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/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/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 #include <asm/olpc_ofw.h>
 107
 108 #include <asm/percpu.h>
 109 #include <asm/topology.h>
 110 #include <asm/apicdef.h>
 111 #include <asm/k8.h>
 112 #ifdef CONFIG_X86_64
 113 #include <asm/numa_64.h>
 114 #endif
 115 #include <asm/mce.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 unsigned int boot_cpu_id __read_mostly;
 130
 131 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
 132 unsigned long _brk_end = (unsigned long)__brk_base;
 133
 134 #ifdef CONFIG_X86_64
 135 int default_cpu_present_to_apicid(int mps_cpu)
 136 {
 137         return __default_cpu_present_to_apicid(mps_cpu);
 138 }
 139
 140 int default_check_phys_apicid_present(int phys_apicid)
 141 {
 142         return __default_check_phys_apicid_present(phys_apicid);
 143 }
 144 #endif
 145
 146 #ifndef CONFIG_DEBUG_BOOT_PARAMS
 147 struct boot_params __initdata boot_params;
 148 #else
 149 struct boot_params boot_params;
 150 #endif
 151
 152 /*
 153  * Machine setup..
 154  */
 155 static struct resource data_resource = {
 156         .name   = "Kernel data",
 157         .start  = 0,
 158         .end    = 0,
 159         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 160 };
 161
 162 static struct resource code_resource = {
 163         .name   = "Kernel code",
 164         .start  = 0,
 165         .end    = 0,
 166         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 167 };
 168
 169 static struct resource bss_resource = {
 170         .name   = "Kernel bss",
 171         .start  = 0,
 172         .end    = 0,
 173         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 174 };
 175
 176
 177 #ifdef CONFIG_X86_32
 178 /* cpu data as detected by the assembly code in head.S */
 179 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
 180 /* common cpu data for all cpus */
 181 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
 182 EXPORT_SYMBOL(boot_cpu_data);
 183 static void set_mca_bus(int x)
 184 {
 185 #ifdef CONFIG_MCA
 186         MCA_bus = x;
 187 #endif
 188 }
 189
 190 unsigned int def_to_bigsmp;
 191
 192 /* for MCA, but anyone else can use it if they want */
 193 unsigned int machine_id;
 194 unsigned int machine_submodel_id;
 195 unsigned int BIOS_revision;
 196
 197 struct apm_info apm_info;
 198 EXPORT_SYMBOL(apm_info);
 199
 200 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 201         defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 202 struct ist_info ist_info;
 203 EXPORT_SYMBOL(ist_info);
 204 #else
 205 struct ist_info ist_info;
 206 #endif
 207
 208 #else
 209 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
 210         .x86_phys_bits = MAX_PHYSMEM_BITS,
 211 };
 212 EXPORT_SYMBOL(boot_cpu_data);
 213 #endif
 214
 215
 216 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
 217 unsigned long mmu_cr4_features;
 218 #else
 219 unsigned long mmu_cr4_features = X86_CR4_PAE;
 220 #endif
 221
 222 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
 223 int bootloader_type, bootloader_version;
 224
 225 /*
 226  * Setup options
 227  */
 228 struct screen_info screen_info;
 229 EXPORT_SYMBOL(screen_info);
 230 struct edid_info edid_info;
 231 EXPORT_SYMBOL_GPL(edid_info);
 232
 233 extern int root_mountflags;
 234
 235 unsigned long saved_video_mode;
 236
 237 #define RAMDISK_IMAGE_START_MASK        0x07FF
 238 #define RAMDISK_PROMPT_FLAG             0x8000
 239 #define RAMDISK_LOAD_FLAG               0x4000
 240
 241 static char __initdata command_line[COMMAND_LINE_SIZE];
 242 #ifdef CONFIG_CMDLINE_BOOL
 243 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
 244 #endif
 245
 246 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 247 struct edd edd;
 248 #ifdef CONFIG_EDD_MODULE
 249 EXPORT_SYMBOL(edd);
 250 #endif
 251 /**
 252  * copy_edd() - Copy the BIOS EDD information
 253  *              from boot_params into a safe place.
