]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/blob - arch/x86/kernel/setup.c
x86: Make 64 bit use early_res instead of bootmem before slab
[mirror_ubuntu-focal-kernel.git] / arch / x86 / kernel / setup.c
1 /*
2 * Copyright (C) 1995 Linus Torvalds
3 *
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
5 *
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
8 *
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
11 *
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
14 *
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
17 *
18 */
19
20 /*
21 * This file handles the architecture-dependent parts of initialization
22 */
23
24 #include <linux/sched.h>
25 #include <linux/mm.h>
26 #include <linux/mmzone.h>
27 #include <linux/screen_info.h>
28 #include <linux/ioport.h>
29 #include <linux/acpi.h>
30 #include <linux/sfi.h>
31 #include <linux/apm_bios.h>
32 #include <linux/initrd.h>
33 #include <linux/bootmem.h>
34 #include <linux/seq_file.h>
35 #include <linux/console.h>
36 #include <linux/mca.h>
37 #include <linux/root_dev.h>
38 #include <linux/highmem.h>
39 #include <linux/module.h>
40 #include <linux/efi.h>
41 #include <linux/init.h>
42 #include <linux/edd.h>
43 #include <linux/iscsi_ibft.h>
44 #include <linux/nodemask.h>
45 #include <linux/kexec.h>
46 #include <linux/dmi.h>
47 #include <linux/pfn.h>
48 #include <linux/pci.h>
49 #include <asm/pci-direct.h>
50 #include <linux/init_ohci1394_dma.h>
51 #include <linux/kvm_para.h>
52
53 #include <linux/errno.h>
54 #include <linux/kernel.h>
55 #include <linux/stddef.h>
56 #include <linux/unistd.h>
57 #include <linux/ptrace.h>
58 #include <linux/slab.h>
59 #include <linux/user.h>
60 #include <linux/delay.h>
61
62 #include <linux/kallsyms.h>
63 #include <linux/cpufreq.h>
64 #include <linux/dma-mapping.h>
65 #include <linux/ctype.h>
66 #include <linux/uaccess.h>
67
68 #include <linux/percpu.h>
69 #include <linux/crash_dump.h>
70 #include <linux/tboot.h>
71
72 #include <video/edid.h>
73
74 #include <asm/mtrr.h>
75 #include <asm/apic.h>
76 #include <asm/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/vmi.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/system.h>
95 #include <asm/vsyscall.h>
96 #include <asm/cpu.h>
97 #include <asm/desc.h>
98 #include <asm/dma.h>
99 #include <asm/iommu.h>
100 #include <asm/gart.h>
101 #include <asm/mmu_context.h>
102 #include <asm/proto.h>
103
104 #include <asm/paravirt.h>
105 #include <asm/hypervisor.h>
106
107 #include <asm/percpu.h>
108 #include <asm/topology.h>
109 #include <asm/apicdef.h>
110 #include <asm/k8.h>
111 #ifdef CONFIG_X86_64
112 #include <asm/numa_64.h>
113 #endif
114 #include <asm/mce.h>
115
116 /*
117 * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
118 * The direct mapping extends to max_pfn_mapped, so that we can directly access
119 * apertures, ACPI and other tables without having to play with fixmaps.
120 */
121 unsigned long max_low_pfn_mapped;
122 unsigned long max_pfn_mapped;
123
124 RESERVE_BRK(dmi_alloc, 65536);
125
126 unsigned int boot_cpu_id __read_mostly;
127
128 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
129 unsigned long _brk_end = (unsigned long)__brk_base;
130
131 #ifdef CONFIG_X86_64
132 int default_cpu_present_to_apicid(int mps_cpu)
133 {
134 return __default_cpu_present_to_apicid(mps_cpu);
135 }
136
137 int default_check_phys_apicid_present(int phys_apicid)
138 {
139 return __default_check_phys_apicid_present(phys_apicid);
140 }
141 #endif
142
143 #ifndef CONFIG_DEBUG_BOOT_PARAMS
144 struct boot_params __initdata boot_params;
145 #else
146 struct boot_params boot_params;
147 #endif
148
149 /*
150 * Machine setup..
