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
;
150 /* cpu data as detected by the assembly code in head.S */
151 struct cpuinfo_x86 new_cpu_data
= {
154 /* common cpu data for all cpus */
155 struct cpuinfo_x86 boot_cpu_data __read_mostly
= {
158 EXPORT_SYMBOL(boot_cpu_data
);
160 unsigned int def_to_bigsmp
;
162 /* for MCA, but anyone else can use it if they want */
163 unsigned int machine_id
;
164 unsigned int machine_submodel_id
;
165 unsigned int BIOS_revision
;
167 struct apm_info apm_info
;
168 EXPORT_SYMBOL(apm_info
);
170 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
171 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
172 struct ist_info ist_info
;
173 EXPORT_SYMBOL(ist_info
);
175 struct ist_info ist_info
;
179 struct cpuinfo_x86 boot_cpu_data __read_mostly
= {
180 .x86_phys_bits
= MAX_PHYSMEM_BITS
,
182 EXPORT_SYMBOL(boot_cpu_data
);
186 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
187 __visible
unsigned long mmu_cr4_features
;
189 __visible
unsigned long mmu_cr4_features
= X86_CR4_PAE
;
192 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
193 int bootloader_type
, bootloader_version
;
198 struct screen_info screen_info
;
199 EXPORT_SYMBOL(screen_info
);
200 struct edid_info edid_info
;
201 EXPORT_SYMBOL_GPL(edid_info
);
203 extern int root_mountflags
;
205 unsigned long saved_video_mode
;
207 #define RAMDISK_IMAGE_START_MASK 0x07FF
208 #define RAMDISK_PROMPT_FLAG 0x8000
209 #define RAMDISK_LOAD_FLAG 0x4000
211 static char __initdata command_line
[COMMAND_LINE_SIZE
];
212 #ifdef CONFIG_CMDLINE_BOOL
213 static char __initdata builtin_cmdline
[COMMAND_LINE_SIZE
] = CONFIG_CMDLINE
;
216 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
218 #ifdef CONFIG_EDD_MODULE
222 * copy_edd() - Copy the BIOS EDD information
223 * from boot_params into a safe place.
226 static inline void __init
copy_edd(void)
228 memcpy(edd
.mbr_signature
, boot_params
.edd_mbr_sig_buffer
,
229 sizeof(edd
.mbr_signature
));
230 memcpy(edd
.edd_info
, boot_params
.eddbuf
, sizeof(edd
.edd_info
));
231 edd
.mbr_signature_nr
= boot_params
.edd_mbr_sig_buf_entries
;
232 edd
.edd_info_nr
= boot_params
.eddbuf_entries
;
235 static inline void __init
copy_edd(void)
240 void * __init
extend_brk(size_t size
, size_t align
)
242 size_t mask
= align
- 1;
245 BUG_ON(_brk_start
== 0);
246 BUG_ON(align
& mask
);
248 _brk_end
= (_brk_end
+ mask
) & ~mask
;
249 BUG_ON((char *)(_brk_end
+ size
) > __brk_limit
);
251 ret
= (void *)_brk_end
;
254 memset(ret
, 0, size
);
260 static void __init
cleanup_highmap(void)
265 static void __init
reserve_brk(void)
267 if (_brk_end
> _brk_start
)
268 memblock_reserve(__pa_symbol(_brk_start
),
269 _brk_end
- _brk_start
);
271 /* Mark brk area as locked down and no longer taking any
276 u64 relocated_ramdisk
;
278 #ifdef CONFIG_BLK_DEV_INITRD
280 static u64 __init
get_ramdisk_image(void)
282 u64 ramdisk_image
= boot_params
.hdr
.ramdisk_image
;
284 ramdisk_image
|= (u64
)boot_params
.ext_ramdisk_image
<< 32;
286 return ramdisk_image
;
288 static u64 __init
get_ramdisk_size(void)
290 u64 ramdisk_size
= boot_params
.hdr
.ramdisk_size
;
292 ramdisk_size
|= (u64
)boot_params
.ext_ramdisk_size
<< 32;
297 static void __init
relocate_initrd(void)
299 /* Assume only end is not page aligned */
300 u64 ramdisk_image
= get_ramdisk_image();
301 u64 ramdisk_size
= get_ramdisk_size();
302 u64 area_size
= PAGE_ALIGN(ramdisk_size
);
304 /* We need to move the initrd down into directly mapped mem */
305 relocated_ramdisk
= memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped
),
306 area_size
, PAGE_SIZE
);
308 if (!relocated_ramdisk
)
309 panic("Cannot find place for new RAMDISK of size %lld\n",
312 /* Note: this includes all the mem currently occupied by
313 the initrd, we rely on that fact to keep the data intact. */
