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/tboot.h>
20 #include <linux/usb/xhci-dbgp.h>
22 #include <uapi/linux/mount.h>
27 #include <asm/bios_ebda.h>
32 #include <asm/hypervisor.h>
33 #include <asm/io_apic.h>
34 #include <asm/kasan.h>
35 #include <asm/kaslr.h>
38 #include <asm/olpc_ofw.h>
39 #include <asm/pci-direct.h>
41 #include <asm/proto.h>
42 #include <asm/unwind.h>
43 #include <asm/vsyscall.h>
44 #include <linux/vmalloc.h>
47 * max_low_pfn_mapped: highest directly mapped pfn < 4 GB
48 * max_pfn_mapped: highest directly mapped pfn > 4 GB
50 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
51 * represented by pfn_mapped[].
53 unsigned long max_low_pfn_mapped
;
54 unsigned long max_pfn_mapped
;
57 RESERVE_BRK(dmi_alloc
, 65536);
62 * Range of the BSS area. The size of the BSS area is determined
63 * at link time, with RESERVE_BRK*() facility reserving additional
67 unsigned long _brk_start
= (unsigned long)__brk_base
;
68 unsigned long _brk_end
= (unsigned long)__brk_base
;
70 struct boot_params boot_params
;
73 * These are the four main kernel memory regions, we put them into
74 * the resource tree so that kdump tools and other debugging tools
78 static struct resource rodata_resource
= {
79 .name
= "Kernel rodata",
82 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
85 static struct resource data_resource
= {
86 .name
= "Kernel data",
89 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
92 static struct resource code_resource
= {
93 .name
= "Kernel code",
96 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
99 static struct resource bss_resource
= {
100 .name
= "Kernel bss",
103 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
108 /* CPU data as detected by the assembly code in head_32.S */
109 struct cpuinfo_x86 new_cpu_data
;
111 /* Common CPU data for all CPUs */
112 struct cpuinfo_x86 boot_cpu_data __read_mostly
;
113 EXPORT_SYMBOL(boot_cpu_data
);
115 unsigned int def_to_bigsmp
;
117 /* For MCA, but anyone else can use it if they want */
118 unsigned int machine_id
;
119 unsigned int machine_submodel_id
;
120 unsigned int BIOS_revision
;
122 struct apm_info apm_info
;
123 EXPORT_SYMBOL(apm_info
);
125 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
126 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
127 struct ist_info ist_info
;
128 EXPORT_SYMBOL(ist_info
);
130 struct ist_info ist_info
;
134 struct cpuinfo_x86 boot_cpu_data __read_mostly
;
135 EXPORT_SYMBOL(boot_cpu_data
);
139 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
140 __visible
unsigned long mmu_cr4_features __ro_after_init
;
142 __visible
unsigned long mmu_cr4_features __ro_after_init
= X86_CR4_PAE
;
145 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
146 int bootloader_type
, bootloader_version
;
151 struct screen_info screen_info
;
152 EXPORT_SYMBOL(screen_info
);
153 struct edid_info edid_info
;
154 EXPORT_SYMBOL_GPL(edid_info
);
156 extern int root_mountflags
;
158 unsigned long saved_video_mode
;
160 #define RAMDISK_IMAGE_START_MASK 0x07FF
161 #define RAMDISK_PROMPT_FLAG 0x8000
162 #define RAMDISK_LOAD_FLAG 0x4000
164 static char __initdata command_line
[COMMAND_LINE_SIZE
];
165 #ifdef CONFIG_CMDLINE_BOOL
166 static char __initdata builtin_cmdline
[COMMAND_LINE_SIZE
] = CONFIG_CMDLINE
;
169 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
171 #ifdef CONFIG_EDD_MODULE
175 * copy_edd() - Copy the BIOS EDD information
176 * from boot_params into a safe place.
