xref: /openbmc/linux/arch/x86/kernel/setup.c (revision 8730046c)
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 
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 __ro_after_init;
214 #else
215 __visible unsigned long mmu_cr4_features __ro_after_init = 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 #else
404 static void __init early_reserve_initrd(void)
405 {
406 }
407 static void __init reserve_initrd(void)
408 {
409 }
410 #endif /* CONFIG_BLK_DEV_INITRD */
411 
412 static void __init parse_setup_data(void)
413 {
414 	struct setup_data *data;
415 	u64 pa_data, pa_next;
416 
417 	pa_data = boot_params.hdr.setup_data;
418 	while (pa_data) {
419 		u32 data_len, data_type;
420 
421 		data = early_memremap(pa_data, sizeof(*data));
422 		data_len = data->len + sizeof(struct setup_data);
423 		data_type = data->type;
424 		pa_next = data->next;
425 		early_memunmap(data, sizeof(*data));
426 
427 		switch (data_type) {
428 		case SETUP_E820_EXT:
429 			parse_e820_ext(pa_data, data_len);
430 			break;
431 		case SETUP_DTB:
432 			add_dtb(pa_data);
433 			break;
434 		case SETUP_EFI:
435 			parse_efi_setup(pa_data, data_len);
436 			break;
437 		default:
438 			break;
439 		}
440 		pa_data = pa_next;
441 	}
442 }
443 
444 static void __init e820_reserve_setup_data(void)
445 {
446 	struct setup_data *data;
447 	u64 pa_data;
448 
449 	pa_data = boot_params.hdr.setup_data;
450 	if (!pa_data)
451 		return;
452 
453 	while (pa_data) {
454 		data = early_memremap(pa_data, sizeof(*data));
455 		e820_update_range(pa_data, sizeof(*data)+data->len,
456 			 E820_RAM, E820_RESERVED_KERN);
457 		pa_data = data->next;
458 		early_memunmap(data, sizeof(*data));
459 	}
460 
461 	sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
462 	memcpy(e820_saved, e820, sizeof(struct e820map));
463 	printk(KERN_INFO "extended physical RAM map:\n");
464 	e820_print_map("reserve setup_data");
465 }
466 
467 static void __init memblock_x86_reserve_range_setup_data(void)
468 {
469 	struct setup_data *data;
470 	u64 pa_data;
471 
472 	pa_data = boot_params.hdr.setup_data;
473 	while (pa_data) {
474 		data = early_memremap(pa_data, sizeof(*data));
475 		memblock_reserve(pa_data, sizeof(*data) + data->len);
476 		pa_data = data->next;
477 		early_memunmap(data, sizeof(*data));
478 	}
479 }
480 
481 /*
482  * --------- Crashkernel reservation ------------------------------
483  */
484 
485 #ifdef CONFIG_KEXEC_CORE
486 
487 /* 16M alignment for crash kernel regions */
488 #define CRASH_ALIGN		(16 << 20)
489 
490 /*
491  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
492  * would limit the kernel to the low 512 MiB due to mapping restrictions.
493  * On 64bit, old kexec-tools need to under 896MiB.
494  */
495 #ifdef CONFIG_X86_32
496 # define CRASH_ADDR_LOW_MAX	(512 << 20)
497 # define CRASH_ADDR_HIGH_MAX	(512 << 20)
498 #else
499 # define CRASH_ADDR_LOW_MAX	(896UL << 20)
500 # define CRASH_ADDR_HIGH_MAX	MAXMEM
501 #endif
502 
503 static int __init reserve_crashkernel_low(void)
504 {
505 #ifdef CONFIG_X86_64
506 	unsigned long long base, low_base = 0, low_size = 0;
507 	unsigned long total_low_mem;
508 	int ret;
509 
510 	total_low_mem = memblock_mem_size(1UL << (32 - PAGE_SHIFT));
511 
512 	/* crashkernel=Y,low */
513 	ret = parse_crashkernel_low(boot_command_line, total_low_mem, &low_size, &base);
514 	if (ret) {
515 		/*
516 		 * two parts from lib/swiotlb.c:
517 		 * -swiotlb size: user-specified with swiotlb= or default.
518 		 *
519 		 * -swiotlb overflow buffer: now hardcoded to 32k. We round it
520 		 * to 8M for other buffers that may need to stay low too. Also
521 		 * make sure we allocate enough extra low memory so that we
522 		 * don't run out of DMA buffers for 32-bit devices.
