1 /*
2 * Copyright (C) 1995 Linus Torvalds
3 *
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
5 *
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
8 *
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
11 *
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
14 *
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
17 *
18 */
19
20 /*
21 * This file handles the architecture-dependent parts of initialization
22 */
23
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>
53
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>
61
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>
67
68 #include <linux/percpu.h>
69 #include <linux/crash_dump.h>
70 #include <linux/tboot.h>
71 #include <linux/jiffies.h>
72
73 #include <video/edid.h>
74
75 #include <asm/mtrr.h>
76 #include <asm/apic.h>
77 #include <asm/realmode.h>
78 #include <asm/e820.h>
79 #include <asm/mpspec.h>
80 #include <asm/setup.h>
81 #include <asm/efi.h>
82 #include <asm/timer.h>
83 #include <asm/i8259.h>
84 #include <asm/sections.h>
85 #include <asm/io_apic.h>
86 #include <asm/ist.h>
87 #include <asm/setup_arch.h>
88 #include <asm/bios_ebda.h>
89 #include <asm/cacheflush.h>
90 #include <asm/processor.h>
91 #include <asm/bugs.h>
92 #include <asm/kasan.h>
93
94 #include <asm/vsyscall.h>
95 #include <asm/cpu.h>
96 #include <asm/desc.h>
97 #include <asm/dma.h>
98 #include <asm/iommu.h>
99 #include <asm/gart.h>
100 #include <asm/mmu_context.h>
101 #include <asm/proto.h>
102
103 #include <asm/paravirt.h>
104 #include <asm/hypervisor.h>
105 #include <asm/olpc_ofw.h>
106
107 #include <asm/percpu.h>
108 #include <asm/topology.h>
109 #include <asm/apicdef.h>
110 #include <asm/amd_nb.h>
111 #include <asm/mce.h>
112 #include <asm/alternative.h>
113 #include <asm/prom.h>
114 #include <asm/microcode.h>
115 #include <asm/mmu_context.h>
116 #include <asm/kaslr.h>
117 #include <asm/kaiser.h>
118
119 /*
120 * max_low_pfn_mapped: highest direct mapped pfn under 4GB
121 * max_pfn_mapped: highest direct mapped pfn over 4GB
122 *
123 * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
124 * represented by pfn_mapped
125 */
126 unsigned long max_low_pfn_mapped;
127 unsigned long max_pfn_mapped;
128
129 #ifdef CONFIG_DMI
130 RESERVE_BRK(dmi_alloc, 65536);
131 #endif
132
133
134 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
135 unsigned long _brk_end = (unsigned long)__brk_base;
136
137 #ifdef CONFIG_X86_64
default_cpu_present_to_apicid(int mps_cpu)138 int default_cpu_present_to_apicid(int mps_cpu)
139 {
140 return __default_cpu_present_to_apicid(mps_cpu);
141 }
142
default_check_phys_apicid_present(int phys_apicid)143 int default_check_phys_apicid_present(int phys_apicid)
144 {
145 return __default_check_phys_apicid_present(phys_apicid);
146 }
147 #endif
148
149 struct boot_params boot_params;
150
151 /*
152 * Machine setup..
153 */
154 static struct resource data_resource = {
155 .name = "Kernel data",
156 .start = 0,
157 .end = 0,
158 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
159 };
160
161 static struct resource code_resource = {
162 .name = "Kernel code",
163 .start = 0,
164 .end = 0,
165 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
166 };
167
168 static struct resource bss_resource = {
169 .name = "Kernel bss",
170 .start = 0,
171 .end = 0,
172 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
173 };
174
175
176 #ifdef CONFIG_X86_32
177 /* cpu data as detected by the assembly code in head.S */
178 struct cpuinfo_x86 new_cpu_data = {
179 .wp_works_ok = -1,
180 };
181 /* common cpu data for all cpus */
182 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
183 .wp_works_ok = -1,
184 };
185 EXPORT_SYMBOL(boot_cpu_data);
186
187 unsigned int def_to_bigsmp;
188
189 /* for MCA, but anyone else can use it if they want */
190 unsigned int machine_id;
191 unsigned int machine_submodel_id;
192 unsigned int BIOS_revision;
193
194 struct apm_info apm_info;
195 EXPORT_SYMBOL(apm_info);
196
197 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
198 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
199 struct ist_info ist_info;
200 EXPORT_SYMBOL(ist_info);
201 #else
202 struct ist_info ist_info;
203 #endif
204
205 #else
206 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
207 .x86_phys_bits = MAX_PHYSMEM_BITS,
208 };
209 EXPORT_SYMBOL(boot_cpu_data);
210 #endif
211
212
213 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
214 __visible unsigned long mmu_cr4_features __ro_after_init;
215 #else
216 __visible unsigned long mmu_cr4_features __ro_after_init = X86_CR4_PAE;
217 #endif
218
219 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
220 int bootloader_type, bootloader_version;
221
222 /*
223 * Setup options
224 */
225 struct screen_info screen_info;
226 EXPORT_SYMBOL(screen_info);
227 struct edid_info edid_info;
228 EXPORT_SYMBOL_GPL(edid_info);
229
230 extern int root_mountflags;
231
232 unsigned long saved_video_mode;
233
234 #define RAMDISK_IMAGE_START_MASK 0x07FF
235 #define RAMDISK_PROMPT_FLAG 0x8000
236 #define RAMDISK_LOAD_FLAG 0x4000
237
238 static char __initdata command_line[COMMAND_LINE_SIZE];
239 #ifdef CONFIG_CMDLINE_BOOL
240 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
241 #endif
242
243 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
244 struct edd edd;
245 #ifdef CONFIG_EDD_MODULE
246 EXPORT_SYMBOL(edd);
247 #endif
248 /**
249 * copy_edd() - Copy the BIOS EDD information
250 * from boot_params into a safe place.
