xref: /linux/arch/x86/kernel/setup.c (revision 0d456bad36d42d16022be045c8a53ddbb59ee478)
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 struct boot_params boot_params;
147 
148 /*
149  * Machine setup..
150  */
151 static struct resource data_resource = {
152 	.name	= "Kernel data",
153 	.start	= 0,
154 	.end	= 0,
155 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
156 };
157 
158 static struct resource code_resource = {
159 	.name	= "Kernel code",
160 	.start	= 0,
161 	.end	= 0,
162 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
163 };
164 
165 static struct resource bss_resource = {
166 	.name	= "Kernel bss",
167 	.start	= 0,
168 	.end	= 0,
169 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
170 };
171 
172 
173 #ifdef CONFIG_X86_32
174 /* cpu data as detected by the assembly code in head.S */
175 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
176 /* common cpu data for all cpus */
177 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
178 EXPORT_SYMBOL(boot_cpu_data);
179 
180 unsigned int def_to_bigsmp;
181 
182 /* for MCA, but anyone else can use it if they want */
183 unsigned int machine_id;
184 unsigned int machine_submodel_id;
185 unsigned int BIOS_revision;
186 
187 struct apm_info apm_info;
188 EXPORT_SYMBOL(apm_info);
189 
190 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
191 	defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
192 struct ist_info ist_info;
193 EXPORT_SYMBOL(ist_info);
194 #else
195 struct ist_info ist_info;
196 #endif
197 
198 #else
199 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
200 	.x86_phys_bits = MAX_PHYSMEM_BITS,
201 };
202 EXPORT_SYMBOL(boot_cpu_data);
203 #endif
204 
205 
206 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
207 unsigned long mmu_cr4_features;
208 #else
209 unsigned long mmu_cr4_features = X86_CR4_PAE;
210 #endif
211 
212 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
213 int bootloader_type, bootloader_version;
214 
215 /*
216  * Setup options
217  */
218 struct screen_info screen_info;
219 EXPORT_SYMBOL(screen_info);
220 struct edid_info edid_info;
221 EXPORT_SYMBOL_GPL(edid_info);
222 
223 extern int root_mountflags;
224 
225 unsigned long saved_video_mode;
226 
227 #define RAMDISK_IMAGE_START_MASK	0x07FF
228 #define RAMDISK_PROMPT_FLAG		0x8000
229 #define RAMDISK_LOAD_FLAG		0x4000
230 
231 static char __initdata command_line[COMMAND_LINE_SIZE];
232 #ifdef CONFIG_CMDLINE_BOOL
233 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
234 #endif
235 
236 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
237 struct edd edd;
238 #ifdef CONFIG_EDD_MODULE
239 EXPORT_SYMBOL(edd);
240 #endif
241 /**
242  * copy_edd() - Copy the BIOS EDD information
243  *              from boot_params into a safe place.
