xref: /openbmc/linux/arch/x86/kernel/setup.c (revision 3805e6a1)
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 
117 /*
118  * max_low_pfn_mapped: highest direct mapped pfn under 4GB
119  * max_pfn_mapped:     highest direct mapped pfn over 4GB
120  *
121  * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
122  * represented by pfn_mapped
123  */
124 unsigned long max_low_pfn_mapped;
125 unsigned long max_pfn_mapped;
126 
127 #ifdef CONFIG_DMI
128 RESERVE_BRK(dmi_alloc, 65536);
129 #endif
130 
131 
132 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
133 unsigned long _brk_end = (unsigned long)__brk_base;
134 
135 #ifdef CONFIG_X86_64
136 int default_cpu_present_to_apicid(int mps_cpu)
137 {
138 	return __default_cpu_present_to_apicid(mps_cpu);
139 }
140 
141 int default_check_phys_apicid_present(int phys_apicid)
142 {
143 	return __default_check_phys_apicid_present(phys_apicid);
144 }
145 #endif
146 
147 struct boot_params boot_params;
148 
149 /*
150  * Machine setup..
151  */
152 static struct resource data_resource = {
153 	.name	= "Kernel data",
154 	.start	= 0,
155 	.end	= 0,
156 	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
157 };
158 
159 static struct resource code_resource = {
160 	.name	= "Kernel code",
161 	.start	= 0,
162 	.end	= 0,
163 	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
164 };
165 
166 static struct resource bss_resource = {
167 	.name	= "Kernel bss",
168 	.start	= 0,
169 	.end	= 0,
170 	.flags	= IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM
171 };
172 
173 
174 #ifdef CONFIG_X86_32
175 /* cpu data as detected by the assembly code in head.S */
176 struct cpuinfo_x86 new_cpu_data = {
177 	.wp_works_ok = -1,
178 };
179 /* common cpu data for all cpus */
180 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
181 	.wp_works_ok = -1,
182 };
183 EXPORT_SYMBOL(boot_cpu_data);
184 
185 unsigned int def_to_bigsmp;
186 
187 /* for MCA, but anyone else can use it if they want */
188 unsigned int machine_id;
189 unsigned int machine_submodel_id;
190 unsigned int BIOS_revision;
191 
192 struct apm_info apm_info;
193 EXPORT_SYMBOL(apm_info);
194 
195 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
196 	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
197 struct ist_info ist_info;
198 EXPORT_SYMBOL(ist_info);
199 #else
200 struct ist_info ist_info;
201 #endif
202 
203 #else
204 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
205 	.x86_phys_bits = MAX_PHYSMEM_BITS,
206 };
207 EXPORT_SYMBOL(boot_cpu_data);
208 #endif
209 
210 
211 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
212 __visible unsigned long mmu_cr4_features;
213 #else
214 __visible unsigned long mmu_cr4_features = X86_CR4_PAE;
215 #endif
216 
217 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
218 int bootloader_type, bootloader_version;
219 
220 /*
221  * Setup options
222  */
223 struct screen_info screen_info;
224 EXPORT_SYMBOL(screen_info);
225 struct edid_info edid_info;
226 EXPORT_SYMBOL_GPL(edid_info);
227 
228 extern int root_mountflags;
229 
230 unsigned long saved_video_mode;
231 
232 #define RAMDISK_IMAGE_START_MASK	0x07FF
233 #define RAMDISK_PROMPT_FLAG		0x8000
234 #define RAMDISK_LOAD_FLAG		0x4000
235 
236 static char __initdata command_line[COMMAND_LINE_SIZE];
237 #ifdef CONFIG_CMDLINE_BOOL
238 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
239 #endif
240 
241 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
242 struct edd edd;
243 #ifdef CONFIG_EDD_MODULE
244 EXPORT_SYMBOL(edd);
245 #endif
246 /**
247  * copy_edd() - Copy the BIOS EDD information
248  *              from boot_params into a safe place.