251 *
252 */
copy_edd(void)253 static inline void __init copy_edd(void)
254 {
255 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
256 sizeof(edd.mbr_signature));
257 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
258 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
259 edd.edd_info_nr = boot_params.eddbuf_entries;
260 }
261 #else
copy_edd(void)262 static inline void __init copy_edd(void)
263 {
264 }
265 #endif
266
extend_brk(size_t size,size_t align)267 void * __init extend_brk(size_t size, size_t align)
268 {
269 size_t mask = align - 1;
270 void *ret;
271
272 BUG_ON(_brk_start == 0);
273 BUG_ON(align & mask);
274
275 _brk_end = (_brk_end + mask) & ~mask;
276 BUG_ON((char *)(_brk_end + size) > __brk_limit);
277
278 ret = (void *)_brk_end;
279 _brk_end += size;
280
281 memset(ret, 0, size);
282
283 return ret;
284 }
285
286 #ifdef CONFIG_X86_32
cleanup_highmap(void)287 static void __init cleanup_highmap(void)
288 {
289 }
290 #endif
291
reserve_brk(void)292 static void __init reserve_brk(void)
293 {
294 if (_brk_end > _brk_start)
295 memblock_reserve(__pa_symbol(_brk_start),
296 _brk_end - _brk_start);
297
298 /* Mark brk area as locked down and no longer taking any
299 new allocations */
300 _brk_start = 0;
301 }
302
303 u64 relocated_ramdisk;
304
305 #ifdef CONFIG_BLK_DEV_INITRD
306
get_ramdisk_image(void)307 static u64 __init get_ramdisk_image(void)
308 {
309 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
310
311 ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
312
313 return ramdisk_image;
314 }
get_ramdisk_size(void)315 static u64 __init get_ramdisk_size(void)
316 {
317 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
318
319 ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
320
321 return ramdisk_size;
322 }
323
relocate_initrd(void)324 static void __init relocate_initrd(void)
325 {
326 /* Assume only end is not page aligned */
327 u64 ramdisk_image = get_ramdisk_image();
328 u64 ramdisk_size = get_ramdisk_size();
329 u64 area_size = PAGE_ALIGN(ramdisk_size);
330
331 /* We need to move the initrd down into directly mapped mem */
332 relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
333 area_size, PAGE_SIZE);
334
335 if (!relocated_ramdisk)
336 panic("Cannot find place for new RAMDISK of size %lld\n",
337 ramdisk_size);
338
339 /* Note: this includes all the mem currently occupied by
340 the initrd, we rely on that fact to keep the data intact. */
341 memblock_reserve(relocated_ramdisk, area_size);
342 initrd_start = relocated_ramdisk + PAGE_OFFSET;
343 initrd_end = initrd_start + ramdisk_size;
344 printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
345 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
346
347 copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
348
349 printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
350 " [mem %#010llx-%#010llx]\n",
351 ramdisk_image, ramdisk_image + ramdisk_size - 1,
352 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
353 }
354
early_reserve_initrd(void)355 static void __init early_reserve_initrd(void)
356 {
357 /* Assume only end is not page aligned */
358 u64 ramdisk_image = get_ramdisk_image();
359 u64 ramdisk_size = get_ramdisk_size();
360 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
361
362 if (!boot_params.hdr.type_of_loader ||
363 !ramdisk_image || !ramdisk_size)
364 return; /* No initrd provided by bootloader */
365
366 memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
367 }
reserve_initrd(void)368 static void __init reserve_initrd(void)
369 {
370 /* Assume only end is not page aligned */
371 u64 ramdisk_image = get_ramdisk_image();
372 u64 ramdisk_size = get_ramdisk_size();
373 u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size);
374 u64 mapped_size;
375
376 if (!boot_params.hdr.type_of_loader ||
377 !ramdisk_image || !ramdisk_size)
378 return; /* No initrd provided by bootloader */
