xref: /openbmc/linux/arch/x86/kernel/setup.c (revision 63dc02bd)
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/mca.h>
38 #include <linux/root_dev.h>
39 #include <linux/highmem.h>
40 #include <linux/module.h>
41 #include <linux/efi.h>
42 #include <linux/init.h>
43 #include <linux/edd.h>
44 #include <linux/iscsi_ibft.h>
45 #include <linux/nodemask.h>
46 #include <linux/kexec.h>
47 #include <linux/dmi.h>
48 #include <linux/pfn.h>
49 #include <linux/pci.h>
50 #include <asm/pci-direct.h>
51 #include <linux/init_ohci1394_dma.h>
52 #include <linux/kvm_para.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 
72 #include <video/edid.h>
73 
74 #include <asm/mtrr.h>
75 #include <asm/apic.h>
76 #include <asm/trampoline.h>
77 #include <asm/e820.h>
78 #include <asm/mpspec.h>
79 #include <asm/setup.h>
80 #include <asm/efi.h>
81 #include <asm/timer.h>
82 #include <asm/i8259.h>
83 #include <asm/sections.h>
84 #include <asm/dmi.h>
85 #include <asm/io_apic.h>
86 #include <asm/ist.h>
87 #include <asm/setup_arch.h>
88 #include <asm/bios_ebda.h>
89 #include <asm/cacheflush.h>
90 #include <asm/processor.h>
91 #include <asm/bugs.h>
92 
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 #include <asm/olpc_ofw.h>
105 
106 #include <asm/percpu.h>
107 #include <asm/topology.h>
108 #include <asm/apicdef.h>
109 #include <asm/amd_nb.h>
110 #ifdef CONFIG_X86_64
111 #include <asm/numa_64.h>
112 #endif
113 #include <asm/mce.h>
114 #include <asm/alternative.h>
115 #include <asm/prom.h>
116 
117 /*
118  * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
119  * The direct mapping extends to max_pfn_mapped, so that we can directly access
120  * apertures, ACPI and other tables without having to play with fixmaps.
121  */
122 unsigned long max_low_pfn_mapped;
123 unsigned long max_pfn_mapped;
124 
125 #ifdef CONFIG_DMI
126 RESERVE_BRK(dmi_alloc, 65536);
127 #endif
128 
129 
130 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
131 unsigned long _brk_end = (unsigned long)__brk_base;
132 
133 #ifdef CONFIG_X86_64
134 int default_cpu_present_to_apicid(int mps_cpu)
135 {
136 	return __default_cpu_present_to_apicid(mps_cpu);
137 }
138 
139 int default_check_phys_apicid_present(int phys_apicid)
140 {
141 	return __default_check_phys_apicid_present(phys_apicid);
142 }
143 #endif
144 
145 #ifndef CONFIG_DEBUG_BOOT_PARAMS
146 struct boot_params __initdata boot_params;
147 #else
148 struct boot_params boot_params;
149 #endif
150 
151 /*
152  * Machine setup..
