xref: /openbmc/linux/arch/x86/kernel/setup.c (revision 78c99ba1)
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 
70 #include <video/edid.h>
71 
72 #include <asm/mtrr.h>
73 #include <asm/apic.h>
74 #include <asm/e820.h>
75 #include <asm/mpspec.h>
76 #include <asm/setup.h>
77 #include <asm/efi.h>
78 #include <asm/timer.h>
79 #include <asm/i8259.h>
80 #include <asm/sections.h>
81 #include <asm/dmi.h>
82 #include <asm/io_apic.h>
83 #include <asm/ist.h>
84 #include <asm/vmi.h>
85 #include <asm/setup_arch.h>
86 #include <asm/bios_ebda.h>
87 #include <asm/cacheflush.h>
88 #include <asm/processor.h>
89 #include <asm/bugs.h>
90 
91 #include <asm/system.h>
92 #include <asm/vsyscall.h>
93 #include <asm/cpu.h>
94 #include <asm/desc.h>
95 #include <asm/dma.h>
96 #include <asm/iommu.h>
97 #include <asm/gart.h>
98 #include <asm/mmu_context.h>
99 #include <asm/proto.h>
100 
101 #include <asm/paravirt.h>
102 #include <asm/hypervisor.h>
103 
104 #include <asm/percpu.h>
105 #include <asm/topology.h>
106 #include <asm/apicdef.h>
107 #ifdef CONFIG_X86_64
108 #include <asm/numa_64.h>
109 #endif
110 
111 #ifndef ARCH_SETUP
112 #define ARCH_SETUP
113 #endif
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 boot_cpu_physical_apicid)
137 {
138 	return __default_check_phys_apicid_present(boot_cpu_physical_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 static struct resource video_ram_resource = {
175 	.name	= "Video RAM area",
176 	.start	= 0xa0000,
177 	.end	= 0xbffff,
178 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
179 };
180 
181 /* cpu data as detected by the assembly code in head.S */
182 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
183 /* common cpu data for all cpus */
184 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
185 EXPORT_SYMBOL(boot_cpu_data);
186 static void set_mca_bus(int x)
187 {
188 #ifdef CONFIG_MCA
189 	MCA_bus = x;
190 #endif
191 }
192 
193 unsigned int def_to_bigsmp;
194 
195 /* for MCA, but anyone else can use it if they want */
196 unsigned int machine_id;
197 unsigned int machine_submodel_id;
198 unsigned int BIOS_revision;
199 
200 struct apm_info apm_info;
201 EXPORT_SYMBOL(apm_info);
202 
203 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
204 	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
205 struct ist_info ist_info;
206 EXPORT_SYMBOL(ist_info);
207 #else
208 struct ist_info ist_info;
209 #endif
210 
211 #else
212 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
213 	.x86_phys_bits = MAX_PHYSMEM_BITS,
214 };
215 EXPORT_SYMBOL(boot_cpu_data);
216 #endif
217 
218 
219 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
220 unsigned long mmu_cr4_features;
221 #else
222 unsigned long mmu_cr4_features = X86_CR4_PAE;
223 #endif
224 
225 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
226 int bootloader_type, bootloader_version;
227 
228 /*
229  * Setup options
230  */
231 struct screen_info screen_info;
232 EXPORT_SYMBOL(screen_info);
233 struct edid_info edid_info;
234 EXPORT_SYMBOL_GPL(edid_info);
235 
236 extern int root_mountflags;
237 
238 unsigned long saved_video_mode;
239 
240 #define RAMDISK_IMAGE_START_MASK	0x07FF
241 #define RAMDISK_PROMPT_FLAG		0x8000
242 #define RAMDISK_LOAD_FLAG		0x4000
243 
244 static char __initdata command_line[COMMAND_LINE_SIZE];
245 #ifdef CONFIG_CMDLINE_BOOL
246 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
247 #endif
248 
249 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
250 struct edd edd;
251 #ifdef CONFIG_EDD_MODULE
252 EXPORT_SYMBOL(edd);
253 #endif
254 /**
255  * copy_edd() - Copy the BIOS EDD information
256  *              from boot_params into a safe place.
