xref: /linux/arch/x86/kernel/setup.c (revision 08ec212c0f92cbf30e3ecc7349f18151714041d6)
1 /*
2  *  Copyright (C) 1995  Linus Torvalds
3  *
4  *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
5  *
6  *  Memory region support
7  *	David Parsons <orc@pell.chi.il.us>, July-August 1999
8  *
9  *  Added E820 sanitization routine (removes overlapping memory regions);
10  *  Brian Moyle <bmoyle@mvista.com>, February 2001
11  *
12  * Moved CPU detection code to cpu/${cpu}.c
13  *    Patrick Mochel <mochel@osdl.org>, March 2002
14  *
15  *  Provisions for empty E820 memory regions (reported by certain BIOSes).
16  *  Alex Achenbach <xela@slit.de>, December 2002.
17  *
18  */
19 
20 /*
21  * This file handles the architecture-dependent parts of initialization
22  */
23 
24 #include <linux/sched.h>
25 #include <linux/mm.h>
26 #include <linux/mmzone.h>
27 #include <linux/screen_info.h>
28 #include <linux/ioport.h>
29 #include <linux/acpi.h>
30 #include <linux/sfi.h>
31 #include <linux/apm_bios.h>
32 #include <linux/initrd.h>
33 #include <linux/bootmem.h>
34 #include <linux/memblock.h>
35 #include <linux/seq_file.h>
36 #include <linux/console.h>
37 #include <linux/root_dev.h>
38 #include <linux/highmem.h>
39 #include <linux/module.h>
40 #include <linux/efi.h>
41 #include <linux/init.h>
42 #include <linux/edd.h>
43 #include <linux/iscsi_ibft.h>
44 #include <linux/nodemask.h>
45 #include <linux/kexec.h>
46 #include <linux/dmi.h>
47 #include <linux/pfn.h>
48 #include <linux/pci.h>
49 #include <asm/pci-direct.h>
50 #include <linux/init_ohci1394_dma.h>
51 #include <linux/kvm_para.h>
52 #include <linux/dma-contiguous.h>
53 
54 #include <linux/errno.h>
55 #include <linux/kernel.h>
56 #include <linux/stddef.h>
57 #include <linux/unistd.h>
58 #include <linux/ptrace.h>
59 #include <linux/user.h>
60 #include <linux/delay.h>
61 
62 #include <linux/kallsyms.h>
63 #include <linux/cpufreq.h>
64 #include <linux/dma-mapping.h>
65 #include <linux/ctype.h>
66 #include <linux/uaccess.h>
67 
68 #include <linux/percpu.h>
69 #include <linux/crash_dump.h>
70 #include <linux/tboot.h>
71 #include <linux/jiffies.h>
72 
73 #include <video/edid.h>
74 
75 #include <asm/mtrr.h>
76 #include <asm/apic.h>
77 #include <asm/realmode.h>
78 #include <asm/e820.h>
79 #include <asm/mpspec.h>
80 #include <asm/setup.h>
81 #include <asm/efi.h>
82 #include <asm/timer.h>
83 #include <asm/i8259.h>
84 #include <asm/sections.h>
85 #include <asm/dmi.h>
86 #include <asm/io_apic.h>
87 #include <asm/ist.h>
88 #include <asm/setup_arch.h>
89 #include <asm/bios_ebda.h>
90 #include <asm/cacheflush.h>
91 #include <asm/processor.h>
92 #include <asm/bugs.h>
93 
94 #include <asm/vsyscall.h>
95 #include <asm/cpu.h>
96 #include <asm/desc.h>
97 #include <asm/dma.h>
98 #include <asm/iommu.h>
99 #include <asm/gart.h>
100 #include <asm/mmu_context.h>
101 #include <asm/proto.h>
102 
103 #include <asm/paravirt.h>
104 #include <asm/hypervisor.h>
105 #include <asm/olpc_ofw.h>
106 
107 #include <asm/percpu.h>
108 #include <asm/topology.h>
109 #include <asm/apicdef.h>
110 #include <asm/amd_nb.h>
111 #ifdef CONFIG_X86_64
112 #include <asm/numa_64.h>
113 #endif
114 #include <asm/mce.h>
115 #include <asm/alternative.h>
116 #include <asm/prom.h>
117 
118 /*
119  * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
120  * The direct mapping extends to max_pfn_mapped, so that we can directly access
121  * apertures, ACPI and other tables without having to play with fixmaps.