 254  *
 255  */
 256 static inline void __init copy_edd(void)
 257 {
 258      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 259             sizeof(edd.mbr_signature));
 260      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 261      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 262      edd.edd_info_nr = boot_params.eddbuf_entries;
 263 }
 264 #else
 265 static inline void __init copy_edd(void)
 266 {
 267 }
 268 #endif
 269
 270 void * __init extend_brk(size_t size, size_t align)
 271 {
 272         size_t mask = align - 1;
 273         void *ret;
 274
 275         BUG_ON(_brk_start == 0);
 276         BUG_ON(align & mask);
 277
 278         _brk_end = (_brk_end + mask) & ~mask;
 279         BUG_ON((char *)(_brk_end + size) > __brk_limit);
 280
 281         ret = (void *)_brk_end;
 282         _brk_end += size;
 283
 284         memset(ret, 0, size);
 285
 286         return ret;
 287 }
 288
 289 #ifdef CONFIG_X86_64
 290 static void __init init_gbpages(void)
 291 {
 292         if (direct_gbpages && cpu_has_gbpages)
 293                 printk(KERN_INFO "Using GB pages for direct mapping\n");
 294         else
 295                 direct_gbpages = 0;
 296 }
 297 #else
 298 static inline void init_gbpages(void)
 299 {
 300 }
 301 #endif
 302
 303 static void __init reserve_brk(void)
 304 {
 305         if (_brk_end > _brk_start)
 306                 memblock_x86_reserve_range(__pa(_brk_start), __pa(_brk_end), "BRK");
 307
 308         /* Mark brk area as locked down and no longer taking any
 309            new allocations */
 310         _brk_start = 0;
 311 }
 312
 313 #ifdef CONFIG_BLK_DEV_INITRD
 314
 315 #define MAX_MAP_CHUNK   (NR_FIX_BTMAPS << PAGE_SHIFT)
 316 static void __init relocate_initrd(void)
 317 {
 318         /* Assume only end is not page aligned */
 319         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 320         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 321         u64 area_size     = PAGE_ALIGN(ramdisk_size);
 322         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 323         u64 ramdisk_here;
 324         unsigned long slop, clen, mapaddr;
 325         char *p, *q;
 326
 327         /* We need to move the initrd down into lowmem */
 328         ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
 329                                          PAGE_SIZE);
 330
 331         if (ramdisk_here == MEMBLOCK_ERROR)
 332                 panic("Cannot find place for new RAMDISK of size %lld\n",
 333                          ramdisk_size);
 334
 335         /* Note: this includes all the lowmem currently occupied by
 336            the initrd, we rely on that fact to keep the data intact. */
 337         memblock_x86_reserve_range(ramdisk_here, ramdisk_here + area_size, "NEW RAMDISK");
 338         initrd_start = ramdisk_here + PAGE_OFFSET;
 339         initrd_end   = initrd_start + ramdisk_size;
 340         printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
 341                          ramdisk_here, ramdisk_here + ramdisk_size);
 342
 343         q = (char *)initrd_start;
 344
 345         /* Copy any lowmem portion of the initrd */
 346         if (ramdisk_image < end_of_lowmem) {
 347                 clen = end_of_lowmem - ramdisk_image;
 348                 p = (char *)__va(ramdisk_image);
 349                 memcpy(q, p, clen);
 350                 q += clen;
 351                 ramdisk_image += clen;
 352                 ramdisk_size  -= clen;
 353         }
 354
 355         /* Copy the highmem portion of the initrd */
 356         while (ramdisk_size) {
 357                 slop = ramdisk_image & ~PAGE_MASK;
 358                 clen = ramdisk_size;
 359                 if (clen > MAX_MAP_CHUNK-slop)
 360                         clen = MAX_MAP_CHUNK-slop;
 361                 mapaddr = ramdisk_image & PAGE_MASK;
 362                 p = early_memremap(mapaddr, clen+slop);
 363                 memcpy(q, p+slop, clen);
 364                 early_iounmap(p, clen+slop);
 365                 q += clen;
 366                 ramdisk_image += clen;
 367                 ramdisk_size  -= clen;
 368         }
 369         /* high pages is not converted by early_res_to_bootmem */
 370         ramdisk_image = boot_params.hdr.ramdisk_image;
 371         ramdisk_size  = boot_params.hdr.ramdisk_size;
 372         printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
 373                 " %08llx - %08llx\n",
 374                 ramdisk_image, ramdisk_image + ramdisk_size - 1,