179 static inline void __init
copy_edd(void)
181 memcpy(edd
.mbr_signature
, boot_params
.edd_mbr_sig_buffer
,
182 sizeof(edd
.mbr_signature
));
183 memcpy(edd
.edd_info
, boot_params
.eddbuf
, sizeof(edd
.edd_info
));
184 edd
.mbr_signature_nr
= boot_params
.edd_mbr_sig_buf_entries
;
185 edd
.edd_info_nr
= boot_params
.eddbuf_entries
;
188 static inline void __init
copy_edd(void)
193 void * __init
extend_brk(size_t size
, size_t align
)
195 size_t mask
= align
- 1;
198 BUG_ON(_brk_start
== 0);
199 BUG_ON(align
& mask
);
201 _brk_end
= (_brk_end
+ mask
) & ~mask
;
202 BUG_ON((char *)(_brk_end
+ size
) > __brk_limit
);
204 ret
= (void *)_brk_end
;
207 memset(ret
, 0, size
);
213 static void __init
cleanup_highmap(void)
218 static void __init
reserve_brk(void)
220 if (_brk_end
> _brk_start
)
221 memblock_reserve(__pa_symbol(_brk_start
),
222 _brk_end
- _brk_start
);
224 /* Mark brk area as locked down and no longer taking any
229 u64 relocated_ramdisk
;
231 #ifdef CONFIG_BLK_DEV_INITRD
233 static u64 __init
get_ramdisk_image(void)
235 u64 ramdisk_image
= boot_params
.hdr
.ramdisk_image
;
237 ramdisk_image
|= (u64
)boot_params
.ext_ramdisk_image
<< 32;
239 return ramdisk_image
;
241 static u64 __init
get_ramdisk_size(void)
243 u64 ramdisk_size
= boot_params
.hdr
.ramdisk_size
;
245 ramdisk_size
|= (u64
)boot_params
.ext_ramdisk_size
<< 32;
250 static void __init
relocate_initrd(void)
252 /* Assume only end is not page aligned */
253 u64 ramdisk_image
= get_ramdisk_image();
254 u64 ramdisk_size
= get_ramdisk_size();
255 u64 area_size
= PAGE_ALIGN(ramdisk_size
);
257 /* We need to move the initrd down into directly mapped mem */
258 relocated_ramdisk
= memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped
),
259 area_size
, PAGE_SIZE
);
261 if (!relocated_ramdisk
)
262 panic("Cannot find place for new RAMDISK of size %lld\n",
265 /* Note: this includes all the mem currently occupied by
266 the initrd, we rely on that fact to keep the data intact. */
267 memblock_reserve(relocated_ramdisk
, area_size
);
268 initrd_start
= relocated_ramdisk
+ PAGE_OFFSET
;
269 initrd_end
= initrd_start
+ ramdisk_size
;
270 printk(KERN_INFO
"Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
271 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
273 copy_from_early_mem((void *)initrd_start
, ramdisk_image
, ramdisk_size
);
275 printk(KERN_INFO
"Move RAMDISK from [mem %#010llx-%#010llx] to"
276 " [mem %#010llx-%#010llx]\n",
277 ramdisk_image
, ramdisk_image
+ ramdisk_size
- 1,
278 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
281 static void __init
early_reserve_initrd(void)
283 /* Assume only end is not page aligned */
284 u64 ramdisk_image
= get_ramdisk_image();
285 u64 ramdisk_size
= get_ramdisk_size();
286 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
288 if (!boot_params
.hdr
.type_of_loader
||
289 !ramdisk_image
|| !ramdisk_size
)
290 return; /* No initrd provided by bootloader */
292 memblock_reserve(ramdisk_image
, ramdisk_end
- ramdisk_image
);
294 static void __init
reserve_initrd(void)
296 /* Assume only end is not page aligned */
297 u64 ramdisk_image
= get_ramdisk_image();
298 u64 ramdisk_size
= get_ramdisk_size();
299 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
302 if (!boot_params
.hdr
.type_of_loader
||
303 !ramdisk_image
|| !ramdisk_size
)
304 return; /* No initrd provided by bootloader */
