mm: fix exec activate_mm vs TLB shootdown and lazy tlb switching race
[linux/fpc-iii.git] / arch / x86 / kernel / setup.c
blob4bc12447a50fce87011e278e95184e97e339a6f0
1 /*
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>
25 #include <linux/mm.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/mem_encrypt.h>
74 #include <linux/usb/xhci-dbgp.h>
75 #include <video/edid.h>
77 #include <asm/mtrr.h>
78 #include <asm/apic.h>
79 #include <asm/realmode.h>
80 #include <asm/e820/api.h>
81 #include <asm/mpspec.h>
82 #include <asm/setup.h>
83 #include <asm/efi.h>
84 #include <asm/timer.h>
85 #include <asm/i8259.h>
86 #include <asm/sections.h>
87 #include <asm/io_apic.h>
88 #include <asm/ist.h>
89 #include <asm/setup_arch.h>
90 #include <asm/bios_ebda.h>
91 #include <asm/cacheflush.h>
92 #include <asm/processor.h>
93 #include <asm/bugs.h>
94 #include <asm/kasan.h>
96 #include <asm/vsyscall.h>
97 #include <asm/cpu.h>
98 #include <asm/desc.h>
99 #include <asm/dma.h>
100 #include <asm/iommu.h>
101 #include <asm/gart.h>
102 #include <asm/mmu_context.h>
103 #include <asm/proto.h>
105 #include <asm/paravirt.h>
106 #include <asm/hypervisor.h>
107 #include <asm/olpc_ofw.h>
109 #include <asm/percpu.h>
110 #include <asm/topology.h>
111 #include <asm/apicdef.h>
112 #include <asm/amd_nb.h>
113 #include <asm/mce.h>
114 #include <asm/alternative.h>
115 #include <asm/prom.h>
116 #include <asm/microcode.h>
117 #include <asm/mmu_context.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;
131 #ifdef CONFIG_DMI
132 RESERVE_BRK(dmi_alloc, 65536);
133 #endif
136 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
137 unsigned long _brk_end = (unsigned long)__brk_base;
139 #ifdef CONFIG_X86_64
140 int default_cpu_present_to_apicid(int mps_cpu)
142 return __default_cpu_present_to_apicid(mps_cpu);
145 int default_check_phys_apicid_present(int phys_apicid)
147 return __default_check_phys_apicid_present(phys_apicid);
149 #endif
151 struct boot_params boot_params;
154 * Machine setup..
156 static struct resource data_resource = {
157 .name = "Kernel data",
158 .start = 0,
159 .end = 0,
160 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
163 static struct resource code_resource = {
164 .name = "Kernel code",
165 .start = 0,
166 .end = 0,
167 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
170 static struct resource bss_resource = {
171 .name = "Kernel bss",
172 .start = 0,
173 .end = 0,
174 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
178 #ifdef CONFIG_X86_32
179 /* cpu data as detected by the assembly code in head_32.S */
180 struct cpuinfo_x86 new_cpu_data;
182 /* common cpu data for all cpus */
183 struct cpuinfo_x86 boot_cpu_data __read_mostly;
184 EXPORT_SYMBOL(boot_cpu_data);
186 unsigned int def_to_bigsmp;
188 /* for MCA, but anyone else can use it if they want */
189 unsigned int machine_id;
190 unsigned int machine_submodel_id;
191 unsigned int BIOS_revision;
193 struct apm_info apm_info;
194 EXPORT_SYMBOL(apm_info);
196 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
197 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
198 struct ist_info ist_info;
199 EXPORT_SYMBOL(ist_info);
200 #else
201 struct ist_info ist_info;
202 #endif
204 #else
205 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
206 .x86_phys_bits = MAX_PHYSMEM_BITS,
208 EXPORT_SYMBOL(boot_cpu_data);
209 #endif
212 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
213 __visible unsigned long mmu_cr4_features __ro_after_init;
214 #else
215 __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
216 #endif
218 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
219 int bootloader_type, bootloader_version;
222 * Setup options
224 struct screen_info screen_info;
225 EXPORT_SYMBOL(screen_info);
226 struct edid_info edid_info;
227 EXPORT_SYMBOL_GPL(edid_info);
229 extern int root_mountflags;
231 unsigned long saved_video_mode;
233 #define RAMDISK_IMAGE_START_MASK 0x07FF
234 #define RAMDISK_PROMPT_FLAG 0x8000
235 #define RAMDISK_LOAD_FLAG 0x4000
237 static char __initdata command_line[COMMAND_LINE_SIZE];
238 #ifdef CONFIG_CMDLINE_BOOL
239 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
240 #endif
242 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
243 struct edd edd;
244 #ifdef CONFIG_EDD_MODULE
245 EXPORT_SYMBOL(edd);
246 #endif
248 * copy_edd() - Copy the BIOS EDD information
249 * from boot_params into a safe place.
