xref: /linux/arch/x86/kernel/setup.c (revision 0483e1fa6e09d4948272680f691dccb1edb9677f)
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/module.h>
40 #include <linux/efi.h>
41 #include <linux/init.h>
42 #include <linux/edd.h>
43 #include <linux/iscsi_ibft.h>
44 #include <linux/nodemask.h>
45 #include <linux/kexec.h>
46 #include <linux/dmi.h>
47 #include <linux/pfn.h>
48 #include <linux/pci.h>
49 #include <asm/pci-direct.h>
50 #include <linux/init_ohci1394_dma.h>
51 #include <linux/kvm_para.h>
52 #include <linux/dma-contiguous.h>
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 
118 /*
119  * max_low_pfn_mapped: highest direct mapped pfn under 4GB
120  * max_pfn_mapped:     highest direct mapped pfn over 4GB
121  *
122  * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
123  * represented by pfn_mapped
124  */
125 unsigned long max_low_pfn_mapped;
126 unsigned long max_pfn_mapped;
127 
128 #ifdef CONFIG_DMI
129 RESERVE_BRK(dmi_alloc, 65536);
130 #endif
131 
132 
133 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
134 unsigned long _brk_end = (unsigned long)__brk_base;
135 
136 #ifdef CONFIG_X86_64
137 int default_cpu_present_to_apicid(int mps_cpu)
138 {
139 	return __default_cpu_present_to_apicid(mps_cpu);
140 }
141 
142 int default_check_phys_apicid_present(int phys_apicid)
143 {
144 	return __default_check_phys_apicid_present(phys_apicid);
145 }
146 #endif
147 
148 struct boot_params boot_params;
149 
150 /*
151  * Machine setup..
152  */
153 static struct resource data_resource = {
154 	.name	= "Kernel data",
155 	.start	= 0,
156 	.end	= 0,
157 	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
158 };
159 
160 static struct resource code_resource = {
161 	.name	= "Kernel code",
162 	.start	= 0,
163 	.end	= 0,
164 	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
165 };
166 
167 static struct resource bss_resource = {
168 	.name	= "Kernel bss",
169 	.start	= 0,
170 	.end	= 0,
171 	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
172 };
173 
174 
175 #ifdef CONFIG_X86_32
176 /* cpu data as detected by the assembly code in head.S */
177 struct cpuinfo_x86 new_cpu_data = {
178 	.wp_works_ok = -1,
179 };
180 /* common cpu data for all cpus */
181 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
182 	.wp_works_ok = -1,
183 };
184 EXPORT_SYMBOL(boot_cpu_data);
185 
186 unsigned int def_to_bigsmp;
187 
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;
192 
193 struct apm_info apm_info;
194 EXPORT_SYMBOL(apm_info);
195 
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
203 
204 #else
205 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
206 	.x86_phys_bits = MAX_PHYSMEM_BITS,
207 };
208 EXPORT_SYMBOL(boot_cpu_data);
209 #endif
210 
211 
212 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
213 __visible unsigned long mmu_cr4_features;
214 #else
215 __visible unsigned long mmu_cr4_features = X86_CR4_PAE;
216 #endif
217 
218 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
219 int bootloader_type, bootloader_version;
220 
221 /*
222  * Setup options
223  */
224 struct screen_info screen_info;
225 EXPORT_SYMBOL(screen_info);
226 struct edid_info edid_info;
227 EXPORT_SYMBOL_GPL(edid_info);
228 
229 extern int root_mountflags;
230 
231 unsigned long saved_video_mode;
232 
233 #define RAMDISK_IMAGE_START_MASK	0x07FF
234 #define RAMDISK_PROMPT_FLAG		0x8000
235 #define RAMDISK_LOAD_FLAG		0x4000
236 
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
241 
242 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
243 struct edd edd;
244 #ifdef CONFIG_EDD_MODULE
245 EXPORT_SYMBOL(edd);
246 #endif
247 /**
248  * copy_edd() - Copy the BIOS EDD information
249  *              from boot_params into a safe place.