153  */
154 static struct resource data_resource = {
155 	.name	= "Kernel data",
156 	.start	= 0,
157 	.end	= 0,
158 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
159 };
160 
161 static struct resource code_resource = {
162 	.name	= "Kernel code",
163 	.start	= 0,
164 	.end	= 0,
165 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
166 };
167 
168 static struct resource bss_resource = {
169 	.name	= "Kernel bss",
170 	.start	= 0,
171 	.end	= 0,
172 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
173 };
174 
175 
176 #ifdef CONFIG_X86_32
177 /* cpu data as detected by the assembly code in head.S */
178 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
179 /* common cpu data for all cpus */
180 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
181 EXPORT_SYMBOL(boot_cpu_data);
182 static void set_mca_bus(int x)
183 {
184 #ifdef CONFIG_MCA
185 	MCA_bus = x;
186 #endif
187 }
188 
189 unsigned int def_to_bigsmp;
190 
191 /* for MCA, but anyone else can use it if they want */
192 unsigned int machine_id;
193 unsigned int machine_submodel_id;
194 unsigned int BIOS_revision;
195 
196 struct apm_info apm_info;
197 EXPORT_SYMBOL(apm_info);
198 
199 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
200 	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
201 struct ist_info ist_info;
202 EXPORT_SYMBOL(ist_info);
203 #else
204 struct ist_info ist_info;
205 #endif
206 
207 #else
208 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
209 	.x86_phys_bits = MAX_PHYSMEM_BITS,
210 };
211 EXPORT_SYMBOL(boot_cpu_data);
212 #endif
213 
214 
215 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
216 unsigned long mmu_cr4_features;
217 #else
218 unsigned long mmu_cr4_features = X86_CR4_PAE;
219 #endif
220 
221 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
222 int bootloader_type, bootloader_version;
223 
224 /*
225  * Setup options
226  */
227 struct screen_info screen_info;
228 EXPORT_SYMBOL(screen_info);
229 struct edid_info edid_info;
230 EXPORT_SYMBOL_GPL(edid_info);
231 
232 extern int root_mountflags;
233 
234 unsigned long saved_video_mode;
235 
236 #define RAMDISK_IMAGE_START_MASK	0x07FF
237 #define RAMDISK_PROMPT_FLAG		0x8000
238 #define RAMDISK_LOAD_FLAG		0x4000
239 
240 static char __initdata command_line[COMMAND_LINE_SIZE];
241 #ifdef CONFIG_CMDLINE_BOOL
242 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
243 #endif
244 
245 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
246 struct edd edd;
247 #ifdef CONFIG_EDD_MODULE
248 EXPORT_SYMBOL(edd);
249 #endif
250 /**
251  * copy_edd() - Copy the BIOS EDD information
252  *              from boot_params into a safe place.
253  *
254  */
255 static inline void __init copy_edd(void)
256 {
257      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
258 	    sizeof(edd.mbr_signature));
259      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
260      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
261      edd.edd_info_nr = boot_params.eddbuf_entries;
262 }
263 #else
264 static inline void __init copy_edd(void)
265 {
266 }
267 #endif
268 
269 void * __init extend_brk(size_t size, size_t align)
270 {
271 	size_t mask = align - 1;
272 	void *ret;
273 
274 	BUG_ON(_brk_start == 0);
275 	BUG_ON(align & mask);
276 
277 	_brk_end = (_brk_end + mask) & ~mask;
278 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
279 
280 	ret = (void *)_brk_end;
281 	_brk_end += size;
282 
283 	memset(ret, 0, size);
284 
285 	return ret;
286 }
287 
288 #ifdef CONFIG_X86_64
289 static void __init init_gbpages(void)
290 {
291 	if (direct_gbpages && cpu_has_gbpages)
292 		printk(KERN_INFO "Using GB pages for direct mapping\n");
293 	else
294 		direct_gbpages = 0;
295 }
296 #else
297 static inline void init_gbpages(void)
298 {
299 }
300 static void __init cleanup_highmap(void)
301 {
302 }
303 #endif
304 
305 static void __init reserve_brk(void)
306 {
307 	if (_brk_end > _brk_start)
308 		memblock_reserve(__pa(_brk_start),
309 				 __pa(_brk_end) - __pa(_brk_start));
310 
311 	/* Mark brk area as locked down and no longer taking any
312 	   new allocations */
313 	_brk_start = 0;
314 }
315 
316 #ifdef CONFIG_BLK_DEV_INITRD
317 
318 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
319 static void __init relocate_initrd(void)
320 {
321 	/* Assume only end is not page aligned */
322 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
323 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
