2 * Copyright (C) 1995 Linus Torvalds
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
21 * This file handles the architecture-dependent parts of initialization
24 #include <linux/sched.h>
26 #include <linux/mmzone.h>
27 #include <linux/screen_info.h>
28 #include <linux/ioport.h>
29 #include <linux/acpi.h>
30 #include <linux/sfi.h>
31 #include <linux/apm_bios.h>
32 #include <linux/initrd.h>
33 #include <linux/bootmem.h>
34 #include <linux/memblock.h>
35 #include <linux/seq_file.h>
36 #include <linux/console.h>
37 #include <linux/root_dev.h>
38 #include <linux/highmem.h>
39 #include <linux/module.h>
40 #include <linux/efi.h>
41 #include <linux/init.h>
42 #include <linux/edd.h>
43 #include <linux/iscsi_ibft.h>
44 #include <linux/nodemask.h>
45 #include <linux/kexec.h>
46 #include <linux/dmi.h>
47 #include <linux/pfn.h>
48 #include <linux/pci.h>
49 #include <asm/pci-direct.h>
50 #include <linux/init_ohci1394_dma.h>
51 #include <linux/kvm_para.h>
52 #include <linux/dma-contiguous.h>
54 #include <linux/errno.h>
55 #include <linux/kernel.h>
56 #include <linux/stddef.h>
57 #include <linux/unistd.h>
58 #include <linux/ptrace.h>
59 #include <linux/user.h>
60 #include <linux/delay.h>
62 #include <linux/kallsyms.h>
63 #include <linux/cpufreq.h>
64 #include <linux/dma-mapping.h>
65 #include <linux/ctype.h>
66 #include <linux/uaccess.h>
68 #include <linux/percpu.h>
69 #include <linux/crash_dump.h>
70 #include <linux/tboot.h>
71 #include <linux/jiffies.h>
73 #include <video/edid.h>
77 #include <asm/realmode.h>
79 #include <asm/mpspec.h>
80 #include <asm/setup.h>
82 #include <asm/timer.h>
83 #include <asm/i8259.h>
84 #include <asm/sections.h>
85 #include <asm/io_apic.h>
87 #include <asm/setup_arch.h>
88 #include <asm/bios_ebda.h>
89 #include <asm/cacheflush.h>
90 #include <asm/processor.h>
92 #include <asm/kasan.h>
94 #include <asm/vsyscall.h>
98 #include <asm/iommu.h>
100 #include <asm/mmu_context.h>
101 #include <asm/proto.h>
103 #include <asm/paravirt.h>
104 #include <asm/hypervisor.h>
105 #include <asm/olpc_ofw.h>
107 #include <asm/percpu.h>
108 #include <asm/topology.h>
109 #include <asm/apicdef.h>
110 #include <asm/amd_nb.h>
112 #include <asm/alternative.h>
113 #include <asm/prom.h>
114 #include <asm/microcode.h>
115 #include <asm/mmu_context.h>
118 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
119 * max_pfn_mapped: highest direct mapped pfn over 4GB
121 * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
122 * represented by pfn_mapped
124 unsigned long max_low_pfn_mapped
;
125 unsigned long max_pfn_mapped
;
128 RESERVE_BRK(dmi_alloc
, 65536);
132 static __initdata
unsigned long _brk_start
= (unsigned long)__brk_base
;
133 unsigned long _brk_end
= (unsigned long)__brk_base
;
136 int default_cpu_present_to_apicid(int mps_cpu
)
138 return __default_cpu_present_to_apicid(mps_cpu
);
141 int default_check_phys_apicid_present(int phys_apicid
)
143 return __default_check_phys_apicid_present(phys_apicid
);
147 struct boot_params boot_params
;
152 static struct resource data_resource
= {
153 .name
= "Kernel data",
156 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
159 static struct resource code_resource
= {
160 .name
= "Kernel code",
163 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
166 static struct resource bss_resource
= {
167 .name
= "Kernel bss",
170 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
175 /* cpu data as detected by the assembly code in head.S */
176 struct cpuinfo_x86 new_cpu_data
= {
179 /* common cpu data for all cpus */
180 struct cpuinfo_x86 boot_cpu_data __read_mostly
= {
183 EXPORT_SYMBOL(boot_cpu_data
);
185 unsigned int def_to_bigsmp
;
187 /* for MCA, but anyone else can use it if they want */
