xref: /linux/arch/x86/kernel/setup.c (revision a0b54e256d513ed99e456bea6e4e188ff92e7c46)
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/apm_bios.h>
31 #include <linux/initrd.h>
32 #include <linux/bootmem.h>
33 #include <linux/seq_file.h>
34 #include <linux/console.h>
35 #include <linux/mca.h>
36 #include <linux/root_dev.h>
37 #include <linux/highmem.h>
38 #include <linux/module.h>
39 #include <linux/efi.h>
40 #include <linux/init.h>
41 #include <linux/edd.h>
42 #include <linux/iscsi_ibft.h>
43 #include <linux/nodemask.h>
44 #include <linux/kexec.h>
45 #include <linux/dmi.h>
46 #include <linux/pfn.h>
47 #include <linux/pci.h>
48 #include <asm/pci-direct.h>
49 #include <linux/init_ohci1394_dma.h>
50 #include <linux/kvm_para.h>
51 
52 #include <linux/errno.h>
53 #include <linux/kernel.h>
54 #include <linux/stddef.h>
55 #include <linux/unistd.h>
56 #include <linux/ptrace.h>
57 #include <linux/slab.h>
58 #include <linux/user.h>
59 #include <linux/delay.h>
60 
61 #include <linux/kallsyms.h>
62 #include <linux/cpufreq.h>
63 #include <linux/dma-mapping.h>
64 #include <linux/ctype.h>
65 #include <linux/uaccess.h>
66 
67 #include <linux/percpu.h>
68 #include <linux/crash_dump.h>
69 #include <linux/tboot.h>
70 
71 #include <video/edid.h>
72 
73 #include <asm/mtrr.h>
74 #include <asm/apic.h>
75 #include <asm/e820.h>
76 #include <asm/mpspec.h>
77 #include <asm/setup.h>
78 #include <asm/efi.h>
79 #include <asm/timer.h>
80 #include <asm/i8259.h>
81 #include <asm/sections.h>
82 #include <asm/dmi.h>
83 #include <asm/io_apic.h>
84 #include <asm/ist.h>
85 #include <asm/vmi.h>
86 #include <asm/setup_arch.h>
87 #include <asm/bios_ebda.h>
88 #include <asm/cacheflush.h>
89 #include <asm/processor.h>
90 #include <asm/bugs.h>
91 
92 #include <asm/system.h>
93 #include <asm/vsyscall.h>
94 #include <asm/cpu.h>
95 #include <asm/desc.h>
96 #include <asm/dma.h>
97 #include <asm/iommu.h>
98 #include <asm/gart.h>
99 #include <asm/mmu_context.h>
100 #include <asm/proto.h>
101 
102 #include <asm/paravirt.h>
103 #include <asm/hypervisor.h>
104 
105 #include <asm/percpu.h>
106 #include <asm/topology.h>
107 #include <asm/apicdef.h>
108 #ifdef CONFIG_X86_64
109 #include <asm/numa_64.h>
110 #endif
111 
112 #ifndef ARCH_SETUP
113 #define ARCH_SETUP
114 #endif
115 
116 /*
117  * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
118  * The direct mapping extends to max_pfn_mapped, so that we can directly access
119  * apertures, ACPI and other tables without having to play with fixmaps.