151 */
152 static struct resource data_resource = {
153 .name = "Kernel data",
154 .start = 0,
155 .end = 0,
156 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
157 };
158
159 static struct resource code_resource = {
160 .name = "Kernel code",
161 .start = 0,
162 .end = 0,
163 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
164 };
165
166 static struct resource bss_resource = {
167 .name = "Kernel bss",
168 .start = 0,
169 .end = 0,
170 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
171 };
172
173
174 #ifdef CONFIG_X86_32
175 /* cpu data as detected by the assembly code in head.S */
176 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
177 /* common cpu data for all cpus */
178 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
179 EXPORT_SYMBOL(boot_cpu_data);
180 static void set_mca_bus(int x)
181 {
182 #ifdef CONFIG_MCA
183 MCA_bus = x;
184 #endif
185 }
186
187 unsigned int def_to_bigsmp;
188
189 /* for MCA, but anyone else can use it if they want */
190 unsigned int machine_id;
191 unsigned int machine_submodel_id;
192 unsigned int BIOS_revision;
193
194 struct apm_info apm_info;
195 EXPORT_SYMBOL(apm_info);
196
197 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
198 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
199 struct ist_info ist_info;
200 EXPORT_SYMBOL(ist_info);
201 #else
202 struct ist_info ist_info;
203 #endif
204
205 #else
206 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
207 .x86_phys_bits = MAX_PHYSMEM_BITS,
208 };
209 EXPORT_SYMBOL(boot_cpu_data);
210 #endif
211
212
213 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
214 unsigned long mmu_cr4_features;
215 #else
216 unsigned long mmu_cr4_features = X86_CR4_PAE;
217 #endif
218
219 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
220 int bootloader_type, bootloader_version;
221
222 /*
223 * Setup options
224 */
225 struct screen_info screen_info;
226 EXPORT_SYMBOL(screen_info);
227 struct edid_info edid_info;
228 EXPORT_SYMBOL_GPL(edid_info);
229
230 extern int root_mountflags;
231
232 unsigned long saved_video_mode;
233
234 #define RAMDISK_IMAGE_START_MASK 0x07FF
235 #define RAMDISK_PROMPT_FLAG 0x8000
236 #define RAMDISK_LOAD_FLAG 0x4000
237
238 static char __initdata command_line[COMMAND_LINE_SIZE];
239 #ifdef CONFIG_CMDLINE_BOOL
240 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
241 #endif
242
243 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
244 struct edd edd;
245 #ifdef CONFIG_EDD_MODULE
246 EXPORT_SYMBOL(edd);
247 #endif
248 /**
249 * copy_edd() - Copy the BIOS EDD information
250 * from boot_params into a safe place.