249  *
250  */
251 static inline void __init copy_edd(void)
252 {
253      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
254 	    sizeof(edd.mbr_signature));
255      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
256      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
257      edd.edd_info_nr = boot_params.eddbuf_entries;
258 }
259 #else
260 static inline void __init copy_edd(void)
261 {
262 }
263 #endif
264 
265 void * __init extend_brk(size_t size, size_t align)
266 {
267 	size_t mask = align - 1;
268 	void *ret;
269 
270 	BUG_ON(_brk_start == 0);
271 	BUG_ON(align & mask);
272 
273 	_brk_end = (_brk_end + mask) & ~mask;
274 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
275 
276 	ret = (void *)_brk_end;
277 	_brk_end += size;
278 
279 	memset(ret, 0, size);
280 
281 	return ret;
282 }
283 
284 #ifdef CONFIG_X86_32
285 static void __init cleanup_highmap(void)
286 {
287 }
288 #endif
289 
290 static void __init reserve_brk(void)
291 {
292 	if (_brk_end > _brk_start)
293 		memblock_reserve(__pa_symbol(_brk_start),
294 				 _brk_end - _brk_start);
295 
296 	/* Mark brk area as locked down and no longer taking any
297 	   new allocations */
298 	_brk_start = 0;
299 }
300 
301 u64 relocated_ramdisk;
302 
303 #ifdef CONFIG_BLK_DEV_INITRD
304 
305 static u64 __init get_ramdisk_image(void)
306 {
307 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
308 
309 	ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
310 
311 	return ramdisk_image;
312 }
313 static u64 __init get_ramdisk_size(void)
314 {
315 	u64 ramdisk_size = boot_params.hdr.ramdisk_size;
316 
317 	ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
318 
319 	return ramdisk_size;
320 }
321 
322 static void __init relocate_initrd(void)
323 {
324 	/* Assume only end is not page aligned */
325 	u64 ramdisk_image = get_ramdisk_image();
326 	u64 ramdisk_size  = get_ramdisk_size();
327 	u64 area_size     = PAGE_ALIGN(ramdisk_size);
328 
329 	/* We need to move the initrd down into directly mapped mem */
330 	relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
331 						   area_size, PAGE_SIZE);
332 
333 	if (!relocated_ramdisk)
334 		panic("Cannot find place for new RAMDISK of size %lld\n",
335 		      ramdisk_size);
336 
337 	/* Note: this includes all the mem currently occupied by
338 	   the initrd, we rely on that fact to keep the data intact. */
339 	memblock_reserve(relocated_ramdisk, area_size);
340 	initrd_start = relocated_ramdisk + PAGE_OFFSET;
341 	initrd_end   = initrd_start + ramdisk_size;
342 	printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
343 	       relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
344 
345 	copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
346 
347 	printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
348 		" [mem %#010llx-%#010llx]\n",
349 		ramdisk_image, ramdisk_image + ramdisk_size - 1,
350 		relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
351 }
352 
353 static void __init early_reserve_initrd(void)
354 {
355 	/* Assume only end is not page aligned */
356 	u64 ramdisk_image = get_ramdisk_image();
357 	u64 ramdisk_size  = get_ramdisk_size();
358 	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
359 
360 	if (!boot_params.hdr.type_of_loader ||
361 	    !ramdisk_image || !ramdisk_size)
362 		return;		/* No initrd provided by bootloader */
363 
364 	memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
365 }
366 static void __init reserve_initrd(void)
367 {
368 	/* Assume only end is not page aligned */
369 	u64 ramdisk_image = get_ramdisk_image();
370 	u64 ramdisk_size  = get_ramdisk_size();
371 	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
372 	u64 mapped_size;
373 
374 	if (!boot_params.hdr.type_of_loader ||
375 	    !ramdisk_image || !ramdisk_size)
376 		return;		/* No initrd provided by bootloader */
377 
378 	initrd_start = 0;
379 
380 	mapped_size = memblock_mem_size(max_pfn_mapped);
381 	if (ramdisk_size >= (mapped_size>>1))
382 		panic("initrd too large to handle, "
383 		       "disabling initrd (%lld needed, %lld available)\n",