188 unsigned int machine_id
;
189 unsigned int machine_submodel_id
;
190 unsigned int BIOS_revision
;
192 struct apm_info apm_info
;
193 EXPORT_SYMBOL(apm_info
);
195 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
196 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
197 struct ist_info ist_info
;
198 EXPORT_SYMBOL(ist_info
);
200 struct ist_info ist_info
;
204 struct cpuinfo_x86 boot_cpu_data __read_mostly
= {
205 .x86_phys_bits
= MAX_PHYSMEM_BITS
,
207 EXPORT_SYMBOL(boot_cpu_data
);
211 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
212 __visible
unsigned long mmu_cr4_features
;
214 __visible
unsigned long mmu_cr4_features
= X86_CR4_PAE
;
217 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
218 int bootloader_type
, bootloader_version
;
223 struct screen_info screen_info
;
224 EXPORT_SYMBOL(screen_info
);
225 struct edid_info edid_info
;
226 EXPORT_SYMBOL_GPL(edid_info
);
228 extern int root_mountflags
;
230 unsigned long saved_video_mode
;
232 #define RAMDISK_IMAGE_START_MASK 0x07FF
233 #define RAMDISK_PROMPT_FLAG 0x8000
234 #define RAMDISK_LOAD_FLAG 0x4000
236 static char __initdata command_line
[COMMAND_LINE_SIZE
];
237 #ifdef CONFIG_CMDLINE_BOOL
238 static char __initdata builtin_cmdline
[COMMAND_LINE_SIZE
] = CONFIG_CMDLINE
;
241 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
243 #ifdef CONFIG_EDD_MODULE
247 * copy_edd() - Copy the BIOS EDD information
248 * from boot_params into a safe place.
251 static inline void __init
copy_edd(void)
253 memcpy(edd
.mbr_signature
, boot_params
.edd_mbr_sig_buffer
,
254 sizeof(edd
.mbr_signature
));
255 memcpy(edd
.edd_info
, boot_params
.eddbuf
, sizeof(edd
.edd_info
));
256 edd
.mbr_signature_nr
= boot_params
.edd_mbr_sig_buf_entries
;
257 edd
.edd_info_nr
= boot_params
.eddbuf_entries
;
260 static inline void __init
copy_edd(void)
265 void * __init
extend_brk(size_t size
, size_t align
)
267 size_t mask
= align
- 1;
270 BUG_ON(_brk_start
== 0);
271 BUG_ON(align
& mask
);
273 _brk_end
= (_brk_end
+ mask
) & ~mask
;
274 BUG_ON((char *)(_brk_end
+ size
) > __brk_limit
);
276 ret
= (void *)_brk_end
;
279 memset(ret
, 0, size
);
285 static void __init
cleanup_highmap(void)
290 static void __init
reserve_brk(void)
292 if (_brk_end
> _brk_start
)
293 memblock_reserve(__pa_symbol(_brk_start
),
294 _brk_end
- _brk_start
);
296 /* Mark brk area as locked down and no longer taking any
301 u64 relocated_ramdisk
;
303 #ifdef CONFIG_BLK_DEV_INITRD
305 static u64 __init
get_ramdisk_image(void)
307 u64 ramdisk_image
= boot_params
.hdr
.ramdisk_image
;
309 ramdisk_image
|= (u64
)boot_params
.ext_ramdisk_image
<< 32;
311 return ramdisk_image
;
313 static u64 __init
get_ramdisk_size(void)
315 u64 ramdisk_size
= boot_params
.hdr
.ramdisk_size
;
317 ramdisk_size
|= (u64
)boot_params
.ext_ramdisk_size
<< 32;
322 static void __init
relocate_initrd(void)
324 /* Assume only end is not page aligned */
325 u64 ramdisk_image
= get_ramdisk_image();
326 u64 ramdisk_size
= get_ramdisk_size();
327 u64 area_size
= PAGE_ALIGN(ramdisk_size
);
329 /* We need to move the initrd down into directly mapped mem */
330 relocated_ramdisk
= memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped
),
331 area_size
, PAGE_SIZE
);
333 if (!relocated_ramdisk
)
334 panic("Cannot find place for new RAMDISK of size %lld\n",
337 /* Note: this includes all the mem currently occupied by
338 the initrd, we rely on that fact to keep the data intact. */
339 memblock_reserve(relocated_ramdisk
, area_size
);
340 initrd_start
= relocated_ramdisk
+ PAGE_OFFSET
;
341 initrd_end
= initrd_start
+ ramdisk_size
;
342 printk(KERN_INFO
"Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
343 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