314 memblock_reserve(relocated_ramdisk
, area_size
);
315 initrd_start
= relocated_ramdisk
+ PAGE_OFFSET
;
316 initrd_end
= initrd_start
+ ramdisk_size
;
317 printk(KERN_INFO
"Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
318 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
320 copy_from_early_mem((void *)initrd_start
, ramdisk_image
, ramdisk_size
);
322 printk(KERN_INFO
"Move RAMDISK from [mem %#010llx-%#010llx] to"
323 " [mem %#010llx-%#010llx]\n",
324 ramdisk_image
, ramdisk_image
+ ramdisk_size
- 1,
325 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
328 static void __init
early_reserve_initrd(void)
330 /* Assume only end is not page aligned */
331 u64 ramdisk_image
= get_ramdisk_image();
332 u64 ramdisk_size
= get_ramdisk_size();
333 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
335 if (!boot_params
.hdr
.type_of_loader
||
336 !ramdisk_image
|| !ramdisk_size
)
337 return; /* No initrd provided by bootloader */
339 memblock_reserve(ramdisk_image
, ramdisk_end
- ramdisk_image
);
341 static void __init
reserve_initrd(void)
343 /* Assume only end is not page aligned */
344 u64 ramdisk_image
= get_ramdisk_image();
345 u64 ramdisk_size
= get_ramdisk_size();
346 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
349 if (!boot_params
.hdr
.type_of_loader
||
350 !ramdisk_image
|| !ramdisk_size
)
351 return; /* No initrd provided by bootloader */
355 mapped_size
= memblock_mem_size(max_pfn_mapped
);
356 if (ramdisk_size
>= (mapped_size
>>1))
357 panic("initrd too large to handle, "
358 "disabling initrd (%lld needed, %lld available)\n",
359 ramdisk_size
, mapped_size
>>1);
361 printk(KERN_INFO
"RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image
,
364 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image
),
365 PFN_DOWN(ramdisk_end
))) {
366 /* All are mapped, easy case */
367 initrd_start
= ramdisk_image
+ PAGE_OFFSET
;
368 initrd_end
= initrd_start
+ ramdisk_size
;
374 memblock_free(ramdisk_image
, ramdisk_end
- ramdisk_image
);
377 static void __init
early_reserve_initrd(void)
380 static void __init
reserve_initrd(void)
383 #endif /* CONFIG_BLK_DEV_INITRD */
385 static void __init
parse_setup_data(void)
387 struct setup_data
*data
;
388 u64 pa_data
, pa_next
;
390 pa_data
= boot_params
.hdr
.setup_data
;
392 u32 data_len
, data_type
;
394 data
= early_memremap(pa_data
, sizeof(*data
));
395 data_len
= data
->len
+ sizeof(struct setup_data
);
396 data_type
= data
->type
;
397 pa_next
= data
->next
;
398 early_memunmap(data
, sizeof(*data
));
402 parse_e820_ext(pa_data
, data_len
);
408 parse_efi_setup(pa_data
, data_len
);
417 static void __init
e820_reserve_setup_data(void)
419 struct setup_data
*data
;
422 pa_data
= boot_params
.hdr
.setup_data
;
427 data
= early_memremap(pa_data
, sizeof(*data
));
428 e820_update_range(pa_data
, sizeof(*data
)+data
->len
,
429 E820_RAM
, E820_RESERVED_KERN
);
430 pa_data
= data
->next
;
431 early_memunmap(data
, sizeof(*data
));
434 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
435 memcpy(&e820_saved
, &e820
, sizeof(struct e820map
));
436 printk(KERN_INFO
"extended physical RAM map:\n");
437 e820_print_map("reserve setup_data");
440 static void __init
memblock_x86_reserve_range_setup_data(void)
442 struct setup_data
*data
;
445 pa_data
= boot_params
.hdr
.setup_data
;
447 data
= early_memremap(pa_data
, sizeof(*data
));
448 memblock_reserve(pa_data
, sizeof(*data
) + data
->len
);
449 pa_data
= data
->next
;
450 early_memunmap(data
, sizeof(*data
));
455 * --------- Crashkernel reservation ------------------------------
458 #ifdef CONFIG_KEXEC_CORE
460 /* 16M alignment for crash kernel regions */
461 #define CRASH_ALIGN (16 << 20)
464 * Keep the crash kernel below this limit. On 32 bits earlier kernels
465 * would limit the kernel to the low 512 MiB due to mapping restrictions.