384 		       ramdisk_size, mapped_size>>1);
385 
386 	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
387 			ramdisk_end - 1);
388 
389 	if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
390 				PFN_DOWN(ramdisk_end))) {
391 		/* All are mapped, easy case */
392 		initrd_start = ramdisk_image + PAGE_OFFSET;
393 		initrd_end = initrd_start + ramdisk_size;
394 		return;
395 	}
396 
397 	relocate_initrd();
398 
399 	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
400 }
401 
402 static void __init early_initrd_acpi_init(void)
403 {
404 	early_acpi_table_init((void *)initrd_start, initrd_end - initrd_start);
405 }
406 #else
407 static void __init early_reserve_initrd(void)
408 {
409 }
410 static void __init reserve_initrd(void)
411 {
412 }
413 static void __init early_initrd_acpi_init(void)
414 {
415 }
416 #endif /* CONFIG_BLK_DEV_INITRD */
417 
418 static void __init parse_setup_data(void)
419 {
420 	struct setup_data *data;
421 	u64 pa_data, pa_next;
422 
423 	pa_data = boot_params.hdr.setup_data;
424 	while (pa_data) {
425 		u32 data_len, data_type;
426 
427 		data = early_memremap(pa_data, sizeof(*data));
428 		data_len = data->len + sizeof(struct setup_data);
429 		data_type = data->type;
430 		pa_next = data->next;
431 		early_memunmap(data, sizeof(*data));
432 
433 		switch (data_type) {
434 		case SETUP_E820_EXT:
435 			parse_e820_ext(pa_data, data_len);
436 			break;
437 		case SETUP_DTB:
438 			add_dtb(pa_data);
439 			break;
440 		case SETUP_EFI:
441 			parse_efi_setup(pa_data, data_len);
442 			break;
443 		default:
444 			break;
445 		}
446 		pa_data = pa_next;
447 	}
448 }
449 
450 static void __init e820_reserve_setup_data(void)
451 {
452 	struct setup_data *data;
453 	u64 pa_data;
454 
455 	pa_data = boot_params.hdr.setup_data;
456 	if (!pa_data)
457 		return;
458 
459 	while (pa_data) {
460 		data = early_memremap(pa_data, sizeof(*data));
461 		e820_update_range(pa_data, sizeof(*data)+data->len,
462 			 E820_RAM, E820_RESERVED_KERN);
463 		pa_data = data->next;
464 		early_memunmap(data, sizeof(*data));
465 	}
466 
467 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
468 	memcpy(&e820_saved, &e820, sizeof(struct e820map));
469 	printk(KERN_INFO "extended physical RAM map:\n");
470 	e820_print_map("reserve setup_data");
471 }
472 
473 static void __init memblock_x86_reserve_range_setup_data(void)
474 {
475 	struct setup_data *data;
476 	u64 pa_data;
477 
478 	pa_data = boot_params.hdr.setup_data;
479 	while (pa_data) {
480 		data = early_memremap(pa_data, sizeof(*data));
481 		memblock_reserve(pa_data, sizeof(*data) + data->len);
482 		pa_data = data->next;
483 		early_memunmap(data, sizeof(*data));
484 	}
485 }
486 
487 /*
488  * --------- Crashkernel reservation ------------------------------
489  */
490 
491 #ifdef CONFIG_KEXEC_CORE
492 
493 /* 16M alignment for crash kernel regions */
494 #define CRASH_ALIGN		(16 << 20)
495 
496 /*
497  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
498  * would limit the kernel to the low 512 MiB due to mapping restrictions.
499  * On 64bit, old kexec-tools need to under 896MiB.
500  */
501 #ifdef CONFIG_X86_32
502 # define CRASH_ADDR_LOW_MAX	(512 << 20)
503 # define CRASH_ADDR_HIGH_MAX	(512 << 20)
504 #else
505 # define CRASH_ADDR_LOW_MAX	(896UL << 20)
506 # define CRASH_ADDR_HIGH_MAX	MAXMEM
507 #endif
508 
509 static int __init reserve_crashkernel_low(void)
510 {
511 #ifdef CONFIG_X86_64
512 	unsigned long long base, low_base = 0, low_size = 0;
513 	unsigned long total_low_mem;
514 	int ret;
515 
516 	total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
517 
518 	/* crashkernel=Y,low */
519 	ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
520 	if (ret) {
521 		/*
522 		 * two parts from lib/swiotlb.c:
523 		 * -swiotlb size: user-specified with swiotlb= or default.
524 		 *
525 		 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
526 		 * to 8M for other buffers that may need to stay low too. Also
527 		 * make sure we allocate enough extra low memory so that we
528 		 * don't run out of DMA buffers for 32-bit devices.
