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/setup_arch.h>
  88 #include <asm/bios_ebda.h>
  89 #include <asm/cacheflush.h>
  90 #include <asm/processor.h>
  91 #include <asm/bugs.h>
  92
  93 #include <asm/system.h>
  94 #include <asm/vsyscall.h>
  95 #include <asm/cpu.h>
  96 #include <asm/desc.h>
  97 #include <asm/dma.h>
  98 #include <asm/iommu.h>
  99 #include <asm/gart.h>
 100 #include <asm/mmu_context.h>
 101 #include <asm/proto.h>
 102
 103 #include <asm/paravirt.h>
 104 #include <asm/hypervisor.h>
 105 #include <asm/olpc_ofw.h>
 106
 107 #include <asm/percpu.h>
 108 #include <asm/topology.h>
 109 #include <asm/apicdef.h>
 110 #include <asm/amd_nb.h>
 111 #ifdef CONFIG_X86_64
 112 #include <asm/numa_64.h>
 113 #endif
 114 #include <asm/mce.h>
 115 #include <asm/alternative.h>
 116
 117 /*
 118  * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
 119  * The direct mapping extends to max_pfn_mapped, so that we can directly access
 120  * apertures, ACPI and other tables without having to play with fixmaps.
 121  */
 122 unsigned long max_low_pfn_mapped;
 123 unsigned long max_pfn_mapped;
 124
 125 #ifdef CONFIG_DMI
 126 RESERVE_BRK(dmi_alloc, 65536);
 127 #endif
 128
 129
 130 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
 131 unsigned long _brk_end = (unsigned long)__brk_base;
 132
 133 #ifdef CONFIG_X86_64
 134 int default_cpu_present_to_apicid(int mps_cpu)
 135 {
 136         return __default_cpu_present_to_apicid(mps_cpu);
 137 }
 138
 139 int default_check_phys_apicid_present(int phys_apicid)
 140 {
 141         return __default_check_phys_apicid_present(phys_apicid);
 142 }
 143 #endif
 144
 145 #ifndef CONFIG_DEBUG_BOOT_PARAMS
 146 struct boot_params __initdata boot_params;
 147 #else
 148 struct boot_params boot_params;
 149 #endif
 150
 151 /*
 152  * Machine setup..
 153  */
 154 static struct resource data_resource = {
 155         .name   = "Kernel data",
 156         .start  = 0,
 157         .end    = 0,
 158         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 159 };
 160
 161 static struct resource code_resource = {
 162         .name   = "Kernel code",
 163         .start  = 0,
 164         .end    = 0,
 165         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 166 };
 167
 168 static struct resource bss_resource = {
 169         .name   = "Kernel bss",
 170         .start  = 0,
 171         .end    = 0,
 172         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 173 };
 174
 175
 176 #ifdef CONFIG_X86_32
 177 /* cpu data as detected by the assembly code in head.S */
 178 struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1};
 179 /* common cpu data for all cpus */
 180 struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1};
 181 EXPORT_SYMBOL(boot_cpu_data);
 182 static void set_mca_bus(int x)
 183 {
 184 #ifdef CONFIG_MCA
 185         MCA_bus = x;
 186 #endif
 187 }
 188
 189 unsigned int def_to_bigsmp;
 190
 191 /* for MCA, but anyone else can use it if they want */
 192 unsigned int machine_id;
 193 unsigned int machine_submodel_id;
 194 unsigned int BIOS_revision;
 195
 196 struct apm_info apm_info;
 197 EXPORT_SYMBOL(apm_info);
 198
 199 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 200         defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 201 struct ist_info ist_info;
 202 EXPORT_SYMBOL(ist_info);
 203 #else
 204 struct ist_info ist_info;
 205 #endif
 206
 207 #else
 208 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
 209         .x86_phys_bits = MAX_PHYSMEM_BITS,
 210 };
 211 EXPORT_SYMBOL(boot_cpu_data);
 212 #endif
 213
 214
 215 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
 216 unsigned long mmu_cr4_features;
 217 #else
 218 unsigned long mmu_cr4_features = X86_CR4_PAE;
 219 #endif
 220
 221 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
 222 int bootloader_type, bootloader_version;
 223
 224 /*
 225  * Setup options
 226  */
 227 struct screen_info screen_info;
 228 EXPORT_SYMBOL(screen_info);
 229 struct edid_info edid_info;
 230 EXPORT_SYMBOL_GPL(edid_info);
 231
 232 extern int root_mountflags;
 233
 234 unsigned long saved_video_mode;
 235
 236 #define RAMDISK_IMAGE_START_MASK        0x07FF
 237 #define RAMDISK_PROMPT_FLAG             0x8000
 238 #define RAMDISK_LOAD_FLAG               0x4000
 239
 240 static char __initdata command_line[COMMAND_LINE_SIZE];
 241 #ifdef CONFIG_CMDLINE_BOOL
 242 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
 243 #endif
 244
 245 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 246 struct edd edd;
 247 #ifdef CONFIG_EDD_MODULE
 248 EXPORT_SYMBOL(edd);
 249 #endif
 250 /**
 251  * copy_edd() - Copy the BIOS EDD information
 252  *              from boot_params into a safe place.