252 static inline void __init copy_edd(void)
254 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
255 sizeof(edd.mbr_signature));
256 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
257 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
258 edd.edd_info_nr = boot_params.eddbuf_entries;
260 #else
261 static inline void __init copy_edd(void)
264 #endif
266 void * __init extend_brk(size_t size, size_t align)
268 size_t mask = align - 1;
269 void *ret;
271 BUG_ON(_brk_start == 0);
272 BUG_ON(align & mask);
274 _brk_end = (_brk_end + mask) & ~mask;
275 BUG_ON((char *)(_brk_end + size) > __brk_limit);
277 ret = (void *)_brk_end;
278 _brk_end += size;
280 memset(ret, 0, size);
282 return ret;
285 #ifdef CONFIG_X86_32
286 static void __init cleanup_highmap(void)
289 #endif
291 static void __init reserve_brk(void)
293 if (_brk_end > _brk_start)
294 memblock_reserve(__pa_symbol(_brk_start),
295 _brk_end - _brk_start);
297 /* Mark brk area as locked down and no longer taking any
298 new allocations */
299 _brk_start = 0;
302 u64 relocated_ramdisk;
304 #ifdef CONFIG_BLK_DEV_INITRD
306 static u64 __init get_ramdisk_image(void)
308 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
310 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
312 return ramdisk_image;
314 static u64 __init get_ramdisk_size(void)
316 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
318 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
320 return ramdisk_size;
323 static void __init relocate_initrd(void)
325 /* Assume only end is not page aligned */
326 u64 ramdisk_image = get_ramdisk_image();
327 u64 ramdisk_size = get_ramdisk_size();
328 u64 area_size = PAGE_ALIGN(ramdisk_size);
330 /* We need to move the initrd down into directly mapped mem */
331 relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
332 area_size, PAGE_SIZE);
334 if (!relocated_ramdisk)
335 panic("Cannot find place for new RAMDISK of size %lld\n",
336 ramdisk_size);
338 /* Note: this includes all the mem currently occupied by
339 the initrd, we rely on that fact to keep the data intact. */
340 memblock_reserve(relocated_ramdisk, area_size);
341 initrd_start = relocated_ramdisk + PAGE_OFFSET;
342 initrd_end = initrd_start + ramdisk_size;
343 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
344 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
346 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
348 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
349 " [mem %#010llx-%#010llx]\n",
350 ramdisk_image, ramdisk_image + ramdisk_size - 1,
351 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
354 static void __init early_reserve_initrd(void)
356 /* Assume only end is not page aligned */
357 u64 ramdisk_image = get_ramdisk_image();
358 u64 ramdisk_size = get_ramdisk_size();
359 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 */
365 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
367 static void __init reserve_initrd(void)
369 /* Assume only end is not page aligned */
370 u64 ramdisk_image = get_ramdisk_image();
371 u64 ramdisk_size = get_ramdisk_size();
372 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
373 u64 mapped_size;
375 if (!boot_params.hdr.type_of_loader ||
376 !ramdisk_image || !ramdisk_size)
377 return; /* No initrd provided by bootloader */
379 initrd_start = 0;
381 mapped_size = memblock_mem_size(max_pfn_mapped);
382 if (ramdisk_size >= (mapped_size>>1))
383 panic("initrd too large to handle, "
384 "disabling initrd (%lld needed, %lld available)\n",
385 ramdisk_size, mapped_size>>1);
387 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
388 ramdisk_end - 1);
390 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
391 PFN_DOWN(ramdisk_end))) {
392 /* All are mapped, easy case */
393 initrd_start = ramdisk_image + PAGE_OFFSET;
394 initrd_end = initrd_start + ramdisk_size;
395 return;
398 relocate_initrd();
400 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
403 #else
404 static void __init early_reserve_initrd(void)