379
380 initrd_start = 0;
381
382 mapped_size = memblock_mem_size(max_pfn_mapped);
383 if (ramdisk_size >= (mapped_size>>1))
384 panic("initrd too large to handle, "
385 "disabling initrd (%lld needed, %lld available)\n",
386 ramdisk_size, mapped_size>>1);
387
388 printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
389 ramdisk_end - 1);
390
391 if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
392 PFN_DOWN(ramdisk_end))) {
393 /* All are mapped, easy case */
394 initrd_start = ramdisk_image + PAGE_OFFSET;
395 initrd_end = initrd_start + ramdisk_size;
396 return;
397 }
398
399 relocate_initrd();
400
401 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
402 }
403
404 #else
early_reserve_initrd(void)405 static void __init early_reserve_initrd(void)
406 {
407 }
reserve_initrd(void)408 static void __init reserve_initrd(void)
409 {
410 }
411 #endif /* CONFIG_BLK_DEV_INITRD */
412
parse_setup_data(void)413 static void __init parse_setup_data(void)
414 {
415 struct setup_data *data;
416 u64 pa_data, pa_next;
417
418 pa_data = boot_params.hdr.setup_data;
419 while (pa_data) {
420 u32 data_len, data_type;
421
422 data = early_memremap(pa_data, sizeof(*data));
423 data_len = data->len + sizeof(struct setup_data);
424 data_type = data->type;
425 pa_next = data->next;
426 early_memunmap(data, sizeof(*data));
427
428 switch (data_type) {
429 case SETUP_E820_EXT:
430 parse_e820_ext(pa_data, data_len);
431 break;
432 case SETUP_DTB:
433 add_dtb(pa_data);
434 break;
435 case SETUP_EFI:
436 parse_efi_setup(pa_data, data_len);
437 break;
438 default:
439 break;
440 }
441 pa_data = pa_next;
442 }
443 }
444
e820_reserve_setup_data(void)445 static void __init e820_reserve_setup_data(void)
446 {
447 struct setup_data *data;
448 u64 pa_data;
449
450 pa_data = boot_params.hdr.setup_data;
451 if (!pa_data)
452 return;
453
454 while (pa_data) {
455 data = early_memremap(pa_data, sizeof(*data));
456 e820_update_range(pa_data, sizeof(*data)+data->len,
457 E820_RAM, E820_RESERVED_KERN);
458 pa_data = data->next;
459 early_memunmap(data, sizeof(*data));
460 }
461
462 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
463 memcpy(e820_saved, e820, sizeof(struct e820map));
464 printk(KERN_INFO "extended physical RAM map:\n");
465 e820_print_map("reserve setup_data");
466 }
467
memblock_x86_reserve_range_setup_data(void)468 static void __init memblock_x86_reserve_range_setup_data(void)
469 {
470 struct setup_data *data;
471 u64 pa_data;
472
473 pa_data = boot_params.hdr.setup_data;
474 while (pa_data) {
475 data = early_memremap(pa_data, sizeof(*data));
476 memblock_reserve(pa_data, sizeof(*data) + data->len);
477 pa_data = data->next;
478 early_memunmap(data, sizeof(*data));
479 }
480 }
481
482 /*
483 * --------- Crashkernel reservation ------------------------------
484 */
485
486 #ifdef CONFIG_KEXEC_CORE
487
488 /* 16M alignment for crash kernel regions */
489 #define CRASH_ALIGN (16 << 20)
490
491 /*
492 * Keep the crash kernel below this limit. On 32 bits earlier kernels
493 * would limit the kernel to the low 512 MiB due to mapping restrictions.
494 * On 64bit, old kexec-tools need to under 896MiB.
495 */
496 #ifdef CONFIG_X86_32
497 # define CRASH_ADDR_LOW_MAX (512 << 20)
498 # define CRASH_ADDR_HIGH_MAX (512 << 20)
499 #else
500 # define CRASH_ADDR_LOW_MAX (896UL << 20)
501 # define CRASH_ADDR_HIGH_MAX MAXMEM
502 #endif
503
reserve_crashkernel_low(void)504 static int __init reserve_crashkernel_low(void)
505 {
506 #ifdef CONFIG_X86_64
507 unsigned long long base, low_base = 0, low_size = 0;
508 unsigned long total_low_mem;
509 int ret;
510
511 total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
512
513 /* crashkernel=Y,low */
514 ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
515 if (ret) {
516 /*
517 * two parts from lib/swiotlb.c:
518 * -swiotlb size: user-specified with swiotlb= or default.
