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/memblock.h>
34 #include <linux/seq_file.h>
35 #include <linux/console.h>
36 #include <linux/root_dev.h>
37 #include <linux/highmem.h>
38 #include <linux/export.h>
39 #include <linux/efi.h>
40 #include <linux/init.h>
41 #include <linux/edd.h>
42 #include <linux/iscsi_ibft.h>
43 #include <linux/nodemask.h>
44 #include <linux/kexec.h>
45 #include <linux/dmi.h>
46 #include <linux/pfn.h>
47 #include <linux/pci.h>
48 #include <asm/pci-direct.h>
49 #include <linux/init_ohci1394_dma.h>
50 #include <linux/kvm_para.h>
51 #include <linux/dma-contiguous.h>
53 #include <uapi/linux/mount.h>
55 #include <linux/errno.h>
56 #include <linux/kernel.h>
57 #include <linux/stddef.h>
58 #include <linux/unistd.h>
59 #include <linux/ptrace.h>
60 #include <linux/user.h>
61 #include <linux/delay.h>
63 #include <linux/kallsyms.h>
64 #include <linux/cpufreq.h>
65 #include <linux/dma-mapping.h>
66 #include <linux/ctype.h>
67 #include <linux/uaccess.h>
69 #include <linux/percpu.h>
70 #include <linux/crash_dump.h>
71 #include <linux/tboot.h>
72 #include <linux/jiffies.h>
73 #include <linux/mem_encrypt.h>
75 #include <linux/usb/xhci-dbgp.h>
76 #include <video/edid.h>
80 #include <asm/realmode.h>
81 #include <asm/e820/api.h>
82 #include <asm/mpspec.h>
83 #include <asm/setup.h>
85 #include <asm/timer.h>
86 #include <asm/i8259.h>
87 #include <asm/sections.h>
88 #include <asm/io_apic.h>
90 #include <asm/setup_arch.h>
91 #include <asm/bios_ebda.h>
92 #include <asm/cacheflush.h>
93 #include <asm/processor.h>
95 #include <asm/kasan.h>
97 #include <asm/vsyscall.h>
101 #include <asm/iommu.h>
102 #include <asm/gart.h>
103 #include <asm/mmu_context.h>
104 #include <asm/proto.h>
106 #include <asm/paravirt.h>
107 #include <asm/hypervisor.h>
108 #include <asm/olpc_ofw.h>
110 #include <asm/percpu.h>
111 #include <asm/topology.h>
112 #include <asm/apicdef.h>
113 #include <asm/amd_nb.h>
115 #include <asm/alternative.h>
116 #include <asm/prom.h>
117 #include <asm/microcode.h>
118 #include <asm/kaslr.h>
119 #include <asm/unwind.h>
122 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
123 * max_pfn_mapped: highest direct mapped pfn over 4GB
125 * The direct mapping only covers E820_TYPE_RAM regions, so the ranges and gaps are
126 * represented by pfn_mapped
128 unsigned long max_low_pfn_mapped
;
129 unsigned long max_pfn_mapped
;
132 RESERVE_BRK(dmi_alloc
, 65536);
136 static __initdata
unsigned long _brk_start
= (unsigned long)__brk_base
;
137 unsigned long _brk_end
= (unsigned long)__brk_base
;
139 struct boot_params boot_params
;
144 static struct resource data_resource
= {
145 .name
= "Kernel data",
148 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
151 static struct resource code_resource
= {
152 .name
= "Kernel code",
155 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
158 static struct resource bss_resource
= {
159 .name
= "Kernel bss",
162 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
167 /* cpu data as detected by the assembly code in head_32.S */
168 struct cpuinfo_x86 new_cpu_data
;
170 /* common cpu data for all cpus */
171 struct cpuinfo_x86 boot_cpu_data __read_mostly
;
172 EXPORT_SYMBOL(boot_cpu_data
);
174 unsigned int def_to_bigsmp
;
176 /* for MCA, but anyone else can use it if they want */
177 unsigned int machine_id
;
178 unsigned int machine_submodel_id
;
179 unsigned int BIOS_revision
;
181 struct apm_info apm_info
;
182 EXPORT_SYMBOL(apm_info
);
184 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
185 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
186 struct ist_info ist_info
;
187 EXPORT_SYMBOL(ist_info
);
189 struct ist_info ist_info
;
193 struct cpuinfo_x86 boot_cpu_data __read_mostly
;
194 EXPORT_SYMBOL(boot_cpu_data
);
198 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
199 __visible
unsigned long mmu_cr4_features __ro_after_init
;
201 __visible
unsigned long mmu_cr4_features __ro_after_init
= X86_CR4_PAE
;
204 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
205 int bootloader_type
, bootloader_version
;
210 struct screen_info screen_info
;
211 EXPORT_SYMBOL(screen_info
);
212 struct edid_info edid_info
;
213 EXPORT_SYMBOL_GPL(edid_info
);
215 extern int root_mountflags
;
217 unsigned long saved_video_mode
;
219 #define RAMDISK_IMAGE_START_MASK 0x07FF
220 #define RAMDISK_PROMPT_FLAG 0x8000
221 #define RAMDISK_LOAD_FLAG 0x4000
223 static char __initdata command_line
[COMMAND_LINE_SIZE
];
224 #ifdef CONFIG_CMDLINE_BOOL
225 static char __initdata builtin_cmdline
[COMMAND_LINE_SIZE
] = CONFIG_CMDLINE
;
228 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
230 #ifdef CONFIG_EDD_MODULE
234 * copy_edd() - Copy the BIOS EDD information
235 * from boot_params into a safe place.