407 static void __init reserve_initrd(void)
410 #endif /* CONFIG_BLK_DEV_INITRD */
412 static void __init parse_setup_data(void)
414 struct setup_data *data;
415 u64 pa_data, pa_next;
417 pa_data = boot_params.hdr.setup_data;
418 while (pa_data) {
419 u32 data_len, data_type;
421 data = early_memremap(pa_data, sizeof(*data));
422 data_len = data->len + sizeof(struct setup_data);
423 data_type = data->type;
424 pa_next = data->next;
425 early_memunmap(data, sizeof(*data));
427 switch (data_type) {
428 case SETUP_E820_EXT:
429 e820__memory_setup_extended(pa_data, data_len);
430 break;
431 case SETUP_DTB:
432 add_dtb(pa_data);
433 break;
434 case SETUP_EFI:
435 parse_efi_setup(pa_data, data_len);
436 break;
437 default:
438 break;
440 pa_data = pa_next;
444 static void __init memblock_x86_reserve_range_setup_data(void)
446 struct setup_data *data;
447 u64 pa_data;
449 pa_data = boot_params.hdr.setup_data;
450 while (pa_data) {
451 data = early_memremap(pa_data, sizeof(*data));
452 memblock_reserve(pa_data, sizeof(*data) + data->len);
453 pa_data = data->next;
454 early_memunmap(data, sizeof(*data));
459 * --------- Crashkernel reservation ------------------------------
462 #ifdef CONFIG_KEXEC_CORE
464 /* 16M alignment for crash kernel regions */
465 #define CRASH_ALIGN (16 << 20)
468 * Keep the crash kernel below this limit. On 32 bits earlier kernels
469 * would limit the kernel to the low 512 MiB due to mapping restrictions.
470 * On 64bit, old kexec-tools need to under 896MiB.
472 #ifdef CONFIG_X86_32
473 # define CRASH_ADDR_LOW_MAX (512 << 20)
474 # define CRASH_ADDR_HIGH_MAX (512 << 20)
475 #else
476 # define CRASH_ADDR_LOW_MAX (896UL << 20)
477 # define CRASH_ADDR_HIGH_MAX MAXMEM
478 #endif
480 static int __init reserve_crashkernel_low(void)
482 #ifdef CONFIG_X86_64
483 unsigned long long base, low_base = 0, low_size = 0;
484 unsigned long total_low_mem;
485 int ret;
487 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
489 /* crashkernel=Y,low */
490 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
491 if (ret) {
493 * two parts from lib/swiotlb.c:
494 * -swiotlb size: user-specified with swiotlb= or default.
496 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
497 * to 8M for other buffers that may need to stay low too. Also
498 * make sure we allocate enough extra low memory so that we
499 * don't run out of DMA buffers for 32-bit devices.
501 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
502 } else {
503 /* passed with crashkernel=0,low ? */
504 if (!low_size)
505 return 0;
508 low_base = memblock_find_in_range(0, 1ULL << 32, low_size, CRASH_ALIGN);
509 if (!low_base) {
510 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
511 (unsigned long)(low_size >> 20));
512 return -ENOMEM;
515 ret = memblock_reserve(low_base, low_size);
516 if (ret) {
517 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
518 return ret;
521 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
522 (unsigned long)(low_size >> 20),
523 (unsigned long)(low_base >> 20),
524 (unsigned long)(total_low_mem >> 20));
526 crashk_low_res.start = low_base;
527 crashk_low_res.end = low_base + low_size - 1;
528 insert_resource(&iomem_resource, &crashk_low_res);
529 #endif
530 return 0;
533 static void __init reserve_crashkernel(void)
535 unsigned long long crash_size, crash_base, total_mem;
536 bool high = false;
537 int ret;
539 total_mem = memblock_phys_mem_size();
541 /* crashkernel=XM */
542 ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
543 if (ret != 0 || crash_size <= 0) {
544 /* crashkernel=X,high */
545 ret = parse_crashkernel_high(boot_command_line, total_mem,
546 &crash_size, &crash_base);
547 if (ret != 0 || crash_size <= 0)
548 return;
549 high = true;
552 /* 0 means: find the address automatically */
553 if (crash_base <= 0) {
555 * Set CRASH_ADDR_LOW_MAX upper bound for crash memory,
556 * as old kexec-tools loads bzImage below that, unless
557 * "crashkernel=size[KMG],high" is specified.
