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/export.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>
72 #include <linux/security.h>
74 #include <linux/fips.h>
75 #include <linux/cred.h>
76 #include <linux/sysrq.h>
77 #include <linux/init_task.h>
79 #include <linux/usb/xhci-dbgp.h>
80 #include <video/edid.h>
84 #include <asm/realmode.h>
85 #include <asm/e820/api.h>
86 #include <asm/mpspec.h>
87 #include <asm/setup.h>
89 #include <asm/timer.h>
90 #include <asm/i8259.h>
91 #include <asm/sections.h>
92 #include <asm/io_apic.h>
94 #include <asm/setup_arch.h>
95 #include <asm/bios_ebda.h>
96 #include <asm/cacheflush.h>
97 #include <asm/processor.h>
99 #include <asm/kasan.h>
101 #include <asm/vsyscall.h>
103 #include <asm/desc.h>
105 #include <asm/iommu.h>
106 #include <asm/gart.h>
107 #include <asm/mmu_context.h>
108 #include <asm/proto.h>
110 #include <asm/paravirt.h>
111 #include <asm/hypervisor.h>
112 #include <asm/olpc_ofw.h>
114 #include <asm/percpu.h>
115 #include <asm/topology.h>
116 #include <asm/apicdef.h>
117 #include <asm/amd_nb.h>
119 #include <asm/alternative.h>
120 #include <asm/prom.h>
121 #include <asm/microcode.h>
122 #include <asm/mmu_context.h>
123 #include <asm/kaslr.h>
126 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
127 * max_pfn_mapped: highest direct mapped pfn over 4GB
129 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
130 * represented by pfn_mapped
132 unsigned long max_low_pfn_mapped
;
133 unsigned long max_pfn_mapped
;
136 RESERVE_BRK(dmi_alloc
, 65536);
140 static __initdata
unsigned long _brk_start
= (unsigned long)__brk_base
;
141 unsigned long _brk_end
= (unsigned long)__brk_base
;
144 int default_cpu_present_to_apicid(int mps_cpu
)
146 return __default_cpu_present_to_apicid(mps_cpu
);
149 int default_check_phys_apicid_present(int phys_apicid
)
151 return __default_check_phys_apicid_present(phys_apicid
);
155 struct boot_params boot_params
;
160 static struct resource data_resource
= {
161 .name
= "Kernel data",
164 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
167 static struct resource code_resource
= {
168 .name
= "Kernel code",
171 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
174 static struct resource bss_resource
= {
175 .name
= "Kernel bss",
178 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
183 /* cpu data as detected by the assembly code in head_32.S */
184 struct cpuinfo_x86 new_cpu_data
;
186 /* common cpu data for all cpus */
187 struct cpuinfo_x86 boot_cpu_data __read_mostly
;
188 EXPORT_SYMBOL(boot_cpu_data
);
190 unsigned int def_to_bigsmp
;
192 /* for MCA, but anyone else can use it if they want */
193 unsigned int machine_id
;
194 unsigned int machine_submodel_id
;
195 unsigned int BIOS_revision
;
197 struct apm_info apm_info
;
198 EXPORT_SYMBOL(apm_info
);
200 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
201 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
202 struct ist_info ist_info
;
203 EXPORT_SYMBOL(ist_info
);
205 struct ist_info ist_info
;
209 struct cpuinfo_x86 boot_cpu_data __read_mostly
= {
210 .x86_phys_bits
= MAX_PHYSMEM_BITS
,
212 EXPORT_SYMBOL(boot_cpu_data
);
216 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
217 __visible
unsigned long mmu_cr4_features __ro_after_init
;
219 __visible
unsigned long mmu_cr4_features __ro_after_init
= X86_CR4_PAE
;
222 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
223 int bootloader_type
, bootloader_version
;
228 struct screen_info screen_info
;
229 EXPORT_SYMBOL(screen_info
);
230 struct edid_info edid_info
;
231 EXPORT_SYMBOL_GPL(edid_info
);
233 extern int root_mountflags
;
235 unsigned long saved_video_mode
;
237 #define RAMDISK_IMAGE_START_MASK 0x07FF
238 #define RAMDISK_PROMPT_FLAG 0x8000
239 #define RAMDISK_LOAD_FLAG 0x4000
241 static char __initdata command_line
[COMMAND_LINE_SIZE
];
242 #ifdef CONFIG_CMDLINE_BOOL
243 static char __initdata builtin_cmdline
[COMMAND_LINE_SIZE
] = CONFIG_CMDLINE
;
246 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
248 #ifdef CONFIG_EDD_MODULE
252 * copy_edd() - Copy the BIOS EDD information
253 * from boot_params into a safe place.