238 static inline void __init
copy_edd(void)
240 memcpy(edd
.mbr_signature
, boot_params
.edd_mbr_sig_buffer
,
241 sizeof(edd
.mbr_signature
));
242 memcpy(edd
.edd_info
, boot_params
.eddbuf
, sizeof(edd
.edd_info
));
243 edd
.mbr_signature_nr
= boot_params
.edd_mbr_sig_buf_entries
;
244 edd
.edd_info_nr
= boot_params
.eddbuf_entries
;
247 static inline void __init
copy_edd(void)
252 void * __init
extend_brk(size_t size
, size_t align
)
254 size_t mask
= align
- 1;
257 BUG_ON(_brk_start
== 0);
258 BUG_ON(align
& mask
);
260 _brk_end
= (_brk_end
+ mask
) & ~mask
;
261 BUG_ON((char *)(_brk_end
+ size
) > __brk_limit
);
263 ret
= (void *)_brk_end
;
266 memset(ret
, 0, size
);
272 static void __init
cleanup_highmap(void)
277 static void __init
reserve_brk(void)
279 if (_brk_end
> _brk_start
)
280 memblock_reserve(__pa_symbol(_brk_start
),
281 _brk_end
- _brk_start
);
283 /* Mark brk area as locked down and no longer taking any
288 u64 relocated_ramdisk
;
290 #ifdef CONFIG_BLK_DEV_INITRD
292 static u64 __init
get_ramdisk_image(void)
294 u64 ramdisk_image
= boot_params
.hdr
.ramdisk_image
;
296 ramdisk_image
|= (u64
)boot_params
.ext_ramdisk_image
<< 32;
298 return ramdisk_image
;
300 static u64 __init
get_ramdisk_size(void)
302 u64 ramdisk_size
= boot_params
.hdr
.ramdisk_size
;
304 ramdisk_size
|= (u64
)boot_params
.ext_ramdisk_size
<< 32;
309 static void __init
relocate_initrd(void)
311 /* Assume only end is not page aligned */
312 u64 ramdisk_image
= get_ramdisk_image();
313 u64 ramdisk_size
= get_ramdisk_size();
314 u64 area_size
= PAGE_ALIGN(ramdisk_size
);
316 /* We need to move the initrd down into directly mapped mem */
317 relocated_ramdisk
= memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped
),
318 area_size
, PAGE_SIZE
);
320 if (!relocated_ramdisk
)
321 panic("Cannot find place for new RAMDISK of size %lld\n",
324 /* Note: this includes all the mem currently occupied by
325 the initrd, we rely on that fact to keep the data intact. */
326 memblock_reserve(relocated_ramdisk
, area_size
);
327 initrd_start
= relocated_ramdisk
+ PAGE_OFFSET
;
328 initrd_end
= initrd_start
+ ramdisk_size
;
329 printk(KERN_INFO
"Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
330 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
332 copy_from_early_mem((void *)initrd_start
, ramdisk_image
, ramdisk_size
);
334 printk(KERN_INFO
"Move RAMDISK from [mem %#010llx-%#010llx] to"
335 " [mem %#010llx-%#010llx]\n",
336 ramdisk_image
, ramdisk_image
+ ramdisk_size
- 1,
337 relocated_ramdisk
, relocated_ramdisk
+ ramdisk_size
- 1);
340 static void __init
early_reserve_initrd(void)
342 /* Assume only end is not page aligned */
343 u64 ramdisk_image
= get_ramdisk_image();
344 u64 ramdisk_size
= get_ramdisk_size();
345 u64 ramdisk_end
= PAGE_ALIGN(ramdisk_image
+ ramdisk_size
);
347 if (!boot_params
.hdr
.type_of_loader
||
348 !ramdisk_image
|| !ramdisk_size
)
349 return; /* No initrd provided by bootloader */
351 memblock_reserve(ramdisk_image
, ramdisk_end