250  *
251  */
252 static inline void __init copy_edd(void)
253 {
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;
259 }
260 #else
261 static inline void __init copy_edd(void)
262 {
263 }
264 #endif
265 
266 void * __init extend_brk(size_t size, size_t align)
267 {
268 	size_t mask = align - 1;
269 	void *ret;
270 
271 	BUG_ON(_brk_start == 0);
272 	BUG_ON(align & mask);
273 
274 	_brk_end = (_brk_end + mask) & ~mask;
275 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
276 
277 	ret = (void *)_brk_end;
278 	_brk_end += size;
279 
280 	memset(ret, 0, size);
281 
282 	return ret;
283 }
284 
285 #ifdef CONFIG_X86_32
286 static void __init cleanup_highmap(void)
287 {
288 }
289 #endif
290 
291 static void __init reserve_brk(void)
292 {
293 	if (_brk_end > _brk_start)
294 		memblock_reserve(__pa_symbol(_brk_start),
295 				 _brk_end - _brk_start);
296 
297 	/* Mark brk area as locked down and no longer taking any
298 	   new allocations */
299 	_brk_start = 0;
300 }
301 
302 u64 relocated_ramdisk;
303 
304 #ifdef CONFIG_BLK_DEV_INITRD
305 
306 static u64 __init get_ramdisk_image(void)
307 {
308 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
309 
310 	ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
311 
312 	return ramdisk_image;
313 }
314 static u64 __init get_ramdisk_size(void)
315 {
316 	u64 ramdisk_size = boot_params.hdr.ramdisk_size;
317 
318 	ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
319 
320 	return ramdisk_size;
321 }
322 
323 static void __init relocate_initrd(void)
324 {
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);
329 
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);
333 
334 	if (!relocated_ramdisk)
335 		panic("Cannot find place for new RAMDISK of size %lld\n",
336 		      ramdisk_size);
337 
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);
345 
346 	copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
347 
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);
352 }
353 
354 static void __init early_reserve_initrd(void)
355 {
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);
360 
361 	if (!boot_params.hdr.type_of_loader ||
362 	    !ramdisk_image || !ramdisk_size)
363 		return;		/* No initrd provided by bootloader */
364 
365 	memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
366 }
367 static void __init reserve_initrd(void)
368 {
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;
374 
375 	if (!boot_params.hdr.type_of_loader ||
376 	    !ramdisk_image || !ramdisk_size)
377 		return;		/* No initrd provided by bootloader */
378 
379 	initrd_start = 0;
380 
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);
386 
387 	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
388 			ramdisk_end - 1);
389 
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;
396 	}
397 
398 	relocate_initrd();
399 
400 	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
401 }
402 
403 static void __init early_initrd_acpi_init(void)
404 {
405 	early_acpi_table_init((void *)initrd_start, initrd_end - initrd_start);
406 }
407 #else
408 static void __init early_reserve_initrd(void)
409 {
410 }
411 static void __init reserve_initrd(void)
412 {
413 }
414 static void __init early_initrd_acpi_init(void)
415 {
416 }
417 #endif /* CONFIG_BLK_DEV_INITRD */
418 
419 static void __init parse_setup_data(void)
420 {
421 	struct setup_data *data;
422 	u64 pa_data, pa_next;
423 
424 	pa_data = boot_params.hdr.setup_data;
425 	while (pa_data) {
426 		u32 data_len, data_type;
427 
428 		data = early_memremap(pa_data, sizeof(*data));
429 		data_len = data->len + sizeof(struct setup_data);
430 		data_type = data->type;
431 		pa_next = data->next;
432 		early_memunmap(data, sizeof(*data));
433 
434 		switch (data_type) {
435 		case SETUP_E820_EXT:
436 			parse_e820_ext(pa_data, data_len);
437 			break;
438 		case SETUP_DTB:
439 			add_dtb(pa_data);
440 			break;
441 		case SETUP_EFI:
442 			parse_efi_setup(pa_data, data_len);
443 			break;
444 		default:
445 			break;
446 		}
447 		pa_data = pa_next;
448 	}
449 }
450 
451 static void __init e820_reserve_setup_data(void)
452 {
453 	struct setup_data *data;
454 	u64 pa_data;
455 
456 	pa_data = boot_params.hdr.setup_data;
457 	if (!pa_data)
458 		return;
459 
460 	while (pa_data) {
461 		data = early_memremap(pa_data, sizeof(*data));
462 		e820_update_range(pa_data, sizeof(*data)+data->len,
463 			 E820_RAM, E820_RESERVED_KERN);
464 		pa_data = data->next;
465 		early_memunmap(data, sizeof(*data));
466 	}
467 
468 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
469 	memcpy(&e820_saved, &e820, sizeof(struct e820map));
470 	printk(KERN_INFO "extended physical RAM map:\n");
471 	e820_print_map("reserve setup_data");
472 }
473 
474 static void __init memblock_x86_reserve_range_setup_data(void)
475 {
476 	struct setup_data *data;
477 	u64 pa_data;
478 
479 	pa_data = boot_params.hdr.setup_data;
480 	while (pa_data) {
481 		data = early_memremap(pa_data, sizeof(*data));
482 		memblock_reserve(pa_data, sizeof(*data) + data->len);
483 		pa_data = data->next;
484 		early_memunmap(data, sizeof(*data));
485 	}
486 }
487 
488 /*
489  * --------- Crashkernel reservation ------------------------------
490  */
491 
492 #ifdef CONFIG_KEXEC_CORE
493 
494 /* 16M alignment for crash kernel regions */
495 #define CRASH_ALIGN		(16 << 20)
496 
497 /*
498  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
499  * would limit the kernel to the low 512 MiB due to mapping restrictions.