345 copy_from_early_mem((void *)initrd_start
, ramdisk_image
, ramdisk_size
);
347 printk(KERN_INFO
"Move RAMDISK from [mem %#010llx-%#010llx] to"
348 " [mem %#010llx-%#010llx]\n",
349 ramdisk_image
, ramdisk_image
+ ramdisk_size
- 1,
350 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
353 static void __init
early_reserve_initrd(void)
355 /* Assume only end is not page aligned */
356 u64 ramdisk_image
= get_ramdisk_image();
357 u64 ramdisk_size
= get_ramdisk_size();
358 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
360 if (!boot_params
.hdr
.type_of_loader
||
361 !ramdisk_image
|| !ramdisk_size
)
362 return; /* No initrd provided by bootloader */
364 memblock_reserve(ramdisk_image
, ramdisk_end
- ramdisk_image
);
366 static void __init
reserve_initrd(void)
368 /* Assume only end is not page aligned */
369 u64 ramdisk_image
= get_ramdisk_image();
370 u64 ramdisk_size
= get_ramdisk_size();
371 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
374 if (!boot_params
.hdr
.type_of_loader
||
375 !ramdisk_image
|| !ramdisk_size
)
376 return; /* No initrd provided by bootloader */
380 mapped_size
= memblock_mem_size(max_pfn_mapped
);
381 if (ramdisk_size
>= (mapped_size
>>1))
382 panic("initrd too large to handle, "
383 "disabling initrd (%lld needed, %lld available)\n",
384 ramdisk_size
, mapped_size
>>1);
386 printk(KERN_INFO
"RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image
,
389 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image
),
390 PFN_DOWN(ramdisk_end
))) {
391 /* All are mapped, easy case */
392 initrd_start
= ramdisk_image
+ PAGE_OFFSET
;
393 initrd_end
= initrd_start
+ ramdisk_size
;
399 memblock_free(ramdisk_image
, ramdisk_end
- ramdisk_image
);
402 static void __init
early_initrd_acpi_init(void)
404 early_acpi_table_init((void *)initrd_start
, initrd_end
- initrd_start
);
407 static void __init
early_reserve_initrd(void)
410 static void __init
reserve_initrd(void)
413 static void __init
early_initrd_acpi_init(void)
416 #endif /* CONFIG_BLK_DEV_INITRD */
418 static void __init
parse_setup_data(void)
420 struct setup_data
*data
;
421 u64 pa_data
, pa_next
;
423 pa_data
= boot_params
.hdr
.setup_data
;
425 u32 data_len
, data_type
;
427 data
= early_memremap(pa_data
, sizeof(*data
));
428 data_len
= data
->len
+ sizeof(struct setup_data
);
429 data_type
= data
->type
;
430 pa_next
= data
->next
;
431 early_memunmap(data
, sizeof(*data
));
435 parse_e820_ext(pa_data
, data_len
);
441 parse_efi_setup(pa_data
, data_len
);
450 static void __init
e820_reserve_setup_data(void)
452 struct setup_data
*data
;
455 pa_data
= boot_params
.hdr
.setup_data
;
460 data
= early_memremap(pa_data
, sizeof(*data
));
461 e820_update_range(pa_data
, sizeof(*data
)+data
->len
,
462 E820_RAM
, E820_RESERVED_KERN
);
463 pa_data
= data
->next
;
464 early_memunmap(data
, sizeof(*data
));
467 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
468 memcpy(&e820_saved
, &e820
, sizeof(struct e820map
));
469 printk(KERN_INFO
"extended physical RAM map:\n");
470 e820_print_map("reserve setup_data");
473 static void __init
memblock_x86_reserve_range_setup_data(void)
475 struct setup_data
*data
;
478 pa_data
= boot_params
.hdr
.setup_data
;
480 data
= early_memremap(pa_data
, sizeof(*data
));
481 memblock_reserve(pa_data
, sizeof(*data
) + data
->len
);
482 pa_data
= data
->next
;
483 early_memunmap(data
, sizeof(*data
));
488 * --------- Crashkernel reservation ------------------------------
491 #ifdef CONFIG_KEXEC_CORE
493 /* 16M alignment for crash kernel regions */
494 #define CRASH_ALIGN (16 << 20)
497 * Keep the crash kernel below this limit. On 32 bits earlier kernels
498 * would limit the kernel to the low 512 MiB due to mapping restrictions.