244  *
245  */
246 static inline void __init copy_edd(void)
247 {
248      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
249 	    sizeof(edd.mbr_signature));
250      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
251      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
252      edd.edd_info_nr = boot_params.eddbuf_entries;
253 }
254 #else
255 static inline void __init copy_edd(void)
256 {
257 }
258 #endif
259 
260 void * __init extend_brk(size_t size, size_t align)
261 {
262 	size_t mask = align - 1;
263 	void *ret;
264 
265 	BUG_ON(_brk_start == 0);
266 	BUG_ON(align & mask);
267 
268 	_brk_end = (_brk_end + mask) & ~mask;
269 	BUG_ON((char *)(_brk_end + size) > __brk_limit);
270 
271 	ret = (void *)_brk_end;
272 	_brk_end += size;
273 
274 	memset(ret, 0, size);
275 
276 	return ret;
277 }
278 
279 #ifdef CONFIG_X86_64
280 static void __init init_gbpages(void)
281 {
282 	if (direct_gbpages && cpu_has_gbpages)
283 		printk(KERN_INFO "Using GB pages for direct mapping\n");
284 	else
285 		direct_gbpages = 0;
286 }
287 #else
288 static inline void init_gbpages(void)
289 {
290 }
291 static void __init cleanup_highmap(void)
292 {
293 }
294 #endif
295 
296 static void __init reserve_brk(void)
297 {
298 	if (_brk_end > _brk_start)
299 		memblock_reserve(__pa(_brk_start),
300 				 __pa(_brk_end) - __pa(_brk_start));
301 
302 	/* Mark brk area as locked down and no longer taking any
303 	   new allocations */
304 	_brk_start = 0;
305 }
306 
307 #ifdef CONFIG_BLK_DEV_INITRD
308 
309 #define MAX_MAP_CHUNK	(NR_FIX_BTMAPS << PAGE_SHIFT)
310 static void __init relocate_initrd(void)
311 {
312 	/* Assume only end is not page aligned */
313 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
314 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
315 	u64 area_size     = PAGE_ALIGN(ramdisk_size);
316 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
317 	u64 ramdisk_here;
318 	unsigned long slop, clen, mapaddr;
319 	char *p, *q;
320 
321 	/* We need to move the initrd down into lowmem */
322 	ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
323 					 PAGE_SIZE);
324 
325 	if (!ramdisk_here)
326 		panic("Cannot find place for new RAMDISK of size %lld\n",
327 			 ramdisk_size);
328 
329 	/* Note: this includes all the lowmem currently occupied by
330 	   the initrd, we rely on that fact to keep the data intact. */
331 	memblock_reserve(ramdisk_here, area_size);
332 	initrd_start = ramdisk_here + PAGE_OFFSET;
333 	initrd_end   = initrd_start + ramdisk_size;
334 	printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
335 			 ramdisk_here, ramdisk_here + ramdisk_size - 1);
336 
337 	q = (char *)initrd_start;
338 
339 	/* Copy any lowmem portion of the initrd */
340 	if (ramdisk_image < end_of_lowmem) {
341 		clen = end_of_lowmem - ramdisk_image;
342 		p = (char *)__va(ramdisk_image);
343 		memcpy(q, p, clen);
344 		q += clen;
345 		ramdisk_image += clen;
346 		ramdisk_size  -= clen;
347 	}
348 
349 	/* Copy the highmem portion of the initrd */
350 	while (ramdisk_size) {
351 		slop = ramdisk_image & ~PAGE_MASK;
352 		clen = ramdisk_size;
353 		if (clen > MAX_MAP_CHUNK-slop)
354 			clen = MAX_MAP_CHUNK-slop;
355 		mapaddr = ramdisk_image & PAGE_MASK;
356 		p = early_memremap(mapaddr, clen+slop);
357 		memcpy(q, p+slop, clen);
358 		early_iounmap(p, clen+slop);
359 		q += clen;
360 		ramdisk_image += clen;
361 		ramdisk_size  -= clen;
362 	}
363 	/* high pages is not converted by early_res_to_bootmem */
364 	ramdisk_image = boot_params.hdr.ramdisk_image;
365 	ramdisk_size  = boot_params.hdr.ramdisk_size;
366 	printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
367 		" [mem %#010llx-%#010llx]\n",
368 		ramdisk_image, ramdisk_image + ramdisk_size - 1,
369 		ramdisk_here, ramdisk_here + ramdisk_size - 1);
370 }
371 