324 	u64 area_size     = PAGE_ALIGN(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 = memblock_find_in_range(0, end_of_lowmem, area_size,
332 					 PAGE_SIZE);
333 
334 	if (!ramdisk_here)
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 	memblock_reserve(ramdisk_here, area_size);
341 	initrd_start = ramdisk_here + PAGE_OFFSET;
342 	initrd_end   = initrd_start + ramdisk_size;
343 	printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
344 			 ramdisk_here, ramdisk_here + ramdisk_size);
345 
346 	q = (char *)initrd_start;
347 
348 	/* Copy any lowmem portion of the initrd */
349 	if (ramdisk_image < end_of_lowmem) {
350 		clen = end_of_lowmem - ramdisk_image;
351 		p = (char *)__va(ramdisk_image);
352 		memcpy(q, p, clen);
353 		q += clen;
354 		ramdisk_image += clen;
355 		ramdisk_size  -= clen;
356 	}
357 
358 	/* Copy the highmem portion of the initrd */
359 	while (ramdisk_size) {
360 		slop = ramdisk_image & ~PAGE_MASK;
361 		clen = ramdisk_size;
362 		if (clen > MAX_MAP_CHUNK-slop)
363 			clen = MAX_MAP_CHUNK-slop;
364 		mapaddr = ramdisk_image & PAGE_MASK;
365 		p = early_memremap(mapaddr, clen+slop);
366 		memcpy(q, p+slop, clen);
367 		early_iounmap(p, clen+slop);
368 		q += clen;
369 		ramdisk_image += clen;
370 		ramdisk_size  -= clen;
371 	}
372 	/* high pages is not converted by early_res_to_bootmem */
373 	ramdisk_image = boot_params.hdr.ramdisk_image;
374 	ramdisk_size  = boot_params.hdr.ramdisk_size;
375 	printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
376 		" %08llx - %08llx\n",
377 		ramdisk_image, ramdisk_image + ramdisk_size - 1,
378 		ramdisk_here, ramdisk_here + ramdisk_size - 1);
379 }
380 
381 static void __init reserve_initrd(void)
382 {
383 	/* Assume only end is not page aligned */
384 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
385 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
386 	u64 ramdisk_end   = PAGE_ALIGN(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 		memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
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 	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
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 		u32 data_len, map_len;
437 
438 		map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
439 			      (u64)sizeof(struct setup_data));
440 		data = early_memremap(pa_data, map_len);
441 		data_len = data->len + sizeof(struct setup_data);
442 		if (data_len > map_len) {
443 			early_iounmap(data, map_len);
444 			data = early_memremap(pa_data, data_len);
445 			map_len = data_len;
446 		}
447 
448 		switch (data->type) {
449 		case SETUP_E820_EXT:
450 			parse_e820_ext(data);
451 			break;
452 		case SETUP_DTB:
453 			add_dtb(pa_data);
454 			break;
455 		default:
456 			break;
457 		}
458 		pa_data = data->next;
459 		early_iounmap(data, map_len);
460 	}
461 }
462 
463 static void __init e820_reserve_setup_data(void)
464 {
465 	struct setup_data *data;
466 	u64 pa_data;
467 	int found = 0;
468 
469 	if (boot_params.hdr.version < 0x0209)
470 		return;
471 	pa_data = boot_params.hdr.setup_data;
472 	while (pa_data) {
473 		data = early_memremap(pa_data, sizeof(*data));
474 		e820_update_range(pa_data, sizeof(*data)+data->len,
475 			 E820_RAM, E820_RESERVED_KERN);
476 		found = 1;
477 		pa_data = data->next;
478 		early_iounmap(data, sizeof(*data));
479 	}
480 	if (!found)
481 		return;
482 
483 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
484 	memcpy(&e820_saved, &e820, sizeof(struct e820map));
485 	printk(KERN_INFO "extended physical RAM map:\n");
486 	e820_print_map("reserve setup_data");
487 }
488 
489 static void __init memblock_x86_reserve_range_setup_data(void)
490 {
491 	struct setup_data *data;
492 	u64 pa_data;
493 
494 	if (boot_params.hdr.version < 0x0209)
495 		return;
496 	pa_data = boot_params.hdr.setup_data;
497 	while (pa_data) {
498 		data = early_memremap(pa_data, sizeof(*data));
499 		memblock_reserve(pa_data, sizeof(*data) + data->len);
500 		pa_data = data->next;
501 		early_iounmap(data, sizeof(*data));
502 	}
503 }
504 
505 /*
506  * --------- Crashkernel reservation ------------------------------
507  */
508 
509 #ifdef CONFIG_KEXEC
510 
511 /*
512  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
513  * would limit the kernel to the low 512 MiB due to mapping restrictions.
514  * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
515  * limit once kexec-tools are fixed.