120  */
121 unsigned long max_low_pfn_mapped;
122 unsigned long max_pfn_mapped;
123 
124 RESERVE_BRK(dmi_alloc, 65536);
125 
126 unsigned int boot_cpu_id __read_mostly;
127 
128 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
129 unsigned long _brk_end = (unsigned long)__brk_base;
130 
131 #ifdef CONFIG_X86_64
132 int default_cpu_present_to_apicid(int mps_cpu)
133 {
134 	return __default_cpu_present_to_apicid(mps_cpu);
135 }
136 
137 int default_check_phys_apicid_present(int boot_cpu_physical_apicid)
138 {
139 	return __default_check_phys_apicid_present(boot_cpu_physical_apicid);
140 }
141 #endif
142 
143 #ifndef CONFIG_DEBUG_BOOT_PARAMS
144 struct boot_params __initdata boot_params;
145 #else
146 struct boot_params boot_params;
147 #endif
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_MEM
157 };
158 
159 static struct resource code_resource = {
160 	.name	= "Kernel code",
161 	.start	= 0,
162 	.end	= 0,
163 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
164 };
165 
166 static struct resource bss_resource = {
167 	.name	= "Kernel bss",
168 	.start	= 0,
169 	.end	= 0,
170 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
171 };
172 
173 
174 #ifdef CONFIG_X86_32
175 static struct resource video_ram_resource = {
176 	.name	= "Video RAM area",
177 	.start	= 0xa0000,
178 	.end	= 0xbffff,
179 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
180 };
181 
182 /* cpu data as detected by the assembly code in head.S */
183 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
184 /* common cpu data for all cpus */
185 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
186 EXPORT_SYMBOL(boot_cpu_data);
187 static void set_mca_bus(int x)
188 {
189 #ifdef CONFIG_MCA
190 	MCA_bus = x;
191 #endif
192 }
193 
194 unsigned int def_to_bigsmp;
195 
196 /* for MCA, but anyone else can use it if they want */
197 unsigned int machine_id;
198 unsigned int machine_submodel_id;
199 unsigned int BIOS_revision;
200 
201 struct apm_info apm_info;
202 EXPORT_SYMBOL(apm_info);
203 
204 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
205 	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
206 struct ist_info ist_info;
207 EXPORT_SYMBOL(ist_info);
208 #else
209 struct ist_info ist_info;
210 #endif
211 
212 #else
213 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
214 	.x86_phys_bits = MAX_PHYSMEM_BITS,
215 };
216 EXPORT_SYMBOL(boot_cpu_data);
217 #endif
218 
219 
220 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
221 unsigned long mmu_cr4_features;
222 #else
223 unsigned long mmu_cr4_features = X86_CR4_PAE;
224 #endif
225 
226 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
227 int bootloader_type, bootloader_version;
228 
229 /*
230  * Setup options
231  */
232 struct screen_info screen_info;
233 EXPORT_SYMBOL(screen_info);
234 struct edid_info edid_info;
235 EXPORT_SYMBOL_GPL(edid_info);
236 
237 extern int root_mountflags;
238 
239 unsigned long saved_video_mode;
240 
241 #define RAMDISK_IMAGE_START_MASK	0x07FF
242 #define RAMDISK_PROMPT_FLAG		0x8000
243 #define RAMDISK_LOAD_FLAG		0x4000
244 
245 static char __initdata command_line[COMMAND_LINE_SIZE];
246 #ifdef CONFIG_CMDLINE_BOOL
247 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
248 #endif
249 
250 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
251 struct edd edd;
252 #ifdef CONFIG_EDD_MODULE
253 EXPORT_SYMBOL(edd);
254 #endif
255 /**
256  * copy_edd() - Copy the BIOS EDD information
257  *              from boot_params into a safe place.
258  *
259  */
260 static inline void copy_edd(void)
261 {
262      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
263 	    sizeof(edd.mbr_signature));
264      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
265      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
266      edd.edd_info_nr = boot_params.eddbuf_entries;
267 }
268 #else
269 static inline void copy_edd(void)
270 {
271 }
272 #endif
273 
274 void * __init extend_brk(size_t size, size_t align)
275 {
276 	size_t mask = align - 1;
277 	void *ret;
278 
279 	BUG_ON(_brk_start == 0);
280 	BUG_ON(align & mask);
281 
282 	_brk_end = (_brk_end + mask) & ~mask;
283 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
284 
285 	ret = (void *)_brk_end;
286 	_brk_end += size;