- ramdisk_image
);
353 static void __init
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
);
361 if (!boot_params
.hdr
.type_of_loader
||
362 !ramdisk_image
|| !ramdisk_size
)
363 return; /* No initrd provided by bootloader */
367 mapped_size
= memblock_mem_size(max_pfn_mapped
);
368 if (ramdisk_size
>= (mapped_size
>>1))
369 panic("initrd too large to handle, "
370 "disabling initrd (%lld needed, %lld available)\n",
371 ramdisk_size
, mapped_size
>>1);
373 printk(KERN_INFO
"RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image
,
376 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image
),
377 PFN_DOWN(ramdisk_end
))) {
378 /* All are mapped, easy case */
379 initrd_start
= ramdisk_image
+ PAGE_OFFSET
;
380 initrd_end
= initrd_start
+ ramdisk_size
;
386 memblock_free(ramdisk_image
, ramdisk_end
- ramdisk_image
);
390 static void __init
early_reserve_initrd(void)
393 static void __init
reserve_initrd(void)
396 #endif /* CONFIG_BLK_DEV_INITRD */
398 static void __init
parse_setup_data(void)
400 struct setup_data
*data
;
401 u64 pa_data
, pa_next
;
403 pa_data
= boot_params
.hdr
.setup_data
;
405 u32 data_len
, data_type
;
407 data
= early_memremap(pa_data
, sizeof(*data
));
408 data_len
= data
->len
+ sizeof(struct setup_data
);
409 data_type
= data
->type
;
410 pa_next
= data
->next
;
411 early_memunmap(data
, sizeof(*data
));
415 e820__memory_setup_extended(pa_data
, data_len
);
421 parse_efi_setup(pa_data
, data_len
);
430 static void __init
memblock_x86_reserve_range_setup_data(void)
432 struct setup_data
*data
;
435 pa_data
= boot_params
.hdr
.setup_data
;
437 data
= early_memremap(pa_data
, sizeof(*data
));
438 memblock_reserve(pa_data
, sizeof(*data
) + data
->len
);
439 pa_data
= data
->next
;
440 early_memunmap(data
, sizeof(*data
));
445 * --------- Crashkernel reservation ------------------------------
448 #ifdef CONFIG_KEXEC_CORE
450 /* 16M alignment for crash kernel regions */
451 #define CRASH_ALIGN (16 << 20)
454 * Keep the crash kernel below this limit. On 32 bits earlier kernels
455 * would limit the kernel to the low 512 MiB due to mapping restrictions.
456 * On 64bit, old kexec-tools need to under 896MiB.
459 # define CRASH_ADDR_LOW_MAX (512 << 20)
460 # define CRASH_ADDR_HIGH_MAX (512 << 20)
462 # define CRASH_ADDR_LOW_MAX (896UL << 20)
463 # define CRASH_ADDR_HIGH_MAX MAXMEM
466 static int __init
reserve_crashkernel_low(void)
469 unsigned long long base
, low_base
= 0, low_size
= 0;
470 unsigned long total_low_mem
;
473 total_low_mem
= memblock_mem_size(1UL << (32 - PAGE_SHIFT
));
475 /* crashkernel=Y,low */
476 ret
= parse_crashkernel_low(boot_command_line
, total_low_mem
, &low_size
, &base
);
479 * two parts from lib/swiotlb.c:
480 * -swiotlb size: user-specified with swiotlb= or default.
482 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
483 * to 8M for other buffers that may need to stay low too. Also
484 * make sure we allocate enough extra low memory so that we
485 * don't run out of DMA buffers for 32-bit devices.
