xref: /linux/arch/x86/kernel/setup.c (revision 273b281fa22c293963ee3e6eec418f5dda2dbc83)
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/seq_file.h>
35 #include <linux/console.h>
36 #include <linux/mca.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 
53 #include <linux/errno.h>
54 #include <linux/kernel.h>
55 #include <linux/stddef.h>
56 #include <linux/unistd.h>
57 #include <linux/ptrace.h>
58 #include <linux/slab.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 
72 #include <video/edid.h>
73 
74 #include <asm/mtrr.h>
75 #include <asm/apic.h>
76 #include <asm/e820.h>
77 #include <asm/mpspec.h>
78 #include <asm/setup.h>
79 #include <asm/efi.h>
80 #include <asm/timer.h>
81 #include <asm/i8259.h>
82 #include <asm/sections.h>
83 #include <asm/dmi.h>
84 #include <asm/io_apic.h>
85 #include <asm/ist.h>
86 #include <asm/vmi.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 
93 #include <asm/system.h>
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 
106 #include <asm/percpu.h>
107 #include <asm/topology.h>
108 #include <asm/apicdef.h>
109 #include <asm/k8.h>
110 #ifdef CONFIG_X86_64
111 #include <asm/numa_64.h>
112 #endif
113 #include <asm/mce.h>
114 
115 /*
116  * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
117  * The direct mapping extends to max_pfn_mapped, so that we can directly access
118  * apertures, ACPI and other tables without having to play with fixmaps.
119  */
120 unsigned long max_low_pfn_mapped;
121 unsigned long max_pfn_mapped;
122 
123 RESERVE_BRK(dmi_alloc, 65536);
124 
125 unsigned int boot_cpu_id __read_mostly;
126 
127 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
128 unsigned long _brk_end = (unsigned long)__brk_base;
129 
130 #ifdef CONFIG_X86_64
131 int default_cpu_present_to_apicid(int mps_cpu)
132 {
133 	return __default_cpu_present_to_apicid(mps_cpu);
134 }
135 
136 int default_check_phys_apicid_present(int phys_apicid)
137 {
138 	return __default_check_phys_apicid_present(phys_apicid);
139 }
140 #endif
141 
142 #ifndef CONFIG_DEBUG_BOOT_PARAMS
143 struct boot_params __initdata boot_params;
144 #else
145 struct boot_params boot_params;
146 #endif
147 
148 /*
149  * Machine setup..
150  */
151 static struct resource data_resource = {
152 	.name	= "Kernel data",
153 	.start	= 0,
154 	.end	= 0,
155 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
156 };
157 
158 static struct resource code_resource = {
159 	.name	= "Kernel code",
160 	.start	= 0,
161 	.end	= 0,
162 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
163 };
164 
165 static struct resource bss_resource = {
166 	.name	= "Kernel bss",
167 	.start	= 0,
168 	.end	= 0,
169 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
170 };
171 
172 
173 #ifdef CONFIG_X86_32
174 /* cpu data as detected by the assembly code in head.S */
175 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
176 /* common cpu data for all cpus */
177 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
178 EXPORT_SYMBOL(boot_cpu_data);
179 static void set_mca_bus(int x)
180 {
181 #ifdef CONFIG_MCA
182 	MCA_bus = x;
183 #endif
184 }
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 unsigned long mmu_cr4_features;
214 #else
215 unsigned long mmu_cr4_features = X86_CR4_PAE;
216 #endif
217 
218 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
219 int bootloader_type, bootloader_version;
220 
221 /*
222  * Setup options
223  */
224 struct screen_info screen_info;
225 EXPORT_SYMBOL(screen_info);
226 struct edid_info edid_info;
227 EXPORT_SYMBOL_GPL(edid_info);
228 
229 extern int root_mountflags;
230 
231 unsigned long saved_video_mode;
232 
233 #define RAMDISK_IMAGE_START_MASK	0x07FF
234 #define RAMDISK_PROMPT_FLAG		0x8000
235 #define RAMDISK_LOAD_FLAG		0x4000
236 
237 static char __initdata command_line[COMMAND_LINE_SIZE];
238 #ifdef CONFIG_CMDLINE_BOOL
239 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
240 #endif
241 
242 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
243 struct edd edd;
244 #ifdef CONFIG_EDD_MODULE
245 EXPORT_SYMBOL(edd);
246 #endif
247 /**
248  * copy_edd() - Copy the BIOS EDD information
249  *              from boot_params into a safe place.