256 static inline void __init
copy_edd(void)
258 memcpy(edd
.mbr_signature
, boot_params
.edd_mbr_sig_buffer
,
259 sizeof(edd
.mbr_signature
));
260 memcpy(edd
.edd_info
, boot_params
.eddbuf
, sizeof(edd
.edd_info
));
261 edd
.mbr_signature_nr
= boot_params
.edd_mbr_sig_buf_entries
;
262 edd
.edd_info_nr
= boot_params
.eddbuf_entries
;
265 static inline void __init
copy_edd(void)
270 void * __init
extend_brk(size_t size
, size_t align
)
272 size_t mask
= align
- 1;
275 BUG_ON(_brk_start
== 0);
276 BUG_ON(align
& mask
);
278 _brk_end
= (_brk_end
+ mask
) & ~mask
;
279 BUG_ON((char *)(_brk_end
+ size
) > __brk_limit
);
281 ret
= (void *)_brk_end
;
284 memset(ret
, 0, size
);
290 static void __init
cleanup_highmap(void)
295 static void __init
reserve_brk(void)
297 if (_brk_end
> _brk_start
)
298 memblock_reserve(__pa_symbol(_brk_start
),
299 _brk_end
- _brk_start
);
301 /* Mark brk area as locked down and no longer taking any
306 u64 relocated_ramdisk
;
308 #ifdef CONFIG_BLK_DEV_INITRD
310 static u64 __init
get_ramdisk_image(void)
312 u64 ramdisk_image
= boot_params
.hdr
.ramdisk_image
;
314 ramdisk_image
|= (u64
)boot_params
.ext_ramdisk_image
<< 32;
316 return ramdisk_image
;
318 static u64 __init
get_ramdisk_size(void)
320 u64 ramdisk_size
= boot_params
.hdr
.ramdisk_size
;
322 ramdisk_size
|= (u64
)boot_params
.ext_ramdisk_size
<< 32;
327 static void __init
relocate_initrd(void)
329 /* Assume only end is not page aligned */
330 u64 ramdisk_image
= get_ramdisk_image();
331 u64 ramdisk_size
= get_ramdisk_size();
332 u64 area_size
= PAGE_ALIGN(ramdisk_size
);
334 /* We need to move the initrd down into directly mapped mem */
335 relocated_ramdisk
= memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped
),
336 area_size
, PAGE_SIZE
);
338 if (!relocated_ramdisk
)
339 panic("Cannot find place for new RAMDISK of size %lld\n",
342 /* Note: this includes all the mem currently occupied by
343 the initrd, we rely on that fact to keep the data intact. */
344 memblock_reserve(relocated_ramdisk
, area_size
);
345 initrd_start
= relocated_ramdisk
+ PAGE_OFFSET
;
346 initrd_end
= initrd_start
+ ramdisk_size
;
347 printk(KERN_INFO
"Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
348 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
350 copy_from_early_mem((void *)initrd_start
, ramdisk_image
, ramdisk_size
);
352 printk(KERN_INFO
"Move RAMDISK from [mem %#010llx-%#010llx] to"
353 " [mem %#010llx-%#010llx]\n",
354 ramdisk_image
, ramdisk_image
+ ramdisk_size
- 1,
355 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
358 static void __init
early_reserve_initrd(void)
360 /* Assume only end is not page aligned */
361 u64 ramdisk_image
= get_ramdisk_image();
362 u64 ramdisk_size
= get_ramdisk_size();
363 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
365 if (!boot_params
.hdr
.type_of_loader
||
366 !ramdisk_image
|| !ramdisk_size
)
367 return; /* No initrd provided by bootloader */
369 memblock_reserve(ramdisk_image
, ramdisk_end
- ramdisk_image
);
371 static void __init
reserve_initrd(void)
373 /* Assume only end is not page aligned */
374 u64 ramdisk_image
= get_ramdisk_image();
375 u64 ramdisk_size
= get_ramdisk_size();
376 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
379 if (!boot_params
.hdr
.type_of_loader
||
380 !ramdisk_image
|| !ramdisk_size
)
381 return; /* No initrd provided by bootloader */
385 mapped_size
= memblock_mem_size(max_pfn_mapped
);
386 if (ramdisk_size
>= (mapped_size
>>1))
387 panic("initrd too large to handle, "
388 "disabling initrd (%lld needed, %lld available)\n",
389 ramdisk_size
, mapped_size
>>1);
391 printk(KERN_INFO
"RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image
,
394 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image
),
395 PFN_DOWN(ramdisk_end