466 * On 64bit, old kexec-tools need to under 896MiB.
469 # define CRASH_ADDR_LOW_MAX (512 << 20)
470 # define CRASH_ADDR_HIGH_MAX (512 << 20)
472 # define CRASH_ADDR_LOW_MAX (896UL << 20)
473 # define CRASH_ADDR_HIGH_MAX MAXMEM
476 static int __init
reserve_crashkernel_low(void)
479 unsigned long long base
, low_base
= 0, low_size
= 0;
480 unsigned long total_low_mem
;
483 total_low_mem
= memblock_mem_size(1UL << (32 - PAGE_SHIFT
));
485 /* crashkernel=Y,low */
486 ret
= parse_crashkernel_low(boot_command_line
, total_low_mem
, &low_size
, &base
);
489 * two parts from lib/swiotlb.c:
490 * -swiotlb size: user-specified with swiotlb= or default.
492 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
493 * to 8M for other buffers that may need to stay low too. Also
494 * make sure we allocate enough extra low memory so that we
495 * don't run out of DMA buffers for 32-bit devices.
497 low_size
= max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
499 /* passed with crashkernel=0,low ? */
504 low_base
= memblock_find_in_range(low_size
, 1ULL << 32, low_size
, CRASH_ALIGN
);
506 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
507 (unsigned long)(low_size
>> 20));
511 ret
= memblock_reserve(low_base
, low_size
);
513 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__
);
517 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
518 (unsigned long)(low_size
>> 20),
519 (unsigned long)(low_base
>> 20),
520 (unsigned long)(total_low_mem
>> 20));
522 crashk_low_res
.start
= low_base
;
523 crashk_low_res
.end
= low_base
+ low_size
- 1;
524 insert_resource(&iomem_resource
, &crashk_low_res
);
529 static void __init
reserve_crashkernel(void)
531 unsigned long long crash_size
, crash_base
, total_mem
;
535 total_mem
= memblock_phys_mem_size();
538 ret
= parse_crashkernel(boot_command_line
, total_mem
, &crash_size
, &crash_base
);
539 if (ret
!= 0 || crash_size
<= 0) {
540 /* crashkernel=X,high */
541 ret
= parse_crashkernel_high(boot_command_line
, total_mem
,
542 &crash_size
, &crash_base
);
543 if (ret
!= 0 || crash_size
<= 0)
548 /* 0 means: find the address automatically */
549 if (crash_base
<= 0) {
551 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
553 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
554 high
? CRASH_ADDR_HIGH_MAX
555 : CRASH_ADDR_LOW_MAX
,
556 crash_size
, CRASH_ALIGN
);
558 pr_info("crashkernel reservation failed - No suitable area found.\n");
563 unsigned long long start
;
565 start
= memblock_find_in_range(crash_base
,
566 crash_base
+ crash_size
,
567 crash_size
, 1 << 20);
568 if (start
!= crash_base
) {
569 pr_info("crashkernel reservation failed - memory is in use.\n");
573 ret
= memblock_reserve(crash_base
, crash_size
);
575 pr_err("%s: Error reserving crashkernel memblock.\n", __func__
);
579 if (crash_base
>= (1ULL << 32) && reserve_crashkernel_low()) {
580 memblock_free(crash_base
, crash_size
);
584 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
585 (unsigned long)(crash_size
>> 20),
586 (unsigned long)(crash_base
>> 20),
587 (unsigned long)(total_mem
>> 20));
589 crashk_res
.start
= crash_base
;
590 crashk_res
.end
= crash_base
+ crash_size
- 1;
591 insert_resource(&iomem_resource
, &crashk_res
);
594 static void __init
reserve_crashkernel(void)
599 static struct resource standard_io_resources
[] = {
600 { .name
= "dma1", .start
= 0x00, .end
= 0x1f,
601 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
602 { .name
= "pic1", .start
= 0x20, .end
= 0x21,
603 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
604 { .name
= "timer0", .start
= 0x40, .end
= 0x43,
605 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