519 *
520 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
521 * to 8M for other buffers that may need to stay low too. Also
522 * make sure we allocate enough extra low memory so that we
523 * don't run out of DMA buffers for 32-bit devices.
524 */
525 low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
526 } else {
527 /* passed with crashkernel=0,low ? */
528 if (!low_size)
529 return 0;
530 }
531
532 low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN);
533 if (!low_base) {
534 pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
535 (unsigned long)(low_size >> 20));
536 return -ENOMEM;
537 }
538
539 ret = memblock_reserve(low_base, low_size);
540 if (ret) {
541 pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
542 return ret;
543 }
544
545 pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
546 (unsigned long)(low_size >> 20),
547 (unsigned long)(low_base >> 20),
548 (unsigned long)(total_low_mem >> 20));
549
550 crashk_low_res.start = low_base;
551 crashk_low_res.end = low_base + low_size - 1;
552 insert_resource(&iomem_resource, &crashk_low_res);
553 #endif
554 return 0;
555 }
556
reserve_crashkernel(void)557 static void __init reserve_crashkernel(void)
558 {
559 unsigned long long crash_size, crash_base, total_mem;
560 bool high = false;
561 int ret;
562
563 total_mem = memblock_phys_mem_size();
564
565 /* crashkernel=XM */
566 ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
567 if (ret != 0 || crash_size <= 0) {
568 /* crashkernel=X,high */
569 ret = parse_crashkernel_high(boot_command_line, total_mem,
570 &crash_size, &crash_base);
571 if (ret != 0 || crash_size <= 0)
572 return;
573 high = true;
574 }
575
576 /* 0 means: find the address automatically */
577 if (crash_base <= 0) {
578 /*
579 * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
580 */
581 crash_base = memblock_find_in_range(CRASH_ALIGN,
582 high ? CRASH_ADDR_HIGH_MAX
583 : CRASH_ADDR_LOW_MAX,
584 crash_size, CRASH_ALIGN);
585 if (!crash_base) {
586 pr_info("crashkernel reservation failed - No suitable area found.\n");
587 return;
588 }
589
590 } else {
591 unsigned long long start;
592
593 start = memblock_find_in_range(crash_base,
594 crash_base + crash_size,
595 crash_size, 1 << 20);
596 if (start != crash_base) {
597 pr_info("crashkernel reservation failed - memory is in use.\n");
598 return;
599 }
600 }
601 ret = memblock_reserve(crash_base, crash_size);
602 if (ret) {
603 pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
604 return;
605 }
606
607 if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
608 memblock_free(crash_base, crash_size);
609 return;
610 }
611
612 pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
613 (unsigned long)(crash_size >> 20),
614 (unsigned long)(crash_base >> 20),
615 (unsigned long)(total_mem >> 20));
616
617 crashk_res.start = crash_base;
618 crashk_res.end = crash_base + crash_size - 1;
619 insert_resource(&iomem_resource, &crashk_res);
620 }
621 #else
reserve_crashkernel(void)622 static void __init reserve_crashkernel(void)
623 {
624 }
625 #endif
626
627 static struct resource standard_io_resources[] = {
628 { .name = "dma1", .start = 0x00, .end = 0x1f,
629 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
630 { .name = "pic1", .start = 0x20, .end = 0x21,
631 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
632 { .name = "timer0", .start = 0x40, .end = 0x43,
633 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
634 { .name = "timer1", .start = 0x50, .end = 0x53,
635 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
636 { .name = "keyboard", .start = 0x60, .end = 0x60,
637 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
638 { .name = "keyboard", .start = 0x64, .end = 0x64,
639 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
640 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
641 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
642 { .name = "pic2", .start = 0xa0, .end = 0xa1,
643 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
644 { .name = "dma2", .start = 0xc0, .end = 0xdf,
645 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
646 { .name = "fpu", .start = 0xf0, .end = 0xff,
647 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
648 };
649
reserve_standard_io_resources(void)650 void __init reserve_standard_io_resources(void)
651 {
652 int i;
653
654 /* request I/O space for devices used on all i[345]86 PCs */
655 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
656 request_resource(&ioport_resource, &standard_io_resources[i]);
657
658 }
659
reserve_ibft_region(void)660 static __init void reserve_ibft_region(void)
661 {
662 unsigned long addr, size = 0;
663
664 addr = find_ibft_region(&size);
665
666 if (size)
667 memblock_reserve(addr, size);
668 }
669
snb_gfx_workaround_needed(void)670 static bool __init snb_gfx_workaround_needed(void)
671 {
672 #ifdef CONFIG_PCI
673 int i;
674 u16 vendor, devid;
675 static const __initconst u16 snb_ids[] = {
676 0x0102,
677 0x0112,
678 0x0122,
679 0x0106,
680 0x0116,
681 0x0126,
682 0x010a,
683 };
684
685 /* Assume no if something weird is going on with PCI */
686 if (!early_pci_allowed())
687 return false;
688
689 vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
690 if (vendor != 0x8086)
691 return false;
692
693 devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
694 for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
695 if (devid == snb_ids[i])
696 return true;
697 #endif
698
699 return false;
700 }
701
702 /*
703 * Sandy Bridge graphics has trouble with certain ranges, exclude
704 * them from allocation.