287 
288 	memset(ret, 0, size);
289 
290 	return ret;
291 }
292 
293 #ifdef CONFIG_X86_64
294 static void __init init_gbpages(void)
295 {
296 	if (direct_gbpages && cpu_has_gbpages)
297 		printk(KERN_INFO "Using GB pages for direct mapping\n");
298 	else
299 		direct_gbpages = 0;
300 }
301 #else
302 static inline void init_gbpages(void)
303 {
304 }
305 #endif
306 
307 static void __init reserve_brk(void)
308 {
309 	if (_brk_end > _brk_start)
310 		reserve_early(__pa(_brk_start), __pa(_brk_end), "BRK");
311 
312 	/* Mark brk area as locked down and no longer taking any
313 	   new allocations */
314 	_brk_start = 0;
315 }
316 
317 #ifdef CONFIG_BLK_DEV_INITRD
318 
319 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
320 static void __init relocate_initrd(void)
321 {
322 
323 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
324 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
325 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
326 	u64 ramdisk_here;
327 	unsigned long slop, clen, mapaddr;
328 	char *p, *q;
329 
330 	/* We need to move the initrd down into lowmem */
331 	ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
332 					 PAGE_SIZE);
333 
334 	if (ramdisk_here == -1ULL)
335 		panic("Cannot find place for new RAMDISK of size %lld\n",
336 			 ramdisk_size);
337 
338 	/* Note: this includes all the lowmem currently occupied by
339 	   the initrd, we rely on that fact to keep the data intact. */
340 	reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
341 			 "NEW RAMDISK");
342 	initrd_start = ramdisk_here + PAGE_OFFSET;
343 	initrd_end   = initrd_start + ramdisk_size;
344 	printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
345 			 ramdisk_here, ramdisk_here + ramdisk_size);
346 
347 	q = (char *)initrd_start;
348 
349 	/* Copy any lowmem portion of the initrd */
350 	if (ramdisk_image < end_of_lowmem) {
351 		clen = end_of_lowmem - ramdisk_image;
352 		p = (char *)__va(ramdisk_image);
353 		memcpy(q, p, clen);
354 		q += clen;
355 		ramdisk_image += clen;
356 		ramdisk_size  -= clen;
357 	}
358 
359 	/* Copy the highmem portion of the initrd */
360 	while (ramdisk_size) {
361 		slop = ramdisk_image & ~PAGE_MASK;
362 		clen = ramdisk_size;
363 		if (clen > MAX_MAP_CHUNK-slop)
364 			clen = MAX_MAP_CHUNK-slop;
365 		mapaddr = ramdisk_image & PAGE_MASK;
366 		p = early_memremap(mapaddr, clen+slop);
367 		memcpy(q, p+slop, clen);
368 		early_iounmap(p, clen+slop);
369 		q += clen;
370 		ramdisk_image += clen;
371 		ramdisk_size  -= clen;
372 	}
373 	/* high pages is not converted by early_res_to_bootmem */
374 	ramdisk_image = boot_params.hdr.ramdisk_image;
375 	ramdisk_size  = boot_params.hdr.ramdisk_size;
376 	printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
377 		" %08llx - %08llx\n",
378 		ramdisk_image, ramdisk_image + ramdisk_size - 1,
379 		ramdisk_here, ramdisk_here + ramdisk_size - 1);
380 }
381 
382 static void __init reserve_initrd(void)
383 {
384 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
385 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
386 	u64 ramdisk_end   = ramdisk_image + ramdisk_size;
387 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
388 
389 	if (!boot_params.hdr.type_of_loader ||
390 	    !ramdisk_image || !ramdisk_size)
391 		return;		/* No initrd provided by bootloader */
392 
393 	initrd_start = 0;
394 
395 	if (ramdisk_size >= (end_of_lowmem>>1)) {
396 		free_early(ramdisk_image, ramdisk_end);
397 		printk(KERN_ERR "initrd too large to handle, "
398 		       "disabling initrd\n");
399 		return;
400 	}
401 
402 	printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
403 			ramdisk_end);
404 
405 
406 	if (ramdisk_end <= end_of_lowmem) {
407 		/* All in lowmem, easy case */
408 		/*
409 		 * don't need to reserve again, already reserved early
410 		 * in i386_start_kernel
411 		 */
412 		initrd_start = ramdisk_image + PAGE_OFFSET;
413 		initrd_end = initrd_start + ramdisk_size;
414 		return;
415 	}
416 
417 	relocate_initrd();
418 
419 	free_early(ramdisk_image, ramdisk_end);
420 }
421 #else
422 static void __init reserve_initrd(void)
423 {
424 }
425 #endif /* CONFIG_BLK_DEV_INITRD */
426 
427 static void __init parse_setup_data(void)
428 {
429 	struct setup_data *data;