257  *
258  */
259 static inline void copy_edd(void)
260 {
261      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
262 	    sizeof(edd.mbr_signature));
263      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
264      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
265      edd.edd_info_nr = boot_params.eddbuf_entries;
266 }
267 #else
268 static inline void copy_edd(void)
269 {
270 }
271 #endif
272 
273 void * __init extend_brk(size_t size, size_t align)
274 {
275 	size_t mask = align - 1;
276 	void *ret;
277 
278 	BUG_ON(_brk_start == 0);
279 	BUG_ON(align & mask);
280 
281 	_brk_end = (_brk_end + mask) & ~mask;
282 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
283 
284 	ret = (void *)_brk_end;
285 	_brk_end += size;
286 
287 	memset(ret, 0, size);
288 
289 	return ret;
290 }
291 
292 static void __init reserve_brk(void)
293 {
294 	if (_brk_end > _brk_start)
295 		reserve_early(__pa(_brk_start), __pa(_brk_end), "BRK");
296 
297 	/* Mark brk area as locked down and no longer taking any
298 	   new allocations */
299 	_brk_start = 0;
300 }
301 
302 #ifdef CONFIG_BLK_DEV_INITRD
303 
304 #ifdef CONFIG_X86_32
305 
306 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
307 static void __init relocate_initrd(void)
308 {
309 
310 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
311 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
312 	u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
313 	u64 ramdisk_here;
314 	unsigned long slop, clen, mapaddr;
315 	char *p, *q;
316 
317 	/* We need to move the initrd down into lowmem */
318 	ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
319 					 PAGE_SIZE);
320 
321 	if (ramdisk_here == -1ULL)
322 		panic("Cannot find place for new RAMDISK of size %lld\n",
323 			 ramdisk_size);
324 
325 	/* Note: this includes all the lowmem currently occupied by
326 	   the initrd, we rely on that fact to keep the data intact. */
327 	reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
328 			 "NEW RAMDISK");
329 	initrd_start = ramdisk_here + PAGE_OFFSET;
330 	initrd_end   = initrd_start + ramdisk_size;
331 	printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
332 			 ramdisk_here, ramdisk_here + ramdisk_size);
333 
334 	q = (char *)initrd_start;
335 
336 	/* Copy any lowmem portion of the initrd */
337 	if (ramdisk_image < end_of_lowmem) {
338 		clen = end_of_lowmem - ramdisk_image;
339 		p = (char *)__va(ramdisk_image);
340 		memcpy(q, p, clen);
341 		q += clen;
342 		ramdisk_image += clen;
343 		ramdisk_size  -= clen;
344 	}
345 
346 	/* Copy the highmem portion of the initrd */
347 	while (ramdisk_size) {
348 		slop = ramdisk_image & ~PAGE_MASK;
349 		clen = ramdisk_size;
350 		if (clen > MAX_MAP_CHUNK-slop)
351 			clen = MAX_MAP_CHUNK-slop;
352 		mapaddr = ramdisk_image & PAGE_MASK;
353 		p = early_memremap(mapaddr, clen+slop);
354 		memcpy(q, p+slop, clen);
355 		early_iounmap(p, clen+slop);
356 		q += clen;
357 		ramdisk_image += clen;
358 		ramdisk_size  -= clen;
359 	}
360 	/* high pages is not converted by early_res_to_bootmem */
361 	ramdisk_image = boot_params.hdr.ramdisk_image;
362 	ramdisk_size  = boot_params.hdr.ramdisk_size;
363 	printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
364 		" %08llx - %08llx\n",
365 		ramdisk_image, ramdisk_image + ramdisk_size - 1,
366 		ramdisk_here, ramdisk_here + ramdisk_size - 1);
367 }
368 #endif
369 
370 static void __init reserve_initrd(void)
371 {
372 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