705 */
trim_snb_memory(void)706 static void __init trim_snb_memory(void)
707 {
708 static const __initconst unsigned long bad_pages[] = {
709 0x20050000,
710 0x20110000,
711 0x20130000,
712 0x20138000,
713 0x40004000,
714 };
715 int i;
716
717 if (!snb_gfx_workaround_needed())
718 return;
719
720 printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
721
722 /*
723 * Reserve all memory below the 1 MB mark that has not
724 * already been reserved.
725 */
726 memblock_reserve(0, 1<<20);
727
728 for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
729 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
730 printk(KERN_WARNING "failed to reserve 0x%08lx\n",
731 bad_pages[i]);
732 }
733 }
734
735 /*
736 * Here we put platform-specific memory range workarounds, i.e.
737 * memory known to be corrupt or otherwise in need to be reserved on
738 * specific platforms.
739 *
740 * If this gets used more widely it could use a real dispatch mechanism.
741 */
trim_platform_memory_ranges(void)742 static void __init trim_platform_memory_ranges(void)
743 {
744 trim_snb_memory();
745 }
746
trim_bios_range(void)747 static void __init trim_bios_range(void)
748 {
749 /*
750 * A special case is the first 4Kb of memory;
751 * This is a BIOS owned area, not kernel ram, but generally
752 * not listed as such in the E820 table.
753 *
754 * This typically reserves additional memory (64KiB by default)
755 * since some BIOSes are known to corrupt low memory. See the
756 * Kconfig help text for X86_RESERVE_LOW.
757 */
758 e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
759
760 /*
761 * special case: Some BIOSen report the PC BIOS
762 * area (640->1Mb) as ram even though it is not.
763 * take them out.
764 */
765 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
766
767 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
768 }
769
770 /* called before trim_bios_range() to spare extra sanitize */
e820_add_kernel_range(void)771 static void __init e820_add_kernel_range(void)
772 {
773 u64 start = __pa_symbol(_text);
774 u64 size = __pa_symbol(_end) - start;
775
776 /*
777 * Complain if .text .data and .bss are not marked as E820_RAM and
778 * attempt to fix it by adding the range. We may have a confused BIOS,
779 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
780 * exclude kernel range. If we really are running on top non-RAM,
781 * we will crash later anyways.
782 */
783 if (e820_all_mapped(start, start + size, E820_RAM))
784 return;
785
786 pr_warn(".text .data .bss are not marked as E820_RAM!\n");
787 e820_remove_range(start, size, E820_RAM, 0);
788 e820_add_region(start, size, E820_RAM);
789 }
790
791 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
792
parse_reservelow(char * p)793 static int __init parse_reservelow(char *p)
794 {
795 unsigned long long size;
796
797 if (!p)
798 return -EINVAL;
799
800 size = memparse(p, &p);
801
802 if (size < 4096)
803 size = 4096;
804
805 if (size > 640*1024)
806 size = 640*1024;
807
808 reserve_low = size;
809
810 return 0;
811 }
812
813 early_param("reservelow", parse_reservelow);
814
trim_low_memory_range(void)815 static void __init trim_low_memory_range(void)
816 {
817 memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
818 }
819
820 /*
821 * Dump out kernel offset information on panic.
822 */
823 static int
dump_kernel_offset(struct notifier_block * self,unsigned long v,void * p)824 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
825 {
826 if (kaslr_enabled()) {
827 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
828 kaslr_offset(),
829 __START_KERNEL,
830 __START_KERNEL_map,
831 MODULES_VADDR-1);
832 } else {
833 pr_emerg("Kernel Offset: disabled\n");
834 }
835
836 return 0;
837 }
838
839 /*
840 * Determine if we were loaded by an EFI loader. If so, then we have also been
841 * passed the efi memmap, systab, etc., so we should use these data structures
842 * for initialization. Note, the efi init code path is determined by the
843 * global efi_enabled. This allows the same kernel image to be used on existing
844 * systems (with a traditional BIOS) as well as on EFI systems.
845 */
846 /*
847 * setup_arch - architecture-specific boot-time initializations
848 *
849 * Note: On x86_64, fixmaps are ready for use even before this is called.