372 static void __init reserve_initrd(void)
373 {
374 	/* Assume only end is not page aligned */
375 	u64 ramdisk_image = boot_params.hdr.ramdisk_image;
376 	u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
377 	u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
378 	u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
379 
380 	if (!boot_params.hdr.type_of_loader ||
381 	    !ramdisk_image || !ramdisk_size)
382 		return;		/* No initrd provided by bootloader */
383 
384 	initrd_start = 0;
385 
386 	if (ramdisk_size >= (end_of_lowmem>>1)) {
387 		panic("initrd too large to handle, "
388 		       "disabling initrd (%lld needed, %lld available)\n",
389 		       ramdisk_size, end_of_lowmem>>1);
390 	}
391 
392 	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
393 			ramdisk_end - 1);
394 
395 
396 	if (ramdisk_end <= end_of_lowmem) {
397 		/* All in lowmem, easy case */
398 		/*
399 		 * don't need to reserve again, already reserved early
400 		 * in i386_start_kernel
401 		 */
402 		initrd_start = ramdisk_image + PAGE_OFFSET;
403 		initrd_end = initrd_start + ramdisk_size;
404 		return;
405 	}
406 
407 	relocate_initrd();
408 
409 	memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
410 }
411 #else
412 static void __init reserve_initrd(void)
413 {
414 }
415 #endif /* CONFIG_BLK_DEV_INITRD */
416 
417 static void __init parse_setup_data(void)
418 {
419 	struct setup_data *data;
420 	u64 pa_data;
421 
422 	if (boot_params.hdr.version < 0x0209)
423 		return;
424 	pa_data = boot_params.hdr.setup_data;
425 	while (pa_data) {
426 		u32 data_len, map_len;
427 
428 		map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
429 			      (u64)sizeof(struct setup_data));
430 		data = early_memremap(pa_data, map_len);
431 		data_len = data->len + sizeof(struct setup_data);
432 		if (data_len > map_len) {
433 			early_iounmap(data, map_len);
434 			data = early_memremap(pa_data, data_len);
435 			map_len = data_len;
436 		}
437 
438 		switch (data->type) {
439 		case SETUP_E820_EXT:
440 			parse_e820_ext(data);
441 			break;
442 		case SETUP_DTB:
443 			add_dtb(pa_data);
444 			break;
445 		default:
446 			break;
447 		}
448 		pa_data = data->next;
449 		early_iounmap(data, map_len);
450 	}
451 }
452 
453 static void __init e820_reserve_setup_data(void)
454 {
455 	struct setup_data *data;
456 	u64 pa_data;
457 	int found = 0;
458 
459 	if (boot_params.hdr.version < 0x0209)
460 		return;
461 	pa_data = boot_params.hdr.setup_data;
462 	while (pa_data) {
463 		data = early_memremap(pa_data, sizeof(*data));
464 		e820_update_range(pa_data, sizeof(*data)+data->len,
465 			 E820_RAM, E820_RESERVED_KERN);
466 		found = 1;
467 		pa_data = data->next;
468 		early_iounmap(data, sizeof(*data));
469 	}
470 	if (!found)
471 		return;
472 
473 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
474 	memcpy(&e820_saved, &e820, sizeof(struct e820map));
475 	printk(KERN_INFO "extended physical RAM map:\n");
476 	e820_print_map("reserve setup_data");
477 }
478 
479 static void __init memblock_x86_reserve_range_setup_data(void)
480 {
481 	struct setup_data *data;
482 	u64 pa_data;
483 
484 	if (boot_params.hdr.version < 0x0209)
485 		return;
486 	pa_data = boot_params.hdr.setup_data;
487 	while (pa_data) {
488 		data = early_memremap(pa_data, sizeof(*data));
489 		memblock_reserve(pa_data, sizeof(*data) + data->len);
490 		pa_data = data->next;
491 		early_iounmap(data, sizeof(*data));
492 	}
493 }
494 
495 /*
496  * --------- Crashkernel reservation ------------------------------
497  */
498 
499 #ifdef CONFIG_KEXEC
500 
501 /*
502  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
503  * would limit the kernel to the low 512 MiB due to mapping restrictions.
504  * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
505  * limit once kexec-tools are fixed.