373 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
374 	u64 ramdisk_end   = ramdisk_image + ramdisk_size;
375 	u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
376 
377 	if (!boot_params.hdr.type_of_loader ||
378 	    !ramdisk_image || !ramdisk_size)
379 		return;		/* No initrd provided by bootloader */
380 
381 	initrd_start = 0;
382 
383 	if (ramdisk_size >= (end_of_lowmem>>1)) {
384 		free_early(ramdisk_image, ramdisk_end);
385 		printk(KERN_ERR "initrd too large to handle, "
386 		       "disabling initrd\n");
387 		return;
388 	}
389 
390 	printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
391 			ramdisk_end);
392 
393 
394 	if (ramdisk_end <= end_of_lowmem) {
395 		/* All in lowmem, easy case */
396 		/*
397 		 * don't need to reserve again, already reserved early
398 		 * in i386_start_kernel
399 		 */
400 		initrd_start = ramdisk_image + PAGE_OFFSET;
401 		initrd_end = initrd_start + ramdisk_size;
402 		return;
403 	}
404 
405 #ifdef CONFIG_X86_32
406 	relocate_initrd();
407 #else
408 	printk(KERN_ERR "initrd extends beyond end of memory "
409 	       "(0x%08llx > 0x%08llx)\ndisabling initrd\n",
410 	       ramdisk_end, end_of_lowmem);
411 	initrd_start = 0;
412 #endif
413 	free_early(ramdisk_image, ramdisk_end);
414 }
415 #else
416 static void __init reserve_initrd(void)
417 {
418 }
419 #endif /* CONFIG_BLK_DEV_INITRD */
420 
421 static void __init parse_setup_data(void)
422 {
423 	struct setup_data *data;
424 	u64 pa_data;
425 
426 	if (boot_params.hdr.version < 0x0209)
427 		return;
428 	pa_data = boot_params.hdr.setup_data;
429 	while (pa_data) {
430 		data = early_memremap(pa_data, PAGE_SIZE);
431 		switch (data->type) {
432 		case SETUP_E820_EXT:
433 			parse_e820_ext(data, pa_data);
434 			break;
435 		default:
436 			break;
437 		}
438 		pa_data = data->next;
439 		early_iounmap(data, PAGE_SIZE);
440 	}
441 }
442 
443 static void __init e820_reserve_setup_data(void)
444 {
445 	struct setup_data *data;
446 	u64 pa_data;
447 	int found = 0;
448 
449 	if (boot_params.hdr.version < 0x0209)
450 		return;
451 	pa_data = boot_params.hdr.setup_data;
452 	while (pa_data) {
453 		data = early_memremap(pa_data, sizeof(*data));
454 		e820_update_range(pa_data, sizeof(*data)+data->len,
455 			 E820_RAM, E820_RESERVED_KERN);
456 		found = 1;
457 		pa_data = data->next;
458 		early_iounmap(data, sizeof(*data));
459 	}
460 	if (!found)
461 		return;
462 
463 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
464 	memcpy(&e820_saved, &e820, sizeof(struct e820map));
465 	printk(KERN_INFO "extended physical RAM map:\n");
466 	e820_print_map("reserve setup_data");
467 }
468 
469 static void __init reserve_early_setup_data(void)
470 {
471 	struct setup_data *data;
472 	u64 pa_data;
473 	char buf[32];
474 
475 	if (boot_params.hdr.version < 0x0209)
476 		return;
477 	pa_data = boot_params.hdr.setup_data;
478 	while (pa_data) {
479 		data = early_memremap(pa_data, sizeof(*data));
480 		sprintf(buf, "setup data %x", data->type);
481 		reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
482 		pa_data = data->next;
483 		early_iounmap(data, sizeof(*data));
484 	}
485 }
486 
487 /*
488  * --------- Crashkernel reservation ------------------------------
489  */
490 
491 #ifdef CONFIG_KEXEC
492 
493 /**
494  * Reserve @size bytes of crashkernel memory at any suitable offset.
495  *
496  * @size: Size of the crashkernel memory to reserve.
497  * Returns the base address on success, and -1ULL on failure.