559 crash_base = memblock_find_in_range(CRASH_ALIGN,
560 high ? CRASH_ADDR_HIGH_MAX
561 : CRASH_ADDR_LOW_MAX,
562 crash_size, CRASH_ALIGN);
563 if (!crash_base) {
564 pr_info("crashkernel reservation failed - No suitable area found.\n");
565 return;
568 } else {
569 unsigned long long start;
571 start = memblock_find_in_range(crash_base,
572 crash_base + crash_size,
573 crash_size, 1 << 20);
574 if (start != crash_base) {
575 pr_info("crashkernel reservation failed - memory is in use.\n");
576 return;
579 ret = memblock_reserve(crash_base, crash_size);
580 if (ret) {
581 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
582 return;
585 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
586 memblock_free(crash_base, crash_size);
587 return;
590 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
591 (unsigned long)(crash_size >> 20),
592 (unsigned long)(crash_base >> 20),
593 (unsigned long)(total_mem >> 20));
595 crashk_res.start = crash_base;
596 crashk_res.end = crash_base + crash_size - 1;
597 insert_resource(&iomem_resource, &crashk_res);
599 #else
600 static void __init reserve_crashkernel(void)
603 #endif
605 static struct resource standard_io_resources[] = {
606 { .name = "dma1", .start = 0x00, .end = 0x1f,
607 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
608 { .name = "pic1", .start = 0x20, .end = 0x21,
609 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
610 { .name = "timer0", .start = 0x40, .end = 0x43,
611 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
612 { .name = "timer1", .start = 0x50, .end = 0x53,
613 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
614 { .name = "keyboard", .start = 0x60, .end = 0x60,
615 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
616 { .name = "keyboard", .start = 0x64, .end = 0x64,
617 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
618 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
619 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
620 { .name = "pic2", .start = 0xa0, .end = 0xa1,
621 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
622 { .name = "dma2", .start = 0xc0, .end = 0xdf,
623 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
624 { .name = "fpu", .start = 0xf0, .end = 0xff,
625 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
628 void __init reserve_standard_io_resources(void)
630 int i;
632 /* request I/O space for devices used on all i[345]86 PCs */
633 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
634 request_resource(&ioport_resource, &standard_io_resources[i]);
638 static __init void reserve_ibft_region(void)
640 unsigned long addr, size = 0;
642 addr = find_ibft_region(&size);
644 if (size)
645 memblock_reserve(addr, size);
648 static bool __init snb_gfx_workaround_needed(void)
650 #ifdef CONFIG_PCI
651 int i;
652 u16 vendor, devid;
653 static const __initconst u16 snb_ids[] = {
654 0x0102,
655 0x0112,
656 0x0122,
657 0x0106,
658 0x0116,
659 0x0126,
660 0x010a,
663 /* Assume no if something weird is going on with PCI */
664 if (!early_pci_allowed())
665 return false;
667 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
668 if (vendor != 0x8086)
669 return false;
671 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
672 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
673 if (devid == snb_ids[i])
674 return true;
675 #endif
677 return false;
681 * Sandy Bridge graphics has trouble with certain ranges, exclude
682 * them from allocation.
684 static void __init trim_snb_memory(void)
686 static const __initconst unsigned long bad_pages[] = {
687 0x20050000,
688 0x20110000,
689 0x20130000,
690 0x20138000,
691 0x40004000,
693 int i;
695 if (!snb_gfx_workaround_needed())
696 return;
698 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
701 * Reserve all memory below the 1 MB mark that has not
702 * already been reserved.
704 memblock_reserve(0, 1<<20);
706 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
707 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
708 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
709 bad_pages[i]);
714 * Here we put platform-specific memory range workarounds, i.e.
715 * memory known to be corrupt or otherwise in need to be reserved on
716 * specific platforms.
718 * If this gets used more widely it could use a real dispatch mechanism.
720 static void __init trim_platform_memory_ranges(void)
722 trim_snb_memory();
725 static void __init trim_bios_range(void)
728 * A special case is the first 4Kb of memory;
729 * This is a BIOS owned area, not kernel ram, but generally
730 * not listed as such in the E820 table.