529 		 */
530 		low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
531 	} else {
532 		/* passed with crashkernel=0,low ? */
533 		if (!low_size)
534 			return 0;
535 	}
536 
537 	low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN);
538 	if (!low_base) {
539 		pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
540 		       (unsigned long)(low_size >> 20));
541 		return -ENOMEM;
542 	}
543 
544 	ret = memblock_reserve(low_base, low_size);
545 	if (ret) {
546 		pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
547 		return ret;
548 	}
549 
550 	pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
551 		(unsigned long)(low_size >> 20),
552 		(unsigned long)(low_base >> 20),
553 		(unsigned long)(total_low_mem >> 20));
554 
555 	crashk_low_res.start = low_base;
556 	crashk_low_res.end   = low_base + low_size - 1;
557 	insert_resource(&iomem_resource, &crashk_low_res);
558 #endif
559 	return 0;
560 }
561 
562 static void __init reserve_crashkernel(void)
563 {
564 	unsigned long long crash_size, crash_base, total_mem;
565 	bool high = false;
566 	int ret;
567 
568 	total_mem = memblock_phys_mem_size();
569 
570 	/* crashkernel=XM */
571 	ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
572 	if (ret != 0 || crash_size <= 0) {
573 		/* crashkernel=X,high */
574 		ret = parse_crashkernel_high(boot_command_line, total_mem,
575 					     &crash_size, &crash_base);
576 		if (ret != 0 || crash_size <= 0)
577 			return;
578 		high = true;
579 	}
580 
581 	/* 0 means: find the address automatically */
582 	if (crash_base <= 0) {
583 		/*
584 		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
585 		 */
586 		crash_base = memblock_find_in_range(CRASH_ALIGN,
587 						    high ? CRASH_ADDR_HIGH_MAX
588 							 : CRASH_ADDR_LOW_MAX,
589 						    crash_size, CRASH_ALIGN);
590 		if (!crash_base) {
591 			pr_info("crashkernel reservation failed - No suitable area found.\n");
592 			return;
593 		}
594 
595 	} else {
596 		unsigned long long start;
597 
598 		start = memblock_find_in_range(crash_base,
599 					       crash_base + crash_size,
600 					       crash_size, 1 << 20);
601 		if (start != crash_base) {
602 			pr_info("crashkernel reservation failed - memory is in use.\n");
603 			return;
604 		}
605 	}
606 	ret = memblock_reserve(crash_base, crash_size);
607 	if (ret) {
608 		pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
609 		return;
610 	}
611 
612 	if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
613 		memblock_free(crash_base, crash_size);
614 		return;
615 	}
616 
617 	pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
618 		(unsigned long)(crash_size >> 20),
619 		(unsigned long)(crash_base >> 20),
620 		(unsigned long)(total_mem >> 20));
621 
622 	crashk_res.start = crash_base;
623 	crashk_res.end   = crash_base + crash_size - 1;
624 	insert_resource(&iomem_resource, &crashk_res);
625 }
626 #else
627 static void __init reserve_crashkernel(void)
628 {
629 }
630 #endif
631 
632 static struct resource standard_io_resources[] = {
633 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
634 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
635 	{ .name = "pic1", .start = 0x20, .end = 0x21,
636 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
637 	{ .name = "timer0", .start = 0x40, .end = 0x43,
638 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
639 	{ .name = "timer1", .start = 0x50, .end = 0x53,
640 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
641 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
642 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
643 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
644 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
645 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
646 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
647 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
648 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
649 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
650 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
651 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
652 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
653 };
654 
655 void __init reserve_standard_io_resources(void)
656 {
657 	int i;
658 
659 	/* request I/O space for devices used on all i[345]86 PCs */
660 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
661 		request_resource(&ioport_resource, &standard_io_resources[i]);
662 
663 }
664 
665 static __init void reserve_ibft_region(void)
666 {
667 	unsigned long addr, size = 0;
668 
669 	addr = find_ibft_region(&size);
670 
671 	if (size)
672 		memblock_reserve(addr, size);
673 }
674 
675 static bool __init snb_gfx_workaround_needed(void)
676 {
677 #ifdef CONFIG_PCI
678 	int i;
679 	u16 vendor, devid;
680 	static const __initconst u16 snb_ids[] = {
681 		0x0102,
682 		0x0112,
683 		0x0122,
684 		0x0106,
685 		0x0116,
686 		0x0126,
687 		0x010a,
688 	};
689 
690 	/* Assume no if something weird is going on with PCI */
691 	if (!early_pci_allowed())
692 		return false;
693 
694 	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
695 	if (vendor != 0x8086)
696 		return false;
697 
698 	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
699 	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
700 		if (devid == snb_ids[i])
701 			return true;
702 #endif
703 
704 	return false;
705 }
706 
707 /*
708  * Sandy Bridge graphics has trouble with certain ranges, exclude
709  * them from allocation.
710  */
711 static void __init trim_snb_memory(void)
712 {
713 	static const __initconst unsigned long bad_pages[] = {
714 		0x20050000,
715 		0x20110000,
716 		0x20130000,
717 		0x20138000,
718 		0x40004000,
719 	};
720 	int i;
721 
722 	if (!snb_gfx_workaround_needed())
723 		return;
724 
725 	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
726 
727 	/*
728 	 * Reserve all memory below the 1 MB mark that has not
729 	 * already been reserved.