122  */
123 unsigned long max_low_pfn_mapped;
124 unsigned long max_pfn_mapped;
125 
126 #ifdef CONFIG_DMI
127 RESERVE_BRK(dmi_alloc, 65536);
128 #endif
129 
130 
131 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
132 unsigned long _brk_end = (unsigned long)__brk_base;
133 
134 #ifdef CONFIG_X86_64
135 int default_cpu_present_to_apicid(int mps_cpu)
136 {
137 	return __default_cpu_present_to_apicid(mps_cpu);
138 }
139 
140 int default_check_phys_apicid_present(int phys_apicid)
141 {
142 	return __default_check_phys_apicid_present(phys_apicid);
143 }
144 #endif
145 
146 #ifndef CONFIG_DEBUG_BOOT_PARAMS
147 struct boot_params __initdata boot_params;
148 #else
149 struct boot_params boot_params;
150 #endif
151 
152 /*
153  * Machine setup..
154  */
155 static struct resource data_resource = {
156 	.name	= "Kernel data",
157 	.start	= 0,
158 	.end	= 0,
159 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
160 };
161 
162 static struct resource code_resource = {
163 	.name	= "Kernel code",
164 	.start	= 0,
165 	.end	= 0,
166 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
167 };
168 
169 static struct resource bss_resource = {
170 	.name	= "Kernel bss",
171 	.start	= 0,
172 	.end	= 0,
173 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
174 };
175 
176 
177 #ifdef CONFIG_X86_32
178 /* cpu data as detected by the assembly code in head.S */
179 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
180 /* common cpu data for all cpus */
181 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
182 EXPORT_SYMBOL(boot_cpu_data);
183 
184 unsigned int def_to_bigsmp;
185 
186 /* for MCA, but anyone else can use it if they want */
187 unsigned int machine_id;
188 unsigned int machine_submodel_id;
189 unsigned int BIOS_revision;
190 
191 struct apm_info apm_info;
192 EXPORT_SYMBOL(apm_info);
193 
194 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
195 	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
196 struct ist_info ist_info;
197 EXPORT_SYMBOL(ist_info);
198 #else
199 struct ist_info ist_info;
200 #endif
201 
202 #else
203 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
204 	.x86_phys_bits = MAX_PHYSMEM_BITS,
205 };
206 EXPORT_SYMBOL(boot_cpu_data);
207 #endif
208 
209 
210 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
211 unsigned long mmu_cr4_features;
212 #else
213 unsigned long mmu_cr4_features = X86_CR4_PAE;
214 #endif
215 
216 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
217 int bootloader_type, bootloader_version;
218 
219 /*
220  * Setup options
221  */
222 struct screen_info screen_info;
223 EXPORT_SYMBOL(screen_info);
224 struct edid_info edid_info;
225 EXPORT_SYMBOL_GPL(edid_info);
226 
227 extern int root_mountflags;
228 
229 unsigned long saved_video_mode;
230 
231 #define RAMDISK_IMAGE_START_MASK	0x07FF
232 #define RAMDISK_PROMPT_FLAG		0x8000
233 #define RAMDISK_LOAD_FLAG		0x4000
234 
235 static char __initdata command_line[COMMAND_LINE_SIZE];
236 #ifdef CONFIG_CMDLINE_BOOL
237 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
238 #endif
239 
240 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
241 struct edd edd;
242 #ifdef CONFIG_EDD_MODULE
243 EXPORT_SYMBOL(edd);
244 #endif
245 /**
246  * copy_edd() - Copy the BIOS EDD information
247  *              from boot_params into a safe place.