 253  *
 254  */
 255 static inline void __init copy_edd(void)
 256 {
 257      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 258             sizeof(edd.mbr_signature));
 259      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 260      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 261      edd.edd_info_nr = boot_params.eddbuf_entries;
 262 }
 263 #else
 264 static inline void __init copy_edd(void)
 265 {
 266 }
 267 #endif
 268
 269 void * __init extend_brk(size_t size, size_t align)
 270 {
 271         size_t mask = align - 1;
 272         void *ret;
 273
 274         BUG_ON(_brk_start == 0);
 275         BUG_ON(align & mask);
 276
 277         _brk_end = (_brk_end + mask) & ~mask;
 278         BUG_ON((char *)(_brk_end + size) > __brk_limit);
 279
 280         ret = (void *)_brk_end;
 281         _brk_end += size;
 282
 283         memset(ret, 0, size);
 284
 285         return ret;
 286 }
 287
 288 #ifdef CONFIG_X86_64
 289 static void __init init_gbpages(void)
 290 {
 291         if (direct_gbpages && cpu_has_gbpages)
 292                 printk(KERN_INFO "Using GB pages for direct mapping\n");
 293         else
 294                 direct_gbpages = 0;
 295 }
 296 #else
 297 static inline void init_gbpages(void)
 298 {
 299 }
 300 #endif
 301
 302 static void __init reserve_brk(void)
 303 {
 304         if (_brk_end > _brk_start)
 305                 memblock_x86_reserve_range(__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         /* Assume only end is not page aligned */
 318         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 319         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 320         u64 area_size     = PAGE_ALIGN(ramdisk_size);
 321         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 322         u64 ramdisk_here;
 323         unsigned long slop, clen, mapaddr;
 324         char *p, *q;
 325
 326         /* We need to move the initrd down into lowmem */
 327         ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
 328                                          PAGE_SIZE);
 329
 330         if (ramdisk_here == MEMBLOCK_ERROR)
 331                 panic("Cannot find place for new RAMDISK of size %lld\n",
 332                          ramdisk_size);
 333
 334         /* Note: this includes all the lowmem currently occupied by
 335            the initrd, we rely on that fact to keep the data intact. */
 336         memblock_x86_reserve_range(ramdisk_here, ramdisk_here + area_size, "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         /* Assume only end is not page aligned */
 380         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 381         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 382         u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 383         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 384
 385         if (!boot_params.hdr.type_of_loader ||
 386             !ramdisk_image || !ramdisk_size)
 387                 return;         /* No initrd provided by bootloader */
 388
 389         initrd_start = 0;
 390
 391         if (ramdisk_size >= (end_of_lowmem>>1)) {
 392                 memblock_x86_free_range(ramdisk_image, ramdisk_end);
 393                 printk(KERN_ERR "initrd too large to handle, "
 394                        "disabling initrd\n");
 395                 return;
 396         }
 397
 398         printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
 399                         ramdisk_end);
 400
 401
 402         if (ramdisk_end <= end_of_lowmem) {
 403                 /* All in lowmem, easy case */
 404                 /*
 405                  * don't need to reserve again, already reserved early
 406                  * in i386_start_kernel
 407                  */
 408                 initrd_start = ramdisk_image + PAGE_OFFSET;
 409                 initrd_end = initrd_start + ramdisk_size;
 410                 return;
 411         }
 412
 413         relocate_initrd();
 414
 415         memblock_x86_free_range(ramdisk_image, ramdisk_end);
 416 }
 417 #else
 418 static void __init reserve_initrd(void)
 419 {
 420 }
 421 #endif /* CONFIG_BLK_DEV_INITRD */
 422
 423 static void __init parse_setup_data(void)
 424 {
 425         struct setup_data *data;