 375                 ramdisk_here, ramdisk_here + ramdisk_size - 1);
 376 }
 377
 378 static void __init reserve_initrd(void)
 379 {
 380         /* Assume only end is not page aligned */
 381         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 382         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 383         u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 384         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 385
 386         if (!boot_params.hdr.type_of_loader ||
 387             !ramdisk_image || !ramdisk_size)
 388                 return;         /* No initrd provided by bootloader */
 389
 390         initrd_start = 0;
 391
 392         if (ramdisk_size >= (end_of_lowmem>>1)) {
 393                 memblock_x86_free_range(ramdisk_image, ramdisk_end);
 394                 printk(KERN_ERR "initrd too large to handle, "
 395                        "disabling initrd\n");
 396                 return;
 397         }
 398
 399         printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
 400                         ramdisk_end);
 401
 402
 403         if (ramdisk_end <= end_of_lowmem) {
 404                 /* All in lowmem, easy case */
 405                 /*
 406                  * don't need to reserve again, already reserved early
 407                  * in i386_start_kernel
 408                  */
 409                 initrd_start = ramdisk_image + PAGE_OFFSET;
 410                 initrd_end = initrd_start + ramdisk_size;
 411                 return;
 412         }
 413
 414         relocate_initrd();
 415
 416         memblock_x86_free_range(ramdisk_image, ramdisk_end);
 417 }
 418 #else
 419 static void __init reserve_initrd(void)
 420 {
 421 }
 422 #endif /* CONFIG_BLK_DEV_INITRD */
 423
 424 static void __init parse_setup_data(void)
 425 {
 426         struct setup_data *data;
 427         u64 pa_data;
 428
 429         if (boot_params.hdr.version < 0x0209)
 430                 return;
 431         pa_data = boot_params.hdr.setup_data;
 432         while (pa_data) {
 433                 data = early_memremap(pa_data, PAGE_SIZE);
 434                 switch (data->type) {
 435                 case SETUP_E820_EXT:
 436                         parse_e820_ext(data, pa_data);
 437                         break;
 438                 default:
 439                         break;
 440                 }
 441                 pa_data = data->next;
 442                 early_iounmap(data, PAGE_SIZE);
 443         }
 444 }
 445
 446 static void __init e820_reserve_setup_data(void)
 447 {
 448         struct setup_data *data;
 449         u64 pa_data;
 450         int found = 0;
 451
 452         if (boot_params.hdr.version < 0x0209)
 453                 return;
 454         pa_data = boot_params.hdr.setup_data;
 455         while (pa_data) {
 456                 data = early_memremap(pa_data, sizeof(*data));
 457                 e820_update_range(pa_data, sizeof(*data)+data->len,
 458                          E820_RAM, E820_RESERVED_KERN);
 459                 found = 1;
 460                 pa_data = data->next;
 461                 early_iounmap(data, sizeof(*data));
 462         }
 463         if (!found)
 464                 return;
 465
 466         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 467         memcpy(&e820_saved, &e820, sizeof(struct e820map));
 468         printk(KERN_INFO "extended physical RAM map:\n");
 469         e820_print_map("reserve setup_data");
 470 }
 471
 472 static void __init memblock_x86_reserve_range_setup_data(void)
 473 {
 474         struct setup_data *data;
 475         u64 pa_data;
 476         char buf[32];
 477
 478         if (boot_params.hdr.version < 0x0209)
 479                 return;
 480         pa_data = boot_params.hdr.setup_data;
 481         while (pa_data) {
 482                 data = early_memremap(pa_data, sizeof(*data));
 483                 sprintf(buf, "setup data %x", data->type);
 484                 memblock_x86_reserve_range(pa_data, pa_data+sizeof(*data)+data->len, buf);
 485                 pa_data = data->next;
 486                 early_iounmap(data, sizeof(*data));
 487         }
 488 }
 489
 490 /*
 491  * --------- Crashkernel reservation ------------------------------
 492  */
 493
 494 #ifdef CONFIG_KEXEC
 495
 496 static inline unsigned long long get_total_mem(void)
 497 {
 498         unsigned long long total;
 499
 500         total = max_pfn - min_low_pfn;
 501