516  */
517 #ifdef CONFIG_X86_32
518 # define CRASH_KERNEL_ADDR_MAX	(512 << 20)
519 #else
520 # define CRASH_KERNEL_ADDR_MAX	(896 << 20)
521 #endif
522 
523 static void __init reserve_crashkernel(void)
524 {
525 	unsigned long long total_mem;
526 	unsigned long long crash_size, crash_base;
527 	int ret;
528 
529 	total_mem = memblock_phys_mem_size();
530 
531 	ret = parse_crashkernel(boot_command_line, total_mem,
532 			&crash_size, &crash_base);
533 	if (ret != 0 || crash_size <= 0)
534 		return;
535 
536 	/* 0 means: find the address automatically */
537 	if (crash_base <= 0) {
538 		const unsigned long long alignment = 16<<20;	/* 16M */
539 
540 		/*
541 		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
542 		 */
543 		crash_base = memblock_find_in_range(alignment,
544 			       CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
545 
546 		if (!crash_base) {
547 			pr_info("crashkernel reservation failed - No suitable area found.\n");
548 			return;
549 		}
550 	} else {
551 		unsigned long long start;
552 
553 		start = memblock_find_in_range(crash_base,
554 				 crash_base + crash_size, crash_size, 1<<20);
555 		if (start != crash_base) {
556 			pr_info("crashkernel reservation failed - memory is in use.\n");
557 			return;
558 		}
559 	}
560 	memblock_reserve(crash_base, crash_size);
561 
562 	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
563 			"for crashkernel (System RAM: %ldMB)\n",
564 			(unsigned long)(crash_size >> 20),
565 			(unsigned long)(crash_base >> 20),
566 			(unsigned long)(total_mem >> 20));
567 
568 	crashk_res.start = crash_base;
569 	crashk_res.end   = crash_base + crash_size - 1;
570 	insert_resource(&iomem_resource, &crashk_res);
571 }
572 #else
573 static void __init reserve_crashkernel(void)
574 {
575 }
576 #endif
577 
578 static struct resource standard_io_resources[] = {
579 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
580 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
581 	{ .name = "pic1", .start = 0x20, .end = 0x21,
582 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
583 	{ .name = "timer0", .start = 0x40, .end = 0x43,
584 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
585 	{ .name = "timer1", .start = 0x50, .end = 0x53,
586 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
587 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
588 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
589 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
590 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
591 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
592 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
593 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
594 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
595 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
596 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
597 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
598 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
599 };
600 
601 void __init reserve_standard_io_resources(void)
602 {
603 	int i;
604 
605 	/* request I/O space for devices used on all i[345]86 PCs */
606 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
607 		request_resource(&ioport_resource, &standard_io_resources[i]);
608 
609 }
610 
611 static __init void reserve_ibft_region(void)
612 {
613 	unsigned long addr, size = 0;
614 
615 	addr = find_ibft_region(&size);
616 
617 	if (size)
618 		memblock_reserve(addr, size);
619 }
620 
621 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
622 
623 static void __init trim_bios_range(void)
624 {
625 	/*
626 	 * A special case is the first 4Kb of memory;
627 	 * This is a BIOS owned area, not kernel ram, but generally
628 	 * not listed as such in the E820 table.
629 	 *
630 	 * This typically reserves additional memory (64KiB by default)
631 	 * since some BIOSes are known to corrupt low memory.  See the
632 	 * Kconfig help text for X86_RESERVE_LOW.
633 	 */
634 	e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
635 			  E820_RAM, E820_RESERVED);
636 
637 	/*
638 	 * special case: Some BIOSen report the PC BIOS
639 	 * area (640->1Mb) as ram even though it is not.
640 	 * take them out.
641 	 */
642 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
643 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
644 }
645 
646 static int __init parse_reservelow(char *p)
647 {
648 	unsigned long long size;
649 
650 	if (!p)
651 		return -EINVAL;
652 
653 	size = memparse(p, &p);
654 
655 	if (size < 4096)
656 		size = 4096;
657 
658 	if (size > 640*1024)
659 		size = 640*1024;
660 
661 	reserve_low = size;
662 
663 	return 0;
664 }
665 
666 early_param("reservelow", parse_reservelow);
667 
668 /*
669  * Determine if we were loaded by an EFI loader.  If so, then we have also been
670  * passed the efi memmap, systab, etc., so we should use these data structures
671  * for initialization.  Note, the efi init code path is determined by the
672  * global efi_enabled. This allows the same kernel image to be used on existing
673  * systems (with a traditional BIOS) as well as on EFI systems.