606 { .name
= "timer1", .start
= 0x50, .end
= 0x53,
607 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
608 { .name
= "keyboard", .start
= 0x60, .end
= 0x60,
609 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
610 { .name
= "keyboard", .start
= 0x64, .end
= 0x64,
611 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
612 { .name
= "dma page reg", .start
= 0x80, .end
= 0x8f,
613 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
614 { .name
= "pic2", .start
= 0xa0, .end
= 0xa1,
615 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
616 { .name
= "dma2", .start
= 0xc0, .end
= 0xdf,
617 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
618 { .name
= "fpu", .start
= 0xf0, .end
= 0xff,
619 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
}
622 void __init
reserve_standard_io_resources(void)
626 /* request I/O space for devices used on all i[345]86 PCs */
627 for (i
= 0; i
< ARRAY_SIZE(standard_io_resources
); i
++)
628 request_resource(&ioport_resource
, &standard_io_resources
[i
]);
632 static __init
void reserve_ibft_region(void)
634 unsigned long addr
, size
= 0;
636 addr
= find_ibft_region(&size
);
639 memblock_reserve(addr
, size
);
642 static bool __init
snb_gfx_workaround_needed(void)
647 static const __initconst u16 snb_ids
[] = {
657 /* Assume no if something weird is going on with PCI */
658 if (!early_pci_allowed())
661 vendor
= read_pci_config_16(0, 2, 0, PCI_VENDOR_ID
);
662 if (vendor
!= 0x8086)
665 devid
= read_pci_config_16(0, 2, 0, PCI_DEVICE_ID
);
666 for (i
= 0; i
< ARRAY_SIZE(snb_ids
); i
++)
667 if (devid
== snb_ids
[i
])
675 * Sandy Bridge graphics has trouble with certain ranges, exclude
676 * them from allocation.
678 static void __init
trim_snb_memory(void)
680 static const __initconst
unsigned long bad_pages
[] = {
689 if (!snb_gfx_workaround_needed())
692 printk(KERN_DEBUG
"reserving inaccessible SNB gfx pages\n");
695 * Reserve all memory below the 1 MB mark that has not
696 * already been reserved.
698 memblock_reserve(0, 1<<20);
700 for (i
= 0; i
< ARRAY_SIZE(bad_pages
); i
++) {
701 if (memblock_reserve(bad_pages
[i
], PAGE_SIZE
))
702 printk(KERN_WARNING
"failed to reserve 0x%08lx\n",
708 * Here we put platform-specific memory range workarounds, i.e.
709 * memory known to be corrupt or otherwise in need to be reserved on
710 * specific platforms.
712 * If this gets used more widely it could use a real dispatch mechanism.
714 static void __init
trim_platform_memory_ranges(void)
719 static void __init
trim_bios_range(void)
722 * A special case is the first 4Kb of memory;
723 * This is a BIOS owned area, not kernel ram, but generally
724 * not listed as such in the E820 table.
726 * This typically reserves additional memory (64KiB by default)
727 * since some BIOSes are known to corrupt low memory. See the
728 * Kconfig help text for X86_RESERVE_LOW.
730 e820_update_range(0, PAGE_SIZE
, E820_RAM
, E820_RESERVED
);
733 * special case: Some BIOSen report the PC BIOS
734 * area (640->1Mb) as ram even though it is not.
737 e820_remove_range(BIOS_BEGIN
, BIOS_END
- BIOS_BEGIN
, E820_RAM
, 1);
739 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
742 /* called before trim_bios_range() to spare extra sanitize */
743 static void __init
e820_add_kernel_range(void)
745 u64 start
= __pa_symbol(_text
);
746 u64 size
= __pa_symbol(_end
) - start
;
749 * Complain if .text .data and .bss are not marked as E820_RAM and
750 * attempt to fix it by adding the range. We may have a confused BIOS,
751 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
752 * exclude kernel range. If we really are running on top non-RAM,
753 * we will crash later anyways.