523 		 */
524 		low_size = max(swiotlb_size_or_default() + (8UL << 20), 256UL << 20);
525 	} else {
526 		/* passed with crashkernel=0,low ? */
527 		if (!low_size)
528 			return 0;
529 	}
530 
531 	low_base = memblock_find_in_range(low_size, 1ULL << 32, low_size, CRASH_ALIGN);
532 	if (!low_base) {
533 		pr_err("Cannot reserve %ldMB crashkernel low memory, please try smaller size.\n",
534 		       (unsigned long)(low_size >> 20));
535 		return -ENOMEM;
536 	}
537 
538 	ret = memblock_reserve(low_base, low_size);
539 	if (ret) {
540 		pr_err("%s: Error reserving crashkernel low memblock.\n", __func__);
541 		return ret;
542 	}
543 
544 	pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
545 		(unsigned long)(low_size >> 20),
546 		(unsigned long)(low_base >> 20),
547 		(unsigned long)(total_low_mem >> 20));
548 
549 	crashk_low_res.start = low_base;
550 	crashk_low_res.end   = low_base + low_size - 1;
551 	insert_resource(&iomem_resource, &crashk_low_res);
552 #endif
553 	return 0;
554 }
555 
556 static void __init reserve_crashkernel(void)
557 {
558 	unsigned long long crash_size, crash_base, total_mem;
559 	bool high = false;
560 	int ret;
561 
562 	total_mem = memblock_phys_mem_size();
563 
564 	/* crashkernel=XM */
565 	ret = parse_crashkernel(boot_command_line, total_mem, &crash_size, &crash_base);
566 	if (ret != 0 || crash_size <= 0) {
567 		/* crashkernel=X,high */
568 		ret = parse_crashkernel_high(boot_command_line, total_mem,
569 					     &crash_size, &crash_base);
570 		if (ret != 0 || crash_size <= 0)
571 			return;
572 		high = true;
573 	}
574 
575 	/* 0 means: find the address automatically */
576 	if (crash_base <= 0) {
577 		/*
578 		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
579 		 */
580 		crash_base = memblock_find_in_range(CRASH_ALIGN,
581 						    high ? CRASH_ADDR_HIGH_MAX
582 							 : CRASH_ADDR_LOW_MAX,
583 						    crash_size, CRASH_ALIGN);
584 		if (!crash_base) {
585 			pr_info("crashkernel reservation failed - No suitable area found.\n");
586 			return;
587 		}
588 
589 	} else {
590 		unsigned long long start;
591 
592 		start = memblock_find_in_range(crash_base,
593 					       crash_base + crash_size,
594 					       crash_size, 1 << 20);
595 		if (start != crash_base) {
596 			pr_info("crashkernel reservation failed - memory is in use.\n");
597 			return;
598 		}
599 	}
600 	ret = memblock_reserve(crash_base, crash_size);
601 	if (ret) {
602 		pr_err("%s: Error reserving crashkernel memblock.\n", __func__);
603 		return;
604 	}
605 
606 	if (crash_base >= (1ULL << 32) && reserve_crashkernel_low()) {
607 		memblock_free(crash_base, crash_size);
608 		return;
609 	}
610 
611 	pr_info("Reserving %ldMB of memory at %ldMB for crashkernel (System RAM: %ldMB)\n",
612 		(unsigned long)(crash_size >> 20),
613 		(unsigned long)(crash_base >> 20),
614 		(unsigned long)(total_mem >> 20));
615 
616 	crashk_res.start = crash_base;
617 	crashk_res.end   = crash_base + crash_size - 1;
618 	insert_resource(&iomem_resource, &crashk_res);
619 }
620 #else
621 static void __init reserve_crashkernel(void)
622 {
623 }
624 #endif
625 
626 static struct resource standard_io_resources[] = {
627 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
628 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
629 	{ .name = "pic1", .start = 0x20, .end = 0x21,
630 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
631 	{ .name = "timer0", .start = 0x40, .end = 0x43,
632 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
633 	{ .name = "timer1", .start = 0x50, .end = 0x53,
634 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
635 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
636 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
637 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
638 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
639 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
640 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
641 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
642 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
643 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
644 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
645 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
646 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
647 };
648 
649 void __init reserve_standard_io_resources(void)
650 {
651 	int i;
652 
653 	/* request I/O space for devices used on all i[345]86 PCs */
