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 #ifdef CONFIG_DMI 125 RESERVE_BRK(dmi_alloc, 65536); 126 #endif 127 128 unsigned int boot_cpu_id __read_mostly; 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 #endif 301 302 static void __init reserve_brk(void) 303 { 304 if (_brk_end > _brk_start) 305 reserve_early(__pa(_brk_start), __pa(_brk_end), "BRK"); 306 307 /* Mark brk area as locked down and no longer taking any 308 new allocations */ 309 _brk_start = 0; 310 } 311 312 #ifdef CONFIG_BLK_DEV_INITRD 313 314 #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT) 315 static void __init relocate_initrd(void) 316 { 317 318 u64 ramdisk_image = boot_params.hdr.ramdisk_image; 319 u64 ramdisk_size = boot_params.hdr.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 = find_e820_area(0, end_of_lowmem, ramdisk_size, 327 PAGE_SIZE); 328 329 if (ramdisk_here == -1ULL) 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 reserve_early(ramdisk_here, ramdisk_here + ramdisk_size, 336 "NEW RAMDISK"); 337 initrd_start = ramdisk_here + PAGE_OFFSET; 338 initrd_end = initrd_start + ramdisk_size; 339 printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n", 340 ramdisk_here, ramdisk_here + ramdisk_size); 341 342 q = (char *)initrd_start; 343 344 /* Copy any lowmem portion of the initrd */ 345 if (ramdisk_image < end_of_lowmem) { 346 clen = end_of_lowmem - ramdisk_image; 347 p = (char *)__va(ramdisk_image); 348 memcpy(q, p, clen); 349 q += clen; 350 ramdisk_image += clen; 351 ramdisk_size -= clen; 352 } 353 354 /* Copy the highmem portion of the initrd */ 355 while (ramdisk_size) { 356 slop = ramdisk_image & ~PAGE_MASK; 357 clen = ramdisk_size; 358 if (clen > MAX_MAP_CHUNK-slop) 359 clen = MAX_MAP_CHUNK-slop; 360 mapaddr = ramdisk_image & PAGE_MASK; 361 p = early_memremap(mapaddr, clen+slop); 362 memcpy(q, p+slop, clen); 363 early_iounmap(p, clen+slop); 364 q += clen; 365 ramdisk_image += clen; 366 ramdisk_size -= clen; 367 } 368 /* high pages is not converted by early_res_to_bootmem */ 369 ramdisk_image = boot_params.hdr.ramdisk_image; 370 ramdisk_size = boot_params.hdr.ramdisk_size; 371 printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to" 372 " %08llx - %08llx\n", 373 ramdisk_image, ramdisk_image + ramdisk_size - 1, 374 ramdisk_here, ramdisk_here + ramdisk_size - 1); 375 } 376 377 static void __init reserve_initrd(void) 378 { 379 u64 ramdisk_image = boot_params.hdr.ramdisk_image; 380 u64 ramdisk_size = boot_params.hdr.ramdisk_size; 381 u64 ramdisk_end = 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 free_early(ramdisk_image, ramdisk_end); 392 printk(KERN_ERR "initrd too large to handle, " 393 "disabling initrd\n"); 394 return; 395 } 396 397 printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image, 398 ramdisk_end); 399 400 401 if (ramdisk_end <= end_of_lowmem) { 402 /* All in lowmem, easy case */ 403 /* 404 * don't need to reserve again, already reserved early 405 * in i386_start_kernel 406 */ 407 initrd_start = ramdisk_image + PAGE_OFFSET; 408 initrd_end = initrd_start + ramdisk_size; 409 return; 410 } 411 412 relocate_initrd(); 413 414 free_early(ramdisk_image, ramdisk_end); 415 } 416 #else 417 static void __init reserve_initrd(void) 418 { 419 } 420 #endif /* CONFIG_BLK_DEV_INITRD */ 421 422 static void __init parse_setup_data(void) 423 { 424 struct setup_data *data; 425 u64 pa_data; 426 427 if (boot_params.hdr.version < 0x0209) 428 return; 429 pa_data = boot_params.hdr.setup_data; 430 while (pa_data) { 431 data = early_memremap(pa_data, PAGE_SIZE); 432 switch (data->type) { 433 case SETUP_E820_EXT: 434 parse_e820_ext(data, pa_data); 435 break; 436 default: 437 break; 438 } 439 pa_data = data->next; 440 early_iounmap(data, PAGE_SIZE); 441 } 442 } 443 444 static void __init e820_reserve_setup_data(void) 445 { 446 struct setup_data *data; 447 u64 pa_data; 448 int found = 0; 449 450 if (boot_params.hdr.version < 0x0209) 451 return; 452 pa_data = boot_params.hdr.setup_data; 453 while (pa_data) { 454 data = early_memremap(pa_data, sizeof(*data)); 455 e820_update_range(pa_data, sizeof(*data)+data->len, 456 E820_RAM, E820_RESERVED_KERN); 457 found = 1; 458 pa_data = data->next; 459 early_iounmap(data, sizeof(*data)); 460 } 461 if (!found) 462 return; 463 464 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); 465 memcpy(&e820_saved, &e820, sizeof(struct e820map)); 466 printk(KERN_INFO "extended physical RAM map:\n"); 467 e820_print_map("reserve setup_data"); 468 } 469 470 static void __init reserve_early_setup_data(void) 471 { 472 struct setup_data *data; 473 u64 pa_data; 474 char buf[32]; 475 476 if (boot_params.hdr.version < 0x0209) 477 return; 478 pa_data = boot_params.hdr.setup_data; 479 while (pa_data) { 480 data = early_memremap(pa_data, sizeof(*data)); 481 sprintf(buf, "setup data %x", data->type); 482 reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf); 483 pa_data = data->next; 484 early_iounmap(data, sizeof(*data)); 485 } 486 } 487 488 /* 489 * --------- Crashkernel reservation ------------------------------ 490 */ 491 492 #ifdef CONFIG_KEXEC 493 494 static inline unsigned long long get_total_mem(void) 495 { 496 unsigned long long total; 497 498 total = max_pfn - min_low_pfn; 499 500 return total << PAGE_SHIFT; 501 } 502 503 static void __init reserve_crashkernel(void) 504 { 505 unsigned long long total_mem; 506 unsigned long long crash_size, crash_base; 507 int ret; 508 509 total_mem = get_total_mem(); 510 511 ret = parse_crashkernel(boot_command_line, total_mem, 512 &crash_size, &crash_base); 513 if (ret != 0 || crash_size <= 0) 514 return; 515 516 /* 0 means: find the address automatically */ 517 if (crash_base <= 0) { 518 const unsigned long long alignment = 16<<20; /* 16M */ 519 520 crash_base = find_e820_area(alignment, ULONG_MAX, crash_size, 521 alignment); 522 if (crash_base == -1ULL) { 523 pr_info("crashkernel reservation failed - No suitable area found.\n"); 524 return; 525 } 526 } else { 527 unsigned long long start; 528 529 start = find_e820_area(crash_base, ULONG_MAX, crash_size, 530 1<<20); 531 if (start != crash_base) { 532 pr_info("crashkernel reservation failed - memory is in use.\n"); 533 return; 534 } 535 } 536 reserve_early(crash_base, crash_base + crash_size, "CRASH KERNEL"); 537 538 printk(KERN_INFO "Reserving %ldMB of memory at %ldMB " 539 "for crashkernel (System RAM: %ldMB)\n", 540 (unsigned long)(crash_size >> 20), 541 (unsigned long)(crash_base >> 20), 542 (unsigned long)(total_mem >> 20)); 543 544 crashk_res.start = crash_base; 545 crashk_res.end = crash_base + crash_size - 1; 546 insert_resource(&iomem_resource, &crashk_res); 547 } 548 #else 549 static void __init reserve_crashkernel(void) 550 { 551 } 552 #endif 553 554 static struct resource standard_io_resources[] = { 555 { .name = "dma1", .start = 0x00, .end = 0x1f, 556 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 557 { .name = "pic1", .start = 0x20, .end = 0x21, 558 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 559 { .name = "timer0", .start = 0x40, .end = 0x43, 560 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 561 { .name = "timer1", .start = 0x50, .end = 0x53, 562 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 563 { .name = "keyboard", .start = 0x60, .end = 0x60, 564 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 565 { .name = "keyboard", .start = 0x64, .end = 0x64, 566 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 567 { .name = "dma page reg", .start = 0x80, .end = 0x8f, 568 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 569 { .name = "pic2", .start = 0xa0, .end = 0xa1, 570 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 571 { .name = "dma2", .start = 0xc0, .end = 0xdf, 572 .flags = IORESOURCE_BUSY | IORESOURCE_IO }, 573 { .name = "fpu", .start = 0xf0, .end = 0xff, 574 .flags = IORESOURCE_BUSY | IORESOURCE_IO } 575 }; 576 577 void __init reserve_standard_io_resources(void) 578 { 579 int i; 580 581 /* request I/O space for devices used on all i[345]86 PCs */ 582 for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) 583 request_resource(&ioport_resource, &standard_io_resources[i]); 584 585 } 586 587 /* 588 * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by 589 * is_kdump_kernel() to determine if we are booting after a panic. Hence 590 * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE. 591 */ 592 593 #ifdef CONFIG_CRASH_DUMP 594 /* elfcorehdr= specifies the location of elf core header 595 * stored by the crashed kernel. This option will be passed 596 * by kexec loader to the capture kernel. 597 */ 598 static int __init setup_elfcorehdr(char *arg) 599 { 600 char *end; 601 if (!arg) 602 return -EINVAL; 603 elfcorehdr_addr = memparse(arg, &end); 604 return end > arg ? 0 : -EINVAL; 605 } 606 early_param("elfcorehdr", setup_elfcorehdr); 607 #endif 608 609 #ifdef CONFIG_X86_RESERVE_LOW_64K 610 static int __init dmi_low_memory_corruption(const struct dmi_system_id *d) 611 { 612 printk(KERN_NOTICE 613 "%s detected: BIOS may corrupt low RAM, working around it.\n", 614 d->ident); 615 616 e820_update_range(0, 0x10000, E820_RAM, E820_RESERVED); 617 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); 618 619 return 0; 620 } 621 #endif 622 623 /* List of systems that have known low memory corruption BIOS problems */ 624 static struct dmi_system_id __initdata bad_bios_dmi_table[] = { 625 #ifdef CONFIG_X86_RESERVE_LOW_64K 626 { 627 .callback = dmi_low_memory_corruption, 628 .ident = "AMI BIOS", 629 .matches = { 630 DMI_MATCH(DMI_BIOS_VENDOR, "American Megatrends Inc."), 631 }, 632 }, 633 { 634 .callback = dmi_low_memory_corruption, 635 .ident = "Phoenix BIOS", 636 .matches = { 637 DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix Technologies"), 638 }, 639 }, 640 { 641 .callback = dmi_low_memory_corruption, 642 .ident = "Phoenix/MSC BIOS", 643 .matches = { 644 DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix/MSC"), 645 }, 646 }, 647 /* 648 * AMI BIOS with low memory corruption was found on Intel DG45ID and 649 * DG45FC boards. 650 * It has a different DMI_BIOS_VENDOR = "Intel Corp.", for now we will 651 * match only DMI_BOARD_NAME and see if there is more bad products 652 * with this vendor. 653 */ 654 { 655 .callback = dmi_low_memory_corruption, 656 .ident = "AMI BIOS", 657 .matches = { 658 DMI_MATCH(DMI_BOARD_NAME, "DG45ID"), 659 }, 660 }, 661 { 662 .callback = dmi_low_memory_corruption, 663 .ident = "AMI BIOS", 664 .matches = { 665 DMI_MATCH(DMI_BOARD_NAME, "DG45FC"), 666 }, 667 }, 668 #endif 669 {} 670 }; 671 672 static void __init trim_bios_range(void) 673 { 674 /* 675 * A special case is the first 4Kb of memory; 676 * This is a BIOS owned area, not kernel ram, but generally 677 * not listed as such in the E820 table. 