))) {
396 /* All are mapped, easy case */
397 initrd_start
= ramdisk_image
+ PAGE_OFFSET
;
398 initrd_end
= initrd_start
+ ramdisk_size
;
404 memblock_free(ramdisk_image
, ramdisk_end
- ramdisk_image
);
408 static void __init
early_reserve_initrd(void)
411 static void __init
reserve_initrd(void)
414 #endif /* CONFIG_BLK_DEV_INITRD */
416 static void __init
parse_setup_data(void)
418 struct setup_data
*data
;
419 u64 pa_data
, pa_next
;
421 pa_data
= boot_params
.hdr
.setup_data
;
423 u32 data_len
, data_type
;
425 data
= early_memremap(pa_data
, sizeof(*data
));
426 data_len
= data
->len
+ sizeof(struct setup_data
);
427 data_type
= data
->type
;
428 pa_next
= data
->next
;
429 early_memunmap(data
, sizeof(*data
));
433 e820__memory_setup_extended(pa_data
, data_len
);
439 parse_efi_setup(pa_data
, data_len
);
448 static void __init
memblock_x86_reserve_range_setup_data(void)
450 struct setup_data
*data
;
453 pa_data
= boot_params
.hdr
.setup_data
;
455 data
= early_memremap(pa_data
, sizeof(*data
));
456 memblock_reserve(pa_data
, sizeof(*data
) + data
->len
);
457 pa_data
= data
->next
;
458 early_memunmap(data
, sizeof(*data
));
463 * --------- Crashkernel reservation ------------------------------
466 #ifdef CONFIG_KEXEC_CORE
468 /* 16M alignment for crash kernel regions */
469 #define CRASH_ALIGN (16 << 20)
472 * Keep the crash kernel below this limit. On 32 bits earlier kernels
473 * would limit the kernel to the low 512 MiB due to mapping restrictions.
474 * On 64bit, old kexec-tools need to under 896MiB.
477 # define CRASH_ADDR_LOW_MAX (512 << 20)
478 # define CRASH_ADDR_HIGH_MAX (512 << 20)
480 # define CRASH_ADDR_LOW_MAX (896UL << 20)
481 # define CRASH_ADDR_HIGH_MAX MAXMEM
484 static int __init
reserve_crashkernel_low(void)
487 unsigned long long base
, low_base
= 0, low_size
= 0;
488 unsigned long total_low_mem
;
491 total_low_mem
= memblock_mem_size(1UL << (32 - PAGE_SHIFT
));
493 /* crashkernel=Y,low */
494 ret
= parse_crashkernel_low(boot_command_line
, total_low_mem
, &low_size
, &base
);
497 * two parts from lib/swiotlb.c:
498 * -swiotlb size: user-specified with swiotlb= or default.
500 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
501 * to 8M for other buffers that may need to stay low too. Also
502 * make sure we allocate enough extra low memory so that we
503 * don't run out of DMA buffers for 32-bit devices.
505 low_size
= max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
507 /* passed with crashkernel=0,low ? */
512 low_base
= memblock_find_in_range(0, 1ULL << 32, low_size
, CRASH_ALIGN
);
514 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
515 (unsigned long)(low_size
>> 20));
519 ret
= memblock_reserve(low_base
, low_size
);
521 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__
);
525 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
526 (unsigned long)(low_size
>> 20),
527 (unsigned long)(low_base
>> 20),
528 (unsigned long)(total_low_mem
>> 20));
530 crashk_low_res
.start
= low_base
;
531 crashk_low_res
.end
= low_base
+ low_size
- 1;
532 insert_resource(&iomem_resource
, &crashk_low_res
);
537 static void __init
reserve_crashkernel(void)
539 unsigned long long crash_size
, crash_base
, total_mem
;
543 total_mem
= memblock_phys_mem_size();
546 ret
= parse_crashkernel(boot_command_line
, total_mem
, &crash_size
, &crash_base
);
547 if (ret
!= 0 || crash_size
<= 0) {
548 /* crashkernel=X,high */
549 ret
= parse_crashkernel_high(boot_command_line
, total_mem
,
550 &crash_size
, &crash_base
);
551 if (ret
!= 0 || crash_size
<= 0)
556 /* 0 means: find the address automatically */
557 if (crash_base
<= 0) {
559 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
560 * as old kexec-tools loads bzImage below that, unless
561 * "crashkernel=size[KMG],high" is specified.