654 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
655 		request_resource(&ioport_resource, &standard_io_resources[i]);
656 
657 }
658 
659 static __init void reserve_ibft_region(void)
660 {
661 	unsigned long addr, size = 0;
662 
663 	addr = find_ibft_region(&size);
664 
665 	if (size)
666 		memblock_reserve(addr, size);
667 }
668 
669 static bool __init snb_gfx_workaround_needed(void)
670 {
671 #ifdef CONFIG_PCI
672 	int i;
673 	u16 vendor, devid;
674 	static const __initconst u16 snb_ids[] = {
675 		0x0102,
676 		0x0112,
677 		0x0122,
678 		0x0106,
679 		0x0116,
680 		0x0126,
681 		0x010a,
682 	};
683 
684 	/* Assume no if something weird is going on with PCI */
685 	if (!early_pci_allowed())
686 		return false;
687 
688 	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
689 	if (vendor != 0x8086)
690 		return false;
691 
692 	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
693 	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
694 		if (devid == snb_ids[i])
695 			return true;
696 #endif
697 
698 	return false;
699 }
700 
701 /*
702  * Sandy Bridge graphics has trouble with certain ranges, exclude
703  * them from allocation.
704  */
705 static void __init trim_snb_memory(void)
706 {
707 	static const __initconst unsigned long bad_pages[] = {
708 		0x20050000,
709 		0x20110000,
710 		0x20130000,
711 		0x20138000,
712 		0x40004000,
713 	};
714 	int i;
715 
716 	if (!snb_gfx_workaround_needed())
717 		return;
718 
719 	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
720 
721 	/*
722 	 * Reserve all memory below the 1 MB mark that has not
723 	 * already been reserved.
724 	 */
725 	memblock_reserve(0, 1<<20);
726 
727 	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
728 		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
729 			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
730 			       bad_pages[i]);
731 	}
732 }
733 
734 /*
735  * Here we put platform-specific memory range workarounds, i.e.
736  * memory known to be corrupt or otherwise in need to be reserved on
737  * specific platforms.
738  *
739  * If this gets used more widely it could use a real dispatch mechanism.
740  */
741 static void __init trim_platform_memory_ranges(void)
742 {
743 	trim_snb_memory();
744 }
745 
746 static void __init trim_bios_range(void)
747 {
748 	/*
749 	 * A special case is the first 4Kb of memory;
750 	 * This is a BIOS owned area, not kernel ram, but generally
751 	 * not listed as such in the E820 table.
752 	 *
753 	 * This typically reserves additional memory (64KiB by default)
754 	 * since some BIOSes are known to corrupt low memory.  See the
755 	 * Kconfig help text for X86_RESERVE_LOW.
756 	 */
757 	e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
758 
759 	/*
760 	 * special case: Some BIOSen report the PC BIOS
761 	 * area (640->1Mb) as ram even though it is not.
762 	 * take them out.
763 	 */
764 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
765 
766 	sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
767 }
768 
769 /* called before trim_bios_range() to spare extra sanitize */
770 static void __init e820_add_kernel_range(void)
771 {
772 	u64 start = __pa_symbol(_text);
773 	u64 size = __pa_symbol(_end) - start;
774 
775 	/*
776 	 * Complain if .text .data and .bss are not marked as E820_RAM and
777 	 * attempt to fix it by adding the range. We may have a confused BIOS,
778 	 * or the user may have used memmap=exactmap or memmap=xxM$yyM to
779 	 * exclude kernel range. If we really are running on top non-RAM,
780 	 * we will crash later anyways.
781 	 */
782 	if (e820_all_mapped(start, start + size, E820_RAM))
783 		return;
784 
785 	pr_warn(".text .data .bss are not marked as E820_RAM!\n");
786 	e820_remove_range(start, size, E820_RAM, 0);
787 	e820_add_region(start, size, E820_RAM);
788 }
789 
790 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
791 
792 static int __init parse_reservelow(char *p)
793 {
794 	unsigned long long size;
795 
796 	if (!p)
797 		return -EINVAL;
798 
799 	size = memparse(p, &p);
800 
801 	if (size < 4096)
802 		size = 4096;
803 
804 	if (size > 640*1024)
805 		size = 640*1024;
806 
807 	reserve_low = size;
808 
809 	return 0;
810 }
811 
812 early_param("reservelow", parse_reservelow);
813 
814 static void __init trim_low_memory_range(void)
815 {
816 	memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
817 }
818 
819 /*
820  * Dump out kernel offset information on panic.