251 *
252 */
253 static inline void __init copy_edd(void)
254 {
255 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
256 sizeof(edd.mbr_signature));
257 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
258 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
259 edd.edd_info_nr = boot_params.eddbuf_entries;
260 }
261 #else
262 static inline void __init copy_edd(void)
263 {
264 }
265 #endif
266
267 void * __init extend_brk(size_t size, size_t align)
268 {
269 size_t mask = align - 1;
270 void *ret;
271
272 BUG_ON(_brk_start == 0);
273 BUG_ON(align & mask);
274
275 _brk_end = (_brk_end + mask) & ~mask;
276 BUG_ON((char *)(_brk_end + size) > __brk_limit);
277
278 ret = (void *)_brk_end;
279 _brk_end += size;
280
281 memset(ret, 0, size);
282
283 return ret;
284 }
285
286 #ifdef CONFIG_X86_64
287 static void __init init_gbpages(void)
288 {
289 if (direct_gbpages && cpu_has_gbpages)
290 printk(KERN_INFO "Using GB pages for direct mapping\n");
291 else
292 direct_gbpages = 0;
293 }
294 #else
295 static inline void init_gbpages(void)
296 {
297 }
298 #endif
299
300 static void __init reserve_brk(void)
301 {
302 if (_brk_end > _brk_start)
303 reserve_early(__pa(_brk_start), __pa(_brk_end), "BRK");
304
305 /* Mark brk area as locked down and no longer taking any
306 new allocations */
307 _brk_start = 0;
308 }
309
310 #ifdef CONFIG_BLK_DEV_INITRD
311
312 #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
313 static void __init relocate_initrd(void)
314 {
315
316 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
317 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
318 u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
319 u64 ramdisk_here;
320 unsigned long slop, clen, mapaddr;
321 char *p, *q;
322
323 /* We need to move the initrd down into lowmem */
324 ramdisk_here = find_e820_area(0, end_of_lowmem, ramdisk_size,
325 PAGE_SIZE);
326
327 if (ramdisk_here == -1ULL)
328 panic("Cannot find place for new RAMDISK of size %lld\n",
329 ramdisk_size);
330
331 /* Note: this includes all the lowmem currently occupied by
332 the initrd, we rely on that fact to keep the data intact. */
333 reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
334 "NEW RAMDISK");
335 initrd_start = ramdisk_here + PAGE_OFFSET;
336 initrd_end = initrd_start + ramdisk_size;
337 printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
338 ramdisk_here, ramdisk_here + ramdisk_size);
339
340 q = (char *)initrd_start;
341
342 /* Copy any lowmem portion of the initrd */
343 if (ramdisk_image < end_of_lowmem) {
344 clen = end_of_lowmem - ramdisk_image;
345 p = (char *)__va(ramdisk_image);
346 memcpy(q, p, clen);
347 q += clen;
348 ramdisk_image += clen;
349 ramdisk_size -= clen;
350 }
351
352 /* Copy the highmem portion of the initrd */
353 while (ramdisk_size) {
354 slop = ramdisk_image & ~PAGE_MASK;
355 clen = ramdisk_size;
356 if (clen > MAX_MAP_CHUNK-slop)
357 clen = MAX_MAP_CHUNK-slop;
358 mapaddr = ramdisk_image & PAGE_MASK;
359 p = early_memremap(mapaddr, clen+slop);
360 memcpy(q, p+slop, clen);
361 early_iounmap(p, clen+slop);
362 q += clen;
363 ramdisk_image += clen;
364 ramdisk_size -= clen;
365 }
366 /* high pages is not converted by early_res_to_bootmem */
367 ramdisk_image = boot_params.hdr.ramdisk_image;
368 ramdisk_size = boot_params.hdr.ramdisk_size;
369 printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
370 " %08llx - %08llx\n",
371 ramdisk_image, ramdisk_image + ramdisk_size - 1,
372 ramdisk_here, ramdisk_here + ramdisk_size - 1);
373 }
374
375 static void __init reserve_initrd(void)
376 {
377 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
378 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
379 u64 ramdisk_end = ramdisk_image + ramdisk_size;
380 u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
381
382 if (!boot_params.hdr.type_of_loader ||
383 !ramdisk_image || !ramdisk_size)