678 */ 679 e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED); 680 /* 681 * special case: Some BIOSen report the PC BIOS 682 * area (640->1Mb) as ram even though it is not. 683 * take them out. 684 */ 685 e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1); 686 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); 687 } 688 689 /* 690 * Determine if we were loaded by an EFI loader. If so, then we have also been 691 * passed the efi memmap, systab, etc., so we should use these data structures 692 * for initialization. Note, the efi init code path is determined by the 693 * global efi_enabled. This allows the same kernel image to be used on existing 694 * systems (with a traditional BIOS) as well as on EFI systems. 695 */ 696 /* 697 * setup_arch - architecture-specific boot-time initializations 698 * 699 * Note: On x86_64, fixmaps are ready for use even before this is called. 700 */ 701 702 void __init setup_arch(char **cmdline_p) 703 { 704 int acpi = 0; 705 int k8 = 0; 706 707 #ifdef CONFIG_X86_32 708 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); 709 visws_early_detect(); 710 #else 711 printk(KERN_INFO "Command line: %s\n", boot_command_line); 712 #endif 713 714 /* VMI may relocate the fixmap; do this before touching ioremap area */ 715 vmi_init(); 716 717 early_cpu_init(); 718 early_ioremap_init(); 719 720 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); 721 screen_info = boot_params.screen_info; 722 edid_info = boot_params.edid_info; 723 #ifdef CONFIG_X86_32 724 apm_info.bios = boot_params.apm_bios_info; 725 ist_info = boot_params.ist_info; 726 if (boot_params.sys_desc_table.length != 0) { 727 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2); 728 machine_id = boot_params.sys_desc_table.table[0]; 729 machine_submodel_id = boot_params.sys_desc_table.table[1]; 730 BIOS_revision = boot_params.sys_desc_table.table[2]; 731 } 732 #endif 733 saved_video_mode = boot_params.hdr.vid_mode; 734 bootloader_type = boot_params.hdr.type_of_loader; 735 if ((bootloader_type >> 4) == 0xe) { 736 bootloader_type &= 0xf; 737 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4; 738 } 739 bootloader_version = bootloader_type & 0xf; 740 bootloader_version |= boot_params.hdr.ext_loader_ver << 4; 741 742 #ifdef CONFIG_BLK_DEV_RAM 743 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; 744 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0); 745 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0); 746 #endif 747 #ifdef CONFIG_EFI 748 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, 749 #ifdef CONFIG_X86_32 750 "EL32", 751 #else 752 "EL64", 753 #endif 754 4)) { 755 efi_enabled = 1; 756 efi_reserve_early(); 757 } 758 #endif 759 760 x86_init.oem.arch_setup(); 761 762 setup_memory_map(); 763 parse_setup_data(); 764 /* update the e820_saved too */ 765 e820_reserve_setup_data(); 766 767 copy_edd(); 768 769 if (!boot_params.hdr.root_flags) 770 root_mountflags &= ~MS_RDONLY; 771 init_mm.start_code = (unsigned long) _text; 772 init_mm.end_code = (unsigned long) _etext; 773 init_mm.end_data = (unsigned long) _edata; 774 init_mm.brk = _brk_end; 775 776 code_resource.start = virt_to_phys(_text); 777 code_resource.end = virt_to_phys(_etext)-1; 778 data_resource.start = virt_to_phys(_etext); 779 data_resource.end = virt_to_phys(_edata)-1; 780 