732 * This typically reserves additional memory (64KiB by default)
733 * since some BIOSes are known to corrupt low memory. See the
734 * Kconfig help text for X86_RESERVE_LOW.
736 e820__range_update(0, PAGE_SIZE, E820_TYPE_RAM, E820_TYPE_RESERVED);
739 * special case: Some BIOSen report the PC BIOS
740 * area (640->1Mb) as ram even though it is not.
741 * take them out.
743 e820__range_remove(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_TYPE_RAM, 1);
745 e820__update_table(e820_table);
748 /* called before trim_bios_range() to spare extra sanitize */
749 static void __init e820_add_kernel_range(void)
751 u64 start = __pa_symbol(_text);
752 u64 size = __pa_symbol(_end) - start;
755 * Complain if .text .data and .bss are not marked as E820_TYPE_RAM and
756 * attempt to fix it by adding the range. We may have a confused BIOS,
757 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
758 * exclude kernel range. If we really are running on top non-RAM,
759 * we will crash later anyways.
761 if (e820__mapped_all(start, start + size, E820_TYPE_RAM))
762 return;
764 pr_warn(".text .data .bss are not marked as E820_TYPE_RAM!\n");
765 e820__range_remove(start, size, E820_TYPE_RAM, 0);
766 e820__range_add(start, size, E820_TYPE_RAM);
769 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
771 static int __init parse_reservelow(char *p)
773 unsigned long long size;
775 if (!p)
776 return -EINVAL;
778 size = memparse(p, &p);
780 if (size < 4096)
781 size = 4096;
783 if (size > 640*1024)
784 size = 640*1024;
786 reserve_low = size;
788 return 0;
791 early_param("reservelow", parse_reservelow);
793 static void __init trim_low_memory_range(void)
795 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
799 * Dump out kernel offset information on panic.
801 static int
802 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
804 if (kaslr_enabled()) {
805 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
806 kaslr_offset(),
807 __START_KERNEL,
808 __START_KERNEL_map,
809 MODULES_VADDR-1);
810 } else {
811 pr_emerg("Kernel Offset: disabled\n");
814 return 0;
817 static void __init simple_udelay_calibration(void)
819 unsigned int tsc_khz, cpu_khz;
820 unsigned long lpj;
822 if (!boot_cpu_has(X86_FEATURE_TSC))
823 return;
825 cpu_khz = x86_platform.calibrate_cpu();
826 tsc_khz = x86_platform.calibrate_tsc();
828 tsc_khz = tsc_khz ? : cpu_khz;
829 if (!tsc_khz)
830 return;
832 lpj = tsc_khz * 1000;
833 do_div(lpj, HZ);
834 loops_per_jiffy = lpj;
838 * Determine if we were loaded by an EFI loader. If so, then we have also been
839 * passed the efi memmap, systab, etc., so we should use these data structures
840 * for initialization. Note, the efi init code path is determined by the
841 * global efi_enabled. This allows the same kernel image to be used on existing
842 * systems (with a traditional BIOS) as well as on EFI systems.
845 * setup_arch - architecture-specific boot-time initializations
847 * Note: On x86_64, fixmaps are ready for use even before this is called.
850 void __init setup_arch(char **cmdline_p)
852 memblock_reserve(__pa_symbol(_text),
853 (unsigned long)__bss_stop - (unsigned long)_text);
856 * Make sure page 0 is always reserved because on systems with
857 * L1TF its contents can be leaked to user processes.
859 memblock_reserve(0, PAGE_SIZE);
861 early_reserve_initrd();
864 * At this point everything still needed from the boot loader
865 * or BIOS or kernel text should be early reserved or marked not
866 * RAM in e820. All other memory is free game.
869 #ifdef CONFIG_X86_32
870 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
873 * copy kernel address range established so far and switch
874 * to the proper swapper page table
876 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
877 initial_page_table + KERNEL_PGD_BOUNDARY,
878 KERNEL_PGD_PTRS);
880 load_cr3(swapper_pg_dir);
882 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
883 * a cr3 based tlb flush, so the following __flush_tlb_all()
884 * will not flush anything because the cpu quirk which clears
885 * X86_FEATURE_PGE has not been invoked yet. Though due to the
886 * load_cr3() above the TLB has been flushed already. The
887 * quirk is invoked before subsequent calls to __flush_tlb_all()
888 * so proper operation is guaranteed.