248  *
249  */
250 static inline void __init copy_edd(void)
251 {
252      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
253 	    sizeof(edd.mbr_signature));
254      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
255      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
256      edd.edd_info_nr = boot_params.eddbuf_entries;
257 }
258 #else
259 static inline void __init copy_edd(void)
260 {
261 }
262 #endif
263 
264 void * __init extend_brk(size_t size, size_t align)
265 {
266 	size_t mask = align - 1;
267 	void *ret;
268 
269 	BUG_ON(_brk_start == 0);
270 	BUG_ON(align & mask);
271 
272 	_brk_end = (_brk_end + mask) & ~mask;
273 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
274 
275 	ret = (void *)_brk_end;
276 	_brk_end += size;
277 
278 	memset(ret, 0, size);
279 
280 	return ret;
281 }
282 
283 #ifdef CONFIG_X86_64
284 static void __init init_gbpages(void)
285 {
286 	if (direct_gbpages && cpu_has_gbpages)
287 		printk(KERN_INFO "Using GB pages for direct mapping\n");
288 	else
289 		direct_gbpages = 0;
290 }
291 #else
292 static inline void init_gbpages(void)
293 {
294 }
295 static void __init cleanup_highmap(void)
296 {
297 }
298 #endif
299 
300 static void __init reserve_brk(void)
301 {
302 	if (_brk_end > _brk_start)
303 		memblock_reserve(__pa(_brk_start),
304 				 __pa(_brk_end) - __pa(_brk_start));
305 
306 	/* Mark brk area as locked down and no longer taking any
307 	   new allocations */
308 	_brk_start = 0;
309 }
310 
311 #ifdef CONFIG_BLK_DEV_INITRD
312 
313 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
314 static void __init relocate_initrd(void)
315 {
316 	/* Assume only end is not page aligned */
317 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
318 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
319 	u64 area_size     = PAGE_ALIGN(ramdisk_size);
320 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
321 	u64 ramdisk_here;
322 	unsigned long slop, clen, mapaddr;
323 	char *p, *q;
324 
325 	/* We need to move the initrd down into lowmem */
326 	ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
327 					 PAGE_SIZE);
328 
329 	if (!ramdisk_here)
330 		panic("Cannot find place for new RAMDISK of size %lld\n",
331 			 ramdisk_size);
332 
333 	/* Note: this includes all the lowmem currently occupied by
334 	   the initrd, we rely on that fact to keep the data intact. */
335 	memblock_reserve(ramdisk_here, area_size);
336 	initrd_start = ramdisk_here + PAGE_OFFSET;
337 	initrd_end   = initrd_start + ramdisk_size;
338 	printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
339 			 ramdisk_here, ramdisk_here + ramdisk_size - 1);
340 
341 	q = (char *)initrd_start;
342 
343 	/* Copy any lowmem portion of the initrd */
344 	if (ramdisk_image < end_of_lowmem) {
345 		clen = end_of_lowmem - ramdisk_image;
346 		p = (char *)__va(ramdisk_image);
347 		memcpy(q, p, clen);
348 		q += clen;
349 		ramdisk_image += clen;
350 		ramdisk_size  -= clen;
351 	}
352 
353 	/* Copy the highmem portion of the initrd */
354 	while (ramdisk_size) {
355 		slop = ramdisk_image & ~PAGE_MASK;
356 		clen = ramdisk_size;
357 		if (clen > MAX_MAP_CHUNK-slop)
358 			clen = MAX_MAP_CHUNK-slop;
359 		mapaddr = ramdisk_image & PAGE_MASK;
360 		p = early_memremap(mapaddr, clen+slop);
361 		memcpy(q, p+slop, clen);
362 		early_iounmap(p, clen+slop);
363 		q += clen;
364 		ramdisk_image += clen;
365 		ramdisk_size  -= clen;
366 	}
367 	/* high pages is not converted by early_res_to_bootmem */
368 	ramdisk_image = boot_params.hdr.ramdisk_image;
369 	ramdisk_size  = boot_params.hdr.ramdisk_size;
370 	printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