500  * On 64bit, old kexec-tools need to under 896MiB.
501  */
502 #ifdef CONFIG_X86_32
503 # define CRASH_ADDR_LOW_MAX	(512 << 20)
504 # define CRASH_ADDR_HIGH_MAX	(512 << 20)
505 #else
506 # define CRASH_ADDR_LOW_MAX	(896UL << 20)
507 # define CRASH_ADDR_HIGH_MAX	MAXMEM
508 #endif
509 
510 static int __init reserve_crashkernel_low(void)
511 {
512 #ifdef CONFIG_X86_64
513 	unsigned long long base, low_base = 0, low_size = 0;
514 	unsigned long total_low_mem;
515 	int ret;
516 
517 	total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
518 
519 	/* crashkernel=Y,low */
520 	ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
521 	if (ret) {
522 		/*
523 		 * two parts from lib/swiotlb.c:
524 		 * -swiotlb size: user-specified with swiotlb= or default.
525 		 *
526 		 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
527 		 * to 8M for other buffers that may need to stay low too. Also
528 		 * make sure we allocate enough extra low memory so that we
529 		 * don't run out of DMA buffers for 32-bit devices.
530 		 */
531 		low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
532 	} else {
533 		/* passed with crashkernel=0,low ? */
534 		if (!low_size)
535 			return 0;
536 	}
537 
538 	low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN);
539 	if (!low_base) {
540 		pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
541 		       (unsigned long)(low_size >> 20));
542 		return -ENOMEM;
543 	}
544 
545 	ret = memblock_reserve(low_base, low_size);
546 	if (ret) {
547 		pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
548 		return ret;
549 	}
550 
551 	pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
552 		(unsigned long)(low_size >> 20),
553 		(unsigned long)(low_base >> 20),
554 		(unsigned long)(total_low_mem >> 20));
555 
556 	crashk_low_res.start = low_base;
557 	crashk_low_res.end   = low_base + low_size - 1;
558 	insert_resource(&iomem_resource, &crashk_low_res);
559 #endif
560 	return 0;
561 }
562 
563 static void __init reserve_crashkernel(void)
564 {
565 	unsigned long long crash_size, crash_base, total_mem;
566 	bool high = false;
567 	int ret;
568 
569 	total_mem = memblock_phys_mem_size();
570 
571 	/* crashkernel=XM */
572 	ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
573 	if (ret != 0 || crash_size <= 0) {
574 		/* crashkernel=X,high */
575 		ret = parse_crashkernel_high(boot_command_line, total_mem,
576 					     &crash_size, &crash_base);
577 		if (ret != 0 || crash_size <= 0)
578 			return;
579 		high = true;
580 	}
581 
582 	/* 0 means: find the address automatically */
583 	if (crash_base <= 0) {
584 		/*
585 		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
586 		 */
587 		crash_base = memblock_find_in_range(CRASH_ALIGN,
588 						    high ? CRASH_ADDR_HIGH_MAX
589 							 : CRASH_ADDR_LOW_MAX,
590 						    crash_size, CRASH_ALIGN);
591 		if (!crash_base) {
592 			pr_info("crashkernel reservation failed - No suitable area found.\n");
593 			return;
594 		}
595 
596 	} else {
597 		unsigned long long start;
598 
599 		start = memblock_find_in_range(crash_base,
600 					       crash_base + crash_size,
601 					       crash_size, 1 << 20);
602 		if (start != crash_base) {
603 			pr_info("crashkernel reservation failed - memory is in use.\n");
604 			return;
605 		}
606 	}
607 	ret = memblock_reserve(crash_base, crash_size);
608 	if (ret) {
609 		pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
610 		return;
611 	}
612 
613 	if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
614 		memblock_free(crash_base, crash_size);
615 		return;
616 	}
617 
618 	pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
619 		(unsigned long)(crash_size >> 20),
620 		(unsigned long)(crash_base >> 20),
621 		(unsigned long)(total_mem >> 20));
622 
623 	crashk_res.start = crash_base;
624 	crashk_res.end   = crash_base + crash_size - 1;
625 	insert_resource(&iomem_resource, &crashk_res);
626 }