563 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
564 high
? CRASH_ADDR_HIGH_MAX
565 : CRASH_ADDR_LOW_MAX
,
566 crash_size
, CRASH_ALIGN
);
568 pr_info("crashkernel reservation failed - No suitable area found.\n");
573 unsigned long long start
;
575 start
= memblock_find_in_range(crash_base
,
576 crash_base
+ crash_size
,
577 crash_size
, 1 << 20);
578 if (start
!= crash_base
) {
579 pr_info("crashkernel reservation failed - memory is in use.\n");
583 ret
= memblock_reserve(crash_base
, crash_size
);
585 pr_err("%s: Error reserving crashkernel memblock.\n", __func__
);
589 if (crash_base
>= (1ULL << 32) && reserve_crashkernel_low()) {
590 memblock_free(crash_base
, crash_size
);
594 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
595 (unsigned long)(crash_size
>> 20),
596 (unsigned long)(crash_base
>> 20),
597 (unsigned long)(total_mem
>> 20));
599 crashk_res
.start
= crash_base
;
600 crashk_res
.end
= crash_base
+ crash_size
- 1;
601 insert_resource(&iomem_resource
, &crashk_res
);
604 static void __init
reserve_crashkernel(void)
609 static struct resource standard_io_resources
[] = {
610 { .name
= "dma1", .start
= 0x00, .end
= 0x1f,
611 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
612 { .name
= "pic1", .start
= 0x20, .end
= 0x21,
613 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
614 { .name
= "timer0", .start
= 0x40, .end
= 0x43,
615 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
616 { .name
= "timer1", .start
= 0x50, .end
= 0x53,
617 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
618 { .name
= "keyboard", .start
= 0x60, .end
= 0x60,
619 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
620 { .name
= "keyboard", .start
= 0x64, .end
= 0x64,
621 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
622 { .name
= "dma page reg", .start
= 0x80, .end
= 0x8f,
623 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
624 { .name
= "pic2", .start
= 0xa0, .end
= 0xa1,
625 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
626 { .name
= "dma2", .start
= 0xc0, .end
= 0xdf,
627 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
628 { .name
= "fpu", .start
= 0xf0, .end
= 0xff,
629 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
}
632 void __init
reserve_standard_io_resources(void)
636 /* request I/O space for devices used on all i[345]86 PCs */
637 for (i
= 0; i
< ARRAY_SIZE(standard_io_resources
); i
++)
638 request_resource(&ioport_resource
, &standard_io_resources
[i
]);
642 static __init
void reserve_ibft_region(void)
644 unsigned long addr
, size
= 0;
646 addr
= find_ibft_region(&size
);
649 memblock_reserve(addr
, size
);
652 static bool __init
snb_gfx_workaround_needed(void)
657 static const __initconst u16 snb_ids
[] = {
667 /* Assume no if something weird is going on with PCI */
668 if (!early_pci_allowed())
671 vendor
= read_pci_config_16(0, 2, 0, PCI_VENDOR_ID
);
672 if (vendor
!= 0x8086)
675 devid
= read_pci_config_16(0, 2, 0, PCI_DEVICE_ID
);
676 for (i
= 0; i
< ARRAY_SIZE(snb_ids
); i
++)
677 if (devid
== snb_ids
[i
])
685 * Sandy Bridge graphics has trouble with certain ranges, exclude
686 * them from allocation.
688 static void __init
trim_snb_memory(void)
690 static const __initconst
unsigned long bad_pages
[] = {
699 if (!snb_gfx_workaround_needed())
702 printk(KERN_DEBUG
"reserving inaccessible SNB gfx pages\n");
705 * Reserve all memory below the 1 MB mark that has not
706 * already been reserved.
708 memblock_reserve(0, 1<<20);
710 for (i
= 0; i
< ARRAY_SIZE(bad_pages
); i
++) {
711 if (memblock_reserve(bad_pages
[i
], PAGE_SIZE
))
712 printk(KERN_WARNING
"failed to reserve 0x%08lx\n",
718 * Here we put platform-specific memory range workarounds, i.e.
719 * memory known to be corrupt or otherwise in need to be reserved on
720 * specific platforms.
722 * If this gets used more widely it could use a real dispatch mechanism.
724 static void __init
trim_platform_memory_ranges(void)
729 static void __init
trim_bios_range(void)
732 * A special case is the first 4Kb of memory;
733 * This is a BIOS owned area, not kernel ram, but generally
734 * not listed as such in the E820 table.