371 		" [mem %#010llx-%#010llx]\n",
372 		ramdisk_image, ramdisk_image + ramdisk_size - 1,
373 		ramdisk_here, ramdisk_here + ramdisk_size - 1);
374 }
375 
376 static void __init reserve_initrd(void)
377 {
378 	/* Assume only end is not page aligned */
379 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
380 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
381 	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
382 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
383 
384 	if (!boot_params.hdr.type_of_loader ||
385 	    !ramdisk_image || !ramdisk_size)
386 		return;		/* No initrd provided by bootloader */
387 
388 	initrd_start = 0;
389 
390 	if (ramdisk_size >= (end_of_lowmem>>1)) {
391 		panic("initrd too large to handle, "
392 		       "disabling initrd (%lld needed, %lld available)\n",
393 		       ramdisk_size, end_of_lowmem>>1);
394 	}
395 
396 	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
397 			ramdisk_end - 1);
398 
399 
400 	if (ramdisk_end <= end_of_lowmem) {
401 		/* All in lowmem, easy case */
402 		/*
403 		 * don't need to reserve again, already reserved early
404 		 * in i386_start_kernel
405 		 */
406 		initrd_start = ramdisk_image + PAGE_OFFSET;
407 		initrd_end = initrd_start + ramdisk_size;
408 		return;
409 	}
410 
411 	relocate_initrd();
412 
413 	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
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 		u32 data_len, map_len;
431 
432 		map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
433 			      (u64)sizeof(struct setup_data));
434 		data = early_memremap(pa_data, map_len);
435 		data_len = data->len + sizeof(struct setup_data);
436 		if (data_len > map_len) {
437 			early_iounmap(data, map_len);
438 			data = early_memremap(pa_data, data_len);
439 			map_len = data_len;
440 		}
441 
442 		switch (data->type) {
443 		case SETUP_E820_EXT:
444 			parse_e820_ext(data);
445 			break;
446 		case SETUP_DTB:
447 			add_dtb(pa_data);
448 			break;
449 		default:
450 			break;
451 		}
452 		pa_data = data->next;
453 		early_iounmap(data, map_len);
454 	}
455 }
456 
457 static void __init e820_reserve_setup_data(void)
458 {
459 	struct setup_data *data;
460 	u64 pa_data;
461 	int found = 0;
462 
463 	if (boot_params.hdr.version < 0x0209)
464 		return;
465 	pa_data = boot_params.hdr.setup_data;
466 	while (pa_data) {
467 		data = early_memremap(pa_data, sizeof(*data));
468 		e820_update_range(pa_data, sizeof(*data)+data->len,
469 			 E820_RAM, E820_RESERVED_KERN);
470 		found = 1;
471 		pa_data = data->next;
472 		early_iounmap(data, sizeof(*data));
473 	}
474 	if (!found)
475 		return;
476 
477 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
478 	memcpy(&e820_saved, &e820, sizeof(struct e820map));
479 	printk(KERN_INFO "extended physical RAM map:\n");
480 	e820_print_map("reserve setup_data");
481 }
482 
483 static void __init memblock_x86_reserve_range_setup_data(void)
484 {
485 	struct setup_data *data;
486 	u64 pa_data;
487 
488 	if (boot_params.hdr.version < 0x0209)
489 		return;
490 	pa_data = boot_params.hdr.setup_data;
491 	while (pa_data) {
492 		data = early_memremap(pa_data, sizeof(*data));
493 		memblock_reserve(pa_data, sizeof(*data) + data->len);
494 		pa_data = data->next;
495 		early_iounmap(data, sizeof(*data));
496 	}
497 }
498 
499 /*
500  * --------- Crashkernel reservation ------------------------------
501  */
502 
503 #ifdef CONFIG_KEXEC
504 
505 /*
506  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
507  * would limit the kernel to the low 512 MiB due to mapping restrictions.
508  * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
509  * limit once kexec-tools are fixed.
