619 if (vendor
!= 0x8086)
622 devid
= read_pci_config_16(0, 2, 0, PCI_DEVICE_ID
);
623 for (i
= 0; i
< ARRAY_SIZE(snb_ids
); i
++)
624 if (devid
== snb_ids
[i
])
632 * Sandy Bridge graphics has trouble with certain ranges, exclude
633 * them from allocation.
635 static void __init
trim_snb_memory(void)
637 static const __initconst
unsigned long bad_pages
[] = {
646 if (!snb_gfx_workaround_needed())
649 printk(KERN_DEBUG
"reserving inaccessible SNB gfx pages\n");
652 * Reserve all memory below the 1 MB mark that has not
653 * already been reserved.
655 memblock_reserve(0, 1<<20);
657 for (i
= 0; i
< ARRAY_SIZE(bad_pages
); i
++) {
658 if (memblock_reserve(bad_pages
[i
], PAGE_SIZE
))
659 printk(KERN_WARNING
"failed to reserve 0x%08lx\n",
665 * Here we put platform-specific memory range workarounds, i.e.
666 * memory known to be corrupt or otherwise in need to be reserved on
667 * specific platforms.
669 * If this gets used more widely it could use a real dispatch mechanism.
671 static void __init
trim_platform_memory_ranges(void)
676 static void __init
trim_bios_range(void)
679 * A special case is the first 4Kb of memory;
680 * This is a BIOS owned area, not kernel ram, but generally
681 * not listed as such in the E820 table.
683 * This typically reserves additional memory (64KiB by default)
684 * since some BIOSes are known to corrupt low memory. See the
685 * Kconfig help text for X86_RESERVE_LOW.
687 e820__range_update(0, PAGE_SIZE
, E820_TYPE_RAM
, E820_TYPE_RESERVED
);
690 * special case: Some BIOSes report the PC BIOS
691 * area (640Kb -> 1Mb) as RAM even though it is not.
694 e820__range_remove(BIOS_BEGIN
, BIOS_END
- BIOS_BEGIN
, E820_TYPE_RAM
, 1);
696 e820__update_table(e820_table
);
699 /* called before trim_bios_range() to spare extra sanitize */
700 static void __init
e820_add_kernel_range(void)
702 u64 start
= __pa_symbol(_text
);
703 u64 size
= __pa_symbol(_end
) - start
;
706 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
707 * attempt to fix it by adding the range. We may have a confused BIOS,
708 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
709 * exclude kernel range. If we really are running on top non-RAM,
710 * we will crash later anyways.
712 if (e820__mapped_all(start
, start
+ size
, E820_TYPE_RAM
))
715 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
716 e820__range_remove(start
, size
, E820_TYPE_RAM
, 0);
717 e820__range_add(start
, size
, E820_TYPE_RAM
);
720 static unsigned reserve_low
= CONFIG_X86_RESERVE_LOW
<< 10;
722 static int __init
parse_reservelow(char *p
)
724 unsigned long long size
;
729 size
= memparse(p
, &p
);
742 early_param("reservelow", parse_reservelow
);
744 static void __init
trim_low_memory_range(void)
746 memblock_reserve(0, ALIGN(reserve_low
, PAGE_SIZE
));
750 * Dump out kernel offset information on panic.
753 dump_kernel_offset(struct notifier_block
*self
, unsigned long v
, void *p
)
755 if (kaslr_enabled()) {
756 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
762 pr_emerg("Kernel Offset: disabled\n");
769 * Determine if we were loaded by an EFI loader. If so, then we have also been
770 * passed the efi memmap, systab, etc., so we should use these data structures
771 * for initialization. Note, the efi init code path is determined by the
772 * global efi_enabled. This allows the same kernel image to be used on existing
773 * systems (with a traditional BIOS) as well as on EFI systems.
776 * setup_arch - architecture-specific boot-time initializations
778 * Note: On x86_64, fixmaps are ready for use even before this is called.
781 void __init
setup_arch(char **cmdline_p
)
784 * Reserve the memory occupied by the kernel between _text and
785 * __end_of_kernel_reserve symbols. Any kernel sections after the
786 * __end_of_kernel_reserve symbol must be explicitly reserved with a
787 * separate memblock_reserve() or they will be discarded.