850 */
851
setup_arch(char ** cmdline_p)852 void __init setup_arch(char **cmdline_p)
853 {
854 memblock_reserve(__pa_symbol(_text),
855 (unsigned long)__bss_stop - (unsigned long)_text);
856
857 early_reserve_initrd();
858
859 /*
860 * At this point everything still needed from the boot loader
861 * or BIOS or kernel text should be early reserved or marked not
862 * RAM in e820. All other memory is free game.
863 */
864
865 #ifdef CONFIG_X86_32
866 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
867
868 /*
869 * copy kernel address range established so far and switch
870 * to the proper swapper page table
871 */
872 clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY,
873 initial_page_table + KERNEL_PGD_BOUNDARY,
874 KERNEL_PGD_PTRS);
875
876 load_cr3(swapper_pg_dir);
877 /*
878 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
879 * a cr3 based tlb flush, so the following __flush_tlb_all()
880 * will not flush anything because the cpu quirk which clears
881 * X86_FEATURE_PGE has not been invoked yet. Though due to the
882 * load_cr3() above the TLB has been flushed already. The
883 * quirk is invoked before subsequent calls to __flush_tlb_all()
884 * so proper operation is guaranteed.
885 */
886 __flush_tlb_all();
887 #else
888 printk(KERN_INFO "Command line: %s\n", boot_command_line);
889 #endif
890
891 /*
892 * If we have OLPC OFW, we might end up relocating the fixmap due to
893 * reserve_top(), so do this before touching the ioremap area.
894 */
895 olpc_ofw_detect();
896
897 early_trap_init();
898 early_cpu_init();
899 early_ioremap_init();
900
901 setup_olpc_ofw_pgd();
902
903 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
904 screen_info = boot_params.screen_info;
905 edid_info = boot_params.edid_info;
906 #ifdef CONFIG_X86_32
907 apm_info.bios = boot_params.apm_bios_info;
908 ist_info = boot_params.ist_info;
909 #endif
910 saved_video_mode = boot_params.hdr.vid_mode;
911 bootloader_type = boot_params.hdr.type_of_loader;
912 if ((bootloader_type >> 4) == 0xe) {
913 bootloader_type &= 0xf;
914 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
915 }
916 bootloader_version = bootloader_type & 0xf;
917 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
918
919 #ifdef CONFIG_BLK_DEV_RAM
920 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
921 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
922 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
923 #endif
924 #ifdef CONFIG_EFI
925 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
926 EFI32_LOADER_SIGNATURE, 4)) {
927 set_bit(EFI_BOOT, &efi.flags);
928 } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
929 EFI64_LOADER_SIGNATURE, 4)) {
930 set_bit(EFI_BOOT, &efi.flags);
931 set_bit(EFI_64BIT, &efi.flags);
932 }
933
934 if (efi_enabled(EFI_BOOT))
935 efi_memblock_x86_reserve_range();
936 #endif
937
938 x86_init.oem.arch_setup();
939
940 iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
941 setup_memory_map();
942 parse_setup_data();
943
944 copy_edd();
945
946 if (!boot_params.hdr.root_flags)
947 root_mountflags &= ~MS_RDONLY;
948 init_mm.start_code = (unsigned long) _text;
949 init_mm.end_code = (unsigned long) _etext;
950 init_mm.end_data = (unsigned long) _edata;
951 init_mm.brk = _brk_end;
952
953 mpx_mm_init(&init_mm);
954
955 code_resource.start = __pa_symbol(_text);
956 code_resource.end = __pa_symbol(_etext)-1;
957 data_resource.start = __pa_symbol(_etext);
958 data_resource.end = __pa_symbol(_edata)-1;
959 bss_resource.start = __pa_symbol(__bss_start);
960 bss_resource.end = __pa_symbol(__bss_stop)-1;
961
962 #ifdef CONFIG_CMDLINE_BOOL
963 #ifdef CONFIG_CMDLINE_OVERRIDE
964 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
965 #else
966 if (builtin_cmdline[0]) {
967 /* append boot loader cmdline to builtin */
968 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
969 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
970 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
971 }
972 #endif
973 #endif
974
975 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
976 *cmdline_p = command_line;
977
978 /*
979 * x86_configure_nx() is called before parse_early_param() to detect
980 * whether hardware doesn't support NX (so that the early EHCI debug
981 * console setup can safely call set_fixmap()). It may then be called
982 * again from within noexec_setup() during parsing early parameters
983 * to honor the respective command line option.