308 mapped_size
= memblock_mem_size(max_pfn_mapped
);
309 if (ramdisk_size
>= (mapped_size
>>1))
310 panic("initrd too large to handle, "
311 "disabling initrd (%lld needed, %lld available)\n",
312 ramdisk_size
, mapped_size
>>1);
314 printk(KERN_INFO
"RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image
,
317 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image
),
318 PFN_DOWN(ramdisk_end
))) {
319 /* All are mapped, easy case */
320 initrd_start
= ramdisk_image
+ PAGE_OFFSET
;
321 initrd_end
= initrd_start
+ ramdisk_size
;
327 memblock_free(ramdisk_image
, ramdisk_end
- ramdisk_image
);
331 static void __init
early_reserve_initrd(void)
334 static void __init
reserve_initrd(void)
337 #endif /* CONFIG_BLK_DEV_INITRD */
339 static void __init
parse_setup_data(void)
341 struct setup_data
*data
;
342 u64 pa_data
, pa_next
;
344 pa_data
= boot_params
.hdr
.setup_data
;
346 u32 data_len
, data_type
;
348 data
= early_memremap(pa_data
, sizeof(*data
));
349 data_len
= data
->len
+ sizeof(struct setup_data
);
350 data_type
= data
->type
;
351 pa_next
= data
->next
;
352 early_memunmap(data
, sizeof(*data
));
356 e820__memory_setup_extended(pa_data
, data_len
);
362 parse_efi_setup(pa_data
, data_len
);
371 static void __init
memblock_x86_reserve_range_setup_data(void)
373 struct setup_data
*data
;
376 pa_data
= boot_params
.hdr
.setup_data
;
378 data
= early_memremap(pa_data
, sizeof(*data
));
379 memblock_reserve(pa_data
, sizeof(*data
) + data
->len
);
381 if (data
->type
== SETUP_INDIRECT
&&
382 ((struct setup_indirect
*)data
->data
)->type
!= SETUP_INDIRECT
)
383 memblock_reserve(((struct setup_indirect
*)data
->data
)->addr
,
384 ((struct setup_indirect
*)data
->data
)->len
);
386 pa_data
= data
->next
;
387 early_memunmap(data
, sizeof(*data
));
392 * --------- Crashkernel reservation ------------------------------
395 #ifdef CONFIG_KEXEC_CORE
397 /* 16M alignment for crash kernel regions */
398 #define CRASH_ALIGN SZ_16M
401 * Keep the crash kernel below this limit.
403 * Earlier 32-bits kernels would limit the kernel to the low 512 MB range
404 * due to mapping restrictions.
406 * 64-bit kdump kernels need to be restricted to be under 64 TB, which is
407 * the upper limit of system RAM in 4-level paging mode. Since the kdump
408 * jump could be from 5-level paging to 4-level paging, the jump will fail if
409 * the kernel is put above 64 TB, and during the 1st kernel bootup there's
410 * no good way to detect the paging mode of the target kernel which will be
411 * loaded for dumping.
414 # define CRASH_ADDR_LOW_MAX SZ_512M
415 # define CRASH_ADDR_HIGH_MAX SZ_512M
417 # define CRASH_ADDR_LOW_MAX SZ_4G
418 # define CRASH_ADDR_HIGH_MAX SZ_64T
421 static int __init
reserve_crashkernel_low(void)
424 unsigned long long base
, low_base
= 0, low_size
= 0;
425 unsigned long total_low_mem
;
428 total_low_mem
= memblock_mem_size(1UL << (32 - PAGE_SHIFT
));
430 /* crashkernel=Y,low */
431 ret
= parse_crashkernel_low(boot_command_line
, total_low_mem
, &low_size
, &base
);
434 * two parts from kernel/dma/swiotlb.c:
435 * -swiotlb size: user-specified with swiotlb= or default.
437 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
438 * to 8M for other buffers that may need to stay low too. Also
439 * make sure we allocate enough extra low memory so that we
440 * don't run out of DMA buffers for 32-bit devices.