430 	u64 pa_data;
431 
432 	if (boot_params.hdr.version < 0x0209)
433 		return;
434 	pa_data = boot_params.hdr.setup_data;
435 	while (pa_data) {
436 		data = early_memremap(pa_data, PAGE_SIZE);
437 		switch (data->type) {
438 		case SETUP_E820_EXT:
439 			parse_e820_ext(data, pa_data);
440 			break;
441 		default:
442 			break;
443 		}
444 		pa_data = data->next;
445 		early_iounmap(data, PAGE_SIZE);
446 	}
447 }
448 
449 static void __init e820_reserve_setup_data(void)
450 {
451 	struct setup_data *data;
452 	u64 pa_data;
453 	int found = 0;
454 
455 	if (boot_params.hdr.version < 0x0209)
456 		return;
457 	pa_data = boot_params.hdr.setup_data;
458 	while (pa_data) {
459 		data = early_memremap(pa_data, sizeof(*data));
460 		e820_update_range(pa_data, sizeof(*data)+data->len,
461 			 E820_RAM, E820_RESERVED_KERN);
462 		found = 1;
463 		pa_data = data->next;
464 		early_iounmap(data, sizeof(*data));
465 	}
466 	if (!found)
467 		return;
468 
469 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
470 	memcpy(&e820_saved, &e820, sizeof(struct e820map));
471 	printk(KERN_INFO "extended physical RAM map:\n");
472 	e820_print_map("reserve setup_data");
473 }
474 
475 static void __init reserve_early_setup_data(void)
476 {
477 	struct setup_data *data;
478 	u64 pa_data;
479 	char buf[32];
480 
481 	if (boot_params.hdr.version < 0x0209)
482 		return;
483 	pa_data = boot_params.hdr.setup_data;
484 	while (pa_data) {
485 		data = early_memremap(pa_data, sizeof(*data));
486 		sprintf(buf, "setup data %x", data->type);
487 		reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
488 		pa_data = data->next;
489 		early_iounmap(data, sizeof(*data));
490 	}
491 }
492 
493 /*
494  * --------- Crashkernel reservation ------------------------------
495  */
496 
497 #ifdef CONFIG_KEXEC
498 
499 /**
500  * Reserve @size bytes of crashkernel memory at any suitable offset.
501  *
502  * @size: Size of the crashkernel memory to reserve.
503  * Returns the base address on success, and -1ULL on failure.
504  */
505 static
506 unsigned long long __init find_and_reserve_crashkernel(unsigned long long size)
507 {
508 	const unsigned long long alignment = 16<<20; 	/* 16M */
509 	unsigned long long start = 0LL;
510 
511 	while (1) {
512 		int ret;
513 
514 		start = find_e820_area(start, ULONG_MAX, size, alignment);
515 		if (start == -1ULL)
516 			return start;
517 
518 		/* try to reserve it */
519 		ret = reserve_bootmem_generic(start, size, BOOTMEM_EXCLUSIVE);
520 		if (ret >= 0)
521 			return start;
522 
523 		start += alignment;
524 	}
525 }
526 
527 static inline unsigned long long get_total_mem(void)
528 {
529 	unsigned long long total;
530 
531 	total = max_low_pfn - min_low_pfn;
532 #ifdef CONFIG_HIGHMEM
533 	total += highend_pfn - highstart_pfn;
534 #endif
535 
536 	return total << PAGE_SHIFT;
537 }
538 
539 static void __init reserve_crashkernel(void)
540 {
541 	unsigned long long total_mem;
542 	unsigned long long crash_size, crash_base;
543 	int ret;
544 
545 	total_mem = get_total_mem();
546 
547 	ret = parse_crashkernel(boot_command_line, total_mem,
548 			&crash_size, &crash_base);
549 	if (ret != 0 || crash_size <= 0)
550 		return;
551 
552 	/* 0 means: find the address automatically */
553 	if (crash_base <= 0) {
554 		crash_base = find_and_reserve_crashkernel(crash_size);
555 		if (crash_base == -1ULL) {
556 			pr_info("crashkernel reservation failed. "
557 				"No suitable area found.\n");
558 			return;
559 		}
560 	} else {
561 		ret = reserve_bootmem_generic(crash_base, crash_size,
562 					BOOTMEM_EXCLUSIVE);
563 		if (ret < 0) {
564 			pr_info("crashkernel reservation failed - "
565 				"memory is in use\n");
566 			return;
567 		}
568 	}
569 
570 	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
571 			"for crashkernel (System RAM: %ldMB)\n",
572 			(unsigned long)(crash_size >> 20),
573 			(unsigned long)(crash_base >> 20),
574 			(unsigned long)(total_mem >> 20));
575 
576 	crashk_res.start = crash_base;
577 	crashk_res.end   = crash_base + crash_size - 1;
578 	insert_resource(&iomem_resource, &crashk_res);
579 }
580 #else
581 static void __init reserve_crashkernel(void)
582 {
583 }
584 #endif
585 
586 static struct resource standard_io_resources[] = {