506  */
507 #ifdef CONFIG_X86_32
508 # define CRASH_KERNEL_ADDR_MAX	(512 << 20)
509 #else
510 # define CRASH_KERNEL_ADDR_MAX	(896 << 20)
511 #endif
512 
513 static void __init reserve_crashkernel(void)
514 {
515 	unsigned long long total_mem;
516 	unsigned long long crash_size, crash_base;
517 	int ret;
518 
519 	total_mem = memblock_phys_mem_size();
520 
521 	ret = parse_crashkernel(boot_command_line, total_mem,
522 			&crash_size, &crash_base);
523 	if (ret != 0 || crash_size <= 0)
524 		return;
525 
526 	/* 0 means: find the address automatically */
527 	if (crash_base <= 0) {
528 		const unsigned long long alignment = 16<<20;	/* 16M */
529 
530 		/*
531 		 *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
532 		 */
533 		crash_base = memblock_find_in_range(alignment,
534 			       CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
535 
536 		if (!crash_base) {
537 			pr_info("crashkernel reservation failed - No suitable area found.\n");
538 			return;
539 		}
540 	} else {
541 		unsigned long long start;
542 
543 		start = memblock_find_in_range(crash_base,
544 				 crash_base + crash_size, crash_size, 1<<20);
545 		if (start != crash_base) {
546 			pr_info("crashkernel reservation failed - memory is in use.\n");
547 			return;
548 		}
549 	}
550 	memblock_reserve(crash_base, crash_size);
551 
552 	printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
553 			"for crashkernel (System RAM: %ldMB)\n",
554 			(unsigned long)(crash_size >> 20),
555 			(unsigned long)(crash_base >> 20),
556 			(unsigned long)(total_mem >> 20));
557 
558 	crashk_res.start = crash_base;
559 	crashk_res.end   = crash_base + crash_size - 1;
560 	insert_resource(&iomem_resource, &crashk_res);
561 }
562 #else
563 static void __init reserve_crashkernel(void)
564 {
565 }
566 #endif
567 
568 static struct resource standard_io_resources[] = {
569 	{ .name = "dma1", .start = 0x00, .end = 0x1f,
570 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
571 	{ .name = "pic1", .start = 0x20, .end = 0x21,
572 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
573 	{ .name = "timer0", .start = 0x40, .end = 0x43,
574 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
575 	{ .name = "timer1", .start = 0x50, .end = 0x53,
576 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
577 	{ .name = "keyboard", .start = 0x60, .end = 0x60,
578 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
579 	{ .name = "keyboard", .start = 0x64, .end = 0x64,
580 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
581 	{ .name = "dma page reg", .start = 0x80, .end = 0x8f,
582 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
583 	{ .name = "pic2", .start = 0xa0, .end = 0xa1,
584 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
585 	{ .name = "dma2", .start = 0xc0, .end = 0xdf,
586 		.flags = IORESOURCE_BUSY | IORESOURCE_IO },
587 	{ .name = "fpu", .start = 0xf0, .end = 0xff,
588 		.flags = IORESOURCE_BUSY | IORESOURCE_IO }
589 };
590 
591 void __init reserve_standard_io_resources(void)
592 {
593 	int i;
594 
595 	/* request I/O space for devices used on all i[345]86 PCs */
596 	for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
597 		request_resource(&ioport_resource, &standard_io_resources[i]);
598 
599 }
600 
601 static __init void reserve_ibft_region(void)
602 {
603 	unsigned long addr, size = 0;
604 
605 	addr = find_ibft_region(&size);
606 
607 	if (size)
608 		memblock_reserve(addr, size);
609 }
610 
611 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
612 
613 static void __init trim_bios_range(void)
614 {
615 	/*
616 	 * A special case is the first 4Kb of memory;
617 	 * This is a BIOS owned area, not kernel ram, but generally
618 	 * not listed as such in the E820 table.
619 	 *
620 	 * This typically reserves additional memory (64KiB by default)
621 	 * since some BIOSes are known to corrupt low memory.  See the
622 	 * Kconfig help text for X86_RESERVE_LOW.
623 	 */
624 	e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
625 			  E820_RAM, E820_RESERVED);
626 
627 	/*
628 	 * special case: Some BIOSen report the PC BIOS
629 	 * area (640->1Mb) as ram even though it is not.
630 	 * take them out.