487 low_size
= max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
489 /* passed with crashkernel=0,low ? */
494 low_base
= memblock_find_in_range(0, 1ULL << 32, low_size
, CRASH_ALIGN
);
496 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
497 (unsigned long)(low_size
>> 20));
501 ret
= memblock_reserve(low_base
, low_size
);
503 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__
);
507 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
508 (unsigned long)(low_size
>> 20),
509 (unsigned long)(low_base
>> 20),
510 (unsigned long)(total_low_mem
>> 20));
512 crashk_low_res
.start
= low_base
;
513 crashk_low_res
.end
= low_base
+ low_size
- 1;
514 insert_resource(&iomem_resource
, &crashk_low_res
);
519 static void __init
reserve_crashkernel(void)
521 unsigned long long crash_size
, crash_base
, total_mem
;
525 total_mem
= memblock_phys_mem_size();
528 ret
= parse_crashkernel(boot_command_line
, total_mem
, &crash_size
, &crash_base
);
529 if (ret
!= 0 || crash_size
<= 0) {
530 /* crashkernel=X,high */
531 ret
= parse_crashkernel_high(boot_command_line
, total_mem
,
532 &crash_size
, &crash_base
);
533 if (ret
!= 0 || crash_size
<= 0)
538 if (xen_pv_domain()) {
539 pr_info("Ignoring crashkernel for a Xen PV domain\n");
543 /* 0 means: find the address automatically */
544 if (crash_base
<= 0) {
546 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
547 * as old kexec-tools loads bzImage below that, unless
548 * "crashkernel=size[KMG],high" is specified.
550 crash_base
= memblock_find_in_range(CRASH_ALIGN
,
551 high
? CRASH_ADDR_HIGH_MAX
552 : CRASH_ADDR_LOW_MAX
,
553 crash_size
, CRASH_ALIGN
);
555 pr_info("crashkernel reservation failed - No suitable area found.\n");
560 unsigned long long start
;
562 start
= memblock_find_in_range(crash_base
,
563 crash_base
+ crash_size
,
564 crash_size
, 1 << 20);
565 if (start
!= crash_base
) {
566 pr_info("crashkernel reservation failed - memory is in use.\n");
570 ret
= memblock_reserve(crash_base
, crash_size
);
572 pr_err("%s: Error reserving crashkernel memblock.\n", __func__
);
576 if (crash_base
>= (1ULL << 32) && reserve_crashkernel_low()) {
577 memblock_free(crash_base
, crash_size
);
581 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
582 (unsigned long)(crash_size
>> 20),
583 (unsigned long)(crash_base
>> 20),
584 (unsigned long)(total_mem
>> 20));
586 crashk_res
.start
= crash_base
;
587 crashk_res
.end
= crash_base
+ crash_size
- 1;
588 insert_resource(&iomem_resource
, &crashk_res
);
591 static void __init
reserve_crashkernel(void)
596 static struct resource standard_io_resources
[] = {
597 { .name
= "dma1", .start
= 0x00, .end
= 0x1f,
598 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
599 { .name
= "pic1", .start
= 0x20, .end
= 0x21,
600 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
601 { .name
= "timer0", .start
= 0x40, .end
= 0x43,
602 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
603 { .name
= "timer1", .start
= 0x50, .end
= 0x53,
604 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
605 { .name
= "keyboard", .start
= 0x60, .end
= 0x60,
606 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
607 { .name
= "keyboard", .start
= 0x64, .end
= 0x64,
608 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
609 { .name
= "dma page reg", .start
= 0x80, .end
= 0x8f,
610 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
611 { .name
= "pic2", .start
= 0xa0, .end
= 0xa1,
612 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
613 { .name
= "dma2", .start
= 0xc0, .end
= 0xdf,
614 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
},
615 { .name
= "fpu", .start
= 0xf0, .end
= 0xff,
616 .flags
= IORESOURCE_BUSY
| IORESOURCE_IO
}
619 void __init
reserve_standard_io_resources(void)
623 /* request I/O space for devices used on all i[345]86 PCs */
624 for (i
= 0; i
< ARRAY_SIZE(standard_io_resources
); i
++)
625 request_resource(&ioport_resource
, &standard_io_resources
[i
]);
629 static __init
void reserve_ibft_region(void)
631 unsigned long addr
, size
= 0;
633 addr
= find_ibft_region(&size
);
636 memblock_reserve(addr
, size
);
639 static bool __init
snb_gfx_workaround_needed(void)
644 static const __initconst u16 snb_ids
[] = {
654 /* Assume no if something weird is going on with PCI */
655 if (!early_pci_allowed())
658 vendor
= read_pci_config_16(0, 2, 0, PCI_VENDOR_ID
);
659 if (vendor
!= 0x8086)
662 devid
= read_pci_config_16(0, 2, 0, PCI_DEVICE_ID
);
663 for (i
= 0; i
< ARRAY_SIZE(snb_ids
); i
++)
664 if (devid
== snb_ids
[i
])
672 * Sandy Bridge graphics has trouble with certain ranges, exclude
673 * them from allocation.