587 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
588 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
589 	{ .name = "pic1", .start = 0x20, .end = 0x21,
590 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
591 	{ .name = "timer0", .start = 0x40, .end = 0x43,
592 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
593 	{ .name = "timer1", .start = 0x50, .end = 0x53,
594 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
595 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
596 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
597 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
598 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
599 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
600 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
601 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
602 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
603 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
604 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
605 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
606 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
607 };
608 
609 static void __init reserve_standard_io_resources(void)
610 {
611 	int i;
612 
613 	/* request I/O space for devices used on all i[345]86 PCs */
614 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
615 		request_resource(&ioport_resource, &standard_io_resources[i]);
616 
617 }
618 
619 /*
620  * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
621  * is_kdump_kernel() to determine if we are booting after a panic. Hence
622  * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
623  */
624 
625 #ifdef CONFIG_CRASH_DUMP
626 /* elfcorehdr= specifies the location of elf core header
627  * stored by the crashed kernel. This option will be passed
628  * by kexec loader to the capture kernel.
629  */
630 static int __init setup_elfcorehdr(char *arg)
631 {
632 	char *end;
633 	if (!arg)
634 		return -EINVAL;
635 	elfcorehdr_addr = memparse(arg, &end);
636 	return end > arg ? 0 : -EINVAL;
637 }
638 early_param("elfcorehdr", setup_elfcorehdr);
639 #endif
640 
641 static struct x86_quirks default_x86_quirks __initdata;
642 
643 struct x86_quirks *x86_quirks __initdata = &default_x86_quirks;
644 
645 #ifdef CONFIG_X86_RESERVE_LOW_64K
646 static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
647 {
648 	printk(KERN_NOTICE
649 		"%s detected: BIOS may corrupt low RAM, working around it.\n",
650 		d->ident);
651 
652 	e820_update_range(0, 0x10000, E820_RAM, E820_RESERVED);
653 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
654 
655 	return 0;
656 }
657 #endif
658 
659 /* List of systems that have known low memory corruption BIOS problems */
660 static struct dmi_system_id __initdata bad_bios_dmi_table[] = {
661 #ifdef CONFIG_X86_RESERVE_LOW_64K
662 	{
663 		.callback = dmi_low_memory_corruption,
664 		.ident = "AMI BIOS",
665 		.matches = {
666 			DMI_MATCH(DMI_BIOS_VENDOR, "American Megatrends Inc."),
667 		},
668 	},
669 	{
670 		.callback = dmi_low_memory_corruption,
671 		.ident = "Phoenix BIOS",
672 		.matches = {
673 			DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix Technologies"),
674 		},
675 	},
676 	{
677 	/*
678 	 * AMI BIOS with low memory corruption was found on Intel DG45ID board.
679 	 * It hase different DMI_BIOS_VENDOR = "Intel Corp.", for now we will
680 	 * match only DMI_BOARD_NAME and see if there is more bad products
681 	 * with this vendor.
682 	 */
683 		.callback = dmi_low_memory_corruption,
684 		.ident = "AMI BIOS",
685 		.matches = {
686 			DMI_MATCH(DMI_BOARD_NAME, "DG45ID"),
687 		},
688 	},
689 #endif
690 	{}
691 };
692 
693 /*
694  * Determine if we were loaded by an EFI loader.  If so, then we have also been
695  * passed the efi memmap, systab, etc., so we should use these data structures
696  * for initialization.  Note, the efi init code path is determined by the
697  * global efi_enabled. This allows the same kernel image to be used on existing
698  * systems (with a traditional BIOS) as well as on EFI systems.
699  */
700 /*
701  * setup_arch - architecture-specific boot-time initializations
702  *
703  * Note: On x86_64, fixmaps are ready for use even before this is called.