627 #else
628 static void __init reserve_crashkernel(void)
629 {
630 }
631 #endif
632 
633 static struct resource standard_io_resources[] = {
634 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
635 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
636 	{ .name = "pic1", .start = 0x20, .end = 0x21,
637 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
638 	{ .name = "timer0", .start = 0x40, .end = 0x43,
639 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
640 	{ .name = "timer1", .start = 0x50, .end = 0x53,
641 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
642 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
643 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
644 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
645 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
646 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
647 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
648 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
649 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
650 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
651 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
652 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
653 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
654 };
655 
656 void __init reserve_standard_io_resources(void)
657 {
658 	int i;
659 
660 	/* request I/O space for devices used on all i[345]86 PCs */
661 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
662 		request_resource(&ioport_resource, &standard_io_resources[i]);
663 
664 }
665 
666 static __init void reserve_ibft_region(void)
667 {
668 	unsigned long addr, size = 0;
669 
670 	addr = find_ibft_region(&size);
671 
672 	if (size)
673 		memblock_reserve(addr, size);
674 }
675 
676 static bool __init snb_gfx_workaround_needed(void)
677 {
678 #ifdef CONFIG_PCI
679 	int i;
680 	u16 vendor, devid;
681 	static const __initconst u16 snb_ids[] = {
682 		0x0102,
683 		0x0112,
684 		0x0122,
685 		0x0106,
686 		0x0116,
687 		0x0126,
688 		0x010a,
689 	};
690 
691 	/* Assume no if something weird is going on with PCI */
692 	if (!early_pci_allowed())
693 		return false;
694 
695 	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
696 	if (vendor != 0x8086)
697 		return false;
698 
699 	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
700 	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
701 		if (devid == snb_ids[i])
702 			return true;
703 #endif
704 
705 	return false;
706 }
707 
708 /*
709  * Sandy Bridge graphics has trouble with certain ranges, exclude
710  * them from allocation.
711  */
712 static void __init trim_snb_memory(void)
713 {
714 	static const __initconst unsigned long bad_pages[] = {
715 		0x20050000,
716 		0x20110000,
717 		0x20130000,
718 		0x20138000,
719 		0x40004000,
720 	};
721 	int i;
722 
723 	if (!snb_gfx_workaround_needed())
724 		return;
725 
726 	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
727 
728 	/*
729 	 * Reserve all memory below the 1 MB mark that has not
730 	 * already been reserved.
731 	 */
732 	memblock_reserve(0, 1<<20);
733 
734 	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
735 		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
736 			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
737 			       bad_pages[i]);
738 	}
739 }
740 
741 /*
742  * Here we put platform-specific memory range workarounds, i.e.
743  * memory known to be corrupt or otherwise in need to be reserved on
744  * specific platforms.
745  *
746  * If this gets used more widely it could use a real dispatch mechanism.
747  */
748 static void __init trim_platform_memory_ranges(void)
749 {
750 	trim_snb_memory();
751 }
752 
753 static void __init trim_bios_range(void)
754 {
755 	/*
756 	 * A special case is the first 4Kb of memory;
757 	 * This is a BIOS owned area, not kernel ram, but generally
758 	 * not listed as such in the E820 table.