 426         u64 pa_data;
 427
 428         if (boot_params.hdr.version < 0x0209)
 429                 return;
 430         pa_data = boot_params.hdr.setup_data;
 431         while (pa_data) {
 432                 data = early_memremap(pa_data, PAGE_SIZE);
 433                 switch (data->type) {
 434                 case SETUP_E820_EXT:
 435                         parse_e820_ext(data, pa_data);
 436                         break;
 437                 default:
 438                         break;
 439                 }
 440                 pa_data = data->next;
 441                 early_iounmap(data, PAGE_SIZE);
 442         }
 443 }
 444
 445 static void __init e820_reserve_setup_data(void)
 446 {
 447         struct setup_data *data;
 448         u64 pa_data;
 449         int found = 0;
 450
 451         if (boot_params.hdr.version < 0x0209)
 452                 return;
 453         pa_data = boot_params.hdr.setup_data;
 454         while (pa_data) {
 455                 data = early_memremap(pa_data, sizeof(*data));
 456                 e820_update_range(pa_data, sizeof(*data)+data->len,
 457                          E820_RAM, E820_RESERVED_KERN);
 458                 found = 1;
 459                 pa_data = data->next;
 460                 early_iounmap(data, sizeof(*data));
 461         }
 462         if (!found)
 463                 return;
 464
 465         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 466         memcpy(&e820_saved, &e820, sizeof(struct e820map));
 467         printk(KERN_INFO "extended physical RAM map:\n");
 468         e820_print_map("reserve setup_data");
 469 }
 470
 471 static void __init memblock_x86_reserve_range_setup_data(void)
 472 {
 473         struct setup_data *data;
 474         u64 pa_data;
 475         char buf[32];
 476
 477         if (boot_params.hdr.version < 0x0209)
 478                 return;
 479         pa_data = boot_params.hdr.setup_data;
 480         while (pa_data) {
 481                 data = early_memremap(pa_data, sizeof(*data));
 482                 sprintf(buf, "setup data %x", data->type);
 483                 memblock_x86_reserve_range(pa_data, pa_data+sizeof(*data)+data->len, buf);
 484                 pa_data = data->next;
 485                 early_iounmap(data, sizeof(*data));
 486         }
 487 }
 488
 489 /*
 490  * --------- Crashkernel reservation ------------------------------
 491  */
 492
 493 #ifdef CONFIG_KEXEC
 494
 495 static inline unsigned long long get_total_mem(void)
 496 {
 497         unsigned long long total;
 498
 499         total = max_pfn - min_low_pfn;
 500
 501         return total << PAGE_SHIFT;
 502 }
 503
 504 #define DEFAULT_BZIMAGE_ADDR_MAX 0x37FFFFFF
 505 static void __init reserve_crashkernel(void)
 506 {
 507         unsigned long long total_mem;
 508         unsigned long long crash_size, crash_base;
 509         int ret;
 510
 511         total_mem = get_total_mem();
 512
 513         ret = parse_crashkernel(boot_command_line, total_mem,
 514                         &crash_size, &crash_base);
 515         if (ret != 0 || crash_size <= 0)
 516                 return;
 517
 518         /* 0 means: find the address automatically */
 519         if (crash_base <= 0) {
 520                 const unsigned long long alignment = 16<<20;    /* 16M */
 521
 522                 /*
 523                  *  kexec want bzImage is below DEFAULT_BZIMAGE_ADDR_MAX
 524                  */
 525                 crash_base = memblock_find_in_range(alignment,
 526                                DEFAULT_BZIMAGE_ADDR_MAX, crash_size, alignment);
 527
 528                 if (crash_base == MEMBLOCK_ERROR) {
 529                         pr_info("crashkernel reservation failed - No suitable area found.\n");
 530                         return;
 531                 }
 532         } else {
 533                 unsigned long long start;
 534
 535                 start = memblock_find_in_range(crash_base,
 536                                  crash_base + crash_size, crash_size, 1<<20);
 537                 if (start != crash_base) {
 538                         pr_info("crashkernel reservation failed - memory is in use.\n");
 539                         return;
 540                 }
 541         }