510  */
511 #ifdef CONFIG_X86_32
512 # define CRASH_KERNEL_ADDR_MAX	(512 << 20)
513 #else
514 # define CRASH_KERNEL_ADDR_MAX	(896 << 20)
515 #endif
516 
517 static void __init reserve_crashkernel(void)
518 {
519 	unsigned long long total_mem;
520 	unsigned long long crash_size, crash_base;
521 	int ret;
522 
523 	total_mem = memblock_phys_mem_size();
524 
525 	ret = parse_crashkernel(boot_command_line, total_mem,
526 			&crash_size, &crash_base);
527 	if (ret != 0 || crash_size <= 0)
528 		return;
529 
530 	/* 0 means: find the address automatically */
531 	if (crash_base <= 0) {
532 		const unsigned long long alignment = 16<<20;	/* 16M */
533 
534 		/*
535 		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
536 		 */
537 		crash_base = memblock_find_in_range(alignment,
538 			       CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
539 
540 		if (!crash_base) {
541 			pr_info("crashkernel reservation failed - No suitable area found.\n");
542 			return;
543 		}
544 	} else {
545 		unsigned long long start;
546 
547 		start = memblock_find_in_range(crash_base,
548 				 crash_base + crash_size, crash_size, 1<<20);
549 		if (start != crash_base) {
550 			pr_info("crashkernel reservation failed - memory is in use.\n");
551 			return;
552 		}
553 	}
554 	memblock_reserve(crash_base, crash_size);
555 
556 	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
557 			"for crashkernel (System RAM: %ldMB)\n",
558 			(unsigned long)(crash_size >> 20),
559 			(unsigned long)(crash_base >> 20),
560 			(unsigned long)(total_mem >> 20));
561 
562 	crashk_res.start = crash_base;
563 	crashk_res.end   = crash_base + crash_size - 1;
564 	insert_resource(&iomem_resource, &crashk_res);
565 }
566 #else
567 static void __init reserve_crashkernel(void)
568 {
569 }
570 #endif
571 
572 static struct resource standard_io_resources[] = {
573 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
574 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
575 	{ .name = "pic1", .start = 0x20, .end = 0x21,
576 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
577 	{ .name = "timer0", .start = 0x40, .end = 0x43,
578 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
579 	{ .name = "timer1", .start = 0x50, .end = 0x53,
580 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
581 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
582 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
583 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
584 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
585 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
586 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
587 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
588 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
589 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
590 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
591 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
592 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
593 };
594 
595 void __init reserve_standard_io_resources(void)
596 {
597 	int i;
598 
599 	/* request I/O space for devices used on all i[345]86 PCs */
600 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
601 		request_resource(&ioport_resource, &standard_io_resources[i]);
602 
603 }
604 
605 static __init void reserve_ibft_region(void)
606 {
607 	unsigned long addr, size = 0;
608 
609 	addr = find_ibft_region(&size);
610 
611 	if (size)
612 		memblock_reserve(addr, size);
613 }
614 
615 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
616 
617 static void __init trim_bios_range(void)
618 {
619 	/*
620 	 * A special case is the first 4Kb of memory;
621 	 * This is a BIOS owned area, not kernel ram, but generally
622 	 * not listed as such in the E820 table.
623 	 *
624 	 * This typically reserves additional memory (64KiB by default)
625 	 * since some BIOSes are known to corrupt low memory.  See the
626 	 * Kconfig help text for X86_RESERVE_LOW.
627 	 */
628 	e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
629 			  E820_RAM, E820_RESERVED);
630 
631 	/*
632 	 * special case: Some BIOSen report the PC BIOS
633 	 * area (640->1Mb) as ram even though it is not.
634 	 * take them out.
635 	 */
636 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
637 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
638 }
639 
640 static int __init parse_reservelow(char *p)
641 {
642 	unsigned long long size;
643 
644 	if (!p)
645 		return -EINVAL;
646 
647 	size = memparse(p, &p);
648 
649 	if (size < 4096)
650 		size = 4096;
651 
652 	if (size > 640*1024)
653 		size = 640*1024;
654 
655 	reserve_low = size;
656 
657 	return 0;
658 }
659 
660 early_param("reservelow", parse_reservelow);
661 
662 /*
663  * Determine if we were loaded by an EFI loader.  If so, then we have also been
664  * passed the efi memmap, systab, etc., so we should use these data structures
665  * for initialization.  Note, the efi init code path is determined by the
666  * global efi_enabled. This allows the same kernel image to be used on existing
667  * systems (with a traditional BIOS) as well as on EFI systems.
668  */
669 /*
670  * setup_arch - architecture-specific boot-time initializations
671  *
672  * Note: On x86_64, fixmaps are ready for use even before this is called.