499 * On 64bit, old kexec-tools need to under 896MiB.
502 # define CRASH_ADDR_LOW_MAX (512 << 20)
503 # define CRASH_ADDR_HIGH_MAX (512 << 20)
505 # define CRASH_ADDR_LOW_MAX (896UL << 20)
506 # define CRASH_ADDR_HIGH_MAX MAXMEM
509 static int __init
reserve_crashkernel_low(void)
512 unsigned long long base
, low_base
= 0, low_size
= 0;
513 unsigned long total_low_mem
;
516 total_low_mem
= memblock_mem_size(1UL << (32 - PAGE_SHIFT
));
518 /* crashkernel=Y,low */
519 ret
= parse_crashkernel_low(boot_command_line
, total_low_mem
, &low_size
, &base
);
522 * two parts from lib/swiotlb.c:
523 * -swiotlb size: user-specified with swiotlb= or default.
525 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
526 * to 8M for other buffers that may need to stay low too. Also
527 * make sure we allocate enough extra low memory so that we
528 * don't run out of DMA buffers for 32-bit devices.
530 low_size
= max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
532 /* passed with crashkernel=0,low ? */
537 low_base
= memblock_find_in_range(low_size
, 1ULL << 32, low_size
, CRASH_ALIGN
);
539 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
540 (unsigned long)(low_size
>> 20));
544 ret
= memblock_reserve(low_base
, low_size
);
546 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__
);
550 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
551 (unsigned long)(low_size
>> 20),
552 (unsigned long)(low_base
>> 20),
553 (unsigned long)(total_low_mem
>> 20));
555 crashk_low_res
.start
= low_base
;
556 crashk_low_res
.end
= low_base
+ low_size
- 1;
557 insert_resource(&iomem_resource
, &crashk_low_res
);
562 static void __init
reserve_crashkernel(void)
564 unsigned long long crash_size
, crash_base
, total_mem
;
568 total_mem
= memblock_phys_mem_size();
571 ret
= parse_crashkernel(boot_command_line
, total_mem
, &crash_size
, &crash_base
);
572 if (ret
!= 0 || crash_size
<= 0) {
573 /* crashkernel=X,high */
574 ret
= parse_crashkernel_high(boot_command_line
, total_mem
,
575 &crash_size
, &crash_base
);
576 if (ret
!= 0 || crash_size
<= 0)
581 /* 0 means: find the address automatically */
582 if (crash_base
<= 0) {
584 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
586 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
587 high
? CRASH_ADDR_HIGH_MAX
588 : CRASH_ADDR_LOW_MAX
,
589 crash_size
, CRASH_ALIGN
);
591 pr_info("crashkernel reservation failed - No suitable area found.\n");
596 unsigned long long start
;
598 start
= memblock_find_in_range(crash_base
,
599 crash_base
+ crash_size
,
600 crash_size
, 1 << 20);
601 if (start
!= crash_base
) {
602 pr_info("crashkernel reservation failed - memory is in use.\n");
606 ret
= memblock_reserve(crash_base
, crash_size
);
608 pr_err("%s: Error reserving crashkernel memblock.\n", __func__
);
612 if (crash_base
>= (1ULL << 32) && reserve_crashkernel_low()) {
613 memblock_free(crash_base
, crash_size
);
617 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
618 (unsigned long)(crash_size
>> 20),
619 (unsigned long)(crash_base
>> 20),
620 (unsigned long)(total_mem
>> 20));
622 crashk_res
.start
= crash_base
;
623 crashk_res
.end
= crash_base
+ crash_size
- 1;
624 insert_resource(&iomem_resource
, &crashk_res
);
627 static void __init
reserve_crashkernel(void)
632 static struct resource standard_io_resources
[] = {
633 { .name
= "dma1", .start
= 0x00, .end
= 0x1f,
634 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
635 { .name
= "pic1", .start
= 0x20, .end
= 0x21,
636 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
637 { .name
= "timer0", .start
= 0x40, .end
= 0x43,
638 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