984 */
985 x86_configure_nx();
986
987 parse_early_param();
988
989 x86_report_nx();
990
991 /* after early param, so could get panic from serial */
992 memblock_x86_reserve_range_setup_data();
993
994 if (acpi_mps_check()) {
995 #ifdef CONFIG_X86_LOCAL_APIC
996 disable_apic = 1;
997 #endif
998 setup_clear_cpu_cap(X86_FEATURE_APIC);
999 }
1000
1001 #ifdef CONFIG_PCI
1002 if (pci_early_dump_regs)
1003 early_dump_pci_devices();
1004 #endif
1005
1006 /* update the e820_saved too */
1007 e820_reserve_setup_data();
1008 finish_e820_parsing();
1009
1010 if (efi_enabled(EFI_BOOT))
1011 efi_init();
1012
1013 dmi_scan_machine();
1014 dmi_memdev_walk();
1015 dmi_set_dump_stack_arch_desc();
1016
1017 /*
1018 * VMware detection requires dmi to be available, so this
1019 * needs to be done after dmi_scan_machine, for the BP.
1020 */
1021 init_hypervisor_platform();
1022
1023 /*
1024 * This needs to happen right after XENPV is set on xen and
1025 * kaiser_enabled is checked below in cleanup_highmap().
1026 */
1027 kaiser_check_boottime_disable();
1028
1029 x86_init.resources.probe_roms();
1030
1031 /* after parse_early_param, so could debug it */
1032 insert_resource(&iomem_resource, &code_resource);
1033 insert_resource(&iomem_resource, &data_resource);
1034 insert_resource(&iomem_resource, &bss_resource);
1035
1036 e820_add_kernel_range();
1037 trim_bios_range();
1038 #ifdef CONFIG_X86_32
1039 if (ppro_with_ram_bug()) {
1040 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1041 E820_RESERVED);
1042 sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
1043 printk(KERN_INFO "fixed physical RAM map:\n");
1044 e820_print_map("bad_ppro");
1045 }
1046 #else
1047 early_gart_iommu_check();
1048 #endif
1049
1050 /*
1051 * partially used pages are not usable - thus
1052 * we are rounding upwards:
1053 */
1054 max_pfn = e820_end_of_ram_pfn();
1055
1056 /* update e820 for memory not covered by WB MTRRs */
1057 mtrr_bp_init();
1058 if (mtrr_trim_uncached_memory(max_pfn))
1059 max_pfn = e820_end_of_ram_pfn();
1060
1061 max_possible_pfn = max_pfn;
1062
1063 /*
1064 * This call is required when the CPU does not support PAT. If
1065 * mtrr_bp_init() invoked it already via pat_init() the call has no
1066 * effect.
1067 */
1068 init_cache_modes();
1069
1070 /*
1071 * Define random base addresses for memory sections after max_pfn is
1072 * defined and before each memory section base is used.
1073 */
1074 kernel_randomize_memory();
1075
1076 #ifdef CONFIG_X86_32
1077 /* max_low_pfn get updated here */
1078 find_low_pfn_range();
1079 #else
1080 check_x2apic();
1081
1082 /* How many end-of-memory variables you have, grandma! */
1083 /* need this before calling reserve_initrd */
1084 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1085 max_low_pfn = e820_end_of_low_ram_pfn();
1086 else
1087 max_low_pfn = max_pfn;
1088
1089 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1090 #endif
1091
1092 /*
1093 * Find and reserve possible boot-time SMP configuration:
1094 */
1095 find_smp_config();
1096
1097 reserve_ibft_region();
1098
1099 early_alloc_pgt_buf();
1100
1101 /*
1102 * Need to conclude brk, before memblock_x86_fill()
1103 * it could use memblock_find_in_range, could overlap with
1104 * brk area.
1105 */
1106 reserve_brk();
1107
1108 cleanup_highmap();
1109
1110 memblock_set_current_limit(ISA_END_ADDRESS);
1111 memblock_x86_fill();
1112
1113 reserve_bios_regions();
1114
1115 if (efi_enabled(EFI_MEMMAP)) {
1116 efi_fake_memmap();
1117 efi_find_mirror();
1118 efi_esrt_init();
1119
1120 /*
1121 * The EFI specification says that boot service code won't be
1122 * called after ExitBootServices(). This is, in fact, a lie.
1123 */
1124 efi_reserve_boot_services();
1125 }
1126
1127 /* preallocate 4k for mptable mpc */
1128 early_reserve_e820_mpc_new();
1129
1130 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1131 setup_bios_corruption_check();
1132 #endif
1133
1134 #ifdef CONFIG_X86_32
1135 printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1136 (max_pfn_mapped<<PAGE_SHIFT) - 1);
1137 #endif
1138
1139 reserve_real_mode();
1140
1141 trim_platform_memory_ranges();
1142 trim_low_memory_range();
1143
1144 init_mem_mapping();
1145
1146 early_trap_pf_init();
1147
1148 /*
1149 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1150 * with the current CR4 value. This may not be necessary, but
1151 * auditing all the early-boot CR4 manipulation would be needed to
1152 * rule it out.