384 return; /* No initrd provided by bootloader */
385
386 initrd_start = 0;
387
388 if (ramdisk_size >= (end_of_lowmem>>1)) {
389 free_early(ramdisk_image, ramdisk_end);
390 printk(KERN_ERR "initrd too large to handle, "
391 "disabling initrd\n");
392 return;
393 }
394
395 printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
396 ramdisk_end);
397
398
399 if (ramdisk_end <= end_of_lowmem) {
400 /* All in lowmem, easy case */
401 /*
402 * don't need to reserve again, already reserved early
403 * in i386_start_kernel
404 */
405 initrd_start = ramdisk_image + PAGE_OFFSET;
406 initrd_end = initrd_start + ramdisk_size;
407 return;
408 }
409
410 relocate_initrd();
411
412 free_early(ramdisk_image, ramdisk_end);
413 }
414 #else
415 static void __init reserve_initrd(void)
416 {
417 }
418 #endif /* CONFIG_BLK_DEV_INITRD */
419
420 static void __init parse_setup_data(void)
421 {
422 struct setup_data *data;
423 u64 pa_data;
424
425 if (boot_params.hdr.version < 0x0209)
426 return;
427 pa_data = boot_params.hdr.setup_data;
428 while (pa_data) {
429 data = early_memremap(pa_data, PAGE_SIZE);
430 switch (data->type) {
431 case SETUP_E820_EXT:
432 parse_e820_ext(data, pa_data);
433 break;
434 default:
435 break;
436 }
437 pa_data = data->next;
438 early_iounmap(data, PAGE_SIZE);
439 }
440 }
441
442 static void __init e820_reserve_setup_data(void)
443 {
444 struct setup_data *data;
445 u64 pa_data;
446 int found = 0;
447
448 if (boot_params.hdr.version < 0x0209)
449 return;
450 pa_data = boot_params.hdr.setup_data;
451 while (pa_data) {
452 data = early_memremap(pa_data, sizeof(*data));
453 e820_update_range(pa_data, sizeof(*data)+data->len,
454 E820_RAM, E820_RESERVED_KERN);
455 found = 1;
456 pa_data = data->next;
457 early_iounmap(data, sizeof(*data));
458 }
459 if (!found)
460 return;
461
462 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
463 memcpy(&e820_saved, &e820, sizeof(struct e820map));
464 printk(KERN_INFO "extended physical RAM map:\n");
465 e820_print_map("reserve setup_data");
466 }
467
468 static void __init reserve_early_setup_data(void)
469 {
470 struct setup_data *data;
471 u64 pa_data;
472 char buf[32];
473
474 if (boot_params.hdr.version < 0x0209)
475 return;
476 pa_data = boot_params.hdr.setup_data;
477 while (pa_data) {
478 data = early_memremap(pa_data, sizeof(*data));
479 sprintf(buf, "setup data %x", data->type);
480 reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
481 pa_data = data->next;
482 early_iounmap(data, sizeof(*data));
483 }
484 }
485
486 /*
487 * --------- Crashkernel reservation ------------------------------
488 */
489
490 #ifdef CONFIG_KEXEC
491
492 static inline unsigned long long get_total_mem(void)
493 {
494 unsigned long long total;
495
496 total = max_pfn - min_low_pfn;
497
498 return total << PAGE_SHIFT;
499 }
500
501 static void __init reserve_crashkernel(void)
502 {
503 unsigned long long total_mem;
504 unsigned long long crash_size, crash_base;
505 int ret;
506
507 total_mem = get_total_mem();
508
509 ret = parse_crashkernel(boot_command_line, total_mem,
510 &crash_size, &crash_base);
511 if (ret != 0 || crash_size <= 0)
512 return;
513
514 /* 0 means: find the address automatically */
515 if (crash_base <= 0) {
516 const unsigned long long alignment = 16<<20; /* 16M */
517
518 crash_base = find_e820_area(alignment, ULONG_MAX, crash_size,
519 alignment);
520 if (crash_base == -1ULL) {
521 pr_info("crashkernel reservation failed - No suitable area found.\n");
522 return;
523 }
524 } else {
525 unsigned long long start;
526
527 start = find_e820_area(crash_base, ULONG_MAX, crash_size,
528 1<<20);
529 if (start != crash_base) {
530 pr_info("crashkernel reservation failed - memory is in use.\n");
531 return;
532 }
533 }
534 reserve_early(crash_base, crash_base + crash_size, "CRASH KERNEL");
535
536 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