639 { .name
= "timer1", .start
= 0x50, .end
= 0x53,
640 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
641 { .name
= "keyboard", .start
= 0x60, .end
= 0x60,
642 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
643 { .name
= "keyboard", .start
= 0x64, .end
= 0x64,
644 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
645 { .name
= "dma page reg", .start
= 0x80, .end
= 0x8f,
646 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
647 { .name
= "pic2", .start
= 0xa0, .end
= 0xa1,
648 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
649 { .name
= "dma2", .start
= 0xc0, .end
= 0xdf,
650 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
651 { .name
= "fpu", .start
= 0xf0, .end
= 0xff,
652 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
}
655 void __init
reserve_standard_io_resources(void)
659 /* request I/O space for devices used on all i[345]86 PCs */
660 for (i
= 0; i
< ARRAY_SIZE(standard_io_resources
); i
++)
661 request_resource(&ioport_resource
, &standard_io_resources
[i
]);
665 static __init
void reserve_ibft_region(void)
667 unsigned long addr
, size
= 0;
669 addr
= find_ibft_region(&size
);
672 memblock_reserve(addr
, size
);
675 static bool __init
snb_gfx_workaround_needed(void)
680 static const __initconst u16 snb_ids
[] = {
690 /* Assume no if something weird is going on with PCI */
691 if (!early_pci_allowed())
694 vendor
= read_pci_config_16(0, 2, 0, PCI_VENDOR_ID
);
695 if (vendor
!= 0x8086)
698 devid
= read_pci_config_16(0, 2, 0, PCI_DEVICE_ID
);
699 for (i
= 0; i
< ARRAY_SIZE(snb_ids
); i
++)
700 if (devid
== snb_ids
[i
])
708 * Sandy Bridge graphics has trouble with certain ranges, exclude
709 * them from allocation.
711 static void __init
trim_snb_memory(void)
713 static const __initconst
unsigned long bad_pages
[] = {
722 if (!snb_gfx_workaround_needed())
725 printk(KERN_DEBUG
"reserving inaccessible SNB gfx pages\n");
728 * Reserve all memory below the 1 MB mark that has not
729 * already been reserved.
731 memblock_reserve(0, 1<<20);
733 for (i
= 0; i
< ARRAY_SIZE(bad_pages
); i
++) {
734 if (memblock_reserve(bad_pages
[i
], PAGE_SIZE
))
735 printk(KERN_WARNING
"failed to reserve 0x%08lx\n",
741 * Here we put platform-specific memory range workarounds, i.e.
742 * memory known to be corrupt or otherwise in need to be reserved on
743 * specific platforms.
745 * If this gets used more widely it could use a real dispatch mechanism.
747 static void __init
trim_platform_memory_ranges(void)
752 static void __init
trim_bios_range(void)
755 * A special case is the first 4Kb of memory;
756 * This is a BIOS owned area, not kernel ram, but generally
757 * not listed as such in the E820 table.
759 * This typically reserves additional memory (64KiB by default)
760 * since some BIOSes are known to corrupt low memory. See the
761 * Kconfig help text for X86_RESERVE_LOW.
763 e820_update_range(0, PAGE_SIZE
, E820_RAM
, E820_RESERVED
);
766 * special case: Some BIOSen report the PC BIOS
767 * area (640->1Mb) as ram even though it is not.
770 e820_remove_range(BIOS_BEGIN
, BIOS_END
- BIOS_BEGIN
, E820_RAM
, 1);
772 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
775 /* called before trim_bios_range() to spare extra sanitize */
776 static void __init
e820_add_kernel_range(void)
778 u64 start
= __pa_symbol(_text
);
779 u64 size
= __pa_symbol(_end
) - start
;
782 * Complain if .text .data and .bss are not marked as E820_RAM and
783 * attempt to fix it by adding the range. We may have a confused BIOS,
784 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
785 * exclude kernel range. If we really are running on top non-RAM,
786 * we will crash later anyways.