 502         return total << PAGE_SHIFT;
 503 }
 504
 505 #define DEFAULT_BZIMAGE_ADDR_MAX 0x37FFFFFF
 506 static void __init reserve_crashkernel(void)
 507 {
 508         unsigned long long total_mem;
 509         unsigned long long crash_size, crash_base;
 510         int ret;
 511
 512         total_mem = get_total_mem();
 513
 514         ret = parse_crashkernel(boot_command_line, total_mem,
 515                         &crash_size, &crash_base);
 516         if (ret != 0 || crash_size <= 0)
 517                 return;
 518
 519         /* 0 means: find the address automatically */
 520         if (crash_base <= 0) {
 521                 const unsigned long long alignment = 16<<20;    /* 16M */
 522
 523                 /*
 524                  *  kexec want bzImage is below DEFAULT_BZIMAGE_ADDR_MAX
 525                  */
 526                 crash_base = memblock_find_in_range(alignment,
 527                                DEFAULT_BZIMAGE_ADDR_MAX, crash_size, alignment);
 528
 529                 if (crash_base == MEMBLOCK_ERROR) {
 530                         pr_info("crashkernel reservation failed - No suitable area found.\n");
 531                         return;
 532                 }
 533         } else {
 534                 unsigned long long start;
 535
 536                 start = memblock_find_in_range(crash_base,
 537                                  crash_base + crash_size, crash_size, 1<<20);
 538                 if (start != crash_base) {
 539                         pr_info("crashkernel reservation failed - memory is in use.\n");
 540                         return;
 541                 }
 542         }
 543         memblock_x86_reserve_range(crash_base, crash_base + crash_size, "CRASH KERNEL");
 544
 545         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 546                         "for crashkernel (System RAM: %ldMB)\n",
 547                         (unsigned long)(crash_size >> 20),
 548                         (unsigned long)(crash_base >> 20),
 549                         (unsigned long)(total_mem >> 20));
 550
 551         crashk_res.start = crash_base;
 552         crashk_res.end   = crash_base + crash_size - 1;
 553         insert_resource(&iomem_resource, &crashk_res);
 554 }
 555 #else
 556 static void __init reserve_crashkernel(void)
 557 {
 558 }
 559 #endif
 560
 561 static struct resource standard_io_resources[] = {
 562         { .name = "dma1", .start = 0x00, .end = 0x1f,
 563                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 564         { .name = "pic1", .start = 0x20, .end = 0x21,
 565                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 566         { .name = "timer0", .start = 0x40, .end = 0x43,
 567                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 568         { .name = "timer1", .start = 0x50, .end = 0x53,
 569                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 570         { .name = "keyboard", .start = 0x60, .end = 0x60,
 571                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 572         { .name = "keyboard", .start = 0x64, .end = 0x64,
 573                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 574         { .name = "dma page reg", .start = 0x80, .end = 0x8f,
 575                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 576         { .name = "pic2", .start = 0xa0, .end = 0xa1,
 577                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 578         { .name = "dma2", .start = 0xc0, .end = 0xdf,
 579                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 580         { .name = "fpu", .start = 0xf0, .end = 0xff,
 581                 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
 582 };
 583
 584 void __init reserve_standard_io_resources(void)
 585 {
 586         int i;
 587
 588         /* request I/O space for devices used on all i[345]86 PCs */
 589         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 590                 request_resource(&ioport_resource, &standard_io_resources[i]);
 591
 592 }
 593
 594 /*
 595  * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
 596  * is_kdump_kernel() to determine if we are booting after a panic. Hence
 597  * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
 598  */
 599
 600 #ifdef CONFIG_CRASH_DUMP
 601 /* elfcorehdr= specifies the location of elf core header
 602  * stored by the crashed kernel. This option will be passed
 603  * by kexec loader to the capture kernel.