537 "for crashkernel (System RAM: %ldMB)\n",
538 (unsigned long)(crash_size >> 20),
539 (unsigned long)(crash_base >> 20),
540 (unsigned long)(total_mem >> 20));
541
542 crashk_res.start = crash_base;
543 crashk_res.end = crash_base + crash_size - 1;
544 insert_resource(&iomem_resource, &crashk_res);
545 }
546 #else
547 static void __init reserve_crashkernel(void)
548 {
549 }
550 #endif
551
552 static struct resource standard_io_resources[] = {
553 { .name = "dma1", .start = 0x00, .end = 0x1f,
554 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
555 { .name = "pic1", .start = 0x20, .end = 0x21,
556 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
557 { .name = "timer0", .start = 0x40, .end = 0x43,
558 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
559 { .name = "timer1", .start = 0x50, .end = 0x53,
560 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
561 { .name = "keyboard", .start = 0x60, .end = 0x60,
562 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
563 { .name = "keyboard", .start = 0x64, .end = 0x64,
564 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
565 { .name = "dma page reg", .start = 0x80, .end = 0x8f,
566 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
567 { .name = "pic2", .start = 0xa0, .end = 0xa1,
568 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
569 { .name = "dma2", .start = 0xc0, .end = 0xdf,
570 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
571 { .name = "fpu", .start = 0xf0, .end = 0xff,
572 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
573 };
574
575 void __init reserve_standard_io_resources(void)
576 {
577 int i;
578
579 /* request I/O space for devices used on all i[345]86 PCs */
580 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
581 request_resource(&ioport_resource, &standard_io_resources[i]);
582
583 }
584
585 /*
586 * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
587 * is_kdump_kernel() to determine if we are booting after a panic. Hence
588 * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
589 */
590
591 #ifdef CONFIG_CRASH_DUMP
592 /* elfcorehdr= specifies the location of elf core header
593 * stored by the crashed kernel. This option will be passed
594 * by kexec loader to the capture kernel.
595 */
596 static int __init setup_elfcorehdr(char *arg)
597 {
598 char *end;
599 if (!arg)
600 return -EINVAL;
601 elfcorehdr_addr = memparse(arg, &end);
602 return end > arg ? 0 : -EINVAL;
603 }
604 early_param("elfcorehdr", setup_elfcorehdr);
605 #endif
606
607 #ifdef CONFIG_X86_RESERVE_LOW_64K
608 static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
609 {
610 printk(KERN_NOTICE
611 "%s detected: BIOS may corrupt low RAM, working around it.\n",
612 d->ident);
613
614 e820_update_range(0, 0x10000, E820_RAM, E820_RESERVED);
615 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
616
617 return 0;
618 }
619 #endif
620
621 /* List of systems that have known low memory corruption BIOS problems */
622 static struct dmi_system_id __initdata bad_bios_dmi_table[] = {
623 #ifdef CONFIG_X86_RESERVE_LOW_64K
624 {
625 .callback = dmi_low_memory_corruption,
626 .ident = "AMI BIOS",
627 .matches = {
628 DMI_MATCH(DMI_BIOS_VENDOR, "American Megatrends Inc."),
629 },
630 },
631 {
632 .callback = dmi_low_memory_corruption,
633 .ident = "Phoenix BIOS",
634 .matches = {
635 DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix Technologies"),
636 },
637 },
638 {
639 .callback = dmi_low_memory_corruption,
640 .ident = "Phoenix/MSC BIOS",
641 .matches = {
642 DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix/MSC"),
643 },
644 },
645 /*
646 * AMI BIOS with low memory corruption was found on Intel DG45ID and
647 * DG45FC boards.
648 * It has a different DMI_BIOS_VENDOR = "Intel Corp.", for now we will
649 * match only DMI_BOARD_NAME and see if there is more bad products
650 * with this vendor.