1153 */
1154 mmu_cr4_features = __read_cr4();
1155
1156 memblock_set_current_limit(get_max_mapped());
1157
1158 /*
1159 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1160 */
1161
1162 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1163 if (init_ohci1394_dma_early)
1164 init_ohci1394_dma_on_all_controllers();
1165 #endif
1166 /* Allocate bigger log buffer */
1167 setup_log_buf(1);
1168
1169 reserve_initrd();
1170
1171 acpi_table_upgrade();
1172
1173 vsmp_init();
1174
1175 io_delay_init();
1176
1177 /*
1178 * Parse the ACPI tables for possible boot-time SMP configuration.
1179 */
1180 acpi_boot_table_init();
1181
1182 early_acpi_boot_init();
1183
1184 initmem_init();
1185 dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1186
1187 /*
1188 * Reserve memory for crash kernel after SRAT is parsed so that it
1189 * won't consume hotpluggable memory.
1190 */
1191 reserve_crashkernel();
1192
1193 memblock_find_dma_reserve();
1194
1195 #ifdef CONFIG_KVM_GUEST
1196 kvmclock_init();
1197 #endif
1198
1199 x86_init.paging.pagetable_init();
1200
1201 kasan_init();
1202
1203 #ifdef CONFIG_X86_32
1204 /* sync back kernel address range */
1205 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1206 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1207 KERNEL_PGD_PTRS);
1208
1209 /*
1210 * sync back low identity map too. It is used for example
1211 * in the 32-bit EFI stub.
1212 */
1213 clone_pgd_range(initial_page_table,
1214 swapper_pg_dir + KERNEL_PGD_BOUNDARY,
1215 min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY));
1216 #endif
1217
1218 tboot_probe();
1219
1220 map_vsyscall();
1221
1222 generic_apic_probe();
1223
1224 early_quirks();
1225
1226 /*
1227 * Read APIC and some other early information from ACPI tables.
1228 */
1229 acpi_boot_init();
1230 sfi_init();
1231 x86_dtb_init();
1232
1233 /*
1234 * get boot-time SMP configuration:
1235 */
1236 get_smp_config();
1237
1238 /*
1239 * Systems w/o ACPI and mptables might not have it mapped the local
1240 * APIC yet, but prefill_possible_map() might need to access it.
1241 */
1242 init_apic_mappings();
1243
1244 prefill_possible_map();
1245
1246 init_cpu_to_node();
1247
1248 io_apic_init_mappings();
1249
1250 kvm_guest_init();
1251
1252 e820_reserve_resources();
1253 e820_mark_nosave_regions(max_low_pfn);
1254
1255 x86_init.resources.reserve_resources();
1256
1257 e820_setup_gap();
1258
1259 #ifdef CONFIG_VT
1260 #if defined(CONFIG_VGA_CONSOLE)
1261 if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1262 conswitchp = &vga_con;
1263 #elif defined(CONFIG_DUMMY_CONSOLE)
1264 conswitchp = &dummy_con;
1265 #endif
1266 #endif
1267 x86_init.oem.banner();
1268
1269 x86_init.timers.wallclock_init();
1270
1271 mcheck_init();
1272
1273 arch_init_ideal_nops();
1274
1275 register_refined_jiffies(CLOCK_TICK_RATE);
1276
1277 #ifdef CONFIG_EFI
1278 if (efi_enabled(EFI_BOOT))
1279 efi_apply_memmap_quirks();
1280 #endif
1281 }
1282
1283 #ifdef CONFIG_X86_32
1284
1285 static struct resource video_ram_resource = {
1286 .name = "Video RAM area",
1287 .start = 0xa0000,
1288 .end = 0xbffff,
1289 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1290 };
1291
i386_reserve_resources(void)1292 void __init i386_reserve_resources(void)
1293 {
1294 request_resource(&iomem_resource, &video_ram_resource);
1295 reserve_standard_io_resources();
1296 }
1297
1298 #endif /* CONFIG_X86_32 */
1299
1300 static struct notifier_block kernel_offset_notifier = {
1301 .notifier_call = dump_kernel_offset
1302 };
1303
register_kernel_offset_dumper(void)1304 static int __init register_kernel_offset_dumper(void)
1305 {
1306 atomic_notifier_chain_register(&panic_notifier_list,
1307 &kernel_offset_notifier);
1308 return 0;
1309 }
1310 __initcall(register_kernel_offset_dumper);
1311
arch_show_smap(struct seq_file * m,struct vm_area_struct * vma)1312 void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1313 {
1314 if (!boot_cpu_has(X86_FEATURE_OSPKE))
1315 return;
1316
1317 seq_printf(m, "ProtectionKey: %8u\n", vma_pkey(vma));
1318 }
1319