789 memblock_reserve(__pa_symbol(_text
),
790 (unsigned long)__end_of_kernel_reserve
- (unsigned long)_text
);
793 * Make sure page 0 is always reserved because on systems with
794 * L1TF its contents can be leaked to user processes.
796 memblock_reserve(0, PAGE_SIZE
);
798 early_reserve_initrd();
801 * At this point everything still needed from the boot loader
802 * or BIOS or kernel text should be early reserved or marked not
803 * RAM in e820. All other memory is free game.
807 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
810 * copy kernel address range established so far and switch
811 * to the proper swapper page table
813 clone_pgd_range(swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
814 initial_page_table
+ KERNEL_PGD_BOUNDARY
,
817 load_cr3(swapper_pg_dir
);
819 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
820 * a cr3 based tlb flush, so the following __flush_tlb_all()
821 * will not flush anything because the CPU quirk which clears
822 * X86_FEATURE_PGE has not been invoked yet. Though due to the
823 * load_cr3() above the TLB has been flushed already. The
824 * quirk is invoked before subsequent calls to __flush_tlb_all()
825 * so proper operation is guaranteed.
829 printk(KERN_INFO
"Command line: %s\n", boot_command_line
);
830 boot_cpu_data
.x86_phys_bits
= MAX_PHYSMEM_BITS
;
834 * If we have OLPC OFW, we might end up relocating the fixmap due to
835 * reserve_top(), so do this before touching the ioremap area.
839 idt_setup_early_traps();
841 arch_init_ideal_nops();
844 early_ioremap_init();
846 setup_olpc_ofw_pgd();
848 ROOT_DEV
= old_decode_dev(boot_params
.hdr
.root_dev
);
849 screen_info
= boot_params
.screen_info
;
850 edid_info
= boot_params
.edid_info
;
852 apm_info
.bios
= boot_params
.apm_bios_info
;
853 ist_info
= boot_params
.ist_info
;
855 saved_video_mode
= boot_params
.hdr
.vid_mode
;
856 bootloader_type
= boot_params
.hdr
.type_of_loader
;
857 if ((bootloader_type
>> 4) == 0xe) {
858 bootloader_type
&= 0xf;
859 bootloader_type
|= (boot_params
.hdr
.ext_loader_type
+0x10) << 4;
861 bootloader_version
= bootloader_type
& 0xf;
862 bootloader_version
|= boot_params
.hdr
.ext_loader_ver
<< 4;
864 #ifdef CONFIG_BLK_DEV_RAM
865 rd_image_start
= boot_params
.hdr
.ram_size
& RAMDISK_IMAGE_START_MASK
;
868 if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
869 EFI32_LOADER_SIGNATURE
, 4)) {
870 set_bit(EFI_BOOT
, &efi
.flags
);
871 } else if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
872 EFI64_LOADER_SIGNATURE
, 4)) {
873 set_bit(EFI_BOOT
, &efi
.flags
);
874 set_bit(EFI_64BIT
, &efi
.flags
);
878 x86_init
.oem
.arch_setup();
880 iomem_resource
.end
= (1ULL << boot_cpu_data
.x86_phys_bits
) - 1;
881 e820__memory_setup();
886 if (!boot_params
.hdr
.root_flags
)
887 root_mountflags
&= ~MS_RDONLY
;
888 init_mm
.start_code
= (unsigned long) _text
;
889 init_mm
.end_code
= (unsigned long) _etext
;
890 init_mm
.end_data
= (unsigned long) _edata
;
891 init_mm
.brk
= _brk_end
;
893 code_resource
.start
= __pa_symbol(_text
);
894 code_resource
.end
= __pa_symbol(_etext
)-1;
895 rodata_resource
.start
= __pa_symbol(__start_rodata
);
896 rodata_resource
.end
= __pa_symbol(__end_rodata
)-1;
897 data_resource
.start
= __pa_symbol(_sdata
);
898 data_resource
.end
= __pa_symbol(_edata
)-1;
899 bss_resource
.start
= __pa_symbol(__bss_start
);
900 bss_resource
.end
= __pa_symbol(__bss_stop
)-1;
902 #ifdef CONFIG_CMDLINE_BOOL
903 #ifdef CONFIG_CMDLINE_OVERRIDE
904 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
906 if (builtin_cmdline
[0]) {
907 /* append boot loader cmdline to builtin */
908 strlcat(builtin_cmdline
, " ", COMMAND_LINE_SIZE
);
909 strlcat(builtin_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
910 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
915 strlcpy(command_line
, boot_command_line
, COMMAND_LINE_SIZE
);
916 *cmdline_p
= command_line
;
919 * x86_configure_nx() is called before parse_early_param() to detect
920 * whether hardware doesn't support NX (so that the early EHCI debug
921 * console setup can safely call set_fixmap()). It may then be called
922 * again from within noexec_setup() during parsing early parameters
923 * to honor the respective command line option.