674  */
675 /*
676  * setup_arch - architecture-specific boot-time initializations
677  *
678  * Note: On x86_64, fixmaps are ready for use even before this is called.
679  */
680 
681 void __init setup_arch(char **cmdline_p)
682 {
683 #ifdef CONFIG_X86_32
684 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
685 	visws_early_detect();
686 
687 	/*
688 	 * copy kernel address range established so far and switch
689 	 * to the proper swapper page table
690 	 */
691 	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
692 			initial_page_table + KERNEL_PGD_BOUNDARY,
693 			KERNEL_PGD_PTRS);
694 
695 	load_cr3(swapper_pg_dir);
696 	__flush_tlb_all();
697 #else
698 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
699 #endif
700 
701 	/*
702 	 * If we have OLPC OFW, we might end up relocating the fixmap due to
703 	 * reserve_top(), so do this before touching the ioremap area.
704 	 */
705 	olpc_ofw_detect();
706 
707 	early_trap_init();
708 	early_cpu_init();
709 	early_ioremap_init();
710 
711 	setup_olpc_ofw_pgd();
712 
713 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
714 	screen_info = boot_params.screen_info;
715 	edid_info = boot_params.edid_info;
716 #ifdef CONFIG_X86_32
717 	apm_info.bios = boot_params.apm_bios_info;
718 	ist_info = boot_params.ist_info;
719 	if (boot_params.sys_desc_table.length != 0) {
720 		set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
721 		machine_id = boot_params.sys_desc_table.table[0];
722 		machine_submodel_id = boot_params.sys_desc_table.table[1];
723 		BIOS_revision = boot_params.sys_desc_table.table[2];
724 	}
725 #endif
726 	saved_video_mode = boot_params.hdr.vid_mode;
727 	bootloader_type = boot_params.hdr.type_of_loader;
728 	if ((bootloader_type >> 4) == 0xe) {
729 		bootloader_type &= 0xf;
730 		bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
731 	}
732 	bootloader_version  = bootloader_type & 0xf;
733 	bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
734 
735 #ifdef CONFIG_BLK_DEV_RAM
736 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
737 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
738 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
739 #endif
740 #ifdef CONFIG_EFI
741 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
742 		     "EL32", 4)) {
743 		efi_enabled = 1;
744 		efi_64bit = false;
745 	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
746 		     "EL64", 4)) {
747 		efi_enabled = 1;
748 		efi_64bit = true;
749 	}
750 	if (efi_enabled && efi_memblock_x86_reserve_range())
751 		efi_enabled = 0;
752 #endif
753 
754 	x86_init.oem.arch_setup();
755 
756 	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
757 	setup_memory_map();
758 	parse_setup_data();
759 	/* update the e820_saved too */
760 	e820_reserve_setup_data();
761 
762 	copy_edd();
763 
764 	if (!boot_params.hdr.root_flags)
765 		root_mountflags &= ~MS_RDONLY;
766 	init_mm.start_code = (unsigned long) _text;
767 	init_mm.end_code = (unsigned long) _etext;
768 	init_mm.end_data = (unsigned long) _edata;
769 	init_mm.brk = _brk_end;
770 
771 	code_resource.start = virt_to_phys(_text);
772 	code_resource.end = virt_to_phys(_etext)-1;
773 	data_resource.start = virt_to_phys(_etext);
774 	data_resource.end = virt_to_phys(_edata)-1;
775 	bss_resource.start = virt_to_phys(&__bss_start);
776 	bss_resource.end = virt_to_phys(&__bss_stop)-1;
777 
778 #ifdef CONFIG_CMDLINE_BOOL
779 #ifdef CONFIG_CMDLINE_OVERRIDE
780 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
781 #else
782 	if (builtin_cmdline[0]) {
783 		/* append boot loader cmdline to builtin */
784 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
785 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
786 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
787 	}
788 #endif
789 #endif
790 
791 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
792 	*cmdline_p = command_line;
793 
794 	/*
795 	 * x86_configure_nx() is called before parse_early_param() to detect
796 	 * whether hardware doesn't support NX (so that the early EHCI debug
797 	 * console setup can safely call set_fixmap()). It may then be called
798 	 * again from within noexec_setup() during parsing early parameters
799 	 * to honor the respective command line option.