 604  */
 605 static int __init setup_elfcorehdr(char *arg)
 606 {
 607         char *end;
 608         if (!arg)
 609                 return -EINVAL;
 610         elfcorehdr_addr = memparse(arg, &end);
 611         return end > arg ? 0 : -EINVAL;
 612 }
 613 early_param("elfcorehdr", setup_elfcorehdr);
 614 #endif
 615
 616 static __init void reserve_ibft_region(void)
 617 {
 618         unsigned long addr, size = 0;
 619
 620         addr = find_ibft_region(&size);
 621
 622         if (size)
 623                 memblock_x86_reserve_range(addr, addr + size, "* ibft");
 624 }
 625
 626 #ifdef CONFIG_X86_RESERVE_LOW_64K
 627 static int __init dmi_low_memory_corruption(const struct dmi_system_id *d)
 628 {
 629         printk(KERN_NOTICE
 630                 "%s detected: BIOS may corrupt low RAM, working around it.\n",
 631                 d->ident);
 632
 633         e820_update_range(0, 0x10000, E820_RAM, E820_RESERVED);
 634         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 635
 636         return 0;
 637 }
 638 #endif
 639
 640 /* List of systems that have known low memory corruption BIOS problems */
 641 static struct dmi_system_id __initdata bad_bios_dmi_table[] = {
 642 #ifdef CONFIG_X86_RESERVE_LOW_64K
 643         {
 644                 .callback = dmi_low_memory_corruption,
 645                 .ident = "AMI BIOS",
 646                 .matches = {
 647                         DMI_MATCH(DMI_BIOS_VENDOR, "American Megatrends Inc."),
 648                 },
 649         },
 650         {
 651                 .callback = dmi_low_memory_corruption,
 652                 .ident = "Phoenix BIOS",
 653                 .matches = {
 654                         DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix Technologies"),
 655                 },
 656         },
 657         {
 658                 .callback = dmi_low_memory_corruption,
 659                 .ident = "Phoenix/MSC BIOS",
 660                 .matches = {
 661                         DMI_MATCH(DMI_BIOS_VENDOR, "Phoenix/MSC"),
 662                 },
 663         },
 664         /*
 665          * AMI BIOS with low memory corruption was found on Intel DG45ID and
 666          * DG45FC boards.
 667          * It has a different DMI_BIOS_VENDOR = "Intel Corp.", for now we will
 668          * match only DMI_BOARD_NAME and see if there is more bad products
 669          * with this vendor.
 670          */
 671         {
 672                 .callback = dmi_low_memory_corruption,
 673                 .ident = "AMI BIOS",
 674                 .matches = {
 675                         DMI_MATCH(DMI_BOARD_NAME, "DG45ID"),
 676                 },
 677         },
 678         {
 679                 .callback = dmi_low_memory_corruption,
 680                 .ident = "AMI BIOS",
 681                 .matches = {
 682                         DMI_MATCH(DMI_BOARD_NAME, "DG45FC"),
 683                 },
 684         },
 685         /*
 686          * The Dell Inspiron Mini 1012 has DMI_BIOS_VENDOR = "Dell Inc.", so
 687          * match on the product name.
 688          */
 689         {
 690                 .callback = dmi_low_memory_corruption,
 691                 .ident = "Phoenix BIOS",
 692                 .matches = {
 693                         DMI_MATCH(DMI_PRODUCT_NAME, "Inspiron 1012"),
 694                 },
 695         },
 696 #endif
 697         {}
 698 };
 699
 700 static void __init trim_bios_range(void)
 701 {
 702         /*
 703          * A special case is the first 4Kb of memory;
 704          * This is a BIOS owned area, not kernel ram, but generally
 705          * not listed as such in the E820 table.
 706          */
 707         e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
 708         /*
 709          * special case: Some BIOSen report the PC BIOS
 710          * area (640->1Mb) as ram even though it is not.
 711          * take them out.