890 __flush_tlb_all();
891 #else
892 printk(KERN_INFO "Command line: %s\n", boot_command_line);
893 #endif
896 * If we have OLPC OFW, we might end up relocating the fixmap due to
897 * reserve_top(), so do this before touching the ioremap area.
899 olpc_ofw_detect();
901 idt_setup_early_traps();
902 early_cpu_init();
903 early_ioremap_init();
905 setup_olpc_ofw_pgd();
907 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
908 screen_info = boot_params.screen_info;
909 edid_info = boot_params.edid_info;
910 #ifdef CONFIG_X86_32
911 apm_info.bios = boot_params.apm_bios_info;
912 ist_info = boot_params.ist_info;
913 #endif
914 saved_video_mode = boot_params.hdr.vid_mode;
915 bootloader_type = boot_params.hdr.type_of_loader;
916 if ((bootloader_type >> 4) == 0xe) {
917 bootloader_type &= 0xf;
918 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
920 bootloader_version = bootloader_type & 0xf;
921 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
923 #ifdef CONFIG_BLK_DEV_RAM
924 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
925 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
926 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
927 #endif
928 #ifdef CONFIG_EFI
929 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
930 EFI32_LOADER_SIGNATURE, 4)) {
931 set_bit(EFI_BOOT, &efi.flags);
932 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
933 EFI64_LOADER_SIGNATURE, 4)) {
934 set_bit(EFI_BOOT, &efi.flags);
935 set_bit(EFI_64BIT, &efi.flags);
937 #endif
939 x86_init.oem.arch_setup();
941 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
942 e820__memory_setup();
943 parse_setup_data();
945 copy_edd();
947 if (!boot_params.hdr.root_flags)
948 root_mountflags &= ~MS_RDONLY;
949 init_mm.start_code = (unsigned long) _text;
950 init_mm.end_code = (unsigned long) _etext;
951 init_mm.end_data = (unsigned long) _edata;
952 init_mm.brk = _brk_end;
954 mpx_mm_init(&init_mm);
956 code_resource.start = __pa_symbol(_text);
957 code_resource.end = __pa_symbol(_etext)-1;
958 data_resource.start = __pa_symbol(_etext);
959 data_resource.end = __pa_symbol(_edata)-1;
960 bss_resource.start = __pa_symbol(__bss_start);
961 bss_resource.end = __pa_symbol(__bss_stop)-1;
963 #ifdef CONFIG_CMDLINE_BOOL
964 #ifdef CONFIG_CMDLINE_OVERRIDE
965 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
966 #else
967 if (builtin_cmdline[0]) {
968 /* append boot loader cmdline to builtin */
969 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
970 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
971 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
973 #endif
974 #endif
976 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
977 *cmdline_p = command_line;
980 * x86_configure_nx() is called before parse_early_param() to detect
981 * whether hardware doesn't support NX (so that the early EHCI debug
982 * console setup can safely call set_fixmap()). It may then be called
983 * again from within noexec_setup() during parsing early parameters
984 * to honor the respective command line option.
986 x86_configure_nx();
988 parse_early_param();
990 if (efi_enabled(EFI_BOOT))
991 efi_memblock_x86_reserve_range();
992 #ifdef CONFIG_MEMORY_HOTPLUG
994 * Memory used by the kernel cannot be hot-removed because Linux
995 * cannot migrate the kernel pages. When memory hotplug is
996 * enabled, we should prevent memblock from allocating memory
997 * for the kernel.
999 * ACPI SRAT records all hotpluggable memory ranges. But before
1000 * SRAT is parsed, we don't know about it.
1002 * The kernel image is loaded into memory at very early time. We
1003 * cannot prevent this anyway. So on NUMA system, we set any
1004 * node the kernel resides in as un-hotpluggable.
1006 * Since on modern servers, one node could have double-digit
1007 * gigabytes memory, we can assume the memory around the kernel
1008 * image is also un-hotpluggable. So before SRAT is parsed, just
1009 * allocate memory near the kernel image to try the best to keep
1010 * the kernel away from hotpluggable memory.