821  */
822 static int
823 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
824 {
825 	if (kaslr_enabled()) {
826 		pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
827 			 kaslr_offset(),
828 			 __START_KERNEL,
829 			 __START_KERNEL_map,
830 			 MODULES_VADDR-1);
831 	} else {
832 		pr_emerg("Kernel Offset: disabled\n");
833 	}
834 
835 	return 0;
836 }
837 
838 /*
839  * Determine if we were loaded by an EFI loader.  If so, then we have also been
840  * passed the efi memmap, systab, etc., so we should use these data structures
841  * for initialization.  Note, the efi init code path is determined by the
842  * global efi_enabled. This allows the same kernel image to be used on existing
843  * systems (with a traditional BIOS) as well as on EFI systems.
844  */
845 /*
846  * setup_arch - architecture-specific boot-time initializations
847  *
848  * Note: On x86_64, fixmaps are ready for use even before this is called.
849  */
850 
851 void __init setup_arch(char **cmdline_p)
852 {
853 	memblock_reserve(__pa_symbol(_text),
854 			 (unsigned long)__bss_stop - (unsigned long)_text);
855 
856 	early_reserve_initrd();
857 
858 	/*
859 	 * At this point everything still needed from the boot loader
860 	 * or BIOS or kernel text should be early reserved or marked not
861 	 * RAM in e820. All other memory is free game.
862 	 */
863 
864 #ifdef CONFIG_X86_32
865 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
866 
867 	/*
868 	 * copy kernel address range established so far and switch
869 	 * to the proper swapper page table
870 	 */
871 	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
872 			initial_page_table + KERNEL_PGD_BOUNDARY,
873 			KERNEL_PGD_PTRS);
874 
875 	load_cr3(swapper_pg_dir);
876 	/*
877 	 * Note: Quark X1000 CPUs advertise PGE incorrectly and require
878 	 * a cr3 based tlb flush, so the following __flush_tlb_all()
879 	 * will not flush anything because the cpu quirk which clears
880 	 * X86_FEATURE_PGE has not been invoked yet. Though due to the
881 	 * load_cr3() above the TLB has been flushed already. The
882 	 * quirk is invoked before subsequent calls to __flush_tlb_all()
883 	 * so proper operation is guaranteed.
884 	 */
885 	__flush_tlb_all();
886 #else
887 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
888 #endif
889 
890 	/*
891 	 * If we have OLPC OFW, we might end up relocating the fixmap due to
892 	 * reserve_top(), so do this before touching the ioremap area.
893 	 */
894 	olpc_ofw_detect();
895 
896 	early_trap_init();
897 	early_cpu_init();
898 	early_ioremap_init();
899 
900 	setup_olpc_ofw_pgd();
901 
902 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
903 	screen_info = boot_params.screen_info;
904 	edid_info = boot_params.edid_info;
905 #ifdef CONFIG_X86_32
906 	apm_info.bios = boot_params.apm_bios_info;
907 	ist_info = boot_params.ist_info;
908 #endif
909 	saved_video_mode = boot_params.hdr.vid_mode;
910 	bootloader_type = boot_params.hdr.type_of_loader;
911 	if ((bootloader_type >> 4) == 0xe) {
912 		bootloader_type &= 0xf;
913 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
914 	}
915 	bootloader_version  = bootloader_type & 0xf;
916 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
917 
918 #ifdef CONFIG_BLK_DEV_RAM
919 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
920 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
921 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
922 #endif
923 #ifdef CONFIG_EFI
924 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
925 		     EFI32_LOADER_SIGNATURE, 4)) {
926 		set_bit(EFI_BOOT, &efi.flags);
927 	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
928 		     EFI64_LOADER_SIGNATURE, 4)) {