250  *
251  */
252 static inline void __init copy_edd(void)
253 {
254      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
255 	    sizeof(edd.mbr_signature));
256      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
257      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
258      edd.edd_info_nr = boot_params.eddbuf_entries;
259 }
260 #else
261 static inline void __init copy_edd(void)
262 {
263 }
264 #endif
265 
266 void * __init extend_brk(size_t size, size_t align)
267 {
268 	size_t mask = align - 1;
269 	void *ret;
270 
271 	BUG_ON(_brk_start == 0);
272 	BUG_ON(align & mask);
273 
274 	_brk_end = (_brk_end + mask) & ~mask;
275 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
276 
277 	ret = (void *)_brk_end;
278 	_brk_end += size;
279 
280 	memset(ret, 0, size);
281 
282 	return ret;
283 }
284 
285 #ifdef CONFIG_X86_64
286 static void __init init_gbpages(void)
287 {
288 	if (direct_gbpages && cpu_has_gbpages)
289 		printk(KERN_INFO "Using GB pages for direct mapping\n");
290 	else
291 		direct_gbpages = 0;
292 }
293 #else
294 static inline void init_gbpages(void)
295 {
296 }
297 #endif
298 
299 static void __init reserve_brk(void)
300 {
301 	if (_brk_end > _brk_start)
302 		reserve_early(__pa(_brk_start), __pa(_brk_end), "BRK");
303 
304 	/* Mark brk area as locked down and no longer taking any
305 	   new allocations */
306 	_brk_start = 0;
307 }
308 
309 #ifdef CONFIG_BLK_DEV_INITRD
310 
311 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
312 static void __init relocate_initrd(void)
313 {
314 
315 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
316 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
317 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
318 	u64 ramdisk_here;
319 	unsigned long slop, clen, mapaddr;
320 	char *p, *q;
321 
322 	/* We need to move the initrd down into lowmem */
323 	ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
324 					 PAGE_SIZE);
325 
326 	if (ramdisk_here == -1ULL)
327 		panic("Cannot find place for new RAMDISK of size %lld\n",
328 			 ramdisk_size);
329 
330 	/* Note: this includes all the lowmem currently occupied by
331 	   the initrd, we rely on that fact to keep the data intact. */
332 	reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
333 			 "NEW RAMDISK");
334 	initrd_start = ramdisk_here + PAGE_OFFSET;
335 	initrd_end   = initrd_start + ramdisk_size;
336 	printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
337 			 ramdisk_here, ramdisk_here + ramdisk_size);
338 
339 	q = (char *)initrd_start;
340 
341 	/* Copy any lowmem portion of the initrd */
342 	if (ramdisk_image < end_of_lowmem) {
343 		clen = end_of_lowmem - ramdisk_image;
344 		p = (char *)__va(ramdisk_image);
345 		memcpy(q, p, clen);
346 		q += clen;
347 		ramdisk_image += clen;
348 		ramdisk_size  -= clen;
349 	}
350 
351 	/* Copy the highmem portion of the initrd */
352 	while (ramdisk_size) {
353 		slop = ramdisk_image & ~PAGE_MASK;
354 		clen = ramdisk_size;
355 		if (clen > MAX_MAP_CHUNK-slop)
356 			clen = MAX_MAP_CHUNK-slop;
357 		mapaddr = ramdisk_image & PAGE_MASK;
358 		p = early_memremap(mapaddr, clen+slop);
359 		memcpy(q, p+slop, clen);
360 		early_iounmap(p, clen+slop);
361 		q += clen;
362 		ramdisk_image += clen;
363 		ramdisk_size  -= clen;
364 	}
365 	/* high pages is not converted by early_res_to_bootmem */
366 	ramdisk_image = boot_params.hdr.ramdisk_image;
367 	ramdisk_size  = boot_params.hdr.ramdisk_size;
368 	printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
369 		" %08llx - %08llx\n",
370 		ramdisk_image, ramdisk_image + ramdisk_size - 1,
371 		ramdisk_here, ramdisk_here + ramdisk_size - 1);
372 }
373 
374 static void __init reserve_initrd(void)
375 {
376 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
377 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
378 	u64 ramdisk_end   = ramdisk_image + ramdisk_size;