759 	 *
760 	 * This typically reserves additional memory (64KiB by default)
761 	 * since some BIOSes are known to corrupt low memory.  See the
762 	 * Kconfig help text for X86_RESERVE_LOW.
763 	 */
764 	e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
765 
766 	/*
767 	 * special case: Some BIOSen report the PC BIOS
768 	 * area (640->1Mb) as ram even though it is not.
769 	 * take them out.
770 	 */
771 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
772 
773 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
774 }
775 
776 /* called before trim_bios_range() to spare extra sanitize */
777 static void __init e820_add_kernel_range(void)
778 {
779 	u64 start = __pa_symbol(_text);
780 	u64 size = __pa_symbol(_end) - start;
781 
782 	/*
783 	 * Complain if .text .data and .bss are not marked as E820_RAM and
784 	 * attempt to fix it by adding the range. We may have a confused BIOS,
785 	 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
786 	 * exclude kernel range. If we really are running on top non-RAM,
787 	 * we will crash later anyways.
788 	 */
789 	if (e820_all_mapped(start, start + size, E820_RAM))
790 		return;
791 
792 	pr_warn(".text .data .bss are not marked as E820_RAM!\n");
793 	e820_remove_range(start, size, E820_RAM, 0);
794 	e820_add_region(start, size, E820_RAM);
795 }
796 
797 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
798 
799 static int __init parse_reservelow(char *p)
800 {
801 	unsigned long long size;
802 
803 	if (!p)
804 		return -EINVAL;
805 
806 	size = memparse(p, &p);
807 
808 	if (size < 4096)
809 		size = 4096;
810 
811 	if (size > 640*1024)
812 		size = 640*1024;
813 
814 	reserve_low = size;
815 
816 	return 0;
817 }
818 
819 early_param("reservelow", parse_reservelow);
820 
821 static void __init trim_low_memory_range(void)
822 {
823 	memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
824 }
825 
826 /*
827  * Dump out kernel offset information on panic.
828  */
829 static int
830 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
831 {
832 	if (kaslr_enabled()) {
833 		pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
834 			 kaslr_offset(),
835 			 __START_KERNEL,
836 			 __START_KERNEL_map,
837 			 MODULES_VADDR-1);
838 	} else {
839 		pr_emerg("Kernel Offset: disabled\n");
840 	}
841 
842 	return 0;
843 }
844 
845 /*
846  * Determine if we were loaded by an EFI loader.  If so, then we have also been
847  * passed the efi memmap, systab, etc., so we should use these data structures
848  * for initialization.  Note, the efi init code path is determined by the
849  * global efi_enabled. This allows the same kernel image to be used on existing
850  * systems (with a traditional BIOS) as well as on EFI systems.
851  */
852 /*
853  * setup_arch - architecture-specific boot-time initializations
854  *
855  * Note: On x86_64, fixmaps are ready for use even before this is called.
856  */
857 
858 void __init setup_arch(char **cmdline_p)
859 {
860 	memblock_reserve(__pa_symbol(_text),
861 			 (unsigned long)__bss_stop - (unsigned long)_text);
862 
863 	early_reserve_initrd();
864 
865 	/*
866 	 * At this point everything still needed from the boot loader
867 	 * or BIOS or kernel text should be early reserved or marked not
868 	 * RAM in e820. All other memory is free game.
869 	 */
870 
871 #ifdef CONFIG_X86_32
872 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
873 
874 	/*
875 	 * copy kernel address range established so far and switch
876 	 * to the proper swapper page table
877 	 */
878 	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
879 			initial_page_table + KERNEL_PGD_BOUNDARY,
880 			KERNEL_PGD_PTRS);
881 
882 	load_cr3(swapper_pg_dir);
883 	/*
884 	 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
885 	 * a cr3 based tlb flush, so the following __flush_tlb_all()
886 	 * will not flush anything because the cpu quirk which clears
887 	 * X86_FEATURE_PGE has not been invoked yet. Though due to the
888 	 * load_cr3() above the TLB has been flushed already. The
889 	 * quirk is invoked before subsequent calls to __flush_tlb_all()
890 	 * so proper operation is guaranteed.
891 	 */
892 	__flush_tlb_all();
893 #else
894 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
895 #endif
896 
897 	/*
898 	 * If we have OLPC OFW, we might end up relocating the fixmap due to
899 	 * reserve_top(), so do this before touching the ioremap area.