 542         memblock_x86_reserve_range(crash_base, crash_base + crash_size, "CRASH KERNEL");
 543
 544         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 545                         "for crashkernel (System RAM: %ldMB)\n",
 546                         (unsigned long)(crash_size >> 20),
 547                         (unsigned long)(crash_base >> 20),
 548                         (unsigned long)(total_mem >> 20));
 549
 550         crashk_res.start = crash_base;
 551         crashk_res.end   = crash_base + crash_size - 1;
 552         insert_resource(&iomem_resource, &crashk_res);
 553 }
 554 #else
 555 static void __init reserve_crashkernel(void)
 556 {
 557 }
 558 #endif
 559
 560 static struct resource standard_io_resources[] = {
 561         { .name = "dma1", .start = 0x00, .end = 0x1f,
 562                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 563         { .name = "pic1", .start = 0x20, .end = 0x21,
 564                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 565         { .name = "timer0", .start = 0x40, .end = 0x43,
 566                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 567         { .name = "timer1", .start = 0x50, .end = 0x53,
 568                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 569         { .name = "keyboard", .start = 0x60, .end = 0x60,
 570                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 571         { .name = "keyboard", .start = 0x64, .end = 0x64,
 572                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 573         { .name = "dma page reg", .start = 0x80, .end = 0x8f,
 574                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 575         { .name = "pic2", .start = 0xa0, .end = 0xa1,
 576                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 577         { .name = "dma2", .start = 0xc0, .end = 0xdf,
 578                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 579         { .name = "fpu", .start = 0xf0, .end = 0xff,
 580                 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
 581 };
 582
 583 void __init reserve_standard_io_resources(void)
 584 {
 585         int i;
 586
 587         /* request I/O space for devices used on all i[345]86 PCs */
 588         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 589                 request_resource(&ioport_resource, &standard_io_resources[i]);
 590
 591 }
 592
 593 /*
 594  * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
 595  * is_kdump_kernel() to determine if we are booting after a panic. Hence
 596  * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
 597  */
 598
 599 #ifdef CONFIG_CRASH_DUMP
 600 /* elfcorehdr= specifies the location of elf core header
 601  * stored by the crashed kernel. This option will be passed
 602  * by kexec loader to the capture kernel.
 603  */
 604 static int __init setup_elfcorehdr(char *arg)
 605 {
 606         char *end;
 607         if (!arg)
 608                 return -EINVAL;
 609         elfcorehdr_addr = memparse(arg, &end);
 610         return end > arg ? 0 : -EINVAL;
 611 }
 612 early_param("elfcorehdr", setup_elfcorehdr);
 613 #endif
 614
 615 static __init void reserve_ibft_region(void)
 616 {
 617         unsigned long addr, size = 0;
 618
 619         addr = find_ibft_region(&size);
 620
 621         if (size)
 622                 memblock_x86_reserve_range(addr, addr + size, "* ibft");
 623 }
 624
 625 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 626
 627 static void __init trim_bios_range(void)
 628 {
 629         /*
 630          * A special case is the first 4Kb of memory;
 631          * This is a BIOS owned area, not kernel ram, but generally
 632          * not listed as such in the E820 table.
 633          *
 634          * This typically reserves additional memory (64KiB by default)
 635          * since some BIOSes are known to corrupt low memory.  See the
 636          * Kconfig help text for X86_RESERVE_LOW.
 637          */
 638         e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
 639                           E820_RAM, E820_RESERVED);
 640
 641         /*
 642          * special case: Some BIOSen report the PC BIOS
 643          * area (640->1Mb) as ram even though it is not.