788 if (e820_all_mapped(start
, start
+ size
, E820_RAM
))
791 pr_warn(".text .data .bss are not marked as E820_RAM!\n");
792 e820_remove_range(start
, size
, E820_RAM
, 0);
793 e820_add_region(start
, size
, E820_RAM
);
796 static unsigned reserve_low
= CONFIG_X86_RESERVE_LOW
<< 10;
798 static int __init
parse_reservelow(char *p
)
800 unsigned long long size
;
805 size
= memparse(p
, &p
);
818 early_param("reservelow", parse_reservelow
);
820 static void __init
trim_low_memory_range(void)
822 memblock_reserve(0, ALIGN(reserve_low
, PAGE_SIZE
));
826 * Dump out kernel offset information on panic.
829 dump_kernel_offset(struct notifier_block
*self
, unsigned long v
, void *p
)
831 if (kaslr_enabled()) {
832 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
838 pr_emerg("Kernel Offset: disabled\n");
845 * Determine if we were loaded by an EFI loader. If so, then we have also been
846 * passed the efi memmap, systab, etc., so we should use these data structures
847 * for initialization. Note, the efi init code path is determined by the
848 * global efi_enabled. This allows the same kernel image to be used on existing
849 * systems (with a traditional BIOS) as well as on EFI systems.
852 * setup_arch - architecture-specific boot-time initializations
854 * Note: On x86_64, fixmaps are ready for use even before this is called.
857 void __init
setup_arch(char **cmdline_p
)
859 memblock_reserve(__pa_symbol(_text
),
860 (unsigned long)__bss_stop
- (unsigned long)_text
);
862 early_reserve_initrd();
865 * At this point everything still needed from the boot loader
866 * or BIOS or kernel text should be early reserved or marked not
867 * RAM in e820. All other memory is free game.
871 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
874 * copy kernel address range established so far and switch
875 * to the proper swapper page table
877 clone_pgd_range(swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
878 initial_page_table
+ KERNEL_PGD_BOUNDARY
,
881 load_cr3(swapper_pg_dir
);
883 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
884 * a cr3 based tlb flush, so the following __flush_tlb_all()
885 * will not flush anything because the cpu quirk which clears
886 * X86_FEATURE_PGE has not been invoked yet. Though due to the
887 * load_cr3() above the TLB has been flushed already. The
888 * quirk is invoked before subsequent calls to __flush_tlb_all()
889 * so proper operation is guaranteed.
893 printk(KERN_INFO
"Command line: %s\n", boot_command_line
);
897 * If we have OLPC OFW, we might end up relocating the fixmap due to
898 * reserve_top(), so do this before touching the ioremap area.
904 early_ioremap_init();
906 setup_olpc_ofw_pgd();
908 ROOT_DEV
= old_decode_dev(boot_params
.hdr
.root_dev
);
909 screen_info
= boot_params
.screen_info
;
910 edid_info
= boot_params
.edid_info
;
912 apm_info
.bios
= boot_params
.apm_bios_info
;
913 ist_info
= boot_params
.ist_info
;
915 saved_video_mode
= boot_params
.hdr
.vid_mode
;
916 bootloader_type
= boot_params
.hdr
.type_of_loader
;
917 if ((bootloader_type
>> 4) == 0xe) {
918 bootloader_type
&= 0xf;
919 bootloader_type
|= (boot_params
.hdr
.ext_loader_type
+0x10) << 4;
921 bootloader_version
= bootloader_type
& 0xf;
922 bootloader_version
|= boot_params
.hdr
.ext_loader_ver
<< 4;
924 #ifdef CONFIG_BLK_DEV_RAM
925 rd_image_start
= boot_params
.hdr
.ram_size
& RAMDISK_IMAGE_START_MASK
;
926 rd_prompt
= ((boot_params
.hdr
.ram_size
& RAMDISK_PROMPT_FLAG
) != 0);
927 rd_doload
= ((boot_params
.hdr
.ram_size
& RAMDISK_LOAD_FLAG
) != 0);
930 if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
931 EFI32_LOADER_SIGNATURE
, 4)) {
932 set_bit(EFI_BOOT
, &efi
.flags
);
933 } else if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
934 EFI64_LOADER_SIGNATURE
, 4)) {
935 set_bit(EFI_BOOT
, &efi
.flags
);
936 set_bit(EFI_64BIT
, &efi
.flags
);
939 if (efi_enabled(EFI_BOOT
))