442 low_size
= max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
444 /* passed with crashkernel=0,low ? */
449 low_base
= memblock_find_in_range(0, 1ULL << 32, low_size
, CRASH_ALIGN
);
451 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
452 (unsigned long)(low_size
>> 20));
456 ret
= memblock_reserve(low_base
, low_size
);
458 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__
);
462 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
463 (unsigned long)(low_size
>> 20),
464 (unsigned long)(low_base
>> 20),
465 (unsigned long)(total_low_mem
>> 20));
467 crashk_low_res
.start
= low_base
;
468 crashk_low_res
.end
= low_base
+ low_size
- 1;
469 insert_resource(&iomem_resource
, &crashk_low_res
);
474 static void __init
reserve_crashkernel(void)
476 unsigned long long crash_size
, crash_base
, total_mem
;
480 total_mem
= memblock_phys_mem_size();
483 ret
= parse_crashkernel(boot_command_line
, total_mem
, &crash_size
, &crash_base
);
484 if (ret
!= 0 || crash_size
<= 0) {
485 /* crashkernel=X,high */
486 ret
= parse_crashkernel_high(boot_command_line
, total_mem
,
487 &crash_size
, &crash_base
);
488 if (ret
!= 0 || crash_size
<= 0)
493 if (xen_pv_domain()) {
494 pr_info("Ignoring crashkernel for a Xen PV domain\n");
498 /* 0 means: find the address automatically */
501 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
502 * crashkernel=x,high reserves memory over 4G, also allocates
503 * 256M extra low memory for DMA buffers and swiotlb.
504 * But the extra memory is not required for all machines.
505 * So try low memory first and fall back to high memory
506 * unless "crashkernel=size[KMG],high" is specified.
509 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
511 crash_size
, CRASH_ALIGN
);
513 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
515 crash_size
, CRASH_ALIGN
);
517 pr_info("crashkernel reservation failed - No suitable area found.\n");
521 unsigned long long start
;
523 start
= memblock_find_in_range(crash_base
,
524 crash_base
+ crash_size
,
525 crash_size
, 1 << 20);
526 if (start
!= crash_base
) {
527 pr_info("crashkernel reservation failed - memory is in use.\n");
531 ret
= memblock_reserve(crash_base
, crash_size
);
533 pr_err("%s: Error reserving crashkernel memblock.\n", __func__
);
537 if (crash_base
>= (1ULL << 32) && reserve_crashkernel_low()) {
538 memblock_free(crash_base
, crash_size
);
542 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
543 (unsigned long)(crash_size
>> 20),
544 (unsigned long)(crash_base
>> 20),
545 (unsigned long)(total_mem
>> 20));
547 crashk_res
.start
= crash_base
;
548 crashk_res
.end
= crash_base
+ crash_size
- 1;
549 insert_resource(&iomem_resource
, &crashk_res
);
552 static void __init
reserve_crashkernel(void)
557 static struct resource standard_io_resources
[] = {
558 { .name
= "dma1", .start
= 0x00, .end
= 0x1f,
559 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
560 { .name
= "pic1", .start
= 0x20, .end
= 0x21,
561 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
562 { .name
= "timer0", .start
= 0x40, .end
= 0x43,
563 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
564 { .name
= "timer1", .start
= 0x50, .end
= 0x53,
565 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
566 { .name
= "keyboard", .start
= 0x60, .end
= 0x60,
567 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
568 { .name
= "keyboard", .start
= 0x64, .end
= 0x64,
569 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
570 { .name
= "dma page reg", .start
= 0x80, .end
= 0x8f,
571 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
572 { .name
= "pic2", .start
= 0xa0, .end
= 0xa1,
573 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
574 { .name
= "dma2", .start
= 0xc0, .end
= 0xdf,
575 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
576 { .name
= "fpu", .start
= 0xf0, .end
= 0xff,
577 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
}
580 void __init
reserve_standard_io_resources(void)
584 /* request I/O space for devices used on all i[345]86 PCs */
585 for (i
= 0; i
< ARRAY_SIZE(standard_io_resources
); i
++)
586 request_resource(&ioport_resource
, &standard_io_resources
[i
]);
590 static __init
void reserve_ibft_region(void)
592 unsigned long addr
, size
= 0;
594 addr
= find_ibft_region(&size
);
597 memblock_reserve(addr
, size
);
600 static bool __init
snb_gfx_workaround_needed(void)
605 static const __initconst u16 snb_ids
[] = {
615 /* Assume no if something weird is going on with PCI */
616 if (!early_pci_allowed())
619 vendor
= read_pci_config_16(0, 2, 0, PCI_VENDOR_ID
);
620 if (vendor
!= 0x8086)
623 devid
= read_pci_config_16(0, 2, 0, PCI_DEVICE_ID
);
624 for (i
= 0; i
< ARRAY_SIZE(snb_ids
); i
++)
625 if (devid
== snb_ids
[i
])
633 * Sandy Bridge graphics has trouble with certain ranges, exclude
634 * them from allocation.