 644          * take them out.
 645          */
 646         e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 647         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 648 }
 649
 650 static int __init parse_reservelow(char *p)
 651 {
 652         unsigned long long size;
 653
 654         if (!p)
 655                 return -EINVAL;
 656
 657         size = memparse(p, &p);
 658
 659         if (size < 4096)
 660                 size = 4096;
 661
 662         if (size > 640*1024)
 663                 size = 640*1024;
 664
 665         reserve_low = size;
 666
 667         return 0;
 668 }
 669
 670 early_param("reservelow", parse_reservelow);
 671
 672 static u64 __init get_max_mapped(void)
 673 {
 674         u64 end = max_pfn_mapped;
 675
 676         end <<= PAGE_SHIFT;
 677
 678         return end;
 679 }
 680
 681 /*
 682  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 683  * passed the efi memmap, systab, etc., so we should use these data structures
 684  * for initialization.  Note, the efi init code path is determined by the
 685  * global efi_enabled. This allows the same kernel image to be used on existing
 686  * systems (with a traditional BIOS) as well as on EFI systems.
 687  */
 688 /*
 689  * setup_arch - architecture-specific boot-time initializations
 690  *
 691  * Note: On x86_64, fixmaps are ready for use even before this is called.
 692  */
 693
 694 void __init setup_arch(char **cmdline_p)
 695 {
 696         int acpi = 0;
 697         int amd = 0;
 698         unsigned long flags;
 699
 700 #ifdef CONFIG_X86_32
 701         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 702         visws_early_detect();
 703
 704         /*
 705          * copy kernel address range established so far and switch
 706          * to the proper swapper page table
 707          */
 708         clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 709                         initial_page_table + KERNEL_PGD_BOUNDARY,
 710                         KERNEL_PGD_PTRS);
 711
 712         load_cr3(swapper_pg_dir);
 713         __flush_tlb_all();
 714 #else
 715         printk(KERN_INFO "Command line: %s\n", boot_command_line);
 716 #endif
 717
 718         /*
 719          * If we have OLPC OFW, we might end up relocating the fixmap due to
 720          * reserve_top(), so do this before touching the ioremap area.
 721          */
 722         olpc_ofw_detect();
 723
 724         early_trap_init();
 725         early_cpu_init();
 726         early_ioremap_init();
 727
 728         setup_olpc_ofw_pgd();
 729
 730         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 731         screen_info = boot_params.screen_info;
 732         edid_info = boot_params.edid_info;
 733 #ifdef CONFIG_X86_32
 734         apm_info.bios = boot_params.apm_bios_info;
 735         ist_info = boot_params.ist_info;
 736         if (boot_params.sys_desc_table.length != 0) {
 737                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 738                 machine_id = boot_params.sys_desc_table.table[0];
 739                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 740                 BIOS_revision = boot_params.sys_desc_table.table[2];
 741         }
 742 #endif
 743         saved_video_mode = boot_params.hdr.vid_mode;
 744         bootloader_type = boot_params.hdr.type_of_loader;
 745         if ((bootloader_type >> 4) == 0xe) {
 746                 bootloader_type &= 0xf;
 747                 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 748         }
 749         bootloader_version  = bootloader_type & 0xf;
 750         bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 751
 752 #ifdef CONFIG_BLK_DEV_RAM
 753         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 754         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 755         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 756 #endif
 757 #ifdef CONFIG_EFI
 758         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 759 #ifdef CONFIG_X86_32
 760                      "EL32",
 761 #else
 762                      "EL64",
 763 #endif
 764          4)) {
 765                 efi_enabled = 1;
 766                 efi_memblock_x86_reserve_range();
 767         }
 768 #endif
 769
 770         x86_init.oem.arch_setup();
 771
 772         resource_alloc_from_bottom = 0;
 773         iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 774         setup_memory_map();
 775         parse_setup_data();
 776         /* update the e820_saved too */
 777         e820_reserve_setup_data();
 778
 779         copy_edd();
 780
 781         if (!boot_params.hdr.root_flags)
 782                 root_mountflags &= ~MS_RDONLY;