900 	 */
901 	olpc_ofw_detect();
902 
903 	early_trap_init();
904 	early_cpu_init();
905 	early_ioremap_init();
906 
907 	setup_olpc_ofw_pgd();
908 
909 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
910 	screen_info = boot_params.screen_info;
911 	edid_info = boot_params.edid_info;
912 #ifdef CONFIG_X86_32
913 	apm_info.bios = boot_params.apm_bios_info;
914 	ist_info = boot_params.ist_info;
915 #endif
916 	saved_video_mode = boot_params.hdr.vid_mode;
917 	bootloader_type = boot_params.hdr.type_of_loader;
918 	if ((bootloader_type >> 4) == 0xe) {
919 		bootloader_type &= 0xf;
920 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
921 	}
922 	bootloader_version  = bootloader_type & 0xf;
923 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
924 
925 #ifdef CONFIG_BLK_DEV_RAM
926 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
927 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
928 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
929 #endif
930 #ifdef CONFIG_EFI
931 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
932 		     EFI32_LOADER_SIGNATURE, 4)) {
933 		set_bit(EFI_BOOT, &efi.flags);
934 	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
935 		     EFI64_LOADER_SIGNATURE, 4)) {
936 		set_bit(EFI_BOOT, &efi.flags);
937 		set_bit(EFI_64BIT, &efi.flags);
938 	}
939 
940 	if (efi_enabled(EFI_BOOT))
941 		efi_memblock_x86_reserve_range();
942 #endif
943 
944 	x86_init.oem.arch_setup();
945 
946 	kernel_randomize_memory();
947 
948 	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
949 	setup_memory_map();
950 	parse_setup_data();
951 
952 	copy_edd();
953 
954 	if (!boot_params.hdr.root_flags)
955 		root_mountflags &= ~MS_RDONLY;
956 	init_mm.start_code = (unsigned long) _text;
957 	init_mm.end_code = (unsigned long) _etext;
958 	init_mm.end_data = (unsigned long) _edata;
959 	init_mm.brk = _brk_end;
960 
961 	mpx_mm_init(&init_mm);
962 
963 	code_resource.start = __pa_symbol(_text);
964 	code_resource.end = __pa_symbol(_etext)-1;
965 	data_resource.start = __pa_symbol(_etext);
966 	data_resource.end = __pa_symbol(_edata)-1;
967 	bss_resource.start = __pa_symbol(__bss_start);
968 	bss_resource.end = __pa_symbol(__bss_stop)-1;
969 
970 #ifdef CONFIG_CMDLINE_BOOL
971 #ifdef CONFIG_CMDLINE_OVERRIDE
972 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
973 #else
974 	if (builtin_cmdline[0]) {
975 		/* append boot loader cmdline to builtin */
976 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
977 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
978 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
979 	}
980 #endif
981 #endif
982 
983 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
984 	*cmdline_p = command_line;
985 
986 	/*
987 	 * x86_configure_nx() is called before parse_early_param() to detect
988 	 * whether hardware doesn't support NX (so that the early EHCI debug
989 	 * console setup can safely call set_fixmap()). It may then be called
990 	 * again from within noexec_setup() during parsing early parameters
991 	 * to honor the respective command line option.
992 	 */
993 	x86_configure_nx();
994 
995 	parse_early_param();
996 
997 	x86_report_nx();
998 
999 	/* after early param, so could get panic from serial */
1000 	memblock_x86_reserve_range_setup_data();
1001 
1002 	if (acpi_mps_check()) {
1003 #ifdef CONFIG_X86_LOCAL_APIC
1004 		disable_apic = 1;
1005 #endif
1006 		setup_clear_cpu_cap(X86_FEATURE_APIC);
1007 	}
1008 
1009 #ifdef CONFIG_PCI
1010 	if (pci_early_dump_regs)
1011 		early_dump_pci_devices();
1012 #endif
1013 
1014 	/* update the e820_saved too */
1015 	e820_reserve_setup_data();
1016 	finish_e820_parsing();
1017 
1018 	if (efi_enabled(EFI_BOOT))
1019 		efi_init();
1020 
1021 	dmi_scan_machine();
1022 	dmi_memdev_walk();
1023 	dmi_set_dump_stack_arch_desc();
1024 
1025 	/*
1026 	 * VMware detection requires dmi to be available, so this
1027 	 * needs to be done after dmi_scan_machine, for the BP.