379 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
380 
381 	if (!boot_params.hdr.type_of_loader ||
382 	    !ramdisk_image || !ramdisk_size)
383 		return;		/* No initrd provided by bootloader */
384 
385 	initrd_start = 0;
386 
387 	if (ramdisk_size >= (end_of_lowmem>>1)) {
388 		free_early(ramdisk_image, ramdisk_end);
389 		printk(KERN_ERR "initrd too large to handle, "
390 		       "disabling initrd\n");
391 		return;
392 	}
393 
394 	printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
395 			ramdisk_end);
396 
397 
398 	if (ramdisk_end <= end_of_lowmem) {
399 		/* All in lowmem, easy case */
400 		/*
401 		 * don't need to reserve again, already reserved early
402 		 * in i386_start_kernel
403 		 */
404 		initrd_start = ramdisk_image + PAGE_OFFSET;
405 		initrd_end = initrd_start + ramdisk_size;
406 		return;
407 	}
408 
409 	relocate_initrd();
410 
411 	free_early(ramdisk_image, ramdisk_end);
412 }
413 #else
414 static void __init reserve_initrd(void)
415 {
416 }
417 #endif /* CONFIG_BLK_DEV_INITRD */
418 
419 static void __init parse_setup_data(void)
420 {
421 	struct setup_data *data;
422 	u64 pa_data;
423 
424 	if (boot_params.hdr.version < 0x0209)
425 		return;
426 	pa_data = boot_params.hdr.setup_data;
427 	while (pa_data) {
428 		data = early_memremap(pa_data, PAGE_SIZE);
429 		switch (data->type) {
430 		case SETUP_E820_EXT:
431 			parse_e820_ext(data, pa_data);
432 			break;
433 		default:
434 			break;
435 		}
436 		pa_data = data->next;
437 		early_iounmap(data, PAGE_SIZE);
438 	}
439 }
440 
441 static void __init e820_reserve_setup_data(void)
442 {
443 	struct setup_data *data;
444 	u64 pa_data;
445 	int found = 0;
446 
447 	if (boot_params.hdr.version < 0x0209)
448 		return;
449 	pa_data = boot_params.hdr.setup_data;
450 	while (pa_data) {
451 		data = early_memremap(pa_data, sizeof(*data));
452 		e820_update_range(pa_data, sizeof(*data)+data->len,
453 			 E820_RAM, E820_RESERVED_KERN);
454 		found = 1;
455 		pa_data = data->next;
456 		early_iounmap(data, sizeof(*data));
457 	}
458 	if (!found)
459 		return;
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 reserve_early_setup_data(void)
468 {
469 	struct setup_data *data;
470 	u64 pa_data;
471 	char buf[32];
472 
473 	if (boot_params.hdr.version < 0x0209)
474 		return;
475 	pa_data = boot_params.hdr.setup_data;
476 	while (pa_data) {
477 		data = early_memremap(pa_data, sizeof(*data));
478 		sprintf(buf, "setup data %x", data->type);
479 		reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
480 		pa_data = data->next;
481 		early_iounmap(data, sizeof(*data));
482 	}
483 }
484 
485 /*
486  * --------- Crashkernel reservation ------------------------------
487  */
488 
489 #ifdef CONFIG_KEXEC
490 
491 static inline unsigned long long get_total_mem(void)
492 {
493 	unsigned long long total;
494 
495 	total = max_pfn - min_low_pfn;
496 
497 	return total << PAGE_SHIFT;
498 }
499 
500 static void __init reserve_crashkernel(void)
501 {
502 	unsigned long long total_mem;
503 	unsigned long long crash_size, crash_base;
504 	int ret;
505 
506 	total_mem = get_total_mem();
507 
508 	ret = parse_crashkernel(boot_command_line, total_mem,
509 			&crash_size, &crash_base);
510 	if (ret != 0 || crash_size <= 0)
511 		return;
512 
513 	/* 0 means: find the address automatically */
514 	if (crash_base <= 0) {
515 		const unsigned long long alignment = 16<<20;	/* 16M */
516 
517 		crash_base = find_e820_area(alignment, ULONG_MAX, crash_size,
518 				 alignment);
519 		if (crash_base == -1ULL) {
520 			pr_info("crashkernel reservation failed - No suitable area found.\n");
521 			return;
522 		}
523 	} else {
524 		unsigned long long start;
525 
526 		start = find_e820_area(crash_base, ULONG_MAX, crash_size,
527 				 1<<20);
528 		if (start != crash_base) {
529 			pr_info("crashkernel reservation failed - memory is in use.\n");
530 			return;
531 		}
532 	}
533 	reserve_early(crash_base, crash_base + crash_size, "CRASH KERNEL");