 783         init_mm.start_code = (unsigned long) _text;
 784         init_mm.end_code = (unsigned long) _etext;
 785         init_mm.end_data = (unsigned long) _edata;
 786         init_mm.brk = _brk_end;
 787
 788         code_resource.start = virt_to_phys(_text);
 789         code_resource.end = virt_to_phys(_etext)-1;
 790         data_resource.start = virt_to_phys(_etext);
 791         data_resource.end = virt_to_phys(_edata)-1;
 792         bss_resource.start = virt_to_phys(&__bss_start);
 793         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 794
 795 #ifdef CONFIG_CMDLINE_BOOL
 796 #ifdef CONFIG_CMDLINE_OVERRIDE
 797         strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 798 #else
 799         if (builtin_cmdline[0]) {
 800                 /* append boot loader cmdline to builtin */
 801                 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 802                 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 803                 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 804         }
 805 #endif
 806 #endif
 807
 808         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 809         *cmdline_p = command_line;
 810
 811         /*
 812          * x86_configure_nx() is called before parse_early_param() to detect
 813          * whether hardware doesn't support NX (so that the early EHCI debug
 814          * console setup can safely call set_fixmap()). It may then be called
 815          * again from within noexec_setup() during parsing early parameters
 816          * to honor the respective command line option.
 817          */
 818         x86_configure_nx();
 819
 820         parse_early_param();
 821
 822         x86_report_nx();
 823
 824         /* after early param, so could get panic from serial */
 825         memblock_x86_reserve_range_setup_data();
 826
 827         if (acpi_mps_check()) {
 828 #ifdef CONFIG_X86_LOCAL_APIC
 829                 disable_apic = 1;
 830 #endif
 831                 setup_clear_cpu_cap(X86_FEATURE_APIC);
 832         }
 833
 834 #ifdef CONFIG_PCI
 835         if (pci_early_dump_regs)
 836                 early_dump_pci_devices();
 837 #endif
 838
 839         finish_e820_parsing();
 840
 841         if (efi_enabled)
 842                 efi_init();
 843
 844         dmi_scan_machine();
 845
 846         /*
 847          * VMware detection requires dmi to be available, so this
 848          * needs to be done after dmi_scan_machine, for the BP.
 849          */
 850         init_hypervisor_platform();
 851
 852         x86_init.resources.probe_roms();
 853
 854         /* after parse_early_param, so could debug it */
 855         insert_resource(&iomem_resource, &code_resource);
 856         insert_resource(&iomem_resource, &data_resource);
 857         insert_resource(&iomem_resource, &bss_resource);
 858
 859         trim_bios_range();
 860 #ifdef CONFIG_X86_32
 861         if (ppro_with_ram_bug()) {
 862                 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
 863                                   E820_RESERVED);
 864                 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 865                 printk(KERN_INFO "fixed physical RAM map:\n");
 866                 e820_print_map("bad_ppro");
 867         }
 868 #else
 869         early_gart_iommu_check();
 870 #endif
 871
 872         /*
 873          * partially used pages are not usable - thus
 874          * we are rounding upwards:
 875          */
 876         max_pfn = e820_end_of_ram_pfn();
 877
 878         /* update e820 for memory not covered by WB MTRRs */
 879         mtrr_bp_init();
 880         if (mtrr_trim_uncached_memory(max_pfn))
 881                 max_pfn = e820_end_of_ram_pfn();
 882
 883 #ifdef CONFIG_X86_32
 884         /* max_low_pfn get updated here */
 885         find_low_pfn_range();
 886 #else
 887         num_physpages = max_pfn;
 888
 889         check_x2apic();
 890
 891         /* How many end-of-memory variables you have, grandma! */
 892         /* need this before calling reserve_initrd */
 893         if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
 894                 max_low_pfn = e820_end_of_low_ram_pfn();
 895         else
 896                 max_low_pfn = max_pfn;
 897
 898         high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
 899 #endif
 900
 901         /*
 902          * Find and reserve possible boot-time SMP configuration:
 903          */
 904         find_smp_config();
 905
 906         reserve_ibft_region();
 907
 908         /*
 909          * Need to conclude brk, before memblock_x86_fill()
 910          *  it could use memblock_find_in_range, could overlap with
 911          *  brk area.