498  */
499 static
500 unsigned long long __init find_and_reserve_crashkernel(unsigned long long size)
501 {
502 	const unsigned long long alignment = 16<<20; 	/* 16M */
503 	unsigned long long start = 0LL;
504 
505 	while (1) {
506 		int ret;
507 
508 		start = find_e820_area(start, ULONG_MAX, size, alignment);
509 		if (start == -1ULL)
510 			return start;
511 
512 		/* try to reserve it */
513 		ret = reserve_bootmem_generic(start, size, BOOTMEM_EXCLUSIVE);
514 		if (ret >= 0)
515 			return start;
516 
517 		start += alignment;
518 	}
519 }
520 
521 static inline unsigned long long get_total_mem(void)
522 {
523 	unsigned long long total;
524 
525 	total = max_low_pfn - min_low_pfn;
526 #ifdef CONFIG_HIGHMEM
527 	total += highend_pfn - highstart_pfn;
528 #endif
529 
530 	return total << PAGE_SHIFT;
531 }
532 
533 static void __init reserve_crashkernel(void)
534 {
535 	unsigned long long total_mem;
536 	unsigned long long crash_size, crash_base;
537 	int ret;
538 
539 	total_mem = get_total_mem();
540 
541 	ret = parse_crashkernel(boot_command_line, total_mem,
542 			&crash_size, &crash_base);
543 	if (ret != 0 || crash_size <= 0)
544 		return;
545 
546 	/* 0 means: find the address automatically */
547 	if (crash_base <= 0) {
548 		crash_base = find_and_reserve_crashkernel(crash_size);
549 		if (crash_base == -1ULL) {
550 			pr_info("crashkernel reservation failed. "
551 				"No suitable area found.\n");
552 			return;
553 		}
554 	} else {
555 		ret = reserve_bootmem_generic(crash_base, crash_size,
556 					BOOTMEM_EXCLUSIVE);
557 		if (ret < 0) {
558 			pr_info("crashkernel reservation failed - "
559 				"memory is in use\n");
560 			return;
561 		}
562 	}
563 
564 	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
565 			"for crashkernel (System RAM: %ldMB)\n",
566 			(unsigned long)(crash_size >> 20),
567 			(unsigned long)(crash_base >> 20),
568 			(unsigned long)(total_mem >> 20));
569 
570 	crashk_res.start = crash_base;
571 	crashk_res.end   = crash_base + crash_size - 1;
572 	insert_resource(&iomem_resource, &crashk_res);
573 }
574 #else
575 static void __init reserve_crashkernel(void)
576 {
577 }
578 #endif
579 
580 static struct resource standard_io_resources[] = {
581 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
582 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
583 	{ .name = "pic1", .start = 0x20, .end = 0x21,
584 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
585 	{ .name = "timer0", .start = 0x40, .end = 0x43,
586 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
587 	{ .name = "timer1", .start = 0x50, .end = 0x53,
588 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
589 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
590 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
591 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
592 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
593 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
594 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
595 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
596 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
597 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
598 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
599 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
600 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
601 };
602 
603 static void __init reserve_standard_io_resources(void)
604 {
605 	int i;
606 
607 	/* request I/O space for devices used on all i[345]86 PCs */
608 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
609 		request_resource(&ioport_resource, &standard_io_resources[i]);
610 
611 }
612 
613 /*
614  * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
615  * is_kdump_kernel() to determine if we are booting after a panic. Hence
616  * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
617  */
618 
619 #ifdef CONFIG_CRASH_DUMP
620 /* elfcorehdr= specifies the location of elf core header
621  * stored by the crashed kernel. This option will be passed
622  * by kexec loader to the capture kernel.
623  */
624 static int __init setup_elfcorehdr(char *arg)
625 {
626 	char *end;
627 	if (!arg)
628 		return -EINVAL;
629 	elfcorehdr_addr = memparse(arg, &end);
630 	return end > arg ? 0 : -EINVAL;
631 }
632 early_param("elfcorehdr", setup_elfcorehdr);
633 #endif
634 
635 static struct x86_quirks default_x86_quirks __initdata;
636 
637 struct x86_quirks *x86_quirks __initdata = &default_x86_quirks;
638 
639 #ifdef CONFIG_X86_RESERVE_LOW_64K
640 static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
641 {
642 	printk(KERN_NOTICE
643 		"%s detected: BIOS may corrupt low RAM, working around it.\n",
644 		d->ident);
645 
646 	e820_update_range(0, 0x10000, E820_RAM, E820_RESERVED);
647 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
648 
649 	return 0;
650 }
651 #endif
652 
653 /* List of systems that have known low memory corruption BIOS problems */
654 static struct dmi_system_id __initdata bad_bios_dmi_table[] = {
655 #ifdef CONFIG_X86_RESERVE_LOW_64K
656 	{
657 		.callback = dmi_low_memory_corruption,
658 		.ident = "AMI BIOS",
659 		.matches = {
660 			DMI_MATCH(DMI_BIOS_VENDOR, "American Megatrends Inc."),
661 		},
662 	},
663 	{
664 		.callback = dmi_low_memory_corruption,
665 		.ident = "Phoenix BIOS",
666 		.matches = {
667 			DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix Technologies"),
668 		},
669 	},
670 #endif
671 	{}
672 };
673 
674 /*
675  * Determine if we were loaded by an EFI loader.  If so, then we have also been
676  * passed the efi memmap, systab, etc., so we should use these data structures
677  * for initialization.  Note, the efi init code path is determined by the
678  * global efi_enabled. This allows the same kernel image to be used on existing
679  * systems (with a traditional BIOS) as well as on EFI systems.