800 	 */
801 	x86_configure_nx();
802 
803 	parse_early_param();
804 
805 	x86_report_nx();
806 
807 	/* after early param, so could get panic from serial */
808 	memblock_x86_reserve_range_setup_data();
809 
810 	if (acpi_mps_check()) {
811 #ifdef CONFIG_X86_LOCAL_APIC
812 		disable_apic = 1;
813 #endif
814 		setup_clear_cpu_cap(X86_FEATURE_APIC);
815 	}
816 
817 #ifdef CONFIG_PCI
818 	if (pci_early_dump_regs)
819 		early_dump_pci_devices();
820 #endif
821 
822 	finish_e820_parsing();
823 
824 	if (efi_enabled)
825 		efi_init();
826 
827 	dmi_scan_machine();
828 
829 	/*
830 	 * VMware detection requires dmi to be available, so this
831 	 * needs to be done after dmi_scan_machine, for the BP.
832 	 */
833 	init_hypervisor_platform();
834 
835 	x86_init.resources.probe_roms();
836 
837 	/* after parse_early_param, so could debug it */
838 	insert_resource(&iomem_resource, &code_resource);
839 	insert_resource(&iomem_resource, &data_resource);
840 	insert_resource(&iomem_resource, &bss_resource);
841 
842 	trim_bios_range();
843 #ifdef CONFIG_X86_32
844 	if (ppro_with_ram_bug()) {
845 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
846 				  E820_RESERVED);
847 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
848 		printk(KERN_INFO "fixed physical RAM map:\n");
849 		e820_print_map("bad_ppro");
850 	}
851 #else
852 	early_gart_iommu_check();
853 #endif
854 
855 	/*
856 	 * partially used pages are not usable - thus
857 	 * we are rounding upwards:
858 	 */
859 	max_pfn = e820_end_of_ram_pfn();
860 
861 	/* update e820 for memory not covered by WB MTRRs */
862 	mtrr_bp_init();
863 	if (mtrr_trim_uncached_memory(max_pfn))
864 		max_pfn = e820_end_of_ram_pfn();
865 
866 #ifdef CONFIG_X86_32
867 	/* max_low_pfn get updated here */
868 	find_low_pfn_range();
869 #else
870 	num_physpages = max_pfn;
871 
872 	check_x2apic();
873 
874 	/* How many end-of-memory variables you have, grandma! */
875 	/* need this before calling reserve_initrd */
876 	if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
877 		max_low_pfn = e820_end_of_low_ram_pfn();
878 	else
879 		max_low_pfn = max_pfn;
880 
881 	high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
882 #endif
883 
884 	/*
885 	 * Find and reserve possible boot-time SMP configuration:
886 	 */
887 	find_smp_config();
888 
889 	reserve_ibft_region();
890 
891 	/*
892 	 * Need to conclude brk, before memblock_x86_fill()
893 	 *  it could use memblock_find_in_range, could overlap with
894 	 *  brk area.
895 	 */
896 	reserve_brk();
897 
898 	cleanup_highmap();
899 
900 	memblock.current_limit = get_max_mapped();
901 	memblock_x86_fill();
902 
903 	/*
904 	 * The EFI specification says that boot service code won't be called
905 	 * after ExitBootServices(). This is, in fact, a lie.