704  */
705 
706 void __init setup_arch(char **cmdline_p)
707 {
708 #ifdef CONFIG_X86_32
709 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
710 	visws_early_detect();
711 #else
712 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
713 #endif
714 
715 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
716 	*cmdline_p = command_line;
717 
718 #ifdef CONFIG_X86_64
719 	/*
720 	 * Must call this twice: Once just to detect whether hardware doesn't
721 	 * support NX (so that the early EHCI debug console setup can safely
722 	 * call set_fixmap(), and then again after parsing early parameters to
723 	 * honor the respective command line option.
724 	 */
725 	check_efer();
726 #endif
727 
728 	parse_early_param();
729 
730 	/* VMI may relocate the fixmap; do this before touching ioremap area */
731 	vmi_init();
732 
733 	early_cpu_init();
734 	early_ioremap_init();
735 
736 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
737 	screen_info = boot_params.screen_info;
738 	edid_info = boot_params.edid_info;
739 #ifdef CONFIG_X86_32
740 	apm_info.bios = boot_params.apm_bios_info;
741 	ist_info = boot_params.ist_info;
742 	if (boot_params.sys_desc_table.length != 0) {
743 		set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
744 		machine_id = boot_params.sys_desc_table.table[0];
745 		machine_submodel_id = boot_params.sys_desc_table.table[1];
746 		BIOS_revision = boot_params.sys_desc_table.table[2];
747 	}
748 #endif
749 	saved_video_mode = boot_params.hdr.vid_mode;
750 	bootloader_type = boot_params.hdr.type_of_loader;
751 	if ((bootloader_type >> 4) == 0xe) {
752 		bootloader_type &= 0xf;
753 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
754 	}
755 	bootloader_version  = bootloader_type & 0xf;
756 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
757 
758 #ifdef CONFIG_BLK_DEV_RAM
759 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
760 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
761 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
762 #endif
763 #ifdef CONFIG_EFI
764 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
765 #ifdef CONFIG_X86_32
766 		     "EL32",
767 #else
768 		     "EL64",
769 #endif
770 	 4)) {
771 		efi_enabled = 1;
772 		efi_reserve_early();
773 	}
774 #endif
775 
776 	ARCH_SETUP
777 
778 	setup_memory_map();
779 	parse_setup_data();
780 	/* update the e820_saved too */
781 	e820_reserve_setup_data();
782 
783 	copy_edd();
784 
785 	if (!boot_params.hdr.root_flags)
786 		root_mountflags &= ~MS_RDONLY;
787 	init_mm.start_code = (unsigned long) _text;
788 	init_mm.end_code = (unsigned long) _etext;
789 	init_mm.end_data = (unsigned long) _edata;
790 	init_mm.brk = _brk_end;
791 
792 	code_resource.start = virt_to_phys(_text);
793 	code_resource.end = virt_to_phys(_etext)-1;
794 	data_resource.start = virt_to_phys(_etext);
795 	data_resource.end = virt_to_phys(_edata)-1;
796 	bss_resource.start = virt_to_phys(&__bss_start);
797 	bss_resource.end = virt_to_phys(&__bss_stop)-1;
798 
799 #ifdef CONFIG_CMDLINE_BOOL
800 #ifdef CONFIG_CMDLINE_OVERRIDE
801 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
802 #else
803 	if (builtin_cmdline[0]) {
804 		/* append boot loader cmdline to builtin */
805 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
806 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
807 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
808 	}
809 #endif
810 #endif
811 
812 #ifdef CONFIG_X86_64
813 	check_efer();
814 #endif
815 
816 	/* Must be before kernel pagetables are setup */
817 	vmi_activate();
818 
819 	/* after early param, so could get panic from serial */
820 	reserve_early_setup_data();
821 
822 	if (acpi_mps_check()) {
823 #ifdef CONFIG_X86_LOCAL_APIC
824 		disable_apic = 1;
825 #endif
826 		setup_clear_cpu_cap(X86_FEATURE_APIC);
827 	}
828 
829 #ifdef CONFIG_PCI
830 	if (pci_early_dump_regs)
831 		early_dump_pci_devices();
832 #endif
833 
834 	finish_e820_parsing();
835 
836 	if (efi_enabled)
837 		efi_init();
838 
839 	dmi_scan_machine();
840 
841 	dmi_check_system(bad_bios_dmi_table);
842 
843 	/*
844 	 * VMware detection requires dmi to be available, so this
845 	 * needs to be done after dmi_scan_machine, for the BP.