636 static void __init
trim_snb_memory(void)
638 static const __initconst
unsigned long bad_pages
[] = {
647 if (!snb_gfx_workaround_needed())
650 printk(KERN_DEBUG
"reserving inaccessible SNB gfx pages\n");
653 * Reserve all memory below the 1 MB mark that has not
654 * already been reserved.
656 memblock_reserve(0, 1<<20);
658 for (i
= 0; i
< ARRAY_SIZE(bad_pages
); i
++) {
659 if (memblock_reserve(bad_pages
[i
], PAGE_SIZE
))
660 printk(KERN_WARNING
"failed to reserve 0x%08lx\n",
666 * Here we put platform-specific memory range workarounds, i.e.
667 * memory known to be corrupt or otherwise in need to be reserved on
668 * specific platforms.
670 * If this gets used more widely it could use a real dispatch mechanism.
672 static void __init
trim_platform_memory_ranges(void)
677 static void __init
trim_bios_range(void)
680 * A special case is the first 4Kb of memory;
681 * This is a BIOS owned area, not kernel ram, but generally
682 * not listed as such in the E820 table.
684 * This typically reserves additional memory (64KiB by default)
685 * since some BIOSes are known to corrupt low memory. See the
686 * Kconfig help text for X86_RESERVE_LOW.
688 e820__range_update(0, PAGE_SIZE
, E820_TYPE_RAM
, E820_TYPE_RESERVED
);
691 * special case: Some BIOSes report the PC BIOS
692 * area (640Kb -> 1Mb) as RAM even though it is not.
695 e820__range_remove(BIOS_BEGIN
, BIOS_END
- BIOS_BEGIN
, E820_TYPE_RAM
, 1);
697 e820__update_table(e820_table
);
700 /* called before trim_bios_range() to spare extra sanitize */
701 static void __init
e820_add_kernel_range(void)
703 u64 start
= __pa_symbol(_text
);
704 u64 size
= __pa_symbol(_end
) - start
;
707 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
708 * attempt to fix it by adding the range. We may have a confused BIOS,
709 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
710 * exclude kernel range. If we really are running on top non-RAM,
711 * we will crash later anyways.
713 if (e820__mapped_all(start
, start
+ size
, E820_TYPE_RAM
))
716 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
717 e820__range_remove(start
, size
, E820_TYPE_RAM
, 0);
718 e820__range_add(start
, size
, E820_TYPE_RAM
);
721 static unsigned reserve_low
= CONFIG_X86_RESERVE_LOW
<< 10;
723 static int __init
parse_reservelow(char *p
)
725 unsigned long long size
;
730 size
= memparse(p
, &p
);
743 early_param("reservelow", parse_reservelow
);
745 static void __init
trim_low_memory_range(void)
747 memblock_reserve(0, ALIGN(reserve_low
, PAGE_SIZE
));
751 * Dump out kernel offset information on panic.