736 * This typically reserves additional memory (64KiB by default)
737 * since some BIOSes are known to corrupt low memory. See the
738 * Kconfig help text for X86_RESERVE_LOW.
740 e820__range_update(0, PAGE_SIZE
, E820_TYPE_RAM
, E820_TYPE_RESERVED
);
743 * special case: Some BIOSen report the PC BIOS
744 * area (640->1Mb) as ram even though it is not.
747 e820__range_remove(BIOS_BEGIN
, BIOS_END
- BIOS_BEGIN
, E820_TYPE_RAM
, 1);
749 e820__update_table(e820_table
);
752 /* called before trim_bios_range() to spare extra sanitize */
753 static void __init
e820_add_kernel_range(void)
755 u64 start
= __pa_symbol(_text
);
756 u64 size
= __pa_symbol(_end
) - start
;
759 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
760 * attempt to fix it by adding the range. We may have a confused BIOS,
761 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
762 * exclude kernel range. If we really are running on top non-RAM,
763 * we will crash later anyways.
765 if (e820__mapped_all(start
, start
+ size
, E820_TYPE_RAM
))
768 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
769 e820__range_remove(start
, size
, E820_TYPE_RAM
, 0);
770 e820__range_add(start
, size
, E820_TYPE_RAM
);
773 static unsigned reserve_low
= CONFIG_X86_RESERVE_LOW
<< 10;
775 static int __init
parse_reservelow(char *p
)
777 unsigned long long size
;
782 size
= memparse(p
, &p
);
795 early_param("reservelow", parse_reservelow
);
797 static void __init
trim_low_memory_range(void)
799 memblock_reserve(0, ALIGN(reserve_low
, PAGE_SIZE
));
803 * Dump out kernel offset information on panic.
806 dump_kernel_offset(struct notifier_block
*self
, unsigned long v
, void *p
)
808 if (kaslr_enabled()) {
809 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
815 pr_emerg("Kernel Offset: disabled\n");
821 static void __init
simple_udelay_calibration(void)
823 unsigned int tsc_khz
, cpu_khz
;
826 if (!boot_cpu_has(X86_FEATURE_TSC
))
829 cpu_khz
= x86_platform
.calibrate_cpu();
830 tsc_khz
= x86_platform
.calibrate_tsc();
832 tsc_khz
= tsc_khz
? : cpu_khz
;
836 lpj
= tsc_khz
* 1000;
838 loops_per_jiffy
= lpj
;
842 * Determine if we were loaded by an EFI loader. If so, then we have also been
843 * passed the efi memmap, systab, etc., so we should use these data structures
844 * for initialization. Note, the efi init code path is determined by the
845 * global efi_enabled. This allows the same kernel image to be used on existing
846 * systems (with a traditional BIOS) as well as on EFI systems.
849 * setup_arch - architecture-specific boot-time initializations
851 * Note: On x86_64, fixmaps are ready for use even before this is called.
854 void __init
setup_arch(char **cmdline_p
)
856 memblock_reserve(__pa_symbol(_text
),
857 (unsigned long)__bss_stop
- (unsigned long)_text
);
859 early_reserve_initrd();
862 * At this point everything still needed from the boot loader
863 * or BIOS or kernel text should be early reserved or marked not
864 * RAM in e820. All other memory is free game.
868 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
871 * copy kernel address range established so far and switch
872 * to the proper swapper page table
874 clone_pgd_range(swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
875 initial_page_table
+ KERNEL_PGD_BOUNDARY
,
878 load_cr3(swapper_pg_dir
);
880 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
881 * a cr3 based tlb flush, so the following __flush_tlb_all()
882 * will not flush anything because the cpu quirk which clears
883 * X86_FEATURE_PGE has not been invoked yet. Though due to the
884 * load_cr3() above the TLB has been flushed already. The
885 * quirk is invoked before subsequent calls to __flush_tlb_all()
886 * so proper operation is guaranteed.
890 printk(KERN_INFO
"Command line: %s\n", boot_command_line
);
894 * If we have OLPC OFW, we might end up relocating the fixmap due to
895 * reserve_top(), so do this before touching the ioremap area.