846 	 */
847 	init_hypervisor(&boot_cpu_data);
848 
849 #ifdef CONFIG_X86_32
850 	probe_roms();
851 #endif
852 
853 	/* after parse_early_param, so could debug it */
854 	insert_resource(&iomem_resource, &code_resource);
855 	insert_resource(&iomem_resource, &data_resource);
856 	insert_resource(&iomem_resource, &bss_resource);
857 
858 
859 #ifdef CONFIG_X86_32
860 	if (ppro_with_ram_bug()) {
861 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
862 				  E820_RESERVED);
863 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
864 		printk(KERN_INFO "fixed physical RAM map:\n");
865 		e820_print_map("bad_ppro");
866 	}
867 #else
868 	early_gart_iommu_check();
869 #endif
870 
871 	/*
872 	 * partially used pages are not usable - thus
873 	 * we are rounding upwards:
874 	 */
875 	max_pfn = e820_end_of_ram_pfn();
876 
877 	/* preallocate 4k for mptable mpc */
878 	early_reserve_e820_mpc_new();
879 	/* update e820 for memory not covered by WB MTRRs */
880 	mtrr_bp_init();
881 	if (mtrr_trim_uncached_memory(max_pfn))
882 		max_pfn = e820_end_of_ram_pfn();
883 
884 #ifdef CONFIG_X86_32
885 	/* max_low_pfn get updated here */
886 	find_low_pfn_range();
887 #else
888 	num_physpages = max_pfn;
889 
890 	check_x2apic();
891 
892 	/* How many end-of-memory variables you have, grandma! */
893 	/* need this before calling reserve_initrd */
894 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
895 		max_low_pfn = e820_end_of_low_ram_pfn();
896 	else
897 		max_low_pfn = max_pfn;
898 
899 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
900 	max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT;
901 #endif
902 
903 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
904 	setup_bios_corruption_check();
905 #endif
906 
907 	printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
908 			max_pfn_mapped<<PAGE_SHIFT);
909 
910 	reserve_brk();
911 
912 	init_gbpages();
913 
914 	/* max_pfn_mapped is updated here */
915 	max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
916 	max_pfn_mapped = max_low_pfn_mapped;
917 
918 #ifdef CONFIG_X86_64
919 	if (max_pfn > max_low_pfn) {
920 		max_pfn_mapped = init_memory_mapping(1UL<<32,
921 						     max_pfn<<PAGE_SHIFT);
922 		/* can we preseve max_low_pfn ?*/
923 		max_low_pfn = max_pfn;
924 	}
925 #endif
926 
927 	/*
928 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
929 	 */
930 
931 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
932 	if (init_ohci1394_dma_early)
933 		init_ohci1394_dma_on_all_controllers();
934 #endif
935 
936 	reserve_initrd();
937 
938 	vsmp_init();
939 
940 	io_delay_init();
941 
942 	/*
943 	 * Parse the ACPI tables for possible boot-time SMP configuration.
944 	 */
945 	acpi_boot_table_init();
946 
947 	early_acpi_boot_init();
948 
949 #ifdef CONFIG_ACPI_NUMA
950 	/*
951 	 * Parse SRAT to discover nodes.
952 	 */
953 	acpi_numa_init();
954 #endif
955 
956 	initmem_init(0, max_pfn);
957 
958 #ifdef CONFIG_ACPI_SLEEP
959 	/*
960 	 * Reserve low memory region for sleep support.
961 	 */
962 	acpi_reserve_bootmem();
963 #endif
964 	/*
965 	 * Find and reserve possible boot-time SMP configuration:
966 	 */
967 	find_smp_config();
968 
969 	reserve_crashkernel();
970 
971 #ifdef CONFIG_X86_64
972 	/*
973 	 * dma32_reserve_bootmem() allocates bootmem which may conflict
974 	 * with the crashkernel command line, so do that after
975 	 * reserve_crashkernel()
976 	 */
977 	dma32_reserve_bootmem();
978 #endif
979 
980 	reserve_ibft_region();
981 
982 #ifdef CONFIG_KVM_CLOCK
983 	kvmclock_init();
984 #endif
985 
986 	paravirt_pagetable_setup_start(swapper_pg_dir);
987 	paging_init();
988 	paravirt_pagetable_setup_done(swapper_pg_dir);
989 	paravirt_post_allocator_init();
990 
991 	tboot_probe();
992 
993 #ifdef CONFIG_X86_64
994 	map_vsyscall();
995 #endif
996 
997 	generic_apic_probe();
998 
999 	early_quirks();
1000 
1001 	/*
1002 	 * Read APIC and some other early information from ACPI tables.