534 
535 	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
536 			"for crashkernel (System RAM: %ldMB)\n",
537 			(unsigned long)(crash_size >> 20),
538 			(unsigned long)(crash_base >> 20),
539 			(unsigned long)(total_mem >> 20));
540 
541 	crashk_res.start = crash_base;
542 	crashk_res.end   = crash_base + crash_size - 1;
543 	insert_resource(&iomem_resource, &crashk_res);
544 }
545 #else
546 static void __init reserve_crashkernel(void)
547 {
548 }
549 #endif
550 
551 static struct resource standard_io_resources[] = {
552 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
553 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
554 	{ .name = "pic1", .start = 0x20, .end = 0x21,
555 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
556 	{ .name = "timer0", .start = 0x40, .end = 0x43,
557 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
558 	{ .name = "timer1", .start = 0x50, .end = 0x53,
559 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
560 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
561 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
562 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
563 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
564 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
565 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
566 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
567 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
568 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
569 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
570 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
571 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
572 };
573 
574 void __init reserve_standard_io_resources(void)
575 {
576 	int i;
577 
578 	/* request I/O space for devices used on all i[345]86 PCs */
579 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
580 		request_resource(&ioport_resource, &standard_io_resources[i]);
581 
582 }
583 
584 /*
585  * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
586  * is_kdump_kernel() to determine if we are booting after a panic. Hence
587  * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
588  */
589 
590 #ifdef CONFIG_CRASH_DUMP
591 /* elfcorehdr= specifies the location of elf core header
592  * stored by the crashed kernel. This option will be passed
593  * by kexec loader to the capture kernel.
594  */
595 static int __init setup_elfcorehdr(char *arg)
596 {
597 	char *end;
598 	if (!arg)
599 		return -EINVAL;
600 	elfcorehdr_addr = memparse(arg, &end);
601 	return end > arg ? 0 : -EINVAL;
602 }
603 early_param("elfcorehdr", setup_elfcorehdr);
604 #endif
605 
606 #ifdef CONFIG_X86_RESERVE_LOW_64K
607 static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
608 {
609 	printk(KERN_NOTICE
610 		"%s detected: BIOS may corrupt low RAM, working around it.\n",
611 		d->ident);
612 
613 	e820_update_range(0, 0x10000, E820_RAM, E820_RESERVED);
614 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
615 
616 	return 0;
617 }
618 #endif
619 
620 /* List of systems that have known low memory corruption BIOS problems */
621 static struct dmi_system_id __initdata bad_bios_dmi_table[] = {
622 #ifdef CONFIG_X86_RESERVE_LOW_64K
623 	{
624 		.callback = dmi_low_memory_corruption,
625 		.ident = "AMI BIOS",
626 		.matches = {
627 			DMI_MATCH(DMI_BIOS_VENDOR, "American Megatrends Inc."),
628 		},
629 	},
630 	{
631 		.callback = dmi_low_memory_corruption,
632 		.ident = "Phoenix BIOS",
633 		.matches = {
634 			DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix Technologies"),
635 		},
636 	},
637 	{
638 		.callback = dmi_low_memory_corruption,
639 		.ident = "Phoenix/MSC BIOS",
640 		.matches = {
641 			DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix/MSC"),
642 		},
643 	},
644 	{
645 	/*
646 	 * AMI BIOS with low memory corruption was found on Intel DG45ID board.