 712          */
 713         e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 714         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 715 }
 716
 717 static u64 __init get_max_mapped(void)
 718 {
 719         u64 end = max_pfn_mapped;
 720
 721         end <<= PAGE_SHIFT;
 722
 723         return end;
 724 }
 725
 726 /*
 727  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 728  * passed the efi memmap, systab, etc., so we should use these data structures
 729  * for initialization.  Note, the efi init code path is determined by the
 730  * global efi_enabled. This allows the same kernel image to be used on existing
 731  * systems (with a traditional BIOS) as well as on EFI systems.
 732  */
 733 /*
 734  * setup_arch - architecture-specific boot-time initializations
 735  *
 736  * Note: On x86_64, fixmaps are ready for use even before this is called.
 737  */
 738
 739 void __init setup_arch(char **cmdline_p)
 740 {
 741         int acpi = 0;
 742         int k8 = 0;
 743
 744 #ifdef CONFIG_X86_32
 745         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 746         visws_early_detect();
 747 #else
 748         printk(KERN_INFO "Command line: %s\n", boot_command_line);
 749 #endif
 750
 751         /* VMI may relocate the fixmap; do this before touching ioremap area */
 752         vmi_init();
 753
 754         /* OFW also may relocate the fixmap */
 755         olpc_ofw_detect();
 756
 757         early_trap_init();
 758         early_cpu_init();
 759         early_ioremap_init();
 760
 761         setup_olpc_ofw_pgd();
 762
 763         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 764         screen_info = boot_params.screen_info;
 765         edid_info = boot_params.edid_info;
 766 #ifdef CONFIG_X86_32
 767         apm_info.bios = boot_params.apm_bios_info;
 768         ist_info = boot_params.ist_info;
 769         if (boot_params.sys_desc_table.length != 0) {
 770                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 771                 machine_id = boot_params.sys_desc_table.table[0];
 772                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 773                 BIOS_revision = boot_params.sys_desc_table.table[2];
 774         }
 775 #endif
 776         saved_video_mode = boot_params.hdr.vid_mode;
 777         bootloader_type = boot_params.hdr.type_of_loader;
 778         if ((bootloader_type >> 4) == 0xe) {
 779                 bootloader_type &= 0xf;
 780                 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 781         }
 782         bootloader_version  = bootloader_type & 0xf;
 783         bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 784
 785 #ifdef CONFIG_BLK_DEV_RAM
 786         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 787         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 788         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 789 #endif
 790 #ifdef CONFIG_EFI
 791         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 792 #ifdef CONFIG_X86_32
 793                      "EL32",
 794 #else
 795                      "EL64",
 796 #endif
 797          4)) {
 798                 efi_enabled = 1;
 799                 efi_memblock_x86_reserve_range();
 800         }
 801 #endif
 802
 803         x86_init.oem.arch_setup();
 804
 805         setup_memory_map();
 806         parse_setup_data();
 807         /* update the e820_saved too */
 808         e820_reserve_setup_data();
 809
 810         copy_edd();
 811
 812         if (!boot_params.hdr.root_flags)
 813                 root_mountflags &= ~MS_RDONLY;
 814         init_mm.start_code = (unsigned long) _text;
 815         init_mm.end_code = (unsigned long) _etext;
 816         init_mm.end_data = (unsigned long) _edata;
 817         init_mm.brk = _brk_end;
 818
 819         code_resource.start = virt_to_phys(_text);
 820         code_resource.end = virt_to_phys(_etext)-1;
 821         data_resource.start = virt_to_phys(_etext);
 822         data_resource.end = virt_to_phys(_edata)-1;
 823         bss_resource.start = virt_to_phys(&__bss_start);
 824         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 825
 826 #ifdef CONFIG_CMDLINE_BOOL
 827 #ifdef CONFIG_CMDLINE_OVERRIDE
 828         strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 829 #else
 830         if (builtin_cmdline[0]) {
 831                 /* append boot loader cmdline to builtin */
 832                 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 833                 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 834                 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 835         }
 836 #endif
 837 #endif
 838
 839         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 840         *cmdline_p = command_line;
 841
 842         /*
 843          * x86_configure_nx() is called before parse_early_param() to detect
 844          * whether hardware doesn't support NX (so that the early EHCI debug
 845          * console setup can safely call set_fixmap()). It may then be called
 846          * again from within noexec_setup() during parsing early parameters
 847          * to honor the respective command line option.