651 */
652 {
653 .callback = dmi_low_memory_corruption,
654 .ident = "AMI BIOS",
655 .matches = {
656 DMI_MATCH(DMI_BOARD_NAME, "DG45ID"),
657 },
658 },
659 {
660 .callback = dmi_low_memory_corruption,
661 .ident = "AMI BIOS",
662 .matches = {
663 DMI_MATCH(DMI_BOARD_NAME, "DG45FC"),
664 },
665 },
666 #endif
667 {}
668 };
669
670 static void __init trim_bios_range(void)
671 {
672 /*
673 * A special case is the first 4Kb of memory;
674 * This is a BIOS owned area, not kernel ram, but generally
675 * not listed as such in the E820 table.
676 */
677 e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
678 /*
679 * special case: Some BIOSen report the PC BIOS
680 * area (640->1Mb) as ram even though it is not.
681 * take them out.
682 */
683 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
684 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
685 }
686
687 /*
688 * Determine if we were loaded by an EFI loader. If so, then we have also been
689 * passed the efi memmap, systab, etc., so we should use these data structures
690 * for initialization. Note, the efi init code path is determined by the
691 * global efi_enabled. This allows the same kernel image to be used on existing
692 * systems (with a traditional BIOS) as well as on EFI systems.
693 */
694 /*
695 * setup_arch - architecture-specific boot-time initializations
696 *
697 * Note: On x86_64, fixmaps are ready for use even before this is called.
698 */
699
700 void __init setup_arch(char **cmdline_p)
701 {
702 int acpi = 0;
703 int k8 = 0;
704
705 #ifdef CONFIG_X86_32
706 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
707 visws_early_detect();
708 #else
709 printk(KERN_INFO "Command line: %s\n", boot_command_line);
710 #endif
711
712 /* VMI may relocate the fixmap; do this before touching ioremap area */
713 vmi_init();
714
715 early_cpu_init();
716 early_ioremap_init();
717
718 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
719 screen_info = boot_params.screen_info;
720 edid_info = boot_params.edid_info;
721 #ifdef CONFIG_X86_32
722 apm_info.bios = boot_params.apm_bios_info;
723 ist_info = boot_params.ist_info;
724 if (boot_params.sys_desc_table.length != 0) {
725 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
726 machine_id = boot_params.sys_desc_table.table[0];
727 machine_submodel_id = boot_params.sys_desc_table.table[1];
728 BIOS_revision = boot_params.sys_desc_table.table[2];
729 }
730 #endif
731 saved_video_mode = boot_params.hdr.vid_mode;
732 bootloader_type = boot_params.hdr.type_of_loader;
733 if ((bootloader_type >> 4) == 0xe) {
734 bootloader_type &= 0xf;
735 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
736 }
737 bootloader_version = bootloader_type & 0xf;
738 bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
739
740 #ifdef CONFIG_BLK_DEV_RAM
741 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
742 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
743 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
744 #endif
745 #ifdef CONFIG_EFI
746 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
747 #ifdef CONFIG_X86_32
748 "EL32",
749 #else
750 "EL64",
751 #endif
752 4)) {
753 efi_enabled = 1;
754 efi_reserve_early();
755 }
756 #endif
757
758 x86_init.oem.arch_setup();
759
760 setup_memory_map();
761 parse_setup_data();
762 /* update the e820_saved too */
763 e820_reserve_setup_data();
764
765 copy_edd();
766
767 if (!boot_params.hdr.root_flags)
768 root_mountflags &= ~MS_RDONLY;
769 init_mm.start_code = (unsigned long) _text;
770 init_mm.end_code = (unsigned long) _etext;
771 init_mm.end_data = (unsigned long) _edata;
772 init_mm.brk = _brk_end;
773
774 code_resource.start = virt_to_phys(_text);
775 code_resource.end = virt_to_phys(_etext)-1;
776 data_resource.start = virt_to_phys(_etext);
777 data_resource.end = virt_to_phys(_edata)-1;
778 bss_resource.start = virt_to_phys(&__bss_start);
779 bss_resource.end = virt_to_phys(&__bss_stop)-1;
780
781 #ifdef CONFIG_CMDLINE_BOOL
782 #ifdef CONFIG_CMDLINE_OVERRIDE
783 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
784 #else
785 if (builtin_cmdline[0]) {
786 /* append boot loader cmdline to builtin */
787 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
788 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
789 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
790 }
791 #endif
792 #endif
793
794 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
795 *cmdline_p = command_line;
796
797 /*
798 * x86_configure_nx() is called before parse_early_param() to detect
799 * whether hardware doesn't support NX (so that the early EHCI debug
800 * console setup can safely call set_fixmap()). It may then be called
801 * again from within noexec_setup() during parsing early parameters
802 * to honor the respective command line option.