680  */
681 /*
682  * setup_arch - architecture-specific boot-time initializations
683  *
684  * Note: On x86_64, fixmaps are ready for use even before this is called.
685  */
686 
687 void __init setup_arch(char **cmdline_p)
688 {
689 #ifdef CONFIG_X86_32
690 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
691 	visws_early_detect();
692 #else
693 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
694 #endif
695 
696 	/* VMI may relocate the fixmap; do this before touching ioremap area */
697 	vmi_init();
698 
699 	early_cpu_init();
700 	early_ioremap_init();
701 
702 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
703 	screen_info = boot_params.screen_info;
704 	edid_info = boot_params.edid_info;
705 #ifdef CONFIG_X86_32
706 	apm_info.bios = boot_params.apm_bios_info;
707 	ist_info = boot_params.ist_info;
708 	if (boot_params.sys_desc_table.length != 0) {
709 		set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
710 		machine_id = boot_params.sys_desc_table.table[0];
711 		machine_submodel_id = boot_params.sys_desc_table.table[1];
712 		BIOS_revision = boot_params.sys_desc_table.table[2];
713 	}
714 #endif
715 	saved_video_mode = boot_params.hdr.vid_mode;
716 	bootloader_type = boot_params.hdr.type_of_loader;
717 	if ((bootloader_type >> 4) == 0xe) {
718 		bootloader_type &= 0xf;
719 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
720 	}
721 	bootloader_version  = bootloader_type & 0xf;
722 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
723 
724 #ifdef CONFIG_BLK_DEV_RAM
725 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
726 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
727 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
728 #endif
729 #ifdef CONFIG_EFI
730 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
731 #ifdef CONFIG_X86_32
732 		     "EL32",
733 #else
734 		     "EL64",
735 #endif
736 	 4)) {
737 		efi_enabled = 1;
738 		efi_reserve_early();
739 	}
740 #endif
741 
742 	ARCH_SETUP
743 
744 	setup_memory_map();
745 	parse_setup_data();
746 	/* update the e820_saved too */
747 	e820_reserve_setup_data();
748 
749 	copy_edd();
750 
751 	if (!boot_params.hdr.root_flags)
752 		root_mountflags &= ~MS_RDONLY;
753 	init_mm.start_code = (unsigned long) _text;
754 	init_mm.end_code = (unsigned long) _etext;
755 	init_mm.end_data = (unsigned long) _edata;
756 	init_mm.brk = _brk_end;
757 
758 	code_resource.start = virt_to_phys(_text);
759 	code_resource.end = virt_to_phys(_etext)-1;
760 	data_resource.start = virt_to_phys(_etext);
761 	data_resource.end = virt_to_phys(_edata)-1;
762 	bss_resource.start = virt_to_phys(&__bss_start);
763 	bss_resource.end = virt_to_phys(&__bss_stop)-1;
764 
765 #ifdef CONFIG_CMDLINE_BOOL
766 #ifdef CONFIG_CMDLINE_OVERRIDE
767 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
768 #else
769 	if (builtin_cmdline[0]) {
770 		/* append boot loader cmdline to builtin */
771 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
772 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
773 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
774 	}
775 #endif
776 #endif
777 
778 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
779 	*cmdline_p = command_line;
780 
781 	parse_early_param();
782 
783 #ifdef CONFIG_X86_64
784 	check_efer();
785 #endif
786 
787 	/* Must be before kernel pagetables are setup */
788 	vmi_activate();
789 
790 	/* after early param, so could get panic from serial */
791 	reserve_early_setup_data();
792 
793 	if (acpi_mps_check()) {
794 #ifdef CONFIG_X86_LOCAL_APIC
795 		disable_apic = 1;
796 #endif
797 		setup_clear_cpu_cap(X86_FEATURE_APIC);
798 	}
799 
800 #ifdef CONFIG_PCI
801 	if (pci_early_dump_regs)
802 		early_dump_pci_devices();
803 #endif
804 
805 	finish_e820_parsing();
806 
807 	if (efi_enabled)
808 		efi_init();
809 
810 	dmi_scan_machine();
811 
812 	dmi_check_system(bad_bios_dmi_table);
813 
814 	/*
815 	 * VMware detection requires dmi to be available, so this
816 	 * needs to be done after dmi_scan_machine, for the BP.