754 dump_kernel_offset(struct notifier_block
*self
, unsigned long v
, void *p
)
756 if (kaslr_enabled()) {
757 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
763 pr_emerg("Kernel Offset: disabled\n");
770 * Determine if we were loaded by an EFI loader. If so, then we have also been
771 * passed the efi memmap, systab, etc., so we should use these data structures
772 * for initialization. Note, the efi init code path is determined by the
773 * global efi_enabled. This allows the same kernel image to be used on existing
774 * systems (with a traditional BIOS) as well as on EFI systems.
777 * setup_arch - architecture-specific boot-time initializations
779 * Note: On x86_64, fixmaps are ready for use even before this is called.
782 void __init
setup_arch(char **cmdline_p
)
785 * Reserve the memory occupied by the kernel between _text and
786 * __end_of_kernel_reserve symbols. Any kernel sections after the
787 * __end_of_kernel_reserve symbol must be explicitly reserved with a
788 * separate memblock_reserve() or they will be discarded.
790 memblock_reserve(__pa_symbol(_text
),
791 (unsigned long)__end_of_kernel_reserve
- (unsigned long)_text
);
794 * Make sure page 0 is always reserved because on systems with
795 * L1TF its contents can be leaked to user processes.
797 memblock_reserve(0, PAGE_SIZE
);
799 early_reserve_initrd();
802 * At this point everything still needed from the boot loader
803 * or BIOS or kernel text should be early reserved or marked not
804 * RAM in e820. All other memory is free game.
808 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
811 * copy kernel address range established so far and switch
812 * to the proper swapper page table
814 clone_pgd_range(swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
815 initial_page_table
+ KERNEL_PGD_BOUNDARY
,
818 load_cr3(swapper_pg_dir
);
820 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
821 * a cr3 based tlb flush, so the following __flush_tlb_all()
822 * will not flush anything because the CPU quirk which clears
823 * X86_FEATURE_PGE has not been invoked yet. Though due to the
824 * load_cr3() above the TLB has been flushed already. The
825 * quirk is invoked before subsequent calls to __flush_tlb_all()
826 * so proper operation is guaranteed.
830 printk(KERN_INFO
"Command line: %s\n", boot_command_line
);
831 boot_cpu_data
.x86_phys_bits
= MAX_PHYSMEM_BITS
;
835 * If we have OLPC OFW, we might end up relocating the fixmap due to
836 * reserve_top(), so do this before touching the ioremap area.
840 idt_setup_early_traps();
842 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
;
866 rd_prompt
= ((boot_params
.hdr
.ram_size
& RAMDISK_PROMPT_FLAG
) != 0);
867 rd_doload
= ((boot_params
.hdr
.ram_size
& RAMDISK_LOAD_FLAG
) != 0);
870 if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
871 EFI32_LOADER_SIGNATURE
, 4)) {
872 set_bit(EFI_BOOT
, &efi
.flags
);
873 } else if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
874 EFI64_LOADER_SIGNATURE
, 4)) {
875 set_bit(EFI_BOOT
, &efi
.flags
);
876 set_bit(EFI_64BIT
, &efi
.flags
);
880 x86_init
.oem
.arch_setup();
882 iomem_resource
.end
= (1ULL << boot_cpu_data
.x86_phys_bits
) - 1;
883 e820__memory_setup();
888 if (!boot_params
.hdr
.root_flags
)
889 root_mountflags
&= ~MS_RDONLY
;
890 init_mm
.start_code
= (unsigned long) _text
;
891 init_mm
.end_code
= (unsigned long) _etext
;
892 init_mm
.end_data
= (unsigned long) _edata
;
893 init_mm
.brk
= _brk_end
;
895 mpx_mm_init(&init_mm
);
897 code_resource
.start
= __pa_symbol(_text
);
898 code_resource
.end
= __pa_symbol(_etext
)-1;
899 rodata_resource
.start
= __pa_symbol(__start_rodata
);
900 rodata_resource
.end
= __pa_symbol(__end_rodata
)-1;
901 data_resource
.start
= __pa_symbol(_sdata
);
902 data_resource
.end
= __pa_symbol(_edata
)-1;
903 bss_resource
.start
= __pa_symbol(__bss_start
);
904 bss_resource
.end
= __pa_symbol(__bss_stop
)-1;
906 #ifdef CONFIG_CMDLINE_BOOL
907 #ifdef CONFIG_CMDLINE_OVERRIDE
908 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
910 if (builtin_cmdline
[0]) {
911 /* append boot loader cmdline to builtin */
912 strlcat(builtin_cmdline
, " ", COMMAND_LINE_SIZE
);
913 strlcat(builtin_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
914 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
919 strlcpy(command_line
, boot_command_line
, COMMAND_LINE_SIZE
);
920 *cmdline_p
= command_line
;
923 * x86_configure_nx() is called before parse_early_param() to detect
924 * whether hardware doesn't support NX (so that the early EHCI debug
925 * console setup can safely call set_fixmap()). It may then be called
926 * again from within noexec_setup() during parsing early parameters
927 * to honor the respective command line option.