673  */
674 
675 void __init setup_arch(char **cmdline_p)
676 {
677 #ifdef CONFIG_X86_32
678 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
679 	visws_early_detect();
680 
681 	/*
682 	 * copy kernel address range established so far and switch
683 	 * to the proper swapper page table
684 	 */
685 	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
686 			initial_page_table + KERNEL_PGD_BOUNDARY,
687 			KERNEL_PGD_PTRS);
688 
689 	load_cr3(swapper_pg_dir);
690 	__flush_tlb_all();
691 #else
692 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
693 #endif
694 
695 	/*
696 	 * If we have OLPC OFW, we might end up relocating the fixmap due to
697 	 * reserve_top(), so do this before touching the ioremap area.
698 	 */
699 	olpc_ofw_detect();
700 
701 	early_trap_init();
702 	early_cpu_init();
703 	early_ioremap_init();
704 
705 	setup_olpc_ofw_pgd();
706 
707 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
708 	screen_info = boot_params.screen_info;
709 	edid_info = boot_params.edid_info;
710 #ifdef CONFIG_X86_32
711 	apm_info.bios = boot_params.apm_bios_info;
712 	ist_info = boot_params.ist_info;
713 	if (boot_params.sys_desc_table.length != 0) {
714 		machine_id = boot_params.sys_desc_table.table[0];
715 		machine_submodel_id = boot_params.sys_desc_table.table[1];
716 		BIOS_revision = boot_params.sys_desc_table.table[2];
717 	}
718 #endif
719 	saved_video_mode = boot_params.hdr.vid_mode;
720 	bootloader_type = boot_params.hdr.type_of_loader;
721 	if ((bootloader_type >> 4) == 0xe) {
722 		bootloader_type &= 0xf;
723 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
724 	}
725 	bootloader_version  = bootloader_type & 0xf;
726 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
727 
728 #ifdef CONFIG_BLK_DEV_RAM
729 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
730 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
731 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
732 #endif
733 #ifdef CONFIG_EFI
734 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
735 		     "EL32", 4)) {
736 		efi_enabled = 1;
737 		efi_64bit = false;
738 	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
739 		     "EL64", 4)) {
740 		efi_enabled = 1;
741 		efi_64bit = true;
742 	}
743 	if (efi_enabled && efi_memblock_x86_reserve_range())
744 		efi_enabled = 0;
745 #endif
746 
747 	x86_init.oem.arch_setup();
748 
749 	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
750 	setup_memory_map();
751 	parse_setup_data();
752 	/* update the e820_saved too */
753 	e820_reserve_setup_data();
754 
755 	copy_edd();
756 
757 	if (!boot_params.hdr.root_flags)
758 		root_mountflags &= ~MS_RDONLY;
759 	init_mm.start_code = (unsigned long) _text;
760 	init_mm.end_code = (unsigned long) _etext;
761 	init_mm.end_data = (unsigned long) _edata;
762 	init_mm.brk = _brk_end;
763 
764 	code_resource.start = virt_to_phys(_text);
765 	code_resource.end = virt_to_phys(_etext)-1;
766 	data_resource.start = virt_to_phys(_etext);
767 	data_resource.end = virt_to_phys(_edata)-1;
768 	bss_resource.start = virt_to_phys(&__bss_start);
769 	bss_resource.end = virt_to_phys(&__bss_stop)-1;
770 
771 #ifdef CONFIG_CMDLINE_BOOL
772 #ifdef CONFIG_CMDLINE_OVERRIDE
773 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
774 #else
775 	if (builtin_cmdline[0]) {
776 		/* append boot loader cmdline to builtin */
777 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
778 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
779 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
780 	}
781 #endif
782 #endif
783 
784 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
785 	*cmdline_p = command_line;
786 
787 	/*
788 	 * x86_configure_nx() is called before parse_early_param() to detect
789 	 * whether hardware doesn't support NX (so that the early EHCI debug
790 	 * console setup can safely call set_fixmap()). It may then be called
791 	 * again from within noexec_setup() during parsing early parameters
792 	 * to honor the respective command line option.