730 	 */
731 	memblock_reserve(0, 1<<20);
732 
733 	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
734 		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
735 			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
736 			       bad_pages[i]);
737 	}
738 }
739 
740 /*
741  * Here we put platform-specific memory range workarounds, i.e.
742  * memory known to be corrupt or otherwise in need to be reserved on
743  * specific platforms.
744  *
745  * If this gets used more widely it could use a real dispatch mechanism.
746  */
747 static void __init trim_platform_memory_ranges(void)
748 {
749 	trim_snb_memory();
750 }
751 
752 static void __init trim_bios_range(void)
753 {
754 	/*
755 	 * A special case is the first 4Kb of memory;
756 	 * This is a BIOS owned area, not kernel ram, but generally
757 	 * not listed as such in the E820 table.
758 	 *
759 	 * This typically reserves additional memory (64KiB by default)
760 	 * since some BIOSes are known to corrupt low memory.  See the
761 	 * Kconfig help text for X86_RESERVE_LOW.
762 	 */
763 	e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
764 
765 	/*
766 	 * special case: Some BIOSen report the PC BIOS
767 	 * area (640->1Mb) as ram even though it is not.
768 	 * take them out.
769 	 */
770 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
771 
772 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
773 }
774 
775 /* called before trim_bios_range() to spare extra sanitize */
776 static void __init e820_add_kernel_range(void)
777 {
778 	u64 start = __pa_symbol(_text);
779 	u64 size = __pa_symbol(_end) - start;
780 
781 	/*
782 	 * Complain if .text .data and .bss are not marked as E820_RAM and
783 	 * attempt to fix it by adding the range. We may have a confused BIOS,
784 	 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
785 	 * exclude kernel range. If we really are running on top non-RAM,
786 	 * we will crash later anyways.
787 	 */
788 	if (e820_all_mapped(start, start + size, E820_RAM))
789 		return;
790 
791 	pr_warn(".text .data .bss are not marked as E820_RAM!\n");
792 	e820_remove_range(start, size, E820_RAM, 0);
793 	e820_add_region(start, size, E820_RAM);
794 }
795 
796 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
797 
798 static int __init parse_reservelow(char *p)
799 {
800 	unsigned long long size;
801 
802 	if (!p)
803 		return -EINVAL;
804 
805 	size = memparse(p, &p);
806 
807 	if (size < 4096)
808 		size = 4096;
809 
810 	if (size > 640*1024)
811 		size = 640*1024;
812 
813 	reserve_low = size;
814 
815 	return 0;
816 }
817 
818 early_param("reservelow", parse_reservelow);
819 
820 static void __init trim_low_memory_range(void)
821 {
822 	memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
823 }
824 
825 /*
826  * Dump out kernel offset information on panic.
827  */
828 static int
829 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
830 {
831 	if (kaslr_enabled()) {
832 		pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
833 			 kaslr_offset(),
834 			 __START_KERNEL,
835 			 __START_KERNEL_map,
836 			 MODULES_VADDR-1);
837 	} else {
838 		pr_emerg("Kernel Offset: disabled\n");
839 	}
840 
841 	return 0;
842 }
843 
844 /*
845  * Determine if we were loaded by an EFI loader.  If so, then we have also been
846  * passed the efi memmap, systab, etc., so we should use these data structures
847  * for initialization.  Note, the efi init code path is determined by the
848  * global efi_enabled. This allows the same kernel image to be used on existing
849  * systems (with a traditional BIOS) as well as on EFI systems.
850  */
851 /*
852  * setup_arch - architecture-specific boot-time initializations
853  *
854  * Note: On x86_64, fixmaps are ready for use even before this is called.
855  */
856 
857 void __init setup_arch(char **cmdline_p)
858 {
859 	memblock_reserve(__pa_symbol(_text),
860 			 (unsigned long)__bss_stop - (unsigned long)_text);
861 
862 	early_reserve_initrd();
863 
864 	/*
865 	 * At this point everything still needed from the boot loader
866 	 * or BIOS or kernel text should be early reserved or marked not
867 	 * RAM in e820. All other memory is free game.