901 early_ioremap_init();
903 setup_olpc_ofw_pgd();
905 ROOT_DEV
= old_decode_dev(boot_params
.hdr
.root_dev
);
906 screen_info
= boot_params
.screen_info
;
907 edid_info
= boot_params
.edid_info
;
909 apm_info
.bios
= boot_params
.apm_bios_info
;
910 ist_info
= boot_params
.ist_info
;
912 saved_video_mode
= boot_params
.hdr
.vid_mode
;
913 bootloader_type
= boot_params
.hdr
.type_of_loader
;
914 if ((bootloader_type
>> 4) == 0xe) {
915 bootloader_type
&= 0xf;
916 bootloader_type
|= (boot_params
.hdr
.ext_loader_type
+0x10) << 4;
918 bootloader_version
= bootloader_type
& 0xf;
919 bootloader_version
|= boot_params
.hdr
.ext_loader_ver
<< 4;
921 #ifdef CONFIG_BLK_DEV_RAM
922 rd_image_start
= boot_params
.hdr
.ram_size
& RAMDISK_IMAGE_START_MASK
;
923 rd_prompt
= ((boot_params
.hdr
.ram_size
& RAMDISK_PROMPT_FLAG
) != 0);
924 rd_doload
= ((boot_params
.hdr
.ram_size
& RAMDISK_LOAD_FLAG
) != 0);
927 if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
928 EFI32_LOADER_SIGNATURE
, 4)) {
929 set_bit(EFI_BOOT
, &efi
.flags
);
930 } else if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
931 EFI64_LOADER_SIGNATURE
, 4)) {
932 set_bit(EFI_BOOT
, &efi
.flags
);
933 set_bit(EFI_64BIT
, &efi
.flags
);
936 if (efi_enabled(EFI_BOOT
))
937 efi_memblock_x86_reserve_range();
940 x86_init
.oem
.arch_setup();
942 iomem_resource
.end
= (1ULL << boot_cpu_data
.x86_phys_bits
) - 1;
943 e820__memory_setup();
948 if (!boot_params
.hdr
.root_flags
)
949 root_mountflags
&= ~MS_RDONLY
;
950 init_mm
.start_code
= (unsigned long) _text
;
951 init_mm
.end_code
= (unsigned long) _etext
;
952 init_mm
.end_data
= (unsigned long) _edata
;
953 init_mm
.brk
= _brk_end
;
955 mpx_mm_init(&init_mm
);
957 code_resource
.start
= __pa_symbol(_text
);
958 code_resource
.end
= __pa_symbol(_etext
)-1;
959 data_resource
.start
= __pa_symbol(_etext
);
960 data_resource
.end
= __pa_symbol(_edata
)-1;
961 bss_resource
.start
= __pa_symbol(__bss_start
);
962 bss_resource
.end
= __pa_symbol(__bss_stop
)-1;
964 #ifdef CONFIG_CMDLINE_BOOL
965 #ifdef CONFIG_CMDLINE_OVERRIDE
966 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
968 if (builtin_cmdline
[0]) {
969 /* append boot loader cmdline to builtin */
970 strlcat(builtin_cmdline
, " ", COMMAND_LINE_SIZE
);
971 strlcat(builtin_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
972 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
977 strlcpy(command_line
, boot_command_line
, COMMAND_LINE_SIZE
);
978 *cmdline_p
= command_line
;
981 * x86_configure_nx() is called before parse_early_param() to detect
982 * whether hardware doesn't support NX (so that the early EHCI debug
983 * console setup can safely call set_fixmap()). It may then be called
984 * again from within noexec_setup() during parsing early parameters
985 * to honor the respective command line option.
991 #ifdef CONFIG_MEMORY_HOTPLUG
993 * Memory used by the kernel cannot be hot-removed because Linux
994 * cannot migrate the kernel pages. When memory hotplug is
995 * enabled, we should prevent memblock from allocating memory
998 * ACPI SRAT records all hotpluggable memory ranges. But before
999 * SRAT is parsed, we don't know about it.
1001 * The kernel image is loaded into memory at very early time. We
1002 * cannot prevent this anyway. So on NUMA system, we set any
1003 * node the kernel resides in as un-hotpluggable.
1005 * Since on modern servers, one node could have double-digit
1006 * gigabytes memory, we can assume the memory around the kernel
1007 * image is also un-hotpluggable. So before SRAT is parsed, just
1008 * allocate memory near the kernel image to try the best to keep
1009 * the kernel away from hotpluggable memory.
1011 if (movable_node_is_enabled())
1012 memblock_set_bottom_up(true);
1017 /* after early param, so could get panic from serial */
1018 memblock_x86_reserve_range_setup_data();
1020 if (acpi_mps_check()) {
1021 #ifdef CONFIG_X86_LOCAL_APIC
1024 setup_clear_cpu_cap(X86_FEATURE_APIC
);
1028 if (pci_early_dump_regs
)
1029 early_dump_pci_devices();
1032 e820__reserve_setup_data();
1033 e820__finish_early_params();
1035 if (efi_enabled(EFI_BOOT
))
1040 dmi_set_dump_stack_arch_desc();
1043 * VMware detection requires dmi to be available, so this
1044 * needs to be done after dmi_scan_machine, for the BP.