 848          */
 849         x86_configure_nx();
 850
 851         parse_early_param();
 852
 853         x86_report_nx();
 854
 855         /* Must be before kernel pagetables are setup */
 856         vmi_activate();
 857
 858         /* after early param, so could get panic from serial */
 859         memblock_x86_reserve_range_setup_data();
 860
 861         if (acpi_mps_check()) {
 862 #ifdef CONFIG_X86_LOCAL_APIC
 863                 disable_apic = 1;
 864 #endif
 865                 setup_clear_cpu_cap(X86_FEATURE_APIC);
 866         }
 867
 868 #ifdef CONFIG_PCI
 869         if (pci_early_dump_regs)
 870                 early_dump_pci_devices();
 871 #endif
 872
 873         finish_e820_parsing();
 874
 875         if (efi_enabled)
 876                 efi_init();
 877
 878         dmi_scan_machine();
 879
 880         dmi_check_system(bad_bios_dmi_table);
 881
 882         /*
 883          * VMware detection requires dmi to be available, so this
 884          * needs to be done after dmi_scan_machine, for the BP.
 885          */
 886         init_hypervisor_platform();
 887
 888         x86_init.resources.probe_roms();
 889
 890         /* after parse_early_param, so could debug it */
 891         insert_resource(&iomem_resource, &code_resource);
 892         insert_resource(&iomem_resource, &data_resource);
 893         insert_resource(&iomem_resource, &bss_resource);
 894
 895         trim_bios_range();
 896 #ifdef CONFIG_X86_32
 897         if (ppro_with_ram_bug()) {
 898                 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
 899                                   E820_RESERVED);
 900                 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 901                 printk(KERN_INFO "fixed physical RAM map:\n");
 902                 e820_print_map("bad_ppro");
 903         }
 904 #else
 905         early_gart_iommu_check();
 906 #endif
 907
 908         /*
 909          * partially used pages are not usable - thus
 910          * we are rounding upwards:
 911          */
 912         max_pfn = e820_end_of_ram_pfn();
 913
 914         /* update e820 for memory not covered by WB MTRRs */
 915         mtrr_bp_init();
 916         if (mtrr_trim_uncached_memory(max_pfn))
 917                 max_pfn = e820_end_of_ram_pfn();
 918
 919 #ifdef CONFIG_X86_32
 920         /* max_low_pfn get updated here */
 921         find_low_pfn_range();
 922 #else
 923         num_physpages = max_pfn;
 924
 925         check_x2apic();
 926
 927         /* How many end-of-memory variables you have, grandma! */
 928         /* need this before calling reserve_initrd */
 929         if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
 930                 max_low_pfn = e820_end_of_low_ram_pfn();
 931         else
 932                 max_low_pfn = max_pfn;
 933
 934         high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
 935         max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT;
 936 #endif
 937
 938         /*
 939          * Find and reserve possible boot-time SMP configuration:
 940          */
 941         find_smp_config();
 942
 943         reserve_ibft_region();
 944
 945         /*
 946          * Need to conclude brk, before memblock_x86_fill()
 947          *  it could use memblock_find_in_range, could overlap with
 948          *  brk area.
 949          */
 950         reserve_brk();
 951
 952         memblock.current_limit = get_max_mapped();
 953         memblock_x86_fill();
 954
 955         /* preallocate 4k for mptable mpc */
 956         early_reserve_e820_mpc_new();
 957
 958 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
 959         setup_bios_corruption_check();
 960 #endif
 961
 962         printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
 963                         max_pfn_mapped<<PAGE_SHIFT);
 964
 965         reserve_trampoline_memory();
 966
 967 #ifdef CONFIG_ACPI_SLEEP
 968         /*
 969          * Reserve low memory region for sleep support.