755 if (e820_all_mapped(start
, start
+ size
, E820_RAM
))
758 pr_warn(".text .data .bss are not marked as E820_RAM!\n");
759 e820_remove_range(start
, size
, E820_RAM
, 0);
760 e820_add_region(start
, size
, E820_RAM
);
763 static unsigned reserve_low
= CONFIG_X86_RESERVE_LOW
<< 10;
765 static int __init
parse_reservelow(char *p
)
767 unsigned long long size
;
772 size
= memparse(p
, &p
);
785 early_param("reservelow", parse_reservelow
);
787 static void __init
trim_low_memory_range(void)
789 memblock_reserve(0, ALIGN(reserve_low
, PAGE_SIZE
));
793 * Dump out kernel offset information on panic.
796 dump_kernel_offset(struct notifier_block
*self
, unsigned long v
, void *p
)
798 if (kaslr_enabled()) {
799 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
805 pr_emerg("Kernel Offset: disabled\n");
812 * Determine if we were loaded by an EFI loader. If so, then we have also been
813 * passed the efi memmap, systab, etc., so we should use these data structures
814 * for initialization. Note, the efi init code path is determined by the
815 * global efi_enabled. This allows the same kernel image to be used on existing
816 * systems (with a traditional BIOS) as well as on EFI systems.
819 * setup_arch - architecture-specific boot-time initializations
821 * Note: On x86_64, fixmaps are ready for use even before this is called.
824 void __init
setup_arch(char **cmdline_p
)
826 memblock_reserve(__pa_symbol(_text
),
827 (unsigned long)__bss_stop
- (unsigned long)_text
);
829 early_reserve_initrd();
832 * At this point everything still needed from the boot loader
833 * or BIOS or kernel text should be early reserved or marked not
834 * RAM in e820. All other memory is free game.
838 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
841 * copy kernel address range established so far and switch
842 * to the proper swapper page table
844 clone_pgd_range(swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
845 initial_page_table
+ KERNEL_PGD_BOUNDARY
,
848 load_cr3(swapper_pg_dir
);
850 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
851 * a cr3 based tlb flush, so the following __flush_tlb_all()
852 * will not flush anything because the cpu quirk which clears
853 * X86_FEATURE_PGE has not been invoked yet. Though due to the
854 * load_cr3() above the TLB has been flushed already. The
855 * quirk is invoked before subsequent calls to __flush_tlb_all()
856 * so proper operation is guaranteed.
860 printk(KERN_INFO
"Command line: %s\n", boot_command_line
);
864 * If we have OLPC OFW, we might end up relocating the fixmap due to
865 * reserve_top(), so do this before touching the ioremap area.
871 early_ioremap_init();
873 setup_olpc_ofw_pgd();
875 ROOT_DEV
= old_decode_dev(boot_params
.hdr
.root_dev
);
876 screen_info
= boot_params
.screen_info
;
877 edid_info
= boot_params
.edid_info
;
879 apm_info
.bios
= boot_params
.apm_bios_info
;
880 ist_info
= boot_params
.ist_info
;
882 saved_video_mode
= boot_params
.hdr
.vid_mode
;
883 bootloader_type
= boot_params
.hdr
.type_of_loader
;
884 if ((bootloader_type
>> 4) == 0xe) {
885 bootloader_type
&= 0xf;
886 bootloader_type
|= (boot_params
.hdr
.ext_loader_type
+0x10) << 4;
888 bootloader_version
= bootloader_type
& 0xf;
889 bootloader_version
|= boot_params
.hdr
.ext_loader_ver
<< 4;
891 #ifdef CONFIG_BLK_DEV_RAM
892 rd_image_start
= boot_params
.hdr
.ram_size
& RAMDISK_IMAGE_START_MASK
;
893 rd_prompt
= ((boot_params
.hdr
.ram_size
& RAMDISK_PROMPT_FLAG
) != 0);