868 	 */
869 
870 #ifdef CONFIG_X86_32
871 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
872 
873 	/*
874 	 * copy kernel address range established so far and switch
875 	 * to the proper swapper page table
876 	 */
877 	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
878 			initial_page_table + KERNEL_PGD_BOUNDARY,
879 			KERNEL_PGD_PTRS);
880 
881 	load_cr3(swapper_pg_dir);
882 	/*
883 	 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
884 	 * a cr3 based tlb flush, so the following __flush_tlb_all()
885 	 * will not flush anything because the cpu quirk which clears
886 	 * X86_FEATURE_PGE has not been invoked yet. Though due to the
887 	 * load_cr3() above the TLB has been flushed already. The
888 	 * quirk is invoked before subsequent calls to __flush_tlb_all()
889 	 * so proper operation is guaranteed.
890 	 */
891 	__flush_tlb_all();
892 #else
893 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
894 #endif
895 
896 	/*
897 	 * If we have OLPC OFW, we might end up relocating the fixmap due to
898 	 * reserve_top(), so do this before touching the ioremap area.
899 	 */
900 	olpc_ofw_detect();
901 
902 	early_trap_init();
903 	early_cpu_init();
904 	early_ioremap_init();
905 
906 	setup_olpc_ofw_pgd();
907 
908 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
909 	screen_info = boot_params.screen_info;
910 	edid_info = boot_params.edid_info;
911 #ifdef CONFIG_X86_32
912 	apm_info.bios = boot_params.apm_bios_info;
913 	ist_info = boot_params.ist_info;
914 #endif
915 	saved_video_mode = boot_params.hdr.vid_mode;
916 	bootloader_type = boot_params.hdr.type_of_loader;
917 	if ((bootloader_type >> 4) == 0xe) {
918 		bootloader_type &= 0xf;
919 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
920 	}
921 	bootloader_version  = bootloader_type & 0xf;
922 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
923 
924 #ifdef CONFIG_BLK_DEV_RAM
925 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
926 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
927 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
928 #endif
929 #ifdef CONFIG_EFI
930 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
931 		     EFI32_LOADER_SIGNATURE, 4)) {
932 		set_bit(EFI_BOOT, &efi.flags);
933 	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
934 		     EFI64_LOADER_SIGNATURE, 4)) {
935 		set_bit(EFI_BOOT, &efi.flags);
936 		set_bit(EFI_64BIT, &efi.flags);
937 	}
938 
939 	if (efi_enabled(EFI_BOOT))
940 		efi_memblock_x86_reserve_range();
941 #endif
942 
943 	x86_init.oem.arch_setup();
944 
945 	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
946 	setup_memory_map();
947 	parse_setup_data();
948 
949 	copy_edd();
950 
951 	if (!boot_params.hdr.root_flags)
952 		root_mountflags &= ~MS_RDONLY;
953 	init_mm.start_code = (unsigned long) _text;
954 	init_mm.end_code = (unsigned long) _etext;
955 	init_mm.end_data = (unsigned long) _edata;
956 	init_mm.brk = _brk_end;
957 
958 	mpx_mm_init(&init_mm);
959 
960 	code_resource.start = __pa_symbol(_text);
961 	code_resource.end = __pa_symbol(_etext)-1;
962 	data_resource.start = __pa_symbol(_etext);
963 	data_resource.end = __pa_symbol(_edata)-1;
964 	bss_resource.start = __pa_symbol(__bss_start);
965 	bss_resource.end = __pa_symbol(__bss_stop)-1;
966 
967 #ifdef CONFIG_CMDLINE_BOOL
968 #ifdef CONFIG_CMDLINE_OVERRIDE
969 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
970 #else
971 	if (builtin_cmdline[0]) {
972 		/* append boot loader cmdline to builtin */
973 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
974 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
975 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
976 	}
977 #endif
978 #endif
979 
980 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
981 	*cmdline_p = command_line;
982 
983 	/*
984 	 * x86_configure_nx() is called before parse_early_param() to detect
985 	 * whether hardware doesn't support NX (so that the early EHCI debug
986 	 * console setup can safely call set_fixmap()). It may then be called
987 	 * again from within noexec_setup() during parsing early parameters
988 	 * to honor the respective command line option.
989 	 */
990 	x86_configure_nx();
991 
992 	parse_early_param();
993 
994 	x86_report_nx();
995 
996 	/* after early param, so could get panic from serial */
997 	memblock_x86_reserve_range_setup_data();
998 
999 	if (acpi_mps_check()) {
1000 #ifdef CONFIG_X86_LOCAL_APIC
1001 		disable_apic = 1;
1002 #endif
1003 		setup_clear_cpu_cap(X86_FEATURE_APIC);
1004 	}
1005 
1006 #ifdef CONFIG_PCI
1007 	if (pci_early_dump_regs)
1008 		early_dump_pci_devices();
1009 #endif
1010 
1011 	/* update the e820_saved too */
1012 	e820_reserve_setup_data();
1013 	finish_e820_parsing();
1014 
1015 	if (efi_enabled(EFI_BOOT))
1016 		efi_init();
1017 
1018 	dmi_scan_machine();
1019 	dmi_memdev_walk();
1020 	dmi_set_dump_stack_arch_desc();
1021 
1022 	/*
1023 	 * VMware detection requires dmi to be available, so this
1024 	 * needs to be done after dmi_scan_machine, for the BP.