631 	 */
632 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
633 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
634 }
635 
636 static int __init parse_reservelow(char *p)
637 {
638 	unsigned long long size;
639 
640 	if (!p)
641 		return -EINVAL;
642 
643 	size = memparse(p, &p);
644 
645 	if (size < 4096)
646 		size = 4096;
647 
648 	if (size > 640*1024)
649 		size = 640*1024;
650 
651 	reserve_low = size;
652 
653 	return 0;
654 }
655 
656 early_param("reservelow", parse_reservelow);
657 
658 /*
659  * Determine if we were loaded by an EFI loader.  If so, then we have also been
660  * passed the efi memmap, systab, etc., so we should use these data structures
661  * for initialization.  Note, the efi init code path is determined by the
662  * global efi_enabled. This allows the same kernel image to be used on existing
663  * systems (with a traditional BIOS) as well as on EFI systems.
664  */
665 /*
666  * setup_arch - architecture-specific boot-time initializations
667  *
668  * Note: On x86_64, fixmaps are ready for use even before this is called.
669  */
670 
671 void __init setup_arch(char **cmdline_p)
672 {
673 #ifdef CONFIG_X86_32
674 	memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
675 	visws_early_detect();
676 
677 	/*
678 	 * copy kernel address range established so far and switch
679 	 * to the proper swapper page table
680 	 */
681 	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
682 			initial_page_table + KERNEL_PGD_BOUNDARY,
683 			KERNEL_PGD_PTRS);
684 
685 	load_cr3(swapper_pg_dir);
686 	__flush_tlb_all();
687 #else
688 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
689 #endif
690 
691 	/*
692 	 * If we have OLPC OFW, we might end up relocating the fixmap due to
693 	 * reserve_top(), so do this before touching the ioremap area.
694 	 */
695 	olpc_ofw_detect();
696 
697 	early_trap_init();
698 	early_cpu_init();
699 	early_ioremap_init();
700 
701 	setup_olpc_ofw_pgd();
702 
703 	ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
704 	screen_info = boot_params.screen_info;
705 	edid_info = boot_params.edid_info;
706 #ifdef CONFIG_X86_32
707 	apm_info.bios = boot_params.apm_bios_info;
708 	ist_info = boot_params.ist_info;
709 	if (boot_params.sys_desc_table.length != 0) {
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 		     "EL32", 4)) {
732 		efi_enabled = 1;
733 		efi_64bit = false;
734 	} else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
735 		     "EL64", 4)) {
736 		efi_enabled = 1;
737 		efi_64bit = true;
738 	}
739 	if (efi_enabled && efi_memblock_x86_reserve_range())
740 		efi_enabled = 0;
741 #endif
742 
743 	x86_init.oem.arch_setup();
744 
745 	iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
746 	setup_memory_map();
747 	parse_setup_data();
748 	/* update the e820_saved too */
749 	e820_reserve_setup_data();
750 
751 	copy_edd();
752 
753 	if (!boot_params.hdr.root_flags)
754 		root_mountflags &= ~MS_RDONLY;
755 	init_mm.start_code = (unsigned long) _text;
756 	init_mm.end_code = (unsigned long) _etext;
757 	init_mm.end_data = (unsigned long) _edata;
758 	init_mm.brk = _brk_end;
759 
760 	code_resource.start = virt_to_phys(_text);
761 	code_resource.end = virt_to_phys(_etext)-1;
762 	data_resource.start = virt_to_phys(_etext);
763 	data_resource.end = virt_to_phys(_edata)-1;
764 	bss_resource.start = virt_to_phys(&__bss_start);
765 	bss_resource.end = virt_to_phys(&__bss_stop)-1;
766 
767 #ifdef CONFIG_CMDLINE_BOOL
768 #ifdef CONFIG_CMDLINE_OVERRIDE
769 	strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
770 #else
771 	if (builtin_cmdline[0]) {
772 		/* append boot loader cmdline to builtin */
773 		strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
774 		strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
775 		strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
776 	}
777 #endif
778 #endif
779 
780 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
781 	*cmdline_p = command_line;
782 
783 	/*
784 	 * x86_configure_nx() is called before parse_early_param() to detect
785 	 * whether hardware doesn't support NX (so that the early EHCI debug
786 	 * console setup can safely call set_fixmap()). It may then be called
787 	 * again from within noexec_setup() during parsing early parameters
788 	 * to honor the respective command line option.