940 efi_memblock_x86_reserve_range();
943 x86_init
.oem
.arch_setup();
945 iomem_resource
.end
= (1ULL << boot_cpu_data
.x86_phys_bits
) - 1;
951 if (!boot_params
.hdr
.root_flags
)
952 root_mountflags
&= ~MS_RDONLY
;
953 init_mm
.start_code
= (unsigned long) _text
;
954 init_mm
.end_code
= (unsigned long) _etext
;
955 init_mm
.end_data
= (unsigned long) _edata
;
956 init_mm
.brk
= _brk_end
;
958 mpx_mm_init(&init_mm
);
960 code_resource
.start
= __pa_symbol(_text
);
961 code_resource
.end
= __pa_symbol(_etext
)-1;
962 data_resource
.start
= __pa_symbol(_etext
);
963 data_resource
.end
= __pa_symbol(_edata
)-1;
964 bss_resource
.start
= __pa_symbol(__bss_start
);
965 bss_resource
.end
= __pa_symbol(__bss_stop
)-1;
967 #ifdef CONFIG_CMDLINE_BOOL
968 #ifdef CONFIG_CMDLINE_OVERRIDE
969 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
971 if (builtin_cmdline
[0]) {
972 /* append boot loader cmdline to builtin */
973 strlcat(builtin_cmdline
, " ", COMMAND_LINE_SIZE
);
974 strlcat(builtin_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
975 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
980 strlcpy(command_line
, boot_command_line
, COMMAND_LINE_SIZE
);
981 *cmdline_p
= command_line
;
984 * x86_configure_nx() is called before parse_early_param() to detect
985 * whether hardware doesn't support NX (so that the early EHCI debug
986 * console setup can safely call set_fixmap()). It may then be called
987 * again from within noexec_setup() during parsing early parameters
988 * to honor the respective command line option.
996 /* after early param, so could get panic from serial */
997 memblock_x86_reserve_range_setup_data();
999 if (acpi_mps_check()) {
1000 #ifdef CONFIG_X86_LOCAL_APIC
1003 setup_clear_cpu_cap(X86_FEATURE_APIC
);
1007 if (pci_early_dump_regs
)
1008 early_dump_pci_devices();
1011 /* update the e820_saved too */
1012 e820_reserve_setup_data();
1013 finish_e820_parsing();
1015 if (efi_enabled(EFI_BOOT
))
1020 dmi_set_dump_stack_arch_desc();
1023 * VMware detection requires dmi to be available, so this
1024 * needs to be done after dmi_scan_machine, for the BP.
1026 init_hypervisor_platform();
1028 x86_init
.resources
.probe_roms();
1030 /* after parse_early_param, so could debug it */
1031 insert_resource(&iomem_resource
, &code_resource
);
1032 insert_resource(&iomem_resource
, &data_resource
);
1033 insert_resource(&iomem_resource
, &bss_resource
);
1035 e820_add_kernel_range();
1037 #ifdef CONFIG_X86_32
1038 if (ppro_with_ram_bug()) {
1039 e820_update_range(0x70000000ULL
, 0x40000ULL
, E820_RAM
,
1041 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
1042 printk(KERN_INFO
"fixed physical RAM map:\n");
1043 e820_print_map("bad_ppro");
1046 early_gart_iommu_check();
1050 * partially used pages are not usable - thus
1051 * we are rounding upwards:
1053 max_pfn
= e820_end_of_ram_pfn();
1055 /* update e820 for memory not covered by WB MTRRs */
1057 if (mtrr_trim_uncached_memory(max_pfn
))
1058 max_pfn
= e820_end_of_ram_pfn();
1060 max_possible_pfn
= max_pfn
;
1062 #ifdef CONFIG_X86_32
1063 /* max_low_pfn get updated here */
1064 find_low_pfn_range();
1068 /* How many end-of-memory variables you have, grandma! */
1069 /* need this before calling reserve_initrd */
1070 if (max_pfn
> (1UL<<(32 - PAGE_SHIFT
)))
1071 max_low_pfn
= e820_end_of_low_ram_pfn();
1073 max_low_pfn
= max_pfn
;
1075 high_memory
= (void *)__va(max_pfn
* PAGE_SIZE
- 1) + 1;
1079 * Find and reserve possible boot-time SMP configuration:
1083 reserve_ibft_region();
1085 early_alloc_pgt_buf();
1088 * Need to conclude brk, before memblock_x86_fill()
1089 * it could use memblock_find_in_range, could overlap with
1096 memblock_set_current_limit(ISA_END_ADDRESS
);
1097 memblock_x86_fill();
1099 if (efi_enabled(EFI_BOOT
)) {
1105 * The EFI specification says that boot service code won't be called
1106 * after ExitBootServices(). This is, in fact, a lie.