906 	 */
907 	if (efi_enabled)
908 		efi_reserve_boot_services();
909 
910 	/* preallocate 4k for mptable mpc */
911 	early_reserve_e820_mpc_new();
912 
913 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
914 	setup_bios_corruption_check();
915 #endif
916 
917 	printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
918 			max_pfn_mapped<<PAGE_SHIFT);
919 
920 	setup_trampolines();
921 
922 	init_gbpages();
923 
924 	/* max_pfn_mapped is updated here */
925 	max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
926 	max_pfn_mapped = max_low_pfn_mapped;
927 
928 #ifdef CONFIG_X86_64
929 	if (max_pfn > max_low_pfn) {
930 		max_pfn_mapped = init_memory_mapping(1UL<<32,
931 						     max_pfn<<PAGE_SHIFT);
932 		/* can we preseve max_low_pfn ?*/
933 		max_low_pfn = max_pfn;
934 	}
935 #endif
936 	memblock.current_limit = get_max_mapped();
937 
938 	/*
939 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
940 	 */
941 
942 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
943 	if (init_ohci1394_dma_early)
944 		init_ohci1394_dma_on_all_controllers();
945 #endif
946 	/* Allocate bigger log buffer */
947 	setup_log_buf(1);
948 
949 	reserve_initrd();
950 
951 	reserve_crashkernel();
952 
953 	vsmp_init();
954 
955 	io_delay_init();
956 
957 	/*
958 	 * Parse the ACPI tables for possible boot-time SMP configuration.
959 	 */
960 	acpi_boot_table_init();
961 
962 	early_acpi_boot_init();
963 
964 	initmem_init();
965 	memblock_find_dma_reserve();
966 
967 #ifdef CONFIG_KVM_CLOCK
968 	kvmclock_init();
969 #endif
970 
971 	x86_init.paging.pagetable_setup_start(swapper_pg_dir);
972 	paging_init();
973 	x86_init.paging.pagetable_setup_done(swapper_pg_dir);
974 
975 	if (boot_cpu_data.cpuid_level >= 0) {
976 		/* A CPU has %cr4 if and only if it has CPUID */
977 		mmu_cr4_features = read_cr4();
978 	}
979 
980 #ifdef CONFIG_X86_32
981 	/* sync back kernel address range */
982 	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
983 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
984 			KERNEL_PGD_PTRS);
985 #endif
986 
987 	tboot_probe();
988 
989 #ifdef CONFIG_X86_64
990 	map_vsyscall();
991 #endif
992 
993 	generic_apic_probe();
994 
995 	early_quirks();
996 
997 	/*
998 	 * Read APIC and some other early information from ACPI tables.
999 	 */
1000 	acpi_boot_init();
1001 	sfi_init();
1002 	x86_dtb_init();
1003 
1004 	/*
1005 	 * get boot-time SMP configuration:
1006 	 */
1007 	if (smp_found_config)
1008 		get_smp_config();
1009 
1010 	prefill_possible_map();
1011 
1012 	init_cpu_to_node();
1013 
1014 	init_apic_mappings();
1015 	ioapic_and_gsi_init();
1016 
1017 	kvm_guest_init();
1018 
1019 	e820_reserve_resources();
1020 	e820_mark_nosave_regions(max_low_pfn);
1021 
1022 	x86_init.resources.reserve_resources();
1023 
1024 	e820_setup_gap();
1025 
1026 #ifdef CONFIG_VT
1027 #if defined(CONFIG_VGA_CONSOLE)
1028 	if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1029 		conswitchp = &vga_con;
1030 #elif defined(CONFIG_DUMMY_CONSOLE)
1031 	conswitchp = &dummy_con;
1032 #endif
1033 #endif
1034 	x86_init.oem.banner();
1035 
1036 	x86_init.timers.wallclock_init();
1037 
1038 	x86_platform.wallclock_init();
1039 
1040 	mcheck_init();
1041 
1042 	arch_init_ideal_nops();
1043 }
1044 
1045 #ifdef CONFIG_X86_32
1046 
1047 static struct resource video_ram_resource = {
1048 	.name	= "Video RAM area",
1049 	.start	= 0xa0000,
1050 	.end	= 0xbffff,
1051 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1052 };
1053 
1054 void __init i386_reserve_resources(void)
1055 {
1056 	request_resource(&iomem_resource, &video_ram_resource);
1057 	reserve_standard_io_resources();
1058 }
1059 
1060 #endif /* CONFIG_X86_32 */
1061