789 	 */
790 	x86_configure_nx();
791 
792 	parse_early_param();
793 
794 	x86_report_nx();
795 
796 	/* after early param, so could get panic from serial */
797 	memblock_x86_reserve_range_setup_data();
798 
799 	if (acpi_mps_check()) {
800 #ifdef CONFIG_X86_LOCAL_APIC
801 		disable_apic = 1;
802 #endif
803 		setup_clear_cpu_cap(X86_FEATURE_APIC);
804 	}
805 
806 #ifdef CONFIG_PCI
807 	if (pci_early_dump_regs)
808 		early_dump_pci_devices();
809 #endif
810 
811 	finish_e820_parsing();
812 
813 	if (efi_enabled)
814 		efi_init();
815 
816 	dmi_scan_machine();
817 
818 	/*
819 	 * VMware detection requires dmi to be available, so this
820 	 * needs to be done after dmi_scan_machine, for the BP.
821 	 */
822 	init_hypervisor_platform();
823 
824 	x86_init.resources.probe_roms();
825 
826 	/* after parse_early_param, so could debug it */
827 	insert_resource(&iomem_resource, &code_resource);
828 	insert_resource(&iomem_resource, &data_resource);
829 	insert_resource(&iomem_resource, &bss_resource);
830 
831 	trim_bios_range();
832 #ifdef CONFIG_X86_32
833 	if (ppro_with_ram_bug()) {
834 		e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
835 				  E820_RESERVED);
836 		sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
837 		printk(KERN_INFO "fixed physical RAM map:\n");
838 		e820_print_map("bad_ppro");
839 	}
840 #else
841 	early_gart_iommu_check();
842 #endif
843 
844 	/*
845 	 * partially used pages are not usable - thus
846 	 * we are rounding upwards:
847 	 */
848 	max_pfn = e820_end_of_ram_pfn();
849 
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 #endif
872 
873 	/*
874 	 * Find and reserve possible boot-time SMP configuration:
875 	 */
876 	find_smp_config();
877 
878 	reserve_ibft_region();
879 
880 	/*
881 	 * Need to conclude brk, before memblock_x86_fill()
882 	 *  it could use memblock_find_in_range, could overlap with
883 	 *  brk area.
884 	 */
885 	reserve_brk();
886 
887 	cleanup_highmap();
888 
889 	memblock.current_limit = get_max_mapped();
890 	memblock_x86_fill();
891 
892 	/*
893 	 * The EFI specification says that boot service code won't be called
894 	 * after ExitBootServices(). This is, in fact, a lie.
895 	 */
896 	if (efi_enabled)
897 		efi_reserve_boot_services();
898 
899 	/* preallocate 4k for mptable mpc */
900 	early_reserve_e820_mpc_new();
901 
902 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
903 	setup_bios_corruption_check();
904 #endif
905 
906 	printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
907 			(max_pfn_mapped<<PAGE_SHIFT) - 1);
908 
909 	setup_real_mode();
910 
911 	init_gbpages();
912 
913 	/* max_pfn_mapped is updated here */
914 	max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
915 	max_pfn_mapped = max_low_pfn_mapped;
916 
917 #ifdef CONFIG_X86_64
918 	if (max_pfn > max_low_pfn) {
919 		int i;
920 		unsigned long start, end;
921 		unsigned long start_pfn, end_pfn;
922 
923 		for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn,
924 							 NULL) {
925 
926 			end = PFN_PHYS(end_pfn);
927 			if (end <= (1UL<<32))
928 				continue;
929 
930 			start = PFN_PHYS(start_pfn);
931 			max_pfn_mapped = init_memory_mapping(
932 						max((1UL<<32), start), end);
933 		}
934 
935 		/* can we preseve max_low_pfn ?*/
936 		max_low_pfn = max_pfn;
937 	}
938 #endif
939 	memblock.current_limit = get_max_mapped();
940 	dma_contiguous_reserve(0);
941 
942 	/*
943 	 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
944 	 */
945 
946 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
947 	if (init_ohci1394_dma_early)
948 		init_ohci1394_dma_on_all_controllers();
949 #endif
950 	/* Allocate bigger log buffer */
951 	setup_log_buf(1);
952 
953 	reserve_initrd();
954 
955 #if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
956 	acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
957 #endif
958 
959 	reserve_crashkernel();
960 
961 	vsmp_init();
962 
963 	io_delay_init();
964 
965 	/*
966 	 * Parse the ACPI tables for possible boot-time SMP configuration.