933 if (efi_enabled(EFI_BOOT
))
934 efi_memblock_x86_reserve_range();
935 #ifdef CONFIG_MEMORY_HOTPLUG
937 * Memory used by the kernel cannot be hot-removed because Linux
938 * cannot migrate the kernel pages. When memory hotplug is
939 * enabled, we should prevent memblock from allocating memory
942 * ACPI SRAT records all hotpluggable memory ranges. But before
943 * SRAT is parsed, we don't know about it.
945 * The kernel image is loaded into memory at very early time. We
946 * cannot prevent this anyway. So on NUMA system, we set any
947 * node the kernel resides in as un-hotpluggable.
949 * Since on modern servers, one node could have double-digit
950 * gigabytes memory, we can assume the memory around the kernel
951 * image is also un-hotpluggable. So before SRAT is parsed, just
952 * allocate memory near the kernel image to try the best to keep
953 * the kernel away from hotpluggable memory.
955 if (movable_node_is_enabled())
956 memblock_set_bottom_up(true);
961 /* after early param, so could get panic from serial */
962 memblock_x86_reserve_range_setup_data();
964 if (acpi_mps_check()) {
965 #ifdef CONFIG_X86_LOCAL_APIC
968 setup_clear_cpu_cap(X86_FEATURE_APIC
);
971 e820__reserve_setup_data();
972 e820__finish_early_params();
974 if (efi_enabled(EFI_BOOT
))
980 * VMware detection requires dmi to be available, so this
981 * needs to be done after dmi_setup(), for the boot CPU.
983 init_hypervisor_platform();
986 x86_init
.resources
.probe_roms();
988 /* after parse_early_param, so could debug it */
989 insert_resource(&iomem_resource
, &code_resource
);
990 insert_resource(&iomem_resource
, &rodata_resource
);
991 insert_resource(&iomem_resource
, &data_resource
);
992 insert_resource(&iomem_resource
, &bss_resource
);
994 e820_add_kernel_range();
997 if (ppro_with_ram_bug()) {
998 e820__range_update(0x70000000ULL
, 0x40000ULL
, E820_TYPE_RAM
,
1000 e820__update_table(e820_table
);
1001 printk(KERN_INFO
"fixed physical RAM map:\n");
1002 e820__print_table("bad_ppro");
1005 early_gart_iommu_check();
1009 * partially used pages are not usable - thus
1010 * we are rounding upwards:
1012 max_pfn
= e820__end_of_ram_pfn();
1014 /* update e820 for memory not covered by WB MTRRs */
1016 if (mtrr_trim_uncached_memory(max_pfn
))
1017 max_pfn
= e820__end_of_ram_pfn();
1019 max_possible_pfn
= max_pfn
;
1022 * This call is required when the CPU does not support PAT. If
1023 * mtrr_bp_init() invoked it already via pat_init() the call has no
1029 * Define random base addresses for memory sections after max_pfn is
1030 * defined and before each memory section base is used.