1028 	 */
1029 	init_hypervisor_platform();
1030 
1031 	x86_init.resources.probe_roms();
1032 
1033 	/* after parse_early_param, so could debug it */
1034 	insert_resource(&iomem_resource, &code_resource);
1035 	insert_resource(&iomem_resource, &data_resource);
1036 	insert_resource(&iomem_resource, &bss_resource);
1037 
1038 	e820_add_kernel_range();
1039 	trim_bios_range();
1040 #ifdef CONFIG_X86_32
1041 	if (ppro_with_ram_bug()) {
1042 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1043 				  E820_RESERVED);
1044 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
1045 		printk(KERN_INFO "fixed physical RAM map:\n");
1046 		e820_print_map("bad_ppro");
1047 	}
1048 #else
1049 	early_gart_iommu_check();
1050 #endif
1051 
1052 	/*
1053 	 * partially used pages are not usable - thus
1054 	 * we are rounding upwards:
1055 	 */
1056 	max_pfn = e820_end_of_ram_pfn();
1057 
1058 	/* update e820 for memory not covered by WB MTRRs */
1059 	mtrr_bp_init();
1060 	if (mtrr_trim_uncached_memory(max_pfn))
1061 		max_pfn = e820_end_of_ram_pfn();
1062 
1063 	max_possible_pfn = max_pfn;
1064 
1065 #ifdef CONFIG_X86_32
1066 	/* max_low_pfn get updated here */
1067 	find_low_pfn_range();
1068 #else
1069 	check_x2apic();
1070 
1071 	/* How many end-of-memory variables you have, grandma! */
1072 	/* need this before calling reserve_initrd */
1073 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1074 		max_low_pfn = e820_end_of_low_ram_pfn();
1075 	else
1076 		max_low_pfn = max_pfn;
1077 
1078 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1079 #endif
1080 
1081 	/*
1082 	 * Find and reserve possible boot-time SMP configuration:
1083 	 */
1084 	find_smp_config();
1085 
1086 	reserve_ibft_region();
1087 
1088 	early_alloc_pgt_buf();
1089 
1090 	/*
1091 	 * Need to conclude brk, before memblock_x86_fill()
1092 	 *  it could use memblock_find_in_range, could overlap with
1093 	 *  brk area.
1094 	 */
1095 	reserve_brk();
1096 
1097 	cleanup_highmap();
1098 
1099 	memblock_set_current_limit(ISA_END_ADDRESS);
1100 	memblock_x86_fill();
1101 
1102 	if (efi_enabled(EFI_BOOT)) {
1103 		efi_fake_memmap();
1104 		efi_find_mirror();
1105 	}
1106 
1107 	/*
1108 	 * The EFI specification says that boot service code won't be called
1109 	 * after ExitBootServices(). This is, in fact, a lie.
1110 	 */
1111 	if (efi_enabled(EFI_MEMMAP))
1112 		efi_reserve_boot_services();
1113 
1114 	/* preallocate 4k for mptable mpc */
1115 	early_reserve_e820_mpc_new();
1116 
1117 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1118 	setup_bios_corruption_check();
1119 #endif
1120 
1121 #ifdef CONFIG_X86_32
1122 	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1123 			(max_pfn_mapped<<PAGE_SHIFT) - 1);
1124 #endif
1125 
1126 	reserve_real_mode();
1127 
1128 	trim_platform_memory_ranges();
1129 	trim_low_memory_range();
1130 
1131 	init_mem_mapping();
1132 
1133 	early_trap_pf_init();
1134 
1135 	setup_real_mode();
1136 
1137 	memblock_set_current_limit(get_max_mapped());
1138 
1139 	/*
1140 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1141 	 */
1142 
1143 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1144 	if (init_ohci1394_dma_early)
1145 		init_ohci1394_dma_on_all_controllers();
1146 #endif
1147 	/* Allocate bigger log buffer */
1148 	setup_log_buf(1);
1149 
1150 	reserve_initrd();
1151 
1152 	early_initrd_acpi_init();
1153 
1154 	vsmp_init();
1155 
1156 	io_delay_init();
1157 
1158 	/*
1159 	 * Parse the ACPI tables for possible boot-time SMP configuration.