894 rd_doload
= ((boot_params
.hdr
.ram_size
& RAMDISK_LOAD_FLAG
) != 0);
897 if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
898 EFI32_LOADER_SIGNATURE
, 4)) {
899 set_bit(EFI_BOOT
, &efi
.flags
);
900 } else if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
901 EFI64_LOADER_SIGNATURE
, 4)) {
902 set_bit(EFI_BOOT
, &efi
.flags
);
903 set_bit(EFI_64BIT
, &efi
.flags
);
906 if (efi_enabled(EFI_BOOT
))
907 efi_memblock_x86_reserve_range();
910 x86_init
.oem
.arch_setup();
912 iomem_resource
.end
= (1ULL << boot_cpu_data
.x86_phys_bits
) - 1;
918 if (!boot_params
.hdr
.root_flags
)
919 root_mountflags
&= ~MS_RDONLY
;
920 init_mm
.start_code
= (unsigned long) _text
;
921 init_mm
.end_code
= (unsigned long) _etext
;
922 init_mm
.end_data
= (unsigned long) _edata
;
923 init_mm
.brk
= _brk_end
;
925 mpx_mm_init(&init_mm
);
927 #ifdef CONFIG_CMDLINE_BOOL
928 #ifdef CONFIG_CMDLINE_OVERRIDE
929 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
931 if (builtin_cmdline
[0]) {
932 /* append boot loader cmdline to builtin */
933 strlcat(builtin_cmdline
, " ", COMMAND_LINE_SIZE
);
934 strlcat(builtin_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
935 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
940 strlcpy(command_line
, boot_command_line
, COMMAND_LINE_SIZE
);
941 *cmdline_p
= command_line
;
944 * x86_configure_nx() is called before parse_early_param() to detect
945 * whether hardware doesn't support NX (so that the early EHCI debug
946 * console setup can safely call set_fixmap()). It may then be called
947 * again from within noexec_setup() during parsing early parameters
948 * to honor the respective command line option.
956 /* after early param, so could get panic from serial */
957 memblock_x86_reserve_range_setup_data();
959 if (acpi_mps_check()) {
960 #ifdef CONFIG_X86_LOCAL_APIC
963 setup_clear_cpu_cap(X86_FEATURE_APIC
);
967 if (pci_early_dump_regs
)
968 early_dump_pci_devices();
971 /* update the e820_saved too */
972 e820_reserve_setup_data();
973 finish_e820_parsing();
975 if (efi_enabled(EFI_BOOT
))
980 dmi_set_dump_stack_arch_desc();
983 * VMware detection requires dmi to be available, so this
984 * needs to be done after dmi_scan_machine, for the BP.
986 init_hypervisor_platform();
988 x86_init
.resources
.probe_roms();
990 e820_add_kernel_range();
993 if (ppro_with_ram_bug()) {
994 e820_update_range(0x70000000ULL
, 0x40000ULL
, E820_RAM
,
996 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
997 printk(KERN_INFO
"fixed physical RAM map:\n");
998 e820_print_map("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
;
1017 #ifdef CONFIG_X86_32
1018 /* max_low_pfn get updated here */
1019 find_low_pfn_range();
1023 /* How many end-of-memory variables you have, grandma! */
1024 /* need this before calling reserve_initrd */
1025 if (max_pfn
> (1UL<<(32 - PAGE_SHIFT
)))
1026 max_low_pfn
= e820_end_of_low_ram_pfn();
1028 max_low_pfn
= max_pfn
;
1030 high_memory
= (void *)__va(max_pfn
* PAGE_SIZE
- 1) + 1;
1034 * Find and reserve possible boot-time SMP configuration:
1038 reserve_ibft_region();
1040 early_alloc_pgt_buf();
1043 * Need to conclude brk, before memblock_x86_fill()
1044 * it could use memblock_find_in_range, could overlap with
1051 memblock_set_current_limit(ISA_END_ADDRESS
);
1052 memblock_x86_fill();
1054 if (efi_enabled(EFI_BOOT
)) {
1060 * The EFI specification says that boot service code won't be called
1061 * after ExitBootServices(). This is, in fact, a lie.