929 		set_bit(EFI_BOOT, &efi.flags);
930 		set_bit(EFI_64BIT, &efi.flags);
931 	}
932 
933 	if (efi_enabled(EFI_BOOT))
934 		efi_memblock_x86_reserve_range();
935 #endif
936 
937 	x86_init.oem.arch_setup();
938 
939 	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
940 	setup_memory_map();
941 	parse_setup_data();
942 
943 	copy_edd();
944 
945 	if (!boot_params.hdr.root_flags)
946 		root_mountflags &= ~MS_RDONLY;
947 	init_mm.start_code = (unsigned long) _text;
948 	init_mm.end_code = (unsigned long) _etext;
949 	init_mm.end_data = (unsigned long) _edata;
950 	init_mm.brk = _brk_end;
951 
952 	mpx_mm_init(&init_mm);
953 
954 	code_resource.start = __pa_symbol(_text);
955 	code_resource.end = __pa_symbol(_etext)-1;
956 	data_resource.start = __pa_symbol(_etext);
957 	data_resource.end = __pa_symbol(_edata)-1;
958 	bss_resource.start = __pa_symbol(__bss_start);
959 	bss_resource.end = __pa_symbol(__bss_stop)-1;
960 
961 #ifdef CONFIG_CMDLINE_BOOL
962 #ifdef CONFIG_CMDLINE_OVERRIDE
963 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
964 #else
965 	if (builtin_cmdline[0]) {
966 		/* append boot loader cmdline to builtin */
967 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
968 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
969 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
970 	}
971 #endif
972 #endif
973 
974 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
975 	*cmdline_p = command_line;
976 
977 	/*
978 	 * x86_configure_nx() is called before parse_early_param() to detect
979 	 * whether hardware doesn't support NX (so that the early EHCI debug
980 	 * console setup can safely call set_fixmap()). It may then be called
981 	 * again from within noexec_setup() during parsing early parameters
982 	 * to honor the respective command line option.
983 	 */
984 	x86_configure_nx();
985 
986 	parse_early_param();
987 
988 #ifdef CONFIG_MEMORY_HOTPLUG
989 	/*
990 	 * Memory used by the kernel cannot be hot-removed because Linux
991 	 * cannot migrate the kernel pages. When memory hotplug is
992 	 * enabled, we should prevent memblock from allocating memory
993 	 * for the kernel.
994 	 *
995 	 * ACPI SRAT records all hotpluggable memory ranges. But before
996 	 * SRAT is parsed, we don't know about it.
997 	 *
998 	 * The kernel image is loaded into memory at very early time. We
999 	 * cannot prevent this anyway. So on NUMA system, we set any
1000 	 * node the kernel resides in as un-hotpluggable.
1001 	 *
1002 	 * Since on modern servers, one node could have double-digit
1003 	 * gigabytes memory, we can assume the memory around the kernel
1004 	 * image is also un-hotpluggable. So before SRAT is parsed, just
1005 	 * allocate memory near the kernel image to try the best to keep
1006 	 * the kernel away from hotpluggable memory.
1007 	 */
1008 	if (movable_node_is_enabled())
1009 		memblock_set_bottom_up(true);
1010 #endif
1011 
1012 	x86_report_nx();
1013 
1014 	/* after early param, so could get panic from serial */
1015 	memblock_x86_reserve_range_setup_data();
1016 
1017 	if (acpi_mps_check()) {
1018 #ifdef CONFIG_X86_LOCAL_APIC
1019 		disable_apic = 1;
1020 #endif
1021 		setup_clear_cpu_cap(X86_FEATURE_APIC);
1022 	}
1023 
1024 #ifdef CONFIG_PCI
1025 	if (pci_early_dump_regs)
1026 		early_dump_pci_devices();
1027 #endif
1028 
1029 	/* update the e820_saved too */
1030 	e820_reserve_setup_data();
1031 	finish_e820_parsing();
1032 
1033 	if (efi_enabled(EFI_BOOT))
1034 		efi_init();
1035 
1036 	dmi_scan_machine();
1037 	dmi_memdev_walk();
1038 	dmi_set_dump_stack_arch_desc();
1039 
1040 	/*
1041 	 * VMware detection requires dmi to be available, so this
1042 	 * needs to be done after dmi_scan_machine, for the BP.