1025 	 */
1026 	init_hypervisor_platform();
1027 
1028 	x86_init.resources.probe_roms();
1029 
1030 	/* after parse_early_param, so could debug it */
1031 	insert_resource(&iomem_resource, &code_resource);
1032 	insert_resource(&iomem_resource, &data_resource);
1033 	insert_resource(&iomem_resource, &bss_resource);
1034 
1035 	e820_add_kernel_range();
1036 	trim_bios_range();
1037 #ifdef CONFIG_X86_32
1038 	if (ppro_with_ram_bug()) {
1039 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1040 				  E820_RESERVED);
1041 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
1042 		printk(KERN_INFO "fixed physical RAM map:\n");
1043 		e820_print_map("bad_ppro");
1044 	}
1045 #else
1046 	early_gart_iommu_check();
1047 #endif
1048 
1049 	/*
1050 	 * partially used pages are not usable - thus
1051 	 * we are rounding upwards:
1052 	 */
1053 	max_pfn = e820_end_of_ram_pfn();
1054 
1055 	/* update e820 for memory not covered by WB MTRRs */
1056 	mtrr_bp_init();
1057 	if (mtrr_trim_uncached_memory(max_pfn))
1058 		max_pfn = e820_end_of_ram_pfn();
1059 
1060 	max_possible_pfn = max_pfn;
1061 
1062 #ifdef CONFIG_X86_32
1063 	/* max_low_pfn get updated here */
1064 	find_low_pfn_range();
1065 #else
1066 	check_x2apic();
1067 
1068 	/* How many end-of-memory variables you have, grandma! */
1069 	/* need this before calling reserve_initrd */
1070 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1071 		max_low_pfn = e820_end_of_low_ram_pfn();
1072 	else
1073 		max_low_pfn = max_pfn;
1074 
1075 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1076 #endif
1077 
1078 	/*
1079 	 * Find and reserve possible boot-time SMP configuration:
1080 	 */
1081 	find_smp_config();
1082 
1083 	reserve_ibft_region();
1084 
1085 	early_alloc_pgt_buf();
1086 
1087 	/*
1088 	 * Need to conclude brk, before memblock_x86_fill()
1089 	 *  it could use memblock_find_in_range, could overlap with
1090 	 *  brk area.
1091 	 */
1092 	reserve_brk();
1093 
1094 	cleanup_highmap();
1095 
1096 	memblock_set_current_limit(ISA_END_ADDRESS);
1097 	memblock_x86_fill();
1098 
1099 	if (efi_enabled(EFI_BOOT)) {
1100 		efi_fake_memmap();
1101 		efi_find_mirror();
1102 	}
1103 
1104 	/*
1105 	 * The EFI specification says that boot service code won't be called
1106 	 * after ExitBootServices(). This is, in fact, a lie.
1107 	 */
1108 	if (efi_enabled(EFI_MEMMAP))
1109 		efi_reserve_boot_services();
1110 
1111 	/* preallocate 4k for mptable mpc */
1112 	early_reserve_e820_mpc_new();
1113 
1114 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1115 	setup_bios_corruption_check();
1116 #endif
1117 
1118 #ifdef CONFIG_X86_32
1119 	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1120 			(max_pfn_mapped<<PAGE_SHIFT) - 1);
1121 #endif
1122 
1123 	reserve_real_mode();
1124 
1125 	trim_platform_memory_ranges();
1126 	trim_low_memory_range();
1127 
1128 	init_mem_mapping();
1129 
1130 	early_trap_pf_init();
1131 
1132 	setup_real_mode();
1133 
1134 	memblock_set_current_limit(get_max_mapped());
1135 
1136 	/*
1137 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1138 	 */
1139 
1140 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1141 	if (init_ohci1394_dma_early)
1142 		init_ohci1394_dma_on_all_controllers();
1143 #endif
1144 	/* Allocate bigger log buffer */
1145 	setup_log_buf(1);
1146 
1147 	reserve_initrd();
1148 
1149 	early_initrd_acpi_init();
1150 
1151 	vsmp_init();
1152 
1153 	io_delay_init();
1154 
1155 	/*
1156 	 * Parse the ACPI tables for possible boot-time SMP configuration.