bss_resource.start = virt_to_phys(&__bss_start); 781 bss_resource.end = virt_to_phys(&__bss_stop)-1; 782 783 #ifdef CONFIG_CMDLINE_BOOL 784 #ifdef CONFIG_CMDLINE_OVERRIDE 785 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); 786 #else 787 if (builtin_cmdline[0]) { 788 /* append boot loader cmdline to builtin */ 789 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE); 790 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE); 791 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); 792 } 793 #endif 794 #endif 795 796 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); 797 *cmdline_p = command_line; 798 799 /* 800 * x86_configure_nx() is called before parse_early_param() to detect 801 * whether hardware doesn't support NX (so that the early EHCI debug 802 * console setup can safely call set_fixmap()). It may then be called 803 * again from within noexec_setup() during parsing early parameters 804 * to honor the respective command line option. 805 */ 806 x86_configure_nx(); 807 808 parse_early_param(); 809 810 x86_report_nx(); 811 812 /* Must be before kernel pagetables are setup */ 813 vmi_activate(); 814 815 /* after early param, so could get panic from serial */ 816 reserve_early_setup_data(); 817 818 if (acpi_mps_check()) { 819 #ifdef CONFIG_X86_LOCAL_APIC 820 disable_apic = 1; 821 #endif 822 setup_clear_cpu_cap(X86_FEATURE_APIC); 823 } 824 825 #ifdef CONFIG_PCI 826 if (pci_early_dump_regs) 827 early_dump_pci_devices(); 828 #endif 829 830 finish_e820_parsing(); 831 832 if (efi_enabled) 833 efi_init(); 834 835 dmi_scan_machine(); 836 837 dmi_check_system(bad_bios_dmi_table); 838 839 /* 840 * VMware detection requires dmi to be available, so this 841 * needs to be done after dmi_scan_machine, for the BP. 842 */ 843 init_hypervisor_platform(); 844 845 x86_init.resources.probe_roms(); 846 847 /* after parse_early_param, so could debug it */ 848 insert_resource(&iomem_resource, &code_resource); 849 insert_resource(&iomem_resource, &data_resource); 850 insert_resource(&iomem_resource, &bss_resource); 851 852 trim_bios_range(); 853 #ifdef CONFIG_X86_32 854 if (ppro_with_ram_bug()) { 855 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM, 856 E820_RESERVED); 857 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); 858 printk(KERN_INFO "fixed physical RAM map:\n"); 859 e820_print_map("bad_ppro"); 860 } 861 #else 862 early_gart_iommu_check(); 863 #endif 864 865 /* 866 * partially used pages are not usable - thus 867 * we are rounding upwards: 868 */ 869 max_pfn = e820_end_of_ram_pfn(); 870 871 /* preallocate 4k for mptable mpc */ 872 early_reserve_e820_mpc_new(); 873 /* update e820 for memory not covered by WB MTRRs */ 874 mtrr_bp_init(); 875 if (mtrr_trim_uncached_memory(max_pfn)) 876 max_pfn = e820_end_of_ram_pfn(); 877 878 #ifdef CONFIG_X86_32 879 /* max_low_pfn get updated here */ 880 find_low_pfn_range(); 881 #else 882 num_physpages = max_pfn; 883 884 check_x2apic(); 885 886 /* How many end-of-memory variables you have, grandma! */ 887 /* need this before calling reserve_initrd */ 888 if (max_pfn > (1UL<<(32 - PAGE_SHIFT))) 889 max_low_pfn = e820_end_of_low_ram_pfn(); 890 else 891 max_low_pfn = max_pfn; 892 893 high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1; 894 max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT; 895 #endif 896 897 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION 898 setup_bios_corruption_check(); 899 #endif 900 901 printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n", 902 max_pfn_mapped<<PAGE_SHIFT); 903 904 reserve_brk(); 905 906 /* 907 * Find and reserve possible boot-time SMP configuration: 908 */ 909 find_smp_config(); 910 911 reserve_trampoline_memory(); 912 913 #ifdef CONFIG_ACPI_SLEEP 914 /* 915 * Reserve low memory region for sleep support. 