793 	 */
794 	x86_configure_nx();
795 
796 	parse_early_param();
797 
798 	x86_report_nx();
799 
800 	/* after early param, so could get panic from serial */
801 	memblock_x86_reserve_range_setup_data();
802 
803 	if (acpi_mps_check()) {
804 #ifdef CONFIG_X86_LOCAL_APIC
805 		disable_apic = 1;
806 #endif
807 		setup_clear_cpu_cap(X86_FEATURE_APIC);
808 	}
809 
810 #ifdef CONFIG_PCI
811 	if (pci_early_dump_regs)
812 		early_dump_pci_devices();
813 #endif
814 
815 	finish_e820_parsing();
816 
817 	if (efi_enabled)
818 		efi_init();
819 
820 	dmi_scan_machine();
821 
822 	/*
823 	 * VMware detection requires dmi to be available, so this
824 	 * needs to be done after dmi_scan_machine, for the BP.
825 	 */
826 	init_hypervisor_platform();
827 
828 	x86_init.resources.probe_roms();
829 
830 	/* after parse_early_param, so could debug it */
831 	insert_resource(&iomem_resource, &code_resource);
832 	insert_resource(&iomem_resource, &data_resource);
833 	insert_resource(&iomem_resource, &bss_resource);
834 
835 	trim_bios_range();
836 #ifdef CONFIG_X86_32
837 	if (ppro_with_ram_bug()) {
838 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
839 				  E820_RESERVED);
840 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
841 		printk(KERN_INFO "fixed physical RAM map:\n");
842 		e820_print_map("bad_ppro");
843 	}
844 #else
845 	early_gart_iommu_check();
846 #endif
847 
848 	/*
849 	 * partially used pages are not usable - thus
850 	 * we are rounding upwards:
851 	 */
852 	max_pfn = e820_end_of_ram_pfn();
853 
854 	/* update e820 for memory not covered by WB MTRRs */
855 	mtrr_bp_init();
856 	if (mtrr_trim_uncached_memory(max_pfn))
857 		max_pfn = e820_end_of_ram_pfn();
858 
859 #ifdef CONFIG_X86_32
860 	/* max_low_pfn get updated here */
861 	find_low_pfn_range();
862 #else
863 	num_physpages = max_pfn;
864 
865 	check_x2apic();
866 
867 	/* How many end-of-memory variables you have, grandma! */
868 	/* need this before calling reserve_initrd */
869 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
870 		max_low_pfn = e820_end_of_low_ram_pfn();
871 	else
872 		max_low_pfn = max_pfn;
873 
874 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
875 #endif
876 
877 	/*
878 	 * Find and reserve possible boot-time SMP configuration:
879 	 */
880 	find_smp_config();
881 
882 	reserve_ibft_region();
883 
884 	/*
885 	 * Need to conclude brk, before memblock_x86_fill()
886 	 *  it could use memblock_find_in_range, could overlap with
887 	 *  brk area.
888 	 */
889 	reserve_brk();
890 
891 	cleanup_highmap();
892 
893 	memblock.current_limit = get_max_mapped();
894 	memblock_x86_fill();
895 
896 	/*
897 	 * The EFI specification says that boot service code won't be called
898 	 * after ExitBootServices(). This is, in fact, a lie.