803 */
804 x86_configure_nx();
805
806 parse_early_param();
807
808 x86_report_nx();
809
810 /* Must be before kernel pagetables are setup */
811 vmi_activate();
812
813 /* after early param, so could get panic from serial */
814 reserve_early_setup_data();
815
816 if (acpi_mps_check()) {
817 #ifdef CONFIG_X86_LOCAL_APIC
818 disable_apic = 1;
819 #endif
820 setup_clear_cpu_cap(X86_FEATURE_APIC);
821 }
822
823 #ifdef CONFIG_PCI
824 if (pci_early_dump_regs)
825 early_dump_pci_devices();
826 #endif
827
828 finish_e820_parsing();
829
830 if (efi_enabled)
831 efi_init();
832
833 dmi_scan_machine();
834
835 dmi_check_system(bad_bios_dmi_table);
836
837 /*
838 * VMware detection requires dmi to be available, so this
839 * needs to be done after dmi_scan_machine, for the BP.
840 */
841 init_hypervisor_platform();
842
843 x86_init.resources.probe_roms();
844
845 /* after parse_early_param, so could debug it */
846 insert_resource(&iomem_resource, &code_resource);
847 insert_resource(&iomem_resource, &data_resource);
848 insert_resource(&iomem_resource, &bss_resource);
849
850 trim_bios_range();
851 #ifdef CONFIG_X86_32
852 if (ppro_with_ram_bug()) {
853 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
854 E820_RESERVED);
855 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
856 printk(KERN_INFO "fixed physical RAM map:\n");
857 e820_print_map("bad_ppro");
858 }
859 #else
860 early_gart_iommu_check();
861 #endif
862
863 /*
864 * partially used pages are not usable - thus
865 * we are rounding upwards:
866 */
867 max_pfn = e820_end_of_ram_pfn();
868
869 /* preallocate 4k for mptable mpc */
870 early_reserve_e820_mpc_new();
871 /* update e820 for memory not covered by WB MTRRs */
872 mtrr_bp_init();
873 if (mtrr_trim_uncached_memory(max_pfn))
874 max_pfn = e820_end_of_ram_pfn();
875
876 #ifdef CONFIG_X86_32
877 /* max_low_pfn get updated here */
878 find_low_pfn_range();
879 #else
880 num_physpages = max_pfn;
881
882 check_x2apic();
883
884 /* How many end-of-memory variables you have, grandma! */
885 /* need this before calling reserve_initrd */
886 if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
887 max_low_pfn = e820_end_of_low_ram_pfn();
888 else
889 max_low_pfn = max_pfn;
890
891 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
892 max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT;
893 #endif
894
895 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
896 setup_bios_corruption_check();
897 #endif
898
899 printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
900 max_pfn_mapped<<PAGE_SHIFT);
901
902 reserve_brk();
903
904 /*
905 * Find and reserve possible boot-time SMP configuration:
906 */
907 find_smp_config();
908
909 reserve_trampoline_memory();
910
911 #ifdef CONFIG_ACPI_SLEEP
912 /*
913 * Reserve low memory region for sleep support.