1160 	 */
1161 	acpi_boot_table_init();
1162 
1163 	early_acpi_boot_init();
1164 
1165 	initmem_init();
1166 	dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1167 
1168 	/*
1169 	 * Reserve memory for crash kernel after SRAT is parsed so that it
1170 	 * won't consume hotpluggable memory.
1171 	 */
1172 	reserve_crashkernel();
1173 
1174 	memblock_find_dma_reserve();
1175 
1176 #ifdef CONFIG_KVM_GUEST
1177 	kvmclock_init();
1178 #endif
1179 
1180 	x86_init.paging.pagetable_init();
1181 
1182 	kasan_init();
1183 
1184 	if (boot_cpu_data.cpuid_level >= 0) {
1185 		/* A CPU has %cr4 if and only if it has CPUID */
1186 		mmu_cr4_features = __read_cr4();
1187 		if (trampoline_cr4_features)
1188 			*trampoline_cr4_features = mmu_cr4_features;
1189 	}
1190 
1191 #ifdef CONFIG_X86_32
1192 	/* sync back kernel address range */
1193 	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1194 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1195 			KERNEL_PGD_PTRS);
1196 
1197 	/*
1198 	 * sync back low identity map too.  It is used for example
1199 	 * in the 32-bit EFI stub.
1200 	 */
1201 	clone_pgd_range(initial_page_table,
1202 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1203 			min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY));
1204 #endif
1205 
1206 	tboot_probe();
1207 
1208 	map_vsyscall();
1209 
1210 	generic_apic_probe();
1211 
1212 	early_quirks();
1213 
1214 	/*
1215 	 * Read APIC and some other early information from ACPI tables.
1216 	 */
1217 	acpi_boot_init();
1218 	sfi_init();
1219 	x86_dtb_init();
1220 
1221 	/*
1222 	 * get boot-time SMP configuration:
1223 	 */
1224 	if (smp_found_config)
1225 		get_smp_config();
1226 
1227 	prefill_possible_map();
1228 
1229 	init_cpu_to_node();
1230 
1231 	init_apic_mappings();
1232 	io_apic_init_mappings();
1233 
1234 	kvm_guest_init();
1235 
1236 	e820_reserve_resources();
1237 	e820_mark_nosave_regions(max_low_pfn);
1238 
1239 	x86_init.resources.reserve_resources();
1240 
1241 	e820_setup_gap();
1242 
1243 #ifdef CONFIG_VT
1244 #if defined(CONFIG_VGA_CONSOLE)
1245 	if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1246 		conswitchp = &vga_con;
1247 #elif defined(CONFIG_DUMMY_CONSOLE)
1248 	conswitchp = &dummy_con;
1249 #endif
1250 #endif
1251 	x86_init.oem.banner();
1252 
1253 	x86_init.timers.wallclock_init();
1254 
1255 	mcheck_init();
1256 
1257 	arch_init_ideal_nops();
1258 
1259 	register_refined_jiffies(CLOCK_TICK_RATE);
1260 
1261 #ifdef CONFIG_EFI
1262 	if (efi_enabled(EFI_BOOT))
1263 		efi_apply_memmap_quirks();
1264 #endif
1265 }
1266 
1267 #ifdef CONFIG_X86_32
1268 
1269 static struct resource video_ram_resource = {
1270 	.name	= "Video RAM area",
1271 	.start	= 0xa0000,
1272 	.end	= 0xbffff,
1273 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1274 };
1275 
1276 void __init i386_reserve_resources(void)
1277 {
1278 	request_resource(&iomem_resource, &video_ram_resource);
1279 	reserve_standard_io_resources();
1280 }
1281 
1282 #endif /* CONFIG_X86_32 */
1283 
1284 static struct notifier_block kernel_offset_notifier = {
1285 	.notifier_call = dump_kernel_offset
1286 };
1287 
1288 static int __init register_kernel_offset_dumper(void)
1289 {
1290 	atomic_notifier_chain_register(&panic_notifier_list,
1291 					&kernel_offset_notifier);
1292 	return 0;
1293 }
1294 __initcall(register_kernel_offset_dumper);
1295 
1296 void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1297 {
1298 	if (!boot_cpu_has(X86_FEATURE_OSPKE))
1299 		return;
1300 
1301 	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
1302 }
1303