675 static void __init
trim_snb_memory(void)
677 static const __initconst
unsigned long bad_pages
[] = {
686 if (!snb_gfx_workaround_needed())
689 printk(KERN_DEBUG
"reserving inaccessible SNB gfx pages\n");
692 * Reserve all memory below the 1 MB mark that has not
693 * already been reserved.
695 memblock_reserve(0, 1<<20);
697 for (i
= 0; i
< ARRAY_SIZE(bad_pages
); i
++) {
698 if (memblock_reserve(bad_pages
[i
], PAGE_SIZE
))
699 printk(KERN_WARNING
"failed to reserve 0x%08lx\n",
705 * Here we put platform-specific memory range workarounds, i.e.
706 * memory known to be corrupt or otherwise in need to be reserved on
707 * specific platforms.
709 * If this gets used more widely it could use a real dispatch mechanism.
711 static void __init
trim_platform_memory_ranges(void)
716 static void __init
trim_bios_range(void)
719 * A special case is the first 4Kb of memory;
720 * This is a BIOS owned area, not kernel ram, but generally
721 * not listed as such in the E820 table.
723 * This typically reserves additional memory (64KiB by default)
724 * since some BIOSes are known to corrupt low memory. See the
725 * Kconfig help text for X86_RESERVE_LOW.
727 e820__range_update(0, PAGE_SIZE
, E820_TYPE_RAM
, E820_TYPE_RESERVED
);
730 * special case: Some BIOSen report the PC BIOS
731 * area (640->1Mb) as ram even though it is not.
734 e820__range_remove(BIOS_BEGIN
, BIOS_END
- BIOS_BEGIN
, E820_TYPE_RAM
, 1);
736 e820__update_table(e820_table
);
739 /* called before trim_bios_range() to spare extra sanitize */
740 static void __init
e820_add_kernel_range(void)
742 u64 start
= __pa_symbol(_text
);
743 u64 size
= __pa_symbol(_end
) - start
;
746 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
747 * attempt to fix it by adding the range. We may have a confused BIOS,
748 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
749 * exclude kernel range. If we really are running on top non-RAM,
750 * we will crash later anyways.
752 if (e820__mapped_all(start
, start
+ size
, E820_TYPE_RAM
))
755 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
756 e820__range_remove(start
, size
, E820_TYPE_RAM
, 0);
757 e820__range_add(start
, size
, E820_TYPE_RAM
);
760 static unsigned reserve_low
= CONFIG_X86_RESERVE_LOW
<< 10;
762 static int __init
parse_reservelow(char *p
)
764 unsigned long long size
;
769 size
= memparse(p
, &p
);
782 early_param("reservelow", parse_reservelow
);
784 static void __init
trim_low_memory_range(void)
786 memblock_reserve(0, ALIGN(reserve_low
, PAGE_SIZE
));
790 * Dump out kernel offset information on panic.
793 dump_kernel_offset(struct notifier_block
*self
, unsigned long v
, void *p
)
795 if (kaslr_enabled()) {
796 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
802 pr_emerg("Kernel Offset: disabled\n");
809 * Determine if we were loaded by an EFI loader. If so, then we have also been
810 * passed the efi memmap, systab, etc., so we should use these data structures
811 * for initialization. Note, the efi init code path is determined by the
812 * global efi_enabled. This allows the same kernel image to be used on existing
813 * systems (with a traditional BIOS) as well as on EFI systems.
816 * setup_arch - architecture-specific boot-time initializations
818 * Note: On x86_64, fixmaps are ready for use even before this is called.
821 void __init
setup_arch(char **cmdline_p
)
823 memblock_reserve(__pa_symbol(_text
),
824 (unsigned long)__bss_stop
- (unsigned long)_text
);
827 * Make sure page 0 is always reserved because on systems with
828 * L1TF its contents can be leaked to user processes.
830 memblock_reserve(0, PAGE_SIZE
);
832 early_reserve_initrd();
835 * At this point everything still needed from the boot loader
836 * or BIOS or kernel text should be early reserved or marked not
837 * RAM in e820. All other memory is free game.