899 	 */
900 	if (efi_enabled)
901 		efi_reserve_boot_services();
902 
903 	/* preallocate 4k for mptable mpc */
904 	early_reserve_e820_mpc_new();
905 
906 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
907 	setup_bios_corruption_check();
908 #endif
909 
910 	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
911 			(max_pfn_mapped<<PAGE_SHIFT) - 1);
912 
913 	setup_real_mode();
914 
915 	init_gbpages();
916 
917 	/* max_pfn_mapped is updated here */
918 	max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
919 	max_pfn_mapped = max_low_pfn_mapped;
920 
921 #ifdef CONFIG_X86_64
922 	if (max_pfn > max_low_pfn) {
923 		int i;
924 		for (i = 0; i < e820.nr_map; i++) {
925 			struct e820entry *ei = &e820.map[i];
926 
927 			if (ei->addr + ei->size <= 1UL << 32)
928 				continue;
929 
930 			if (ei->type == E820_RESERVED)
931 				continue;
932 
933 			max_pfn_mapped = init_memory_mapping(
934 				ei->addr < 1UL << 32 ? 1UL << 32 : ei->addr,
935 				ei->addr + ei->size);
936 		}
937 
938 		/* can we preseve max_low_pfn ?*/
939 		max_low_pfn = max_pfn;
940 	}
941 #endif
942 	memblock.current_limit = get_max_mapped();
943 	dma_contiguous_reserve(0);
944 
945 	/*
946 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
947 	 */
948 
949 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
950 	if (init_ohci1394_dma_early)
951 		init_ohci1394_dma_on_all_controllers();
952 #endif
953 	/* Allocate bigger log buffer */
954 	setup_log_buf(1);
955 
956 	reserve_initrd();
957 
958 	reserve_crashkernel();
959 
960 	vsmp_init();
961 
962 	io_delay_init();
963 
964 	/*
965 	 * Parse the ACPI tables for possible boot-time SMP configuration.
966 	 */
967 	acpi_boot_table_init();
968 
969 	early_acpi_boot_init();
970 
971 	initmem_init();
972 	memblock_find_dma_reserve();
973 
974 #ifdef CONFIG_KVM_GUEST
975 	kvmclock_init();
976 #endif
977 
978 	x86_init.paging.pagetable_init();
979 
980 	if (boot_cpu_data.cpuid_level >= 0) {
981 		/* A CPU has %cr4 if and only if it has CPUID */
982 		mmu_cr4_features = read_cr4();
983 		if (trampoline_cr4_features)
984 			*trampoline_cr4_features = mmu_cr4_features;
985 	}
986 
987 #ifdef CONFIG_X86_32
988 	/* sync back kernel address range */
989 	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
990 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
991 			KERNEL_PGD_PTRS);
992 #endif
993 
994 	tboot_probe();
995 
996 #ifdef CONFIG_X86_64
997 	map_vsyscall();
998 #endif
999 
1000 	generic_apic_probe();
1001 
1002 	early_quirks();
1003 
1004 	/*
1005 	 * Read APIC and some other early information from ACPI tables.
1006 	 */
1007 	acpi_boot_init();
1008 	sfi_init();
1009 	x86_dtb_init();
1010 
1011 	/*
1012 	 * get boot-time SMP configuration:
1013 	 */
1014 	if (smp_found_config)
1015 		get_smp_config();
1016 
1017 	prefill_possible_map();
1018 
1019 	init_cpu_to_node();
1020 
1021 	init_apic_mappings();
1022 	if (x86_io_apic_ops.init)
1023 		x86_io_apic_ops.init();
1024 
1025 	kvm_guest_init();
1026 
1027 	e820_reserve_resources();
1028 	e820_mark_nosave_regions(max_low_pfn);
1029 
1030 	x86_init.resources.reserve_resources();
1031 
1032 	e820_setup_gap();
1033 
1034 #ifdef CONFIG_VT
1035 #if defined(CONFIG_VGA_CONSOLE)
1036 	if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1037 		conswitchp = &vga_con;
1038 #elif defined(CONFIG_DUMMY_CONSOLE)
1039 	conswitchp = &dummy_con;
1040 #endif
1041 #endif
1042 	x86_init.oem.banner();
1043 
1044 	x86_init.timers.wallclock_init();
1045 
1046 	mcheck_init();
1047 
1048 	arch_init_ideal_nops();
1049 
1050 	register_refined_jiffies(CLOCK_TICK_RATE);
1051 }
1052 
1053 #ifdef CONFIG_X86_32
1054 
1055 static struct resource video_ram_resource = {
1056 	.name	= "Video RAM area",
1057 	.start	= 0xa0000,
1058 	.end	= 0xbffff,
1059 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1060 };
1061 
1062 void __init i386_reserve_resources(void)
1063 {
1064 	request_resource(&iomem_resource, &video_ram_resource);
1065 	reserve_standard_io_resources();
1066 }
1067 
1068 #endif /* CONFIG_X86_32 */
1069