929 if (efi_enabled(EFI_BOOT
))
930 efi_memblock_x86_reserve_range();
931 #ifdef CONFIG_MEMORY_HOTPLUG
933 * Memory used by the kernel cannot be hot-removed because Linux
934 * cannot migrate the kernel pages. When memory hotplug is
935 * enabled, we should prevent memblock from allocating memory
938 * ACPI SRAT records all hotpluggable memory ranges. But before
939 * SRAT is parsed, we don't know about it.
941 * The kernel image is loaded into memory at very early time. We
942 * cannot prevent this anyway. So on NUMA system, we set any
943 * node the kernel resides in as un-hotpluggable.
945 * Since on modern servers, one node could have double-digit
946 * gigabytes memory, we can assume the memory around the kernel
947 * image is also un-hotpluggable. So before SRAT is parsed, just
948 * allocate memory near the kernel image to try the best to keep
949 * the kernel away from hotpluggable memory.
951 if (movable_node_is_enabled())
952 memblock_set_bottom_up(true);
957 /* after early param, so could get panic from serial */
958 memblock_x86_reserve_range_setup_data();
960 if (acpi_mps_check()) {
961 #ifdef CONFIG_X86_LOCAL_APIC
964 setup_clear_cpu_cap(X86_FEATURE_APIC
);
967 e820__reserve_setup_data();
968 e820__finish_early_params();
970 if (efi_enabled(EFI_BOOT
))
976 * VMware detection requires dmi to be available, so this
977 * needs to be done after dmi_setup(), for the boot CPU.
979 init_hypervisor_platform();
982 x86_init
.resources
.probe_roms();
984 /* after parse_early_param, so could debug it */
985 insert_resource(&iomem_resource
, &code_resource
);
986 insert_resource(&iomem_resource
, &rodata_resource
);
987 insert_resource(&iomem_resource
, &data_resource
);
988 insert_resource(&iomem_resource
, &bss_resource
);
990 e820_add_kernel_range();
993 if (ppro_with_ram_bug()) {
994 e820__range_update(0x70000000ULL
, 0x40000ULL
, E820_TYPE_RAM
,
996 e820__update_table(e820_table
);
997 printk(KERN_INFO
"fixed physical RAM map:\n");
998 e820__print_table("bad_ppro");
1001 early_gart_iommu_check();
1005 * partially used pages are not usable - thus
1006 * we are rounding upwards:
1008 max_pfn
= e820__end_of_ram_pfn();
1010 /* update e820 for memory not covered by WB MTRRs */
1012 if (mtrr_trim_uncached_memory(max_pfn
))
1013 max_pfn
= e820__end_of_ram_pfn();
1015 max_possible_pfn
= max_pfn
;
1018 * This call is required when the CPU does not support PAT. If
1019 * mtrr_bp_init() invoked it already via pat_init() the call has no
1025 * Define random base addresses for memory sections after max_pfn is
1026 * defined and before each memory section base is used.
1028 kernel_randomize_memory();
1030 #ifdef CONFIG_X86_32
1031 /* max_low_pfn get updated here */
1032 find_low_pfn_range();
1036 /* How many end-of-memory variables you have, grandma! */
1037 /* need this before calling reserve_initrd */
1038 if (max_pfn
> (1UL<<(32 - PAGE_SHIFT
)))
1039 max_low_pfn
= e820__end_of_low_ram_pfn();
1041 max_low_pfn
= max_pfn
;
1043 high_memory
= (void *)__va(max_pfn
* PAGE_SIZE
- 1) + 1;
1047 * Find and reserve possible boot-time SMP configuration:
1051 reserve_ibft_region();
1053 early_alloc_pgt_buf();
1056 * Need to conclude brk, before e820__memblock_setup()
1057 * it could use memblock_find_in_range, could overlap with
1064 memblock_set_current_limit(ISA_END_ADDRESS
);
1065 e820__memblock_setup();
1067 reserve_bios_regions();
1072 efi_mokvar_table_init();
1075 * The EFI specification says that boot service code won't be
1076 * called after ExitBootServices(). This is, in fact, a lie.