1012 if (movable_node_is_enabled())
1013 memblock_set_bottom_up(true);
1014 #endif
1016 x86_report_nx();
1018 /* after early param, so could get panic from serial */
1019 memblock_x86_reserve_range_setup_data();
1021 if (acpi_mps_check()) {
1022 #ifdef CONFIG_X86_LOCAL_APIC
1023 disable_apic = 1;
1024 #endif
1025 setup_clear_cpu_cap(X86_FEATURE_APIC);
1028 #ifdef CONFIG_PCI
1029 if (pci_early_dump_regs)
1030 early_dump_pci_devices();
1031 #endif
1033 e820__reserve_setup_data();
1034 e820__finish_early_params();
1036 if (efi_enabled(EFI_BOOT))
1037 efi_init();
1039 dmi_scan_machine();
1040 dmi_memdev_walk();
1041 dmi_set_dump_stack_arch_desc();
1044 * VMware detection requires dmi to be available, so this
1045 * needs to be done after dmi_scan_machine, for the BP.
1047 init_hypervisor_platform();
1049 simple_udelay_calibration();
1051 x86_init.resources.probe_roms();
1053 /* after parse_early_param, so could debug it */
1054 insert_resource(&iomem_resource, &code_resource);
1055 insert_resource(&iomem_resource, &data_resource);
1056 insert_resource(&iomem_resource, &bss_resource);
1058 e820_add_kernel_range();
1059 trim_bios_range();
1060 #ifdef CONFIG_X86_32
1061 if (ppro_with_ram_bug()) {
1062 e820__range_update(0x70000000ULL, 0x40000ULL, E820_TYPE_RAM,
1063 E820_TYPE_RESERVED);
1064 e820__update_table(e820_table);
1065 printk(KERN_INFO "fixed physical RAM map:\n");
1066 e820__print_table("bad_ppro");
1068 #else
1069 early_gart_iommu_check();
1070 #endif
1073 * partially used pages are not usable - thus
1074 * we are rounding upwards:
1076 max_pfn = e820__end_of_ram_pfn();
1078 /* update e820 for memory not covered by WB MTRRs */
1079 mtrr_bp_init();
1080 if (mtrr_trim_uncached_memory(max_pfn))
1081 max_pfn = e820__end_of_ram_pfn();
1083 max_possible_pfn = max_pfn;
1086 * This call is required when the CPU does not support PAT. If
1087 * mtrr_bp_init() invoked it already via pat_init() the call has no
1088 * effect.
1090 init_cache_modes();
1093 * Define random base addresses for memory sections after max_pfn is
1094 * defined and before each memory section base is used.
1096 kernel_randomize_memory();
1098 #ifdef CONFIG_X86_32
1099 /* max_low_pfn get updated here */
1100 find_low_pfn_range();
1101 #else
1102 check_x2apic();
1104 /* How many end-of-memory variables you have, grandma! */
1105 /* need this before calling reserve_initrd */
1106 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1107 max_low_pfn = e820__end_of_low_ram_pfn();
1108 else
1109 max_low_pfn = max_pfn;
1111 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1112 #endif
1115 * Find and reserve possible boot-time SMP configuration:
1117 find_smp_config();
1119 reserve_ibft_region();
1121 early_alloc_pgt_buf();
1124 * Need to conclude brk, before e820__memblock_setup()
1125 * it could use memblock_find_in_range, could overlap with
1126 * brk area.
1128 reserve_brk();
1130 cleanup_highmap();
1132 memblock_set_current_limit(ISA_END_ADDRESS);
1133 e820__memblock_setup();
1135 if (!early_xdbc_setup_hardware())
1136 early_xdbc_register_console();
1138 reserve_bios_regions();
1140 if (efi_enabled(EFI_MEMMAP)) {
1141 efi_fake_memmap();
1142 efi_find_mirror();
1143 efi_esrt_init();
1146 * The EFI specification says that boot service code won't be
1147 * called after ExitBootServices(). This is, in fact, a lie.
1149 efi_reserve_boot_services();
1152 /* preallocate 4k for mptable mpc */
1153 e820__memblock_alloc_reserved_mpc_new();
1155 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1156 setup_bios_corruption_check();
1157 #endif
1159 #ifdef CONFIG_X86_32
1160 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1161 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1162 #endif
1164 reserve_real_mode();
1166 trim_platform_memory_ranges();
1167 trim_low_memory_range();
1169 init_mem_mapping();
1171 idt_setup_early_pf();
1174 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1175 * with the current CR4 value. This may not be necessary, but
1176 * auditing all the early-boot CR4 manipulation would be needed to
1177 * rule it out.