1046 init_hypervisor_platform();
1048 simple_udelay_calibration();
1050 x86_init
.resources
.probe_roms();
1052 /* after parse_early_param, so could debug it */
1053 insert_resource(&iomem_resource
, &code_resource
);
1054 insert_resource(&iomem_resource
, &data_resource
);
1055 insert_resource(&iomem_resource
, &bss_resource
);
1057 e820_add_kernel_range();
1059 #ifdef CONFIG_X86_32
1060 if (ppro_with_ram_bug()) {
1061 e820__range_update(0x70000000ULL
, 0x40000ULL
, E820_TYPE_RAM
,
1062 E820_TYPE_RESERVED
);
1063 e820__update_table(e820_table
);
1064 printk(KERN_INFO
"fixed physical RAM map:\n");
1065 e820__print_table("bad_ppro");
1068 early_gart_iommu_check();
1072 * partially used pages are not usable - thus
1073 * we are rounding upwards:
1075 max_pfn
= e820__end_of_ram_pfn();
1077 /* update e820 for memory not covered by WB MTRRs */
1079 if (mtrr_trim_uncached_memory(max_pfn
))
1080 max_pfn
= e820__end_of_ram_pfn();
1082 max_possible_pfn
= max_pfn
;
1085 * This call is required when the CPU does not support PAT. If
1086 * mtrr_bp_init() invoked it already via pat_init() the call has no
1092 * Define random base addresses for memory sections after max_pfn is
1093 * defined and before each memory section base is used.
1095 kernel_randomize_memory();
1097 #ifdef CONFIG_X86_32
1098 /* max_low_pfn get updated here */
1099 find_low_pfn_range();
1103 /* How many end-of-memory variables you have, grandma! */
1104 /* need this before calling reserve_initrd */
1105 if (max_pfn
> (1UL<<(32 - PAGE_SHIFT
)))
1106 max_low_pfn
= e820__end_of_low_ram_pfn();
1108 max_low_pfn
= max_pfn
;
1110 high_memory
= (void *)__va(max_pfn
* PAGE_SIZE
- 1) + 1;
1114 * Find and reserve possible boot-time SMP configuration:
1118 reserve_ibft_region();
1120 early_alloc_pgt_buf();
1123 * Need to conclude brk, before e820__memblock_setup()
1124 * it could use memblock_find_in_range, could overlap with
1131 memblock_set_current_limit(ISA_END_ADDRESS
);
1132 e820__memblock_setup();
1134 if (!early_xdbc_setup_hardware())
1135 early_xdbc_register_console();
1137 reserve_bios_regions();
1139 if (efi_enabled(EFI_MEMMAP
)) {
1145 * The EFI specification says that boot service code won't be
1146 * called after ExitBootServices(). This is, in fact, a lie.
1148 efi_reserve_boot_services();
1151 /* preallocate 4k for mptable mpc */
1152 e820__memblock_alloc_reserved_mpc_new();
1154 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1155 setup_bios_corruption_check();
1158 #ifdef CONFIG_X86_32
1159 printk(KERN_DEBUG
"initial memory mapped: [mem 0x00000000-%#010lx]\n",
1160 (max_pfn_mapped
<<PAGE_SHIFT
) - 1);
1163 reserve_real_mode();
1165 trim_platform_memory_ranges();
1166 trim_low_memory_range();
1170 early_trap_pf_init();
1173 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1174 * with the current CR4 value. This may not be necessary, but
1175 * auditing all the early-boot CR4 manipulation would be needed to
1178 * Mask off features that don't work outside long mode (just
1181 mmu_cr4_features
= __read_cr4() & ~X86_CR4_PCIDE
;
1183 memblock_set_current_limit(get_max_mapped());
1186 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1189 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1190 if (init_ohci1394_dma_early
)
1191 init_ohci1394_dma_on_all_controllers();
1193 /* Allocate bigger log buffer */
1196 if (efi_enabled(EFI_BOOT
)) {
1197 switch (boot_params
.secure_boot
) {
1198 case efi_secureboot_mode_disabled
:
1199 pr_info("Secure boot disabled\n");
1201 case efi_secureboot_mode_enabled
:
1202 set_bit(EFI_SECURE_BOOT
, &efi
.flags
);
1203 if (IS_ENABLED(CONFIG_EFI_SECURE_BOOT_LOCK_DOWN
)) {
1205 pr_info("Secure boot enabled and kernel locked down\n");
1207 pr_info("Secure boot enabled\n");
1211 pr_info("Secure boot could not be determined\n");
1218 acpi_table_upgrade();
1225 * Parse the ACPI tables for possible boot-time SMP configuration.