841 memcpy(&boot_cpu_data
, &new_cpu_data
, sizeof(new_cpu_data
));
844 * copy kernel address range established so far and switch
845 * to the proper swapper page table
847 clone_pgd_range(swapper_pg_dir
+ KERNEL_PGD_BOUNDARY
,
848 initial_page_table
+ KERNEL_PGD_BOUNDARY
,
851 load_cr3(swapper_pg_dir
);
853 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
854 * a cr3 based tlb flush, so the following __flush_tlb_all()
855 * will not flush anything because the cpu quirk which clears
856 * X86_FEATURE_PGE has not been invoked yet. Though due to the
857 * load_cr3() above the TLB has been flushed already. The
858 * quirk is invoked before subsequent calls to __flush_tlb_all()
859 * so proper operation is guaranteed.
863 printk(KERN_INFO
"Command line: %s\n", boot_command_line
);
864 boot_cpu_data
.x86_phys_bits
= MAX_PHYSMEM_BITS
;
868 * If we have OLPC OFW, we might end up relocating the fixmap due to
869 * reserve_top(), so do this before touching the ioremap area.
873 idt_setup_early_traps();
875 arch_init_ideal_nops();
877 early_ioremap_init();
879 setup_olpc_ofw_pgd();
881 ROOT_DEV
= old_decode_dev(boot_params
.hdr
.root_dev
);
882 screen_info
= boot_params
.screen_info
;
883 edid_info
= boot_params
.edid_info
;
885 apm_info
.bios
= boot_params
.apm_bios_info
;
886 ist_info
= boot_params
.ist_info
;
888 saved_video_mode
= boot_params
.hdr
.vid_mode
;
889 bootloader_type
= boot_params
.hdr
.type_of_loader
;
890 if ((bootloader_type
>> 4) == 0xe) {
891 bootloader_type
&= 0xf;
892 bootloader_type
|= (boot_params
.hdr
.ext_loader_type
+0x10) << 4;
894 bootloader_version
= bootloader_type
& 0xf;
895 bootloader_version
|= boot_params
.hdr
.ext_loader_ver
<< 4;
897 #ifdef CONFIG_BLK_DEV_RAM
898 rd_image_start
= boot_params
.hdr
.ram_size
& RAMDISK_IMAGE_START_MASK
;
899 rd_prompt
= ((boot_params
.hdr
.ram_size
& RAMDISK_PROMPT_FLAG
) != 0);
900 rd_doload
= ((boot_params
.hdr
.ram_size
& RAMDISK_LOAD_FLAG
) != 0);
903 if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
904 EFI32_LOADER_SIGNATURE
, 4)) {
905 set_bit(EFI_BOOT
, &efi
.flags
);
906 } else if (!strncmp((char *)&boot_params
.efi_info
.efi_loader_signature
,
907 EFI64_LOADER_SIGNATURE
, 4)) {
908 set_bit(EFI_BOOT
, &efi
.flags
);
909 set_bit(EFI_64BIT
, &efi
.flags
);
913 x86_init
.oem
.arch_setup();
915 iomem_resource
.end
= (1ULL << boot_cpu_data
.x86_phys_bits
) - 1;
916 e820__memory_setup();
921 if (!boot_params
.hdr
.root_flags
)
922 root_mountflags
&= ~MS_RDONLY
;
923 init_mm
.start_code
= (unsigned long) _text
;
924 init_mm
.end_code
= (unsigned long) _etext
;
925 init_mm
.end_data
= (unsigned long) _edata
;
926 init_mm
.brk
= _brk_end
;
928 mpx_mm_init(&init_mm
);
930 code_resource
.start
= __pa_symbol(_text
);
931 code_resource
.end
= __pa_symbol(_etext
)-1;
932 data_resource
.start
= __pa_symbol(_etext
);
933 data_resource
.end
= __pa_symbol(_edata
)-1;
934 bss_resource
.start
= __pa_symbol(__bss_start
);
935 bss_resource
.end
= __pa_symbol(__bss_stop
)-1;
937 #ifdef CONFIG_CMDLINE_BOOL
938 #ifdef CONFIG_CMDLINE_OVERRIDE
939 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
941 if (builtin_cmdline
[0]) {
942 /* append boot loader cmdline to builtin */
943 strlcat(builtin_cmdline
, " ", COMMAND_LINE_SIZE
);
944 strlcat(builtin_cmdline
, boot_command_line
, COMMAND_LINE_SIZE
);
945 strlcpy(boot_command_line
, builtin_cmdline
, COMMAND_LINE_SIZE
);
950 strlcpy(command_line
, boot_command_line
, COMMAND_LINE_SIZE
);
951 *cmdline_p
= command_line
;
954 * x86_configure_nx() is called before parse_early_param() to detect
955 * whether hardware doesn't support NX (so that the early EHCI debug
956 * console setup can safely call set_fixmap()). It may then be called
957 * again from within noexec_setup() during parsing early parameters
958 * to honor the respective command line option.