817 	 */
818 	init_hypervisor(&boot_cpu_data);
819 
820 #ifdef CONFIG_X86_32
821 	probe_roms();
822 #endif
823 
824 	/* after parse_early_param, so could debug it */
825 	insert_resource(&iomem_resource, &code_resource);
826 	insert_resource(&iomem_resource, &data_resource);
827 	insert_resource(&iomem_resource, &bss_resource);
828 
829 
830 #ifdef CONFIG_X86_32
831 	if (ppro_with_ram_bug()) {
832 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
833 				  E820_RESERVED);
834 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
835 		printk(KERN_INFO "fixed physical RAM map:\n");
836 		e820_print_map("bad_ppro");
837 	}
838 #else
839 	early_gart_iommu_check();
840 #endif
841 
842 	/*
843 	 * partially used pages are not usable - thus
844 	 * we are rounding upwards:
845 	 */
846 	max_pfn = e820_end_of_ram_pfn();
847 
848 	/* preallocate 4k for mptable mpc */
849 	early_reserve_e820_mpc_new();
850 	/* update e820 for memory not covered by WB MTRRs */
851 	mtrr_bp_init();
852 	if (mtrr_trim_uncached_memory(max_pfn))
853 		max_pfn = e820_end_of_ram_pfn();
854 
855 #ifdef CONFIG_X86_32
856 	/* max_low_pfn get updated here */
857 	find_low_pfn_range();
858 #else
859 	num_physpages = max_pfn;
860 
861 	check_x2apic();
862 
863 	/* How many end-of-memory variables you have, grandma! */
864 	/* need this before calling reserve_initrd */
865 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
866 		max_low_pfn = e820_end_of_low_ram_pfn();
867 	else
868 		max_low_pfn = max_pfn;
869 
870 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
871 	max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT;
872 #endif
873 
874 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
875 	setup_bios_corruption_check();
876 #endif
877 
878 	printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
879 			max_pfn_mapped<<PAGE_SHIFT);
880 
881 	reserve_brk();
882 
883 	/* max_pfn_mapped is updated here */
884 	max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
885 	max_pfn_mapped = max_low_pfn_mapped;
886 
887 #ifdef CONFIG_X86_64
888 	if (max_pfn > max_low_pfn) {
889 		max_pfn_mapped = init_memory_mapping(1UL<<32,
890 						     max_pfn<<PAGE_SHIFT);
891 		/* can we preseve max_low_pfn ?*/
892 		max_low_pfn = max_pfn;
893 	}
894 #endif
895 
896 	/*
897 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
898 	 */
899 
900 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
901 	if (init_ohci1394_dma_early)
902 		init_ohci1394_dma_on_all_controllers();
903 #endif
904 
905 	reserve_initrd();
906 
907 	vsmp_init();
908 
909 	io_delay_init();
910 
911 	/*
912 	 * Parse the ACPI tables for possible boot-time SMP configuration.
913 	 */
914 	acpi_boot_table_init();
915 
916 	early_acpi_boot_init();
917 
918 #ifdef CONFIG_ACPI_NUMA
919 	/*
920 	 * Parse SRAT to discover nodes.
921 	 */
922 	acpi_numa_init();
923 #endif
924 
925 	initmem_init(0, max_pfn);
926 
927 #ifdef CONFIG_ACPI_SLEEP
928 	/*
929 	 * Reserve low memory region for sleep support.
930 	 */
931 	acpi_reserve_bootmem();
932 #endif
933 	/*
934 	 * Find and reserve possible boot-time SMP configuration:
935 	 */
936 	find_smp_config();
937 
938 	reserve_crashkernel();
939 
940 #ifdef CONFIG_X86_64
941 	/*
942 	 * dma32_reserve_bootmem() allocates bootmem which may conflict
943 	 * with the crashkernel command line, so do that after
944 	 * reserve_crashkernel()
945 	 */
946 	dma32_reserve_bootmem();
947 #endif
948 
949 	reserve_ibft_region();
950 
951 #ifdef CONFIG_KVM_CLOCK
952 	kvmclock_init();
953 #endif
954 
955 	paravirt_pagetable_setup_start(swapper_pg_dir);
956 	paging_init();
957 	paravirt_pagetable_setup_done(swapper_pg_dir);
958 	paravirt_post_allocator_init();
959 
960 #ifdef CONFIG_X86_64
961 	map_vsyscall();
962 #endif
963 
964 	generic_apic_probe();
965 
966 	early_quirks();
967 
968 	/*
969 	 * Read APIC and some other early information from ACPI tables.