1078 efi_reserve_boot_services();
1080 /* preallocate 4k for mptable mpc */
1081 e820__memblock_alloc_reserved_mpc_new();
1083 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1084 setup_bios_corruption_check();
1087 #ifdef CONFIG_X86_32
1088 printk(KERN_DEBUG
"initial memory mapped: [mem 0x00000000-%#010lx]\n",
1089 (max_pfn_mapped
<<PAGE_SHIFT
) - 1);
1092 reserve_real_mode();
1094 trim_platform_memory_ranges();
1095 trim_low_memory_range();
1099 idt_setup_early_pf();
1102 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1103 * with the current CR4 value. This may not be necessary, but
1104 * auditing all the early-boot CR4 manipulation would be needed to
1107 * Mask off features that don't work outside long mode (just
1110 mmu_cr4_features
= __read_cr4() & ~X86_CR4_PCIDE
;
1112 memblock_set_current_limit(get_max_mapped());
1115 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1118 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1119 if (init_ohci1394_dma_early
)
1120 init_ohci1394_dma_on_all_controllers();
1122 /* Allocate bigger log buffer */
1125 if (efi_enabled(EFI_BOOT
)) {
1126 switch (boot_params
.secure_boot
) {
1127 case efi_secureboot_mode_disabled
:
1128 pr_info("Secure boot disabled\n");
1130 case efi_secureboot_mode_enabled
:
1131 pr_info("Secure boot enabled\n");
1134 pr_info("Secure boot could not be determined\n");
1141 acpi_table_upgrade();
1147 early_platform_quirks();
1150 * Parse the ACPI tables for possible boot-time SMP configuration.
1152 acpi_boot_table_init();
1154 early_acpi_boot_init();
1157 dma_contiguous_reserve(max_pfn_mapped
<< PAGE_SHIFT
);
1159 if (boot_cpu_has(X86_FEATURE_GBPAGES
))
1160 hugetlb_cma_reserve(PUD_SHIFT
- PAGE_SHIFT
);
1163 * Reserve memory for crash kernel after SRAT is parsed so that it
1164 * won't consume hotpluggable memory.
1166 reserve_crashkernel();
1168 memblock_find_dma_reserve();
1170 if (!early_xdbc_setup_hardware())
1171 early_xdbc_register_console();
1173 x86_init
.paging
.pagetable_init();
1178 * Sync back kernel address range.
1180 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1183 sync_initial_page_table();
1189 generic_apic_probe();
1194 * Read APIC and some other early information from ACPI tables.
1201 * get boot-time SMP configuration:
1206 * Systems w/o ACPI and mptables might not have it mapped the local
1207 * APIC yet, but prefill_possible_map() might need to access it.
1209 init_apic_mappings();
1211 prefill_possible_map();
1215 io_apic_init_mappings();
1217 x86_init
.hyper
.guest_late_init();
1219 e820__reserve_resources();
1220 e820__register_nosave_regions(max_pfn
);
1222 x86_init
.resources
.reserve_resources();
1224 e820__setup_pci_gap();
1227 #if defined(CONFIG_VGA_CONSOLE)
1228 if (!efi_enabled(EFI_BOOT
) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY
))
1229 conswitchp
= &vga_con
;
1232 x86_init
.oem
.banner();
1234 x86_init
.timers
.wallclock_init();
1238 register_refined_jiffies(CLOCK_TICK_RATE
);
1241 if (efi_enabled(EFI_BOOT
))
1242 efi_apply_memmap_quirks();
1248 #ifdef CONFIG_X86_32
1250 static struct resource video_ram_resource
= {
1251 .name
= "Video RAM area",
1254 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
1257 void __init
i386_reserve_resources(void)
1259 request_resource(&iomem_resource
, &video_ram_resource
);
1260 reserve_standard_io_resources();
1263 #endif /* CONFIG_X86_32 */
1265 static struct notifier_block kernel_offset_notifier
= {
1266 .notifier_call
= dump_kernel_offset
1269 static int __init
register_kernel_offset_dumper(void)
1271 atomic_notifier_chain_register(&panic_notifier_list
,
1272 &kernel_offset_notifier
);
1275 __initcall(register_kernel_offset_dumper
);