1043 	 */
1044 	init_hypervisor_platform();
1045 
1046 	x86_init.resources.probe_roms();
1047 
1048 	/* after parse_early_param, so could debug it */
1049 	insert_resource(&iomem_resource, &code_resource);
1050 	insert_resource(&iomem_resource, &data_resource);
1051 	insert_resource(&iomem_resource, &bss_resource);
1052 
1053 	e820_add_kernel_range();
1054 	trim_bios_range();
1055 #ifdef CONFIG_X86_32
1056 	if (ppro_with_ram_bug()) {
1057 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1058 				  E820_RESERVED);
1059 		sanitize_e820_map(e820->map, ARRAY_SIZE(e820->map), &e820->nr_map);
1060 		printk(KERN_INFO "fixed physical RAM map:\n");
1061 		e820_print_map("bad_ppro");
1062 	}
1063 #else
1064 	early_gart_iommu_check();
1065 #endif
1066 
1067 	/*
1068 	 * partially used pages are not usable - thus
1069 	 * we are rounding upwards:
1070 	 */
1071 	max_pfn = e820_end_of_ram_pfn();
1072 
1073 	/* update e820 for memory not covered by WB MTRRs */
1074 	mtrr_bp_init();
1075 	if (mtrr_trim_uncached_memory(max_pfn))
1076 		max_pfn = e820_end_of_ram_pfn();
1077 
1078 	max_possible_pfn = max_pfn;
1079 
1080 	/*
1081 	 * Define random base addresses for memory sections after max_pfn is
1082 	 * defined and before each memory section base is used.
1083 	 */
1084 	kernel_randomize_memory();
1085 
1086 #ifdef CONFIG_X86_32
1087 	/* max_low_pfn get updated here */
1088 	find_low_pfn_range();
1089 #else
1090 	check_x2apic();
1091 
1092 	/* How many end-of-memory variables you have, grandma! */
1093 	/* need this before calling reserve_initrd */
1094 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1095 		max_low_pfn = e820_end_of_low_ram_pfn();
1096 	else
1097 		max_low_pfn = max_pfn;
1098 
1099 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1100 #endif
1101 
1102 	/*
1103 	 * Find and reserve possible boot-time SMP configuration:
1104 	 */
1105 	find_smp_config();
1106 
1107 	reserve_ibft_region();
1108 
1109 	early_alloc_pgt_buf();
1110 
1111 	/*
1112 	 * Need to conclude brk, before memblock_x86_fill()
1113 	 *  it could use memblock_find_in_range, could overlap with
1114 	 *  brk area.
1115 	 */
1116 	reserve_brk();
1117 
1118 	cleanup_highmap();
1119 
1120 	memblock_set_current_limit(ISA_END_ADDRESS);
1121 	memblock_x86_fill();
1122 
1123 	reserve_bios_regions();
1124 
1125 	if (efi_enabled(EFI_MEMMAP)) {
1126 		efi_fake_memmap();
1127 		efi_find_mirror();
1128 		efi_esrt_init();
1129 
1130 		/*
1131 		 * The EFI specification says that boot service code won't be
1132 		 * called after ExitBootServices(). This is, in fact, a lie.
1133 		 */
1134 		efi_reserve_boot_services();
1135 	}
1136 
1137 	/* preallocate 4k for mptable mpc */
1138 	early_reserve_e820_mpc_new();
1139 
1140 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1141 	setup_bios_corruption_check();
1142 #endif
1143 
1144 #ifdef CONFIG_X86_32
1145 	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1146 			(max_pfn_mapped<<PAGE_SHIFT) - 1);
1147 #endif
1148 
1149 	reserve_real_mode();
1150 
1151 	trim_platform_memory_ranges();
1152 	trim_low_memory_range();
1153 
1154 	init_mem_mapping();
1155 
1156 	early_trap_pf_init();
1157 
1158 	/*
1159 	 * Update mmu_cr4_features (and, indirectly, trampoline_cr4_features)
1160 	 * with the current CR4 value.  This may not be necessary, but
1161 	 * auditing all the early-boot CR4 manipulation would be needed to
1162 	 * rule it out.
1163 	 */
1164 	mmu_cr4_features = __read_cr4();
1165 
1166 	memblock_set_current_limit(get_max_mapped());
1167 
1168 	/*
1169 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1170 	 */
1171 
1172 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1173 	if (init_ohci1394_dma_early)
1174 		init_ohci1394_dma_on_all_controllers();
1175 #endif
1176 	/* Allocate bigger log buffer */
1177 	setup_log_buf(1);
1178 
1179 	reserve_initrd();
1180 
1181 	acpi_table_upgrade();
1182 
1183 	vsmp_init();
1184 
1185 	io_delay_init();
1186 
1187 	/*
1188 	 * Parse the ACPI tables for possible boot-time SMP configuration.