1003 	 */
1004 	acpi_boot_init();
1005 
1006 #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
1007 	/*
1008 	 * get boot-time SMP configuration:
1009 	 */
1010 	if (smp_found_config)
1011 		get_smp_config();
1012 #endif
1013 
1014 	prefill_possible_map();
1015 
1016 #ifdef CONFIG_X86_64
1017 	init_cpu_to_node();
1018 #endif
1019 
1020 	init_apic_mappings();
1021 	ioapic_init_mappings();
1022 
1023 	/* need to wait for io_apic is mapped */
1024 	probe_nr_irqs_gsi();
1025 
1026 	kvm_guest_init();
1027 
1028 	e820_reserve_resources();
1029 	e820_mark_nosave_regions(max_low_pfn);
1030 
1031 #ifdef CONFIG_X86_32
1032 	request_resource(&iomem_resource, &video_ram_resource);
1033 #endif
1034 	reserve_standard_io_resources();
1035 
1036 	e820_setup_gap();
1037 
1038 #ifdef CONFIG_VT
1039 #if defined(CONFIG_VGA_CONSOLE)
1040 	if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1041 		conswitchp = &vga_con;
1042 #elif defined(CONFIG_DUMMY_CONSOLE)
1043 	conswitchp = &dummy_con;
1044 #endif
1045 #endif
1046 }
1047 
1048 #ifdef CONFIG_X86_32
1049 
1050 /**
1051  * x86_quirk_intr_init - post gate setup interrupt initialisation
1052  *
1053  * Description:
1054  *	Fill in any interrupts that may have been left out by the general
1055  *	init_IRQ() routine.  interrupts having to do with the machine rather
1056  *	than the devices on the I/O bus (like APIC interrupts in intel MP
1057  *	systems) are started here.
1058  **/
1059 void __init x86_quirk_intr_init(void)
1060 {
1061 	if (x86_quirks->arch_intr_init) {
1062 		if (x86_quirks->arch_intr_init())
1063 			return;
1064 	}
1065 }
1066 
1067 /**
1068  * x86_quirk_trap_init - initialise system specific traps
1069  *
1070  * Description:
1071  *	Called as the final act of trap_init().  Used in VISWS to initialise
1072  *	the various board specific APIC traps.
1073  **/
1074 void __init x86_quirk_trap_init(void)
1075 {
1076 	if (x86_quirks->arch_trap_init) {
1077 		if (x86_quirks->arch_trap_init())
1078 			return;
1079 	}
1080 }
1081 
1082 static struct irqaction irq0  = {
1083 	.handler = timer_interrupt,
1084 	.flags = IRQF_DISABLED | IRQF_NOBALANCING | IRQF_IRQPOLL | IRQF_TIMER,
1085 	.name = "timer"
1086 };
1087 
1088 /**
1089  * x86_quirk_pre_time_init - do any specific initialisations before.
1090  *
1091  **/
1092 void __init x86_quirk_pre_time_init(void)
1093 {
1094 	if (x86_quirks->arch_pre_time_init)
1095 		x86_quirks->arch_pre_time_init();
1096 }
1097 
1098 /**
1099  * x86_quirk_time_init - do any specific initialisations for the system timer.
1100  *
1101  * Description:
1102  *	Must plug the system timer interrupt source at HZ into the IRQ listed
1103  *	in irq_vectors.h:TIMER_IRQ
1104  **/
1105 void __init x86_quirk_time_init(void)
1106 {
1107 	if (x86_quirks->arch_time_init) {
1108 		/*
1109 		 * A nonzero return code does not mean failure, it means
1110 		 * that the architecture quirk does not want any
1111 		 * generic (timer) setup to be performed after this:
1112 		 */
1113 		if (x86_quirks->arch_time_init())
1114 			return;
1115 	}
1116 
1117 	irq0.mask = cpumask_of_cpu(0);
1118 	setup_irq(0, &irq0);
1119 }
1120 #endif /* CONFIG_X86_32 */
1121