967 	 */
968 	acpi_boot_table_init();
969 
970 	early_acpi_boot_init();
971 
972 	initmem_init();
973 	memblock_find_dma_reserve();
974 
975 #ifdef CONFIG_KVM_GUEST
976 	kvmclock_init();
977 #endif
978 
979 	x86_init.paging.pagetable_init();
980 
981 	if (boot_cpu_data.cpuid_level >= 0) {
982 		/* A CPU has %cr4 if and only if it has CPUID */
983 		mmu_cr4_features = read_cr4();
984 		if (trampoline_cr4_features)
985 			*trampoline_cr4_features = mmu_cr4_features;
986 	}
987 
988 #ifdef CONFIG_X86_32
989 	/* sync back kernel address range */
990 	clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
991 			swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
992 			KERNEL_PGD_PTRS);
993 #endif
994 
995 	tboot_probe();
996 
997 #ifdef CONFIG_X86_64
998 	map_vsyscall();
999 #endif
1000 
1001 	generic_apic_probe();
1002 
1003 	early_quirks();
1004 
1005 	/*
1006 	 * Read APIC and some other early information from ACPI tables.
1007 	 */
1008 	acpi_boot_init();
1009 	sfi_init();
1010 	x86_dtb_init();
1011 
1012 	/*
1013 	 * get boot-time SMP configuration:
1014 	 */
1015 	if (smp_found_config)
1016 		get_smp_config();
1017 
1018 	prefill_possible_map();
1019 
1020 	init_cpu_to_node();
1021 
1022 	init_apic_mappings();
1023 	if (x86_io_apic_ops.init)
1024 		x86_io_apic_ops.init();
1025 
1026 	kvm_guest_init();
1027 
1028 	e820_reserve_resources();
1029 	e820_mark_nosave_regions(max_low_pfn);
1030 
1031 	x86_init.resources.reserve_resources();
1032 
1033 	e820_setup_gap();
1034 
1035 #ifdef CONFIG_VT
1036 #if defined(CONFIG_VGA_CONSOLE)
1037 	if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1038 		conswitchp = &vga_con;
1039 #elif defined(CONFIG_DUMMY_CONSOLE)
1040 	conswitchp = &dummy_con;
1041 #endif
1042 #endif
1043 	x86_init.oem.banner();
1044 
1045 	x86_init.timers.wallclock_init();
1046 
1047 	mcheck_init();
1048 
1049 	arch_init_ideal_nops();
1050 
1051 	register_refined_jiffies(CLOCK_TICK_RATE);
1052 
1053 #ifdef CONFIG_EFI
1054 	/* Once setup is done above, disable efi_enabled on mismatched
1055 	 * firmware/kernel archtectures since there is no support for
1056 	 * runtime services.
1057 	 */
1058 	if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) {
1059 		pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
1060 		efi_unmap_memmap();
1061 		efi_enabled = 0;
1062 	}
1063 #endif
1064 }
1065 
1066 #ifdef CONFIG_X86_32
1067 
1068 static struct resource video_ram_resource = {
1069 	.name	= "Video RAM area",
1070 	.start	= 0xa0000,
1071 	.end	= 0xbffff,
1072 	.flags	= IORESOURCE_BUSY | IORESOURCE_MEM
1073 };
1074 
1075 void __init i386_reserve_resources(void)
1076 {
1077 	request_resource(&iomem_resource, &video_ram_resource);
1078 	reserve_standard_io_resources();
1079 }
1080 
1081 #endif /* CONFIG_X86_32 */
1082