1032 kernel_randomize_memory();
1034 #ifdef CONFIG_X86_32
1035 /* max_low_pfn get updated here */
1036 find_low_pfn_range();
1040 /* How many end-of-memory variables you have, grandma! */
1041 /* need this before calling reserve_initrd */
1042 if (max_pfn
> (1UL<<(32 - PAGE_SHIFT
)))
1043 max_low_pfn
= e820__end_of_low_ram_pfn();
1045 max_low_pfn
= max_pfn
;
1047 high_memory
= (void *)__va(max_pfn
* PAGE_SIZE
- 1) + 1;
1051 * Find and reserve possible boot-time SMP configuration:
1055 reserve_ibft_region();
1057 early_alloc_pgt_buf();
1060 * Need to conclude brk, before e820__memblock_setup()
1061 * it could use memblock_find_in_range, could overlap with
1068 memblock_set_current_limit(ISA_END_ADDRESS
);
1069 e820__memblock_setup();
1071 reserve_bios_regions();
1078 * The EFI specification says that boot service code won't be
1079 * called after ExitBootServices(). This is, in fact, a lie.
1081 efi_reserve_boot_services();
1083 /* preallocate 4k for mptable mpc */
1084 e820__memblock_alloc_reserved_mpc_new();
1086 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1087 setup_bios_corruption_check();
1090 #ifdef CONFIG_X86_32
1091 printk(KERN_DEBUG
"initial memory mapped: [mem 0x00000000-%#010lx]\n",
1092 (max_pfn_mapped
<<PAGE_SHIFT
) - 1);
1095 reserve_real_mode();
1097 trim_platform_memory_ranges();
1098 trim_low_memory_range();
1102 idt_setup_early_pf();
1105 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1106 * with the current CR4 value. This may not be necessary, but
1107 * auditing all the early-boot CR4 manipulation would be needed to
1110 * Mask off features that don't work outside long mode (just
1113 mmu_cr4_features
= __read_cr4() & ~X86_CR4_PCIDE
;
1115 memblock_set_current_limit(get_max_mapped());
1118 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1121 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1122 if (init_ohci1394_dma_early
)
1123 init_ohci1394_dma_on_all_controllers();
1125 /* Allocate bigger log buffer */
1128 if (efi_enabled(EFI_BOOT
)) {
1129 switch (boot_params
.secure_boot
) {
1130 case efi_secureboot_mode_disabled
:
1131 pr_info("Secure boot disabled\n");
1133 case efi_secureboot_mode_enabled
:
1134 pr_info("Secure boot enabled\n");
1137 pr_info("Secure boot could not be determined\n");
1144 acpi_table_upgrade();
1150 early_platform_quirks();
1153 * Parse the ACPI tables for possible boot-time SMP configuration.
1155 acpi_boot_table_init();
1157 early_acpi_boot_init();
1160 dma_contiguous_reserve(max_pfn_mapped
<< 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
;
1230 #elif defined(CONFIG_DUMMY_CONSOLE)
1231 conswitchp
= &dummy_con
;
1234 x86_init
.oem
.banner();
1236 x86_init
.timers
.wallclock_init();
1240 register_refined_jiffies(CLOCK_TICK_RATE
);
1243 if (efi_enabled(EFI_BOOT
))
1244 efi_apply_memmap_quirks();
1250 #ifdef CONFIG_X86_32
1252 static struct resource video_ram_resource
= {
1253 .name
= "Video RAM area",
1256 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
1259 void __init
i386_reserve_resources(void)
1261 request_resource(&iomem_resource
, &video_ram_resource
);
1262 reserve_standard_io_resources();
1265 #endif /* CONFIG_X86_32 */
1267 static struct notifier_block kernel_offset_notifier
= {
1268 .notifier_call
= dump_kernel_offset
1271 static int __init
register_kernel_offset_dumper(void)
1273 atomic_notifier_chain_register(&panic_notifier_list
,
1274 &kernel_offset_notifier
);
1277 __initcall(register_kernel_offset_dumper
);