1157 	 */
1158 	acpi_boot_table_init();
1159 
1160 	early_acpi_boot_init();
1161 
1162 	initmem_init();
1163 	dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1164 
1165 	/*
1166 	 * Reserve memory for crash kernel after SRAT is parsed so that it
1167 	 * won't consume hotpluggable memory.
1168 	 */
1169 	reserve_crashkernel();
1170 
1171 	memblock_find_dma_reserve();
1172 
1173 #ifdef CONFIG_KVM_GUEST
1174 	kvmclock_init();
1175 #endif
1176 
1177 	x86_init.paging.pagetable_init();
1178 
1179 	kasan_init();
1180 
1181 	if (boot_cpu_data.cpuid_level >= 0) {
1182 		/* A CPU has %cr4 if and only if it has CPUID */
1183 		mmu_cr4_features = __read_cr4();
1184 		if (trampoline_cr4_features)
1185 			*trampoline_cr4_features = mmu_cr4_features;
1186 	}
1187 
1188 #ifdef CONFIG_X86_32
1189 	/* sync back kernel address range */
1190 	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1191 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1192 			KERNEL_PGD_PTRS);
1193 
1194 	/*
1195 	 * sync back low identity map too.  It is used for example
1196 	 * in the 32-bit EFI stub.
1197 	 */
1198 	clone_pgd_range(initial_page_table,
1199 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1200 			min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY));
1201 #endif
1202 
1203 	tboot_probe();
1204 
1205 	map_vsyscall();
1206 
1207 	generic_apic_probe();
1208 
1209 	early_quirks();
1210 
1211 	/*
1212 	 * Read APIC and some other early information from ACPI tables.
1213 	 */
1214 	acpi_boot_init();
1215 	sfi_init();
1216 	x86_dtb_init();
1217 
1218 	/*
1219 	 * get boot-time SMP configuration:
1220 	 */
1221 	if (smp_found_config)
1222 		get_smp_config();
1223 
1224 	prefill_possible_map();
1225 
1226 	init_cpu_to_node();
1227 
1228 	init_apic_mappings();
1229 	io_apic_init_mappings();
1230 
1231 	kvm_guest_init();
1232 
1233 	e820_reserve_resources();
1234 	e820_mark_nosave_regions(max_low_pfn);
1235 
1236 	x86_init.resources.reserve_resources();
1237 
1238 	e820_setup_gap();
1239 
1240 #ifdef CONFIG_VT
1241 #if defined(CONFIG_VGA_CONSOLE)
1242 	if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1243 		conswitchp = &vga_con;
1244 #elif defined(CONFIG_DUMMY_CONSOLE)
1245 	conswitchp = &dummy_con;
1246 #endif
1247 #endif
1248 	x86_init.oem.banner();
1249 
1250 	x86_init.timers.wallclock_init();
1251 
1252 	mcheck_init();
1253 
1254 	arch_init_ideal_nops();
1255 
1256 	register_refined_jiffies(CLOCK_TICK_RATE);
1257 
1258 #ifdef CONFIG_EFI
1259 	if (efi_enabled(EFI_BOOT))
1260 		efi_apply_memmap_quirks();
1261 #endif
1262 }
1263 
1264 #ifdef CONFIG_X86_32
1265 
1266 static struct resource video_ram_resource = {
1267 	.name	= "Video RAM area",
1268 	.start	= 0xa0000,
1269 	.end	= 0xbffff,
1270 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1271 };
1272 
1273 void __init i386_reserve_resources(void)
1274 {
1275 	request_resource(&iomem_resource, &video_ram_resource);
1276 	reserve_standard_io_resources();
1277 }
1278 
1279 #endif /* CONFIG_X86_32 */
1280 
1281 static struct notifier_block kernel_offset_notifier = {
1282 	.notifier_call = dump_kernel_offset
1283 };
1284 
1285 static int __init register_kernel_offset_dumper(void)
1286 {
1287 	atomic_notifier_chain_register(&panic_notifier_list,
1288 					&kernel_offset_notifier);
1289 	return 0;
1290 }
1291 __initcall(register_kernel_offset_dumper);
1292 
1293 void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1294 {
1295 	if (!boot_cpu_has(X86_FEATURE_OSPKE))
1296 		return;
1297 
1298 	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
1299 }
1300