 970          * even before init_memory_mapping
 971          */
 972         acpi_reserve_wakeup_memory();
 973 #endif
 974         init_gbpages();
 975
 976         /* max_pfn_mapped is updated here */
 977         max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
 978         max_pfn_mapped = max_low_pfn_mapped;
 979
 980 #ifdef CONFIG_X86_64
 981         if (max_pfn > max_low_pfn) {
 982                 max_pfn_mapped = init_memory_mapping(1UL<<32,
 983                                                      max_pfn<<PAGE_SHIFT);
 984                 /* can we preseve max_low_pfn ?*/
 985                 max_low_pfn = max_pfn;
 986         }
 987 #endif
 988         memblock.current_limit = get_max_mapped();
 989
 990         /*
 991          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 992          */
 993
 994 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 995         if (init_ohci1394_dma_early)
 996                 init_ohci1394_dma_on_all_controllers();
 997 #endif
 998
 999         reserve_initrd();
1000
1001         reserve_crashkernel();
1002
1003         vsmp_init();
1004
1005         io_delay_init();
1006
1007         /*
1008          * Parse the ACPI tables for possible boot-time SMP configuration.
1009          */
1010         acpi_boot_table_init();
1011
1012         early_acpi_boot_init();
1013
1014 #ifdef CONFIG_ACPI_NUMA
1015         /*
1016          * Parse SRAT to discover nodes.
1017          */
1018         acpi = acpi_numa_init();
1019 #endif
1020
1021 #ifdef CONFIG_K8_NUMA
1022         if (!acpi)
1023                 k8 = !k8_numa_init(0, max_pfn);
1024 #endif
1025
1026         initmem_init(0, max_pfn, acpi, k8);
1027         memblock_find_dma_reserve();
1028         dma32_reserve_bootmem();
1029
1030 #ifdef CONFIG_KVM_CLOCK
1031         kvmclock_init();
1032 #endif
1033
1034         x86_init.paging.pagetable_setup_start(swapper_pg_dir);
1035         paging_init();
1036         x86_init.paging.pagetable_setup_done(swapper_pg_dir);
1037
1038         setup_trampoline_page_table();
1039
1040         tboot_probe();
1041
1042 #ifdef CONFIG_X86_64
1043         map_vsyscall();
1044 #endif
1045
1046         generic_apic_probe();
1047
1048         early_quirks();
1049
1050         /*
1051          * Read APIC and some other early information from ACPI tables.
1052          */
1053         acpi_boot_init();
1054
1055         sfi_init();
1056
1057         /*
1058          * get boot-time SMP configuration:
1059          */
1060         if (smp_found_config)
1061                 get_smp_config();
1062
1063         prefill_possible_map();
1064
1065 #ifdef CONFIG_X86_64
1066         init_cpu_to_node();
1067 #endif
1068
1069         init_apic_mappings();
1070         ioapic_init_mappings();
1071
1072         /* need to wait for io_apic is mapped */
1073         probe_nr_irqs_gsi();
1074
1075         kvm_guest_init();
1076
1077         e820_reserve_resources();
1078         e820_mark_nosave_regions(max_low_pfn);
1079
1080         x86_init.resources.reserve_resources();
1081
1082         e820_setup_gap();
1083
1084 #ifdef CONFIG_VT
1085 #if defined(CONFIG_VGA_CONSOLE)
1086         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1087                 conswitchp = &vga_con;
1088 #elif defined(CONFIG_DUMMY_CONSOLE)
1089         conswitchp = &dummy_con;
1090 #endif
1091 #endif
1092         x86_init.oem.banner();
1093
1094         mcheck_init();
1095 }
1096
1097 #ifdef CONFIG_X86_32
1098
1099 static struct resource video_ram_resource = {
1100         .name   = "Video RAM area",
1101         .start  = 0xa0000,
1102         .end    = 0xbffff,
1103         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
1104 };
1105
1106 void __init i386_reserve_resources(void)
1107 {
1108         request_resource(&iomem_resource, &video_ram_resource);
1109         reserve_standard_io_resources();
1110 }
1111
1112 #endif /* CONFIG_X86_32 */