1063 if (efi_enabled(EFI_MEMMAP
))
1064 efi_reserve_boot_services();
1066 /* preallocate 4k for mptable mpc */
1067 early_reserve_e820_mpc_new();
1069 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1070 setup_bios_corruption_check();
1073 #ifdef CONFIG_X86_32
1074 printk(KERN_DEBUG
"initial memory mapped: [mem 0x00000000-%#010lx]\n",
1075 (max_pfn_mapped
<<PAGE_SHIFT
) - 1);
1078 reserve_real_mode();
1080 trim_platform_memory_ranges();
1081 trim_low_memory_range();
1085 early_trap_pf_init();
1089 memblock_set_current_limit(get_max_mapped());
1092 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1095 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1096 if (init_ohci1394_dma_early
)
1097 init_ohci1394_dma_on_all_controllers();
1099 /* Allocate bigger log buffer */
1104 #if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
1105 acpi_initrd_override((void *)initrd_start
, initrd_end
- initrd_start
);
1113 * Parse the ACPI tables for possible boot-time SMP configuration.
1115 acpi_boot_table_init();
1117 early_acpi_boot_init();
1120 dma_contiguous_reserve(max_pfn_mapped
<< PAGE_SHIFT
);
1123 * Reserve memory for crash kernel after SRAT is parsed so that it
1124 * won't consume hotpluggable memory.
1126 reserve_crashkernel();
1128 memblock_find_dma_reserve();
1130 #ifdef CONFIG_KVM_GUEST
1134 x86_init
.paging
.pagetable_init();
1138 if (boot_cpu_data
.cpuid_level
>= 0) {
1139 /* A CPU has %cr4 if and only if it has CPUID */
1140 mmu_cr4_features
= __read_cr4();
1141 if (trampoline_cr4_features
)
1142 *trampoline_cr4_features
= mmu_cr4_features
;
1145 #ifdef CONFIG_X86_32
1146 /* sync back kernel address range */
1147 clone_pgd_range(initial_page_table
+ KERNEL_PGD_BOUNDARY
,
1148 swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
1152 * sync back low identity map too. It is used for example
1153 * in the 32-bit EFI stub.
1155 clone_pgd_range(initial_page_table
,
1156 swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
1157 min(KERNEL_PGD_PTRS
, KERNEL_PGD_BOUNDARY
));
1164 generic_apic_probe();
1169 * Read APIC and some other early information from ACPI tables.
1176 * get boot-time SMP configuration:
1178 if (smp_found_config
)
1181 prefill_possible_map();
1185 init_apic_mappings();
1186 io_apic_init_mappings();
1190 e820_reserve_resources();
1191 e820_mark_nosave_regions(max_low_pfn
);
1193 x86_init
.resources
.reserve_resources();
1198 #if defined(CONFIG_VGA_CONSOLE)
1199 if (!efi_enabled(EFI_BOOT
) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY
))
1200 conswitchp
= &vga_con
;
1201 #elif defined(CONFIG_DUMMY_CONSOLE)
1202 conswitchp
= &dummy_con
;
1205 x86_init
.oem
.banner();
1207 x86_init
.timers
.wallclock_init();
1211 arch_init_ideal_nops();
1213 register_refined_jiffies(CLOCK_TICK_RATE
);
1216 if (efi_enabled(EFI_BOOT
))
1217 efi_apply_memmap_quirks();
1221 #ifdef CONFIG_X86_32
1223 static struct resource video_ram_resource
= {
1224 .name
= "Video RAM area",
1227 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
1230 void __init
i386_reserve_resources(void)
1232 request_resource(&iomem_resource
, &video_ram_resource
);
1233 reserve_standard_io_resources();
1236 #endif /* CONFIG_X86_32 */
1238 static struct notifier_block kernel_offset_notifier
= {
1239 .notifier_call
= dump_kernel_offset
1242 static int __init
register_kernel_offset_dumper(void)
1244 atomic_notifier_chain_register(&panic_notifier_list
,
1245 &kernel_offset_notifier
);
1248 __initcall(register_kernel_offset_dumper
);
1250 void arch_show_smap(struct seq_file
*m
, struct vm_area_struct
*vma
)
1252 if (!boot_cpu_has(X86_FEATURE_OSPKE
))
1255 seq_printf(m
, "ProtectionKey: %8u\n", vma_pkey(vma
));