647 	 * It hase different DMI_BIOS_VENDOR = "Intel Corp.", for now we will
648 	 * match only DMI_BOARD_NAME and see if there is more bad products
649 	 * with this vendor.
650 	 */
651 		.callback = dmi_low_memory_corruption,
652 		.ident = "AMI BIOS",
653 		.matches = {
654 			DMI_MATCH(DMI_BOARD_NAME, "DG45ID"),
655 		},
656 	},
657 #endif
658 	{}
659 };
660 
661 /*
662  * Determine if we were loaded by an EFI loader.  If so, then we have also been
663  * passed the efi memmap, systab, etc., so we should use these data structures
664  * for initialization.  Note, the efi init code path is determined by the
665  * global efi_enabled. This allows the same kernel image to be used on existing
666  * systems (with a traditional BIOS) as well as on EFI systems.
667  */
668 /*
669  * setup_arch - architecture-specific boot-time initializations
670  *
671  * Note: On x86_64, fixmaps are ready for use even before this is called.
672  */
673 
674 void __init setup_arch(char **cmdline_p)
675 {
676 	int acpi = 0;
677 	int k8 = 0;
678 
679 #ifdef CONFIG_X86_32
680 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
681 	visws_early_detect();
682 #else
683 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
684 #endif
685 
686 	/* VMI may relocate the fixmap; do this before touching ioremap area */
687 	vmi_init();
688 
689 	early_cpu_init();
690 	early_ioremap_init();
691 
692 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
693 	screen_info = boot_params.screen_info;
694 	edid_info = boot_params.edid_info;
695 #ifdef CONFIG_X86_32
696 	apm_info.bios = boot_params.apm_bios_info;
697 	ist_info = boot_params.ist_info;
698 	if (boot_params.sys_desc_table.length != 0) {
699 		set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
700 		machine_id = boot_params.sys_desc_table.table[0];
701 		machine_submodel_id = boot_params.sys_desc_table.table[1];
702 		BIOS_revision = boot_params.sys_desc_table.table[2];
703 	}
704 #endif
705 	saved_video_mode = boot_params.hdr.vid_mode;
706 	bootloader_type = boot_params.hdr.type_of_loader;
707 	if ((bootloader_type >> 4) == 0xe) {
708 		bootloader_type &= 0xf;
709 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
710 	}
711 	bootloader_version  = bootloader_type & 0xf;
712 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
713 
714 #ifdef CONFIG_BLK_DEV_RAM
715 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
716 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
717 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
718 #endif
719 #ifdef CONFIG_EFI
720 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
721 #ifdef CONFIG_X86_32
722 		     "EL32",
723 #else
724 		     "EL64",
725 #endif
726 	 4)) {
727 		efi_enabled = 1;
728 		efi_reserve_early();
729 	}
730 #endif
731 
732 	x86_init.oem.arch_setup();
733 
734 	setup_memory_map();
735 	parse_setup_data();
736 	/* update the e820_saved too */
737 	e820_reserve_setup_data();
738 
739 	copy_edd();
740 
741 	if (!boot_params.hdr.root_flags)
742 		root_mountflags &= ~MS_RDONLY;
743 	init_mm.start_code = (unsigned long) _text;
744 	init_mm.end_code = (unsigned long) _etext;
745 	init_mm.end_data = (unsigned long) _edata;
746 	init_mm.brk = _brk_end;
747 
748 	code_resource.start = virt_to_phys(_text);
749 	code_resource.end = virt_to_phys(_etext)-1;
750 	data_resource.start = virt_to_phys(_etext);
751 	data_resource.end = virt_to_phys(_edata)-1;
752 	bss_resource.start = virt_to_phys(&__bss_start);
753 	bss_resource.end = virt_to_phys(&__bss_stop)-1;
754 
755 #ifdef CONFIG_CMDLINE_BOOL
756 #ifdef CONFIG_CMDLINE_OVERRIDE
757 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
758 #else
759 	if (builtin_cmdline[0]) {
760 		/* append boot loader cmdline to builtin */
761 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
762 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
763 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
764 	}
765 #endif
766 #endif
767 
768 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
769 	*cmdline_p = command_line;
770 
771 	/*
772 	 * x86_configure_nx() is called before parse_early_param() to detect
773 	 * whether hardware doesn't support NX (so that the early EHCI debug
774 	 * console setup can safely call set_fixmap()). It may then be called
775 	 * again from within noexec_setup() during parsing early parameters
776 	 * to honor the respective command line option.