914 * even before init_memory_mapping
915 */
916 acpi_reserve_wakeup_memory();
917 #endif
918 init_gbpages();
919
920 /* max_pfn_mapped is updated here */
921 max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
922 max_pfn_mapped = max_low_pfn_mapped;
923
924 #ifdef CONFIG_X86_64
925 if (max_pfn > max_low_pfn) {
926 max_pfn_mapped = init_memory_mapping(1UL<<32,
927 max_pfn<<PAGE_SHIFT);
928 /* can we preseve max_low_pfn ?*/
929 max_low_pfn = max_pfn;
930 }
931 #endif
932
933 /*
934 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
935 */
936
937 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
938 if (init_ohci1394_dma_early)
939 init_ohci1394_dma_on_all_controllers();
940 #endif
941
942 reserve_initrd();
943
944 reserve_crashkernel();
945
946 vsmp_init();
947
948 io_delay_init();
949
950 /*
951 * Parse the ACPI tables for possible boot-time SMP configuration.
952 */
953 acpi_boot_table_init();
954
955 early_acpi_boot_init();
956
957 #ifdef CONFIG_ACPI_NUMA
958 /*
959 * Parse SRAT to discover nodes.
960 */
961 acpi = acpi_numa_init();
962 #endif
963
964 #ifdef CONFIG_K8_NUMA
965 if (!acpi)
966 k8 = !k8_numa_init(0, max_pfn);
967 #endif
968
969 initmem_init(0, max_pfn, acpi, k8);
970 #ifndef CONFIG_NO_BOOTMEM
971 early_res_to_bootmem(0, max_low_pfn<<PAGE_SHIFT);
972 #endif
973
974 dma32_reserve_bootmem();
975
976 reserve_ibft_region();
977
978 #ifdef CONFIG_KVM_CLOCK
979 kvmclock_init();
980 #endif
981
982 x86_init.paging.pagetable_setup_start(swapper_pg_dir);
983 paging_init();
984 x86_init.paging.pagetable_setup_done(swapper_pg_dir);
985
986 tboot_probe();
987
988 #ifdef CONFIG_X86_64
989 map_vsyscall();
990 #endif
991
992 generic_apic_probe();
993
994 early_quirks();
995
996 /*
997 * Read APIC and some other early information from ACPI tables.
998 */
999 acpi_boot_init();
1000
1001 sfi_init();
1002
1003 /*
1004 * get boot-time SMP configuration:
1005 */
1006 if (smp_found_config)
1007 get_smp_config();
1008
1009 prefill_possible_map();
1010
1011 #ifdef CONFIG_X86_64
1012 init_cpu_to_node();
1013 #endif
1014
1015 init_apic_mappings();
1016 ioapic_init_mappings();
1017
1018 /* need to wait for io_apic is mapped */
1019 probe_nr_irqs_gsi();
1020
1021 kvm_guest_init();
1022
1023 e820_reserve_resources();
1024 e820_mark_nosave_regions(max_low_pfn);
1025
1026 x86_init.resources.reserve_resources();
1027
1028 e820_setup_gap();
1029
1030 #ifdef CONFIG_VT
1031 #if defined(CONFIG_VGA_CONSOLE)
1032 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1033 conswitchp = &vga_con;
1034 #elif defined(CONFIG_DUMMY_CONSOLE)
1035 conswitchp = &dummy_con;
1036 #endif
1037 #endif
1038 x86_init.oem.banner();
1039
1040 mcheck_init();
1041 }
1042
1043 #ifdef CONFIG_X86_32
1044
1045 static struct resource video_ram_resource = {
1046 .name = "Video RAM area",
1047 .start = 0xa0000,
1048 .end = 0xbffff,
1049 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
1050 };
1051
1052 void __init i386_reserve_resources(void)
1053 {
1054 request_resource(&iomem_resource, &video_ram_resource);
1055 reserve_standard_io_resources();
1056 }
1057
1058 #endif /* CONFIG_X86_32 */