1189 	 */
1190 	acpi_boot_table_init();
1191 
1192 	early_acpi_boot_init();
1193 
1194 	initmem_init();
1195 	dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1196 
1197 	/*
1198 	 * Reserve memory for crash kernel after SRAT is parsed so that it
1199 	 * won't consume hotpluggable memory.
1200 	 */
1201 	reserve_crashkernel();
1202 
1203 	memblock_find_dma_reserve();
1204 
1205 #ifdef CONFIG_KVM_GUEST
1206 	kvmclock_init();
1207 #endif
1208 
1209 	x86_init.paging.pagetable_init();
1210 
1211 	kasan_init();
1212 
1213 #ifdef CONFIG_X86_32
1214 	/* sync back kernel address range */
1215 	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1216 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1217 			KERNEL_PGD_PTRS);
1218 
1219 	/*
1220 	 * sync back low identity map too.  It is used for example
1221 	 * in the 32-bit EFI stub.
1222 	 */
1223 	clone_pgd_range(initial_page_table,
1224 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1225 			min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY));
1226 #endif
1227 
1228 	tboot_probe();
1229 
1230 	map_vsyscall();
1231 
1232 	generic_apic_probe();
1233 
1234 	early_quirks();
1235 
1236 	/*
1237 	 * Read APIC and some other early information from ACPI tables.
1238 	 */
1239 	acpi_boot_init();
1240 	sfi_init();
1241 	x86_dtb_init();
1242 
1243 	/*
1244 	 * get boot-time SMP configuration:
1245 	 */
1246 	get_smp_config();
1247 
1248 	/*
1249 	 * Systems w/o ACPI and mptables might not have it mapped the local
1250 	 * APIC yet, but prefill_possible_map() might need to access it.
1251 	 */
1252 	init_apic_mappings();
1253 
1254 	prefill_possible_map();
1255 
1256 	init_cpu_to_node();
1257 
1258 	io_apic_init_mappings();
1259 
1260 	kvm_guest_init();
1261 
1262 	e820_reserve_resources();
1263 	e820_mark_nosave_regions(max_low_pfn);
1264 
1265 	x86_init.resources.reserve_resources();
1266 
1267 	e820_setup_gap();
1268 
1269 #ifdef CONFIG_VT
1270 #if defined(CONFIG_VGA_CONSOLE)
1271 	if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1272 		conswitchp = &vga_con;
1273 #elif defined(CONFIG_DUMMY_CONSOLE)
1274 	conswitchp = &dummy_con;
1275 #endif
1276 #endif
1277 	x86_init.oem.banner();
1278 
1279 	x86_init.timers.wallclock_init();
1280 
1281 	mcheck_init();
1282 
1283 	arch_init_ideal_nops();
1284 
1285 	register_refined_jiffies(CLOCK_TICK_RATE);
1286 
1287 #ifdef CONFIG_EFI
1288 	if (efi_enabled(EFI_BOOT))
1289 		efi_apply_memmap_quirks();
1290 #endif
1291 }
1292 
1293 #ifdef CONFIG_X86_32
1294 
1295 static struct resource video_ram_resource = {
1296 	.name	= "Video RAM area",
1297 	.start	= 0xa0000,
1298 	.end	= 0xbffff,
1299 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1300 };
1301 
1302 void __init i386_reserve_resources(void)
1303 {
1304 	request_resource(&iomem_resource, &video_ram_resource);
1305 	reserve_standard_io_resources();
1306 }
1307 
1308 #endif /* CONFIG_X86_32 */
1309 
1310 static struct notifier_block kernel_offset_notifier = {
1311 	.notifier_call = dump_kernel_offset
1312 };
1313 
1314 static int __init register_kernel_offset_dumper(void)
1315 {
1316 	atomic_notifier_chain_register(&panic_notifier_list,
1317 					&kernel_offset_notifier);
1318 	return 0;
1319 }
1320 __initcall(register_kernel_offset_dumper);
1321 
1322 void arch_show_smap(struct seq_file *m, struct vm_area_struct *vma)
1323 {
1324 	if (!boot_cpu_has(X86_FEATURE_OSPKE))
1325 		return;
1326 
1327 	seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
1328 }
1329