1179 * Mask off features that don't work outside long mode (just
1180 * PCIDE for now).
1182 mmu_cr4_features = __read_cr4() & ~X86_CR4_PCIDE;
1184 memblock_set_current_limit(get_max_mapped());
1187 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1190 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1191 if (init_ohci1394_dma_early)
1192 init_ohci1394_dma_on_all_controllers();
1193 #endif
1194 /* Allocate bigger log buffer */
1195 setup_log_buf(1);
1197 if (efi_enabled(EFI_BOOT)) {
1198 switch (boot_params.secure_boot) {
1199 case efi_secureboot_mode_disabled:
1200 pr_info("Secure boot disabled\n");
1201 break;
1202 case efi_secureboot_mode_enabled:
1203 pr_info("Secure boot enabled\n");
1204 break;
1205 default:
1206 pr_info("Secure boot could not be determined\n");
1207 break;
1211 reserve_initrd();
1213 acpi_table_upgrade();
1215 vsmp_init();
1217 io_delay_init();
1219 early_platform_quirks();
1222 * Parse the ACPI tables for possible boot-time SMP configuration.
1224 acpi_boot_table_init();
1226 early_acpi_boot_init();
1228 initmem_init();
1229 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1232 * Reserve memory for crash kernel after SRAT is parsed so that it
1233 * won't consume hotpluggable memory.
1235 reserve_crashkernel();
1237 memblock_find_dma_reserve();
1239 #ifdef CONFIG_KVM_GUEST
1240 kvmclock_init();
1241 #endif
1243 x86_init.paging.pagetable_init();
1245 kasan_init();
1248 * Sync back kernel address range.
1250 * FIXME: Can the later sync in setup_cpu_entry_areas() replace
1251 * this call?
1253 sync_initial_page_table();
1255 tboot_probe();
1257 map_vsyscall();
1259 generic_apic_probe();
1261 early_quirks();
1264 * Read APIC and some other early information from ACPI tables.
1266 acpi_boot_init();
1267 sfi_init();
1268 x86_dtb_init();
1271 * get boot-time SMP configuration:
1273 get_smp_config();
1276 * Systems w/o ACPI and mptables might not have it mapped the local
1277 * APIC yet, but prefill_possible_map() might need to access it.
1279 init_apic_mappings();
1281 prefill_possible_map();
1283 init_cpu_to_node();
1285 io_apic_init_mappings();
1287 kvm_guest_init();
1289 e820__reserve_resources();
1290 e820__register_nosave_regions(max_pfn);
1292 x86_init.resources.reserve_resources();
1294 e820__setup_pci_gap();
1296 #ifdef CONFIG_VT
1297 #if defined(CONFIG_VGA_CONSOLE)
1298 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1299 conswitchp = &vga_con;
1300 #elif defined(CONFIG_DUMMY_CONSOLE)
1301 conswitchp = &dummy_con;
1302 #endif
1303 #endif
1304 x86_init.oem.banner();
1306 x86_init.timers.wallclock_init();
1308 mcheck_init();
1310 arch_init_ideal_nops();
1312 register_refined_jiffies(CLOCK_TICK_RATE);
1314 #ifdef CONFIG_EFI
1315 if (efi_enabled(EFI_BOOT))
1316 efi_apply_memmap_quirks();
1317 #endif
1319 unwind_init();
1322 #ifdef CONFIG_X86_32
1324 static struct resource video_ram_resource = {
1325 .name = "Video RAM area",
1326 .start = 0xa0000,
1327 .end = 0xbffff,
1328 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1331 void __init i386_reserve_resources(void)
1333 request_resource(&iomem_resource, &video_ram_resource);
1334 reserve_standard_io_resources();
1337 #endif /* CONFIG_X86_32 */
1339 static struct notifier_block kernel_offset_notifier = {
1340 .notifier_call = dump_kernel_offset
1343 static int __init register_kernel_offset_dumper(void)
1345 atomic_notifier_chain_register(&panic_notifier_list,
1346 &kernel_offset_notifier);
1347 return 0;
1349 __initcall(register_kernel_offset_dumper);
1351 void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1353 if (!boot_cpu_has(X86_FEATURE_OSPKE))
1354 return;
1356 seq_printf(m, "ProtectionKey: %8u\n", vma_pkey(vma));