964 if (efi_enabled(EFI_BOOT
))
965 efi_memblock_x86_reserve_range();
966 #ifdef CONFIG_MEMORY_HOTPLUG
968 * Memory used by the kernel cannot be hot-removed because Linux
969 * cannot migrate the kernel pages. When memory hotplug is
970 * enabled, we should prevent memblock from allocating memory
973 * ACPI SRAT records all hotpluggable memory ranges. But before
974 * SRAT is parsed, we don't know about it.
976 * The kernel image is loaded into memory at very early time. We
977 * cannot prevent this anyway. So on NUMA system, we set any
978 * node the kernel resides in as un-hotpluggable.
980 * Since on modern servers, one node could have double-digit
981 * gigabytes memory, we can assume the memory around the kernel
982 * image is also un-hotpluggable. So before SRAT is parsed, just
983 * allocate memory near the kernel image to try the best to keep
984 * the kernel away from hotpluggable memory.
986 if (movable_node_is_enabled())
987 memblock_set_bottom_up(true);
992 /* after early param, so could get panic from serial */
993 memblock_x86_reserve_range_setup_data();
995 if (acpi_mps_check()) {
996 #ifdef CONFIG_X86_LOCAL_APIC
999 setup_clear_cpu_cap(X86_FEATURE_APIC
);
1002 e820__reserve_setup_data();
1003 e820__finish_early_params();
1005 if (efi_enabled(EFI_BOOT
))
1010 dmi_set_dump_stack_arch_desc();
1013 * VMware detection requires dmi to be available, so this
1014 * needs to be done after dmi_scan_machine(), for the boot CPU.
1016 init_hypervisor_platform();
1019 x86_init
.resources
.probe_roms();
1021 /* after parse_early_param, so could debug it */
1022 insert_resource(&iomem_resource
, &code_resource
);
1023 insert_resource(&iomem_resource
, &data_resource
);
1024 insert_resource(&iomem_resource
, &bss_resource
);
1026 e820_add_kernel_range();
1028 #ifdef CONFIG_X86_32
1029 if (ppro_with_ram_bug()) {
1030 e820__range_update(0x70000000ULL
, 0x40000ULL
, E820_TYPE_RAM
,
1031 E820_TYPE_RESERVED
);
1032 e820__update_table(e820_table
);
1033 printk(KERN_INFO
"fixed physical RAM map:\n");
1034 e820__print_table("bad_ppro");
1037 early_gart_iommu_check();
1041 * partially used pages are not usable - thus
1042 * we are rounding upwards:
1044 max_pfn
= e820__end_of_ram_pfn();
1046 /* update e820 for memory not covered by WB MTRRs */
1048 if (mtrr_trim_uncached_memory(max_pfn
))
1049 max_pfn
= e820__end_of_ram_pfn();
1051 max_possible_pfn
= max_pfn
;
1054 * This call is required when the CPU does not support PAT. If
1055 * mtrr_bp_init() invoked it already via pat_init() the call has no
1061 * Define random base addresses for memory sections after max_pfn is
1062 * defined and before each memory section base is used.
1064 kernel_randomize_memory();
1066 #ifdef CONFIG_X86_32
1067 /* max_low_pfn get updated here */
1068 find_low_pfn_range();
1072 /* How many end-of-memory variables you have, grandma! */
1073 /* need this before calling reserve_initrd */
1074 if (max_pfn
> (1UL<<(32 - PAGE_SHIFT
)))
1075 max_low_pfn
= e820__end_of_low_ram_pfn();
1077 max_low_pfn
= max_pfn
;
1079 high_memory
= (void *)__va(max_pfn
* PAGE_SIZE
- 1) + 1;
1083 * Find and reserve possible boot-time SMP configuration:
1087 reserve_ibft_region();
1089 early_alloc_pgt_buf();
1092 * Need to conclude brk, before e820__memblock_setup()
1093 * it could use memblock_find_in_range, could overlap with
1100 memblock_set_current_limit(ISA_END_ADDRESS
);
1101 e820__memblock_setup();
1103 reserve_bios_regions();
1105 if (efi_enabled(EFI_MEMMAP
)) {
1111 * The EFI specification says that boot service code won't be
1112 * called after ExitBootServices(). This is, in fact, a lie.