970 	 */
971 	acpi_boot_init();
972 
973 #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
974 	/*
975 	 * get boot-time SMP configuration:
976 	 */
977 	if (smp_found_config)
978 		get_smp_config();
979 #endif
980 
981 	prefill_possible_map();
982 
983 #ifdef CONFIG_X86_64
984 	init_cpu_to_node();
985 #endif
986 
987 	init_apic_mappings();
988 	ioapic_init_mappings();
989 
990 	/* need to wait for io_apic is mapped */
991 	probe_nr_irqs_gsi();
992 
993 	kvm_guest_init();
994 
995 	e820_reserve_resources();
996 	e820_mark_nosave_regions(max_low_pfn);
997 
998 #ifdef CONFIG_X86_32
999 	request_resource(&iomem_resource, &video_ram_resource);
1000 #endif
1001 	reserve_standard_io_resources();
1002 
1003 	e820_setup_gap();
1004 
1005 #ifdef CONFIG_VT
1006 #if defined(CONFIG_VGA_CONSOLE)
1007 	if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1008 		conswitchp = &vga_con;
1009 #elif defined(CONFIG_DUMMY_CONSOLE)
1010 	conswitchp = &dummy_con;
1011 #endif
1012 #endif
1013 }
1014 
1015 #ifdef CONFIG_X86_32
1016 
1017 /**
1018  * x86_quirk_intr_init - post gate setup interrupt initialisation
1019  *
1020  * Description:
1021  *	Fill in any interrupts that may have been left out by the general
1022  *	init_IRQ() routine.  interrupts having to do with the machine rather
1023  *	than the devices on the I/O bus (like APIC interrupts in intel MP
1024  *	systems) are started here.
1025  **/
1026 void __init x86_quirk_intr_init(void)
1027 {
1028 	if (x86_quirks->arch_intr_init) {
1029 		if (x86_quirks->arch_intr_init())
1030 			return;
1031 	}
1032 }
1033 
1034 /**
1035  * x86_quirk_trap_init - initialise system specific traps
1036  *
1037  * Description:
1038  *	Called as the final act of trap_init().  Used in VISWS to initialise
1039  *	the various board specific APIC traps.
1040  **/
1041 void __init x86_quirk_trap_init(void)
1042 {
1043 	if (x86_quirks->arch_trap_init) {
1044 		if (x86_quirks->arch_trap_init())
1045 			return;
1046 	}
1047 }
1048 
1049 static struct irqaction irq0  = {
1050 	.handler = timer_interrupt,
1051 	.flags = IRQF_DISABLED | IRQF_NOBALANCING | IRQF_IRQPOLL | IRQF_TIMER,
1052 	.name = "timer"
1053 };
1054 
1055 /**
1056  * x86_quirk_pre_time_init - do any specific initialisations before.
1057  *
1058  **/
1059 void __init x86_quirk_pre_time_init(void)
1060 {
1061 	if (x86_quirks->arch_pre_time_init)
1062 		x86_quirks->arch_pre_time_init();
1063 }
1064 
1065 /**
1066  * x86_quirk_time_init - do any specific initialisations for the system timer.
1067  *
1068  * Description:
1069  *	Must plug the system timer interrupt source at HZ into the IRQ listed
1070  *	in irq_vectors.h:TIMER_IRQ
1071  **/
1072 void __init x86_quirk_time_init(void)
1073 {
1074 	if (x86_quirks->arch_time_init) {
1075 		/*
1076 		 * A nonzero return code does not mean failure, it means
1077 		 * that the architecture quirk does not want any
1078 		 * generic (timer) setup to be performed after this:
1079 		 */
1080 		if (x86_quirks->arch_time_init())
1081 			return;
1082 	}
1083 
1084 	irq0.mask = cpumask_of_cpu(0);
1085 	setup_irq(0, &irq0);
1086 }
1087 #endif /* CONFIG_X86_32 */
1088