916 * even before init_memory_mapping 917 */ 918 acpi_reserve_wakeup_memory(); 919 #endif 920 init_gbpages(); 921 922 /* max_pfn_mapped is updated here */ 923 max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT); 924 max_pfn_mapped = max_low_pfn_mapped; 925 926 #ifdef CONFIG_X86_64 927 if (max_pfn > max_low_pfn) { 928 max_pfn_mapped = init_memory_mapping(1UL<<32, 929 max_pfn<<PAGE_SHIFT); 930 /* can we preseve max_low_pfn ?*/ 931 max_low_pfn = max_pfn; 932 } 933 #endif 934 935 /* 936 * NOTE: On x86-32, only from this point on, fixmaps are ready for use. 937 */ 938 939 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT 940 if (init_ohci1394_dma_early) 941 init_ohci1394_dma_on_all_controllers(); 942 #endif 943 944 reserve_initrd(); 945 946 reserve_crashkernel(); 947 948 vsmp_init(); 949 950 io_delay_init(); 951 952 /* 953 * Parse the ACPI tables for possible boot-time SMP configuration. 954 */ 955 acpi_boot_table_init(); 956 957 early_acpi_boot_init(); 958 959 #ifdef CONFIG_ACPI_NUMA 960 /* 961 * Parse SRAT to discover nodes. 962 */ 963 acpi = acpi_numa_init(); 964 #endif 965 966 #ifdef CONFIG_K8_NUMA 967 if (!acpi) 968 k8 = !k8_numa_init(0, max_pfn); 969 #endif 970 971 initmem_init(0, max_pfn, acpi, k8); 972 973 #ifdef CONFIG_X86_64 974 /* 975 * dma32_reserve_bootmem() allocates bootmem which may conflict 976 * with the crashkernel command line, so do that after 977 * reserve_crashkernel() 978 */ 979 dma32_reserve_bootmem(); 980 #endif 981 982 reserve_ibft_region(); 983 984 #ifdef CONFIG_KVM_CLOCK 985 kvmclock_init(); 986 #endif 987 988 x86_init.paging.pagetable_setup_start(swapper_pg_dir); 989 paging_init(); 990 x86_init.paging.pagetable_setup_done(swapper_pg_dir); 991 992 tboot_probe(); 993 994 #ifdef CONFIG_X86_64 995 map_vsyscall(); 996 #endif 997 998 generic_apic_probe(); 999 1000 early_quirks(); 1001 1002 /* 1003 * Read APIC and some other early information from ACPI tables. 1004 */ 1005 acpi_boot_init(); 1006 1007 sfi_init(); 1008 1009 /* 1010 * get boot-time SMP configuration: 1011 */ 1012 if (smp_found_config) 1013 get_smp_config(); 1014 1015 prefill_possible_map(); 1016 1017 #ifdef CONFIG_X86_64 1018 init_cpu_to_node(); 1019 #endif 1020 1021 init_apic_mappings(); 1022 ioapic_init_mappings(); 1023 1024 /* need to wait for io_apic is mapped */ 1025 probe_nr_irqs_gsi(); 1026 1027 kvm_guest_init(); 1028 1029 e820_reserve_resources(); 1030 e820_mark_nosave_regions(max_low_pfn); 1031 1032 x86_init.resources.reserve_resources(); 1033 1034 e820_setup_gap(); 1035 1036 #ifdef CONFIG_VT 1037 #if defined(CONFIG_VGA_CONSOLE) 1038 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) 1039 conswitchp = &vga_con; 1040 #elif defined(CONFIG_DUMMY_CONSOLE) 1041 conswitchp = &dummy_con; 1042 #endif 1043 #endif 1044 x86_init.oem.banner(); 1045 1046 mcheck_init(); 1047 } 1048 1049 #ifdef CONFIG_X86_32 1050 1051 static struct resource video_ram_resource = { 1052 .name = "Video RAM area", 1053 .start = 0xa0000, 1054 .end = 0xbffff, 1055 .flags = IORESOURCE_BUSY | IORESOURCE_MEM 1056 }; 1057 1058 void __init i386_reserve_resources(void) 1059 { 1060 request_resource(&iomem_resource, &video_ram_resource); 1061 reserve_standard_io_resources(); 1062 } 1063 1064 #endif /* CONFIG_X86_32 */ 1065