1227 acpi_boot_table_init();
1229 early_acpi_boot_init();
1232 dma_contiguous_reserve(max_pfn_mapped
<< PAGE_SHIFT
);
1235 * Reserve memory for crash kernel after SRAT is parsed so that it
1236 * won't consume hotpluggable memory.
1238 reserve_crashkernel();
1240 memblock_find_dma_reserve();
1242 #ifdef CONFIG_KVM_GUEST
1246 x86_init
.paging
.pagetable_init();
1250 #ifdef CONFIG_X86_32
1251 /* sync back kernel address range */
1252 clone_pgd_range(initial_page_table
+ KERNEL_PGD_BOUNDARY
,
1253 swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
1257 * sync back low identity map too. It is used for example
1258 * in the 32-bit EFI stub.
1260 clone_pgd_range(initial_page_table
,
1261 swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
1262 min(KERNEL_PGD_PTRS
, KERNEL_PGD_BOUNDARY
));
1269 generic_apic_probe();
1274 * Read APIC and some other early information from ACPI tables.
1281 * get boot-time SMP configuration:
1286 * Systems w/o ACPI and mptables might not have it mapped the local
1287 * APIC yet, but prefill_possible_map() might need to access it.
1289 init_apic_mappings();
1291 prefill_possible_map();
1295 io_apic_init_mappings();
1299 e820__reserve_resources();
1300 e820__register_nosave_regions(max_low_pfn
);
1302 x86_init
.resources
.reserve_resources();
1304 e820__setup_pci_gap();
1307 #if defined(CONFIG_VGA_CONSOLE)
1308 if (!efi_enabled(EFI_BOOT
) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY
))
1309 conswitchp
= &vga_con
;
1310 #elif defined(CONFIG_DUMMY_CONSOLE)
1311 conswitchp
= &dummy_con
;
1314 x86_init
.oem
.banner();
1316 x86_init
.timers
.wallclock_init();
1320 arch_init_ideal_nops();
1322 register_refined_jiffies(CLOCK_TICK_RATE
);
1325 if (efi_enabled(EFI_BOOT
))
1326 efi_apply_memmap_quirks();
1330 #ifdef CONFIG_X86_32
1332 static struct resource video_ram_resource
= {
1333 .name
= "Video RAM area",
1336 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
1339 void __init
i386_reserve_resources(void)
1341 request_resource(&iomem_resource
, &video_ram_resource
);
1342 reserve_standard_io_resources();
1345 #endif /* CONFIG_X86_32 */
1347 #ifdef CONFIG_EFI_ALLOW_SECURE_BOOT_EXIT
1349 static void sysrq_handle_secure_boot(int key
)
1351 if (!efi_enabled(EFI_SECURE_BOOT
))
1354 pr_info("Secure boot disabled\n");
1355 lift_kernel_lockdown();
1357 static struct sysrq_key_op secure_boot_sysrq_op
= {
1358 .handler
= sysrq_handle_secure_boot
,
1359 .help_msg
= "unSB(x)",
1360 .action_msg
= "Disabling Secure Boot restrictions",
1361 .enable_mask
= SYSRQ_DISABLE_USERSPACE
,
1363 static int __init
secure_boot_sysrq(void)
1365 if (efi_enabled(EFI_SECURE_BOOT
))
1366 register_sysrq_key('x', &secure_boot_sysrq_op
);
1369 late_initcall(secure_boot_sysrq
);
1370 #endif /*CONFIG_EFI_ALLOW_SECURE_BOOT_EXIT*/
1373 static struct notifier_block kernel_offset_notifier
= {
1374 .notifier_call
= dump_kernel_offset
1377 static int __init
register_kernel_offset_dumper(void)
1379 atomic_notifier_chain_register(&panic_notifier_list
,
1380 &kernel_offset_notifier
);
1383 __initcall(register_kernel_offset_dumper
);
1385 void arch_show_smap(struct seq_file
*m
, struct vm_area_struct
*vma
)
1387 if (!boot_cpu_has(X86_FEATURE_OSPKE
))
1390 seq_printf(m
, "ProtectionKey: %8u\n", vma_pkey(vma
));