1114 efi_reserve_boot_services();
1117 /* preallocate 4k for mptable mpc */
1118 e820__memblock_alloc_reserved_mpc_new();
1120 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1121 setup_bios_corruption_check();
1124 #ifdef CONFIG_X86_32
1125 printk(KERN_DEBUG
"initial memory mapped: [mem 0x00000000-%#010lx]\n",
1126 (max_pfn_mapped
<<PAGE_SHIFT
) - 1);
1129 reserve_real_mode();
1131 trim_platform_memory_ranges();
1132 trim_low_memory_range();
1136 idt_setup_early_pf();
1139 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1140 * with the current CR4 value. This may not be necessary, but
1141 * auditing all the early-boot CR4 manipulation would be needed to
1144 * Mask off features that don't work outside long mode (just
1147 mmu_cr4_features
= __read_cr4() & ~X86_CR4_PCIDE
;
1149 memblock_set_current_limit(get_max_mapped());
1152 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1155 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1156 if (init_ohci1394_dma_early
)
1157 init_ohci1394_dma_on_all_controllers();
1159 /* Allocate bigger log buffer */
1162 if (efi_enabled(EFI_BOOT
)) {
1163 switch (boot_params
.secure_boot
) {
1164 case efi_secureboot_mode_disabled
:
1165 pr_info("Secure boot disabled\n");
1167 case efi_secureboot_mode_enabled
:
1168 pr_info("Secure boot enabled\n");
1171 pr_info("Secure boot could not be determined\n");
1178 acpi_table_upgrade();
1184 early_platform_quirks();
1187 * Parse the ACPI tables for possible boot-time SMP configuration.
1189 acpi_boot_table_init();
1191 early_acpi_boot_init();
1194 dma_contiguous_reserve(max_pfn_mapped
<< PAGE_SHIFT
);
1197 * Reserve memory for crash kernel after SRAT is parsed so that it
1198 * won't consume hotpluggable memory.
1200 reserve_crashkernel();
1202 memblock_find_dma_reserve();
1204 if (!early_xdbc_setup_hardware())
1205 early_xdbc_register_console();
1207 x86_init
.paging
.pagetable_init();
1212 * Sync back kernel address range.
1214 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1217 sync_initial_page_table();
1223 generic_apic_probe();
1228 * Read APIC and some other early information from ACPI tables.
1235 * get boot-time SMP configuration:
1240 * Systems w/o ACPI and mptables might not have it mapped the local
1241 * APIC yet, but prefill_possible_map() might need to access it.
1243 init_apic_mappings();
1245 prefill_possible_map();
1249 io_apic_init_mappings();
1251 x86_init
.hyper
.guest_late_init();
1253 e820__reserve_resources();
1254 e820__register_nosave_regions(max_pfn
);
1256 x86_init
.resources
.reserve_resources();
1258 e820__setup_pci_gap();
1261 #if defined(CONFIG_VGA_CONSOLE)
1262 if (!efi_enabled(EFI_BOOT
) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY
))
1263 conswitchp
= &vga_con
;
1264 #elif defined(CONFIG_DUMMY_CONSOLE)
1265 conswitchp
= &dummy_con
;
1268 x86_init
.oem
.banner();
1270 x86_init
.timers
.wallclock_init();
1274 register_refined_jiffies(CLOCK_TICK_RATE
);
1277 if (efi_enabled(EFI_BOOT
))
1278 efi_apply_memmap_quirks();
1284 #ifdef CONFIG_X86_32
1286 static struct resource video_ram_resource
= {
1287 .name
= "Video RAM area",
1290 .flags
= IORESOURCE_BUSY
| IORESOURCE_MEM
1293 void __init
i386_reserve_resources(void)
1295 request_resource(&iomem_resource
, &video_ram_resource
);
1296 reserve_standard_io_resources();
1299 #endif /* CONFIG_X86_32 */
1301 static struct notifier_block kernel_offset_notifier
= {
1302 .notifier_call
= dump_kernel_offset
1305 static int __init
register_kernel_offset_dumper(void)
1307 atomic_notifier_chain_register(&panic_notifier_list
,
1308 &kernel_offset_notifier
);
1311 __initcall(register_kernel_offset_dumper
);