1108 if (efi_enabled(EFI_MEMMAP
))
1109 efi_reserve_boot_services();
1111 /* preallocate 4k for mptable mpc */
1112 early_reserve_e820_mpc_new();
1114 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1115 setup_bios_corruption_check();
1118 #ifdef CONFIG_X86_32
1119 printk(KERN_DEBUG
"initial memory mapped: [mem 0x00000000-%#010lx]\n",
1120 (max_pfn_mapped
<<PAGE_SHIFT
) - 1);
1123 reserve_real_mode();
1125 trim_platform_memory_ranges();
1126 trim_low_memory_range();
1130 early_trap_pf_init();
1134 memblock_set_current_limit(get_max_mapped());
1137 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1140 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1141 if (init_ohci1394_dma_early
)
1142 init_ohci1394_dma_on_all_controllers();
1144 /* Allocate bigger log buffer */
1149 early_initrd_acpi_init();
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
);
1166 * Reserve memory for crash kernel after SRAT is parsed so that it
1167 * won't consume hotpluggable memory.
1169 reserve_crashkernel();
1171 memblock_find_dma_reserve();
1173 #ifdef CONFIG_KVM_GUEST
1177 x86_init
.paging
.pagetable_init();
1181 if (boot_cpu_data
.cpuid_level
>= 0) {
1182 /* A CPU has %cr4 if and only if it has CPUID */
1183 mmu_cr4_features
= __read_cr4();
1184 if (trampoline_cr4_features
)
1185 *trampoline_cr4_features
= mmu_cr4_features
;
1188 #ifdef CONFIG_X86_32
1189 /* sync back kernel address range */
1190 clone_pgd_range(initial_page_table
+ KERNEL_PGD_BOUNDARY
,
1191 swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
1195 * sync back low identity map too. It is used for example
1196 * in the 32-bit EFI stub.
1198 clone_pgd_range(initial_page_table
,
1199 swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
1200 min(KERNEL_PGD_PTRS
, KERNEL_PGD_BOUNDARY
));
1207 generic_apic_probe();
1212 * Read APIC and some other early information from ACPI tables.
1219 * get boot-time SMP configuration:
1221 if (smp_found_config
)
1224 prefill_possible_map();
1228 init_apic_mappings();
1229 io_apic_init_mappings();
1233 e820_reserve_resources();
1234 e820_mark_nosave_regions(max_low_pfn
);
1236 x86_init
.resources
.reserve_resources();
1241 #if defined(CONFIG_VGA_CONSOLE)
1242 if (!efi_enabled(EFI_BOOT
) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY
))
1243 conswitchp
= &vga_con
;
1244 #elif defined(CONFIG_DUMMY_CONSOLE)
1245 conswitchp
= &dummy_con
;
1248 x86_init
.oem
.banner();
1250 x86_init
.timers
.wallclock_init();
1254 arch_init_ideal_nops();
1256 register_refined_jiffies(CLOCK_TICK_RATE
);
1259 if (efi_enabled(EFI_BOOT
))
1260 efi_apply_memmap_quirks();
1264 #ifdef CONFIG_X86_32
1266 static struct resource video_ram_resource
= {
1267 .name
= "Video RAM area",
1270 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
1273 void __init
i386_reserve_resources(void)
1275 request_resource(&iomem_resource
, &video_ram_resource
);
1276 reserve_standard_io_resources();
1279 #endif /* CONFIG_X86_32 */
1281 static struct notifier_block kernel_offset_notifier
= {
1282 .notifier_call
= dump_kernel_offset
1285 static int __init
register_kernel_offset_dumper(void)
1287 atomic_notifier_chain_register(&panic_notifier_list
,
1288 &kernel_offset_notifier
);
1291 __initcall(register_kernel_offset_dumper
);
1293 void arch_show_smap(struct seq_file
*m
, struct vm_area_struct
*vma
)
1295 if (!boot_cpu_has(X86_FEATURE_OSPKE
))
1298 seq_printf(m
, "ProtectionKey: %8u\n", vma_pkey(vma
));