777 	 */
778 	x86_configure_nx();
779 
780 	parse_early_param();
781 
782 	x86_report_nx();
783 
784 	/* Must be before kernel pagetables are setup */
785 	vmi_activate();
786 
787 	/* after early param, so could get panic from serial */
788 	reserve_early_setup_data();
789 
790 	if (acpi_mps_check()) {
791 #ifdef CONFIG_X86_LOCAL_APIC
792 		disable_apic = 1;
793 #endif
794 		setup_clear_cpu_cap(X86_FEATURE_APIC);
795 	}
796 
797 #ifdef CONFIG_PCI
798 	if (pci_early_dump_regs)
799 		early_dump_pci_devices();
800 #endif
801 
802 	finish_e820_parsing();
803 
804 	if (efi_enabled)
805 		efi_init();
806 
807 	dmi_scan_machine();
808 
809 	dmi_check_system(bad_bios_dmi_table);
810 
811 	/*
812 	 * VMware detection requires dmi to be available, so this
813 	 * needs to be done after dmi_scan_machine, for the BP.
814 	 */
815 	init_hypervisor_platform();
816 
817 	x86_init.resources.probe_roms();
818 
819 	/* after parse_early_param, so could debug it */
820 	insert_resource(&iomem_resource, &code_resource);
821 	insert_resource(&iomem_resource, &data_resource);
822 	insert_resource(&iomem_resource, &bss_resource);
823 
824 
825 #ifdef CONFIG_X86_32
826 	if (ppro_with_ram_bug()) {
827 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
828 				  E820_RESERVED);
829 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
830 		printk(KERN_INFO "fixed physical RAM map:\n");
831 		e820_print_map("bad_ppro");
832 	}
833 #else
834 	early_gart_iommu_check();
835 #endif
836 
837 	/*
838 	 * partially used pages are not usable - thus
839 	 * we are rounding upwards:
840 	 */
841 	max_pfn = e820_end_of_ram_pfn();
842 
843 	/* preallocate 4k for mptable mpc */
844 	early_reserve_e820_mpc_new();
845 	/* update e820 for memory not covered by WB MTRRs */
846 	mtrr_bp_init();
847 	if (mtrr_trim_uncached_memory(max_pfn))
848 		max_pfn = e820_end_of_ram_pfn();
849 
850 #ifdef CONFIG_X86_32
851 	/* max_low_pfn get updated here */
852 	find_low_pfn_range();
853 #else
854 	num_physpages = max_pfn;
855 
856 	check_x2apic();
857 
858 	/* How many end-of-memory variables you have, grandma! */
859 	/* need this before calling reserve_initrd */
860 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
861 		max_low_pfn = e820_end_of_low_ram_pfn();
862 	else
863 		max_low_pfn = max_pfn;
864 
865 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
866 	max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT;
867 #endif
868 
869 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
870 	setup_bios_corruption_check();
871 #endif
872 
873 	printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
874 			max_pfn_mapped<<PAGE_SHIFT);
875 
876 	reserve_brk();
877 
878 #ifdef CONFIG_ACPI_SLEEP
879 	/*
880 	 * Reserve low memory region for sleep support.
881 	 * even before init_memory_mapping
882 	 */
883 	acpi_reserve_wakeup_memory();
884 #endif
885 	init_gbpages();
886 
887 	/* max_pfn_mapped is updated here */
888 	max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
889 	max_pfn_mapped = max_low_pfn_mapped;
890 
891 #ifdef CONFIG_X86_64
892 	if (max_pfn > max_low_pfn) {
893 		max_pfn_mapped = init_memory_mapping(1UL<<32,
894 						     max_pfn<<PAGE_SHIFT);
895 		/* can we preseve max_low_pfn ?*/
896 		max_low_pfn = max_pfn;
897 	}
898 #endif
899 
900 	/*
901 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
902 	 */
903 
904 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
905 	if (init_ohci1394_dma_early)
906 		init_ohci1394_dma_on_all_controllers();
907 #endif
908 
909 	reserve_initrd();
910 
911 	reserve_crashkernel();
912 
913 	vsmp_init();
914 
915 	io_delay_init();
916 
917 	/*
918 	 * Parse the ACPI tables for possible boot-time SMP configuration.
919 	 */
920 	acpi_boot_table_init();
921 
922 	early_acpi_boot_init();
923 
924 	/*
925 	 * Find and reserve possible boot-time SMP configuration:
926 	 */
927 	find_smp_config();
928 
929 #ifdef CONFIG_ACPI_NUMA
930 	/*
931 	 * Parse SRAT to discover nodes.
932 	 */
933 	acpi = acpi_numa_init();
934 #endif
935 
936 #ifdef CONFIG_K8_NUMA
937 	if (!acpi)
938 		k8 = !k8_numa_init(0, max_pfn);
939 #endif
940 
941 	initmem_init(0, max_pfn, acpi, k8);
942 
943 #ifdef CONFIG_X86_64
944 	/*
945 	 * dma32_reserve_bootmem() allocates bootmem which may conflict
946 	 * with the crashkernel command line, so do that after
947 	 * reserve_crashkernel()
948 	 */
949 	dma32_reserve_bootmem();
950 #endif
951 
952 	reserve_ibft_region();
953 
954 #ifdef CONFIG_KVM_CLOCK
955 	kvmclock_init();
956 #endif
957 
958 	x86_init.paging.pagetable_setup_start(swapper_pg_dir);
959 	paging_init();
960 	x86_init.paging.pagetable_setup_done(swapper_pg_dir);
961 
962 	tboot_probe();
963 
964 #ifdef CONFIG_X86_64
965 	map_vsyscall();
966 #endif
967 
968 	generic_apic_probe();
969 
970 	early_quirks();
971 
972 	/*
973 	 * Read APIC and some other early information from ACPI tables.
974 	 */
975 	acpi_boot_init();
976 
977 	sfi_init();
978 
979 	/*
980 	 * get boot-time SMP configuration:
981 	 */
982 	if (smp_found_config)
983 		get_smp_config();
984 
985 	prefill_possible_map();
986 
987 #ifdef CONFIG_X86_64
988 	init_cpu_to_node();
989 #endif
990 
991 	init_apic_mappings();
992 	ioapic_init_mappings();
993 
994 	/* need to wait for io_apic is mapped */
995 	probe_nr_irqs_gsi();
996 
997 	kvm_guest_init();
998 
999 	e820_reserve_resources();
1000 	e820_mark_nosave_regions(max_low_pfn);
1001 
1002 	x86_init.resources.reserve_resources();
1003 
1004 	e820_setup_gap();
1005 
1006 #ifdef CONFIG_VT
1007 #if defined(CONFIG_VGA_CONSOLE)
1008 	if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1009 		conswitchp = &vga_con;
1010 #elif defined(CONFIG_DUMMY_CONSOLE)
1011 	conswitchp = &dummy_con;
1012 #endif
1013 #endif
1014 	x86_init.oem.banner();
1015 
1016 	mcheck_init();
1017 }
1018 
1019 #ifdef CONFIG_X86_32
1020 
1021 static struct resource video_ram_resource = {
1022 	.name	= "Video RAM area",
1023 	.start	= 0xa0000,
1024 	.end	= 0xbffff,
1025 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1026 };
1027 
1028 void __init i386_reserve_resources(void)
1029 {
1030 	request_resource(&iomem_resource, &video_ram_resource);
1031 	reserve_standard_io_resources();
1032 }
1033 
1034 #endif /* CONFIG_X86_32 */
1035