 912          */
 913         reserve_brk();
 914
 915         memblock.current_limit = get_max_mapped();
 916         memblock_x86_fill();
 917
 918         /* preallocate 4k for mptable mpc */
 919         early_reserve_e820_mpc_new();
 920
 921 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
 922         setup_bios_corruption_check();
 923 #endif
 924
 925         printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
 926                         max_pfn_mapped<<PAGE_SHIFT);
 927
 928         reserve_trampoline_memory();
 929
 930 #ifdef CONFIG_ACPI_SLEEP
 931         /*
 932          * Reserve low memory region for sleep support.
 933          * even before init_memory_mapping
 934          */
 935         acpi_reserve_wakeup_memory();
 936 #endif
 937         init_gbpages();
 938
 939         /* max_pfn_mapped is updated here */
 940         max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
 941         max_pfn_mapped = max_low_pfn_mapped;
 942
 943 #ifdef CONFIG_X86_64
 944         if (max_pfn > max_low_pfn) {
 945                 max_pfn_mapped = init_memory_mapping(1UL<<32,
 946                                                      max_pfn<<PAGE_SHIFT);
 947                 /* can we preseve max_low_pfn ?*/
 948                 max_low_pfn = max_pfn;
 949         }
 950 #endif
 951         memblock.current_limit = get_max_mapped();
 952
 953         /*
 954          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 955          */
 956
 957 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 958         if (init_ohci1394_dma_early)
 959                 init_ohci1394_dma_on_all_controllers();
 960 #endif
 961
 962         reserve_initrd();
 963
 964         reserve_crashkernel();
 965
 966         vsmp_init();
 967
 968         io_delay_init();
 969
 970         /*
 971          * Parse the ACPI tables for possible boot-time SMP configuration.
 972          */
 973         acpi_boot_table_init();
 974
 975         early_acpi_boot_init();
 976
 977 #ifdef CONFIG_ACPI_NUMA
 978         /*
 979          * Parse SRAT to discover nodes.
 980          */
 981         acpi = acpi_numa_init();
 982 #endif
 983
 984 #ifdef CONFIG_AMD_NUMA
 985         if (!acpi)
 986                 amd = !amd_numa_init(0, max_pfn);
 987 #endif
 988
 989         initmem_init(0, max_pfn, acpi, amd);
 990         memblock_find_dma_reserve();
 991         dma32_reserve_bootmem();
 992
 993 #ifdef CONFIG_KVM_CLOCK
 994         kvmclock_init();
 995 #endif
 996
 997         x86_init.paging.pagetable_setup_start(swapper_pg_dir);
 998         paging_init();
 999         x86_init.paging.pagetable_setup_done(swapper_pg_dir);
1000
1001 #ifdef CONFIG_X86_32
1002         /* sync back kernel address range */
1003         clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1004                         swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1005                         KERNEL_PGD_PTRS);
1006 #endif
1007
1008         tboot_probe();
1009
1010 #ifdef CONFIG_X86_64
1011         map_vsyscall();
1012 #endif
1013
1014         generic_apic_probe();
1015
1016         early_quirks();
1017
1018         /*
1019          * Read APIC and some other early information from ACPI tables.
1020          */
1021         acpi_boot_init();
1022
1023         sfi_init();
1024
1025         /*
1026          * get boot-time SMP configuration:
1027          */
1028         if (smp_found_config)
1029                 get_smp_config();
1030
1031         prefill_possible_map();
1032
1033 #ifdef CONFIG_X86_64
1034         init_cpu_to_node();
1035 #endif
1036
1037         init_apic_mappings();
1038         ioapic_init_mappings();
1039
1040         /* need to wait for io_apic is mapped */
1041         probe_nr_irqs_gsi();
1042
1043         kvm_guest_init();
1044
1045         e820_reserve_resources();
1046         e820_mark_nosave_regions(max_low_pfn);
1047
1048         x86_init.resources.reserve_resources();
1049
1050         e820_setup_gap();
1051
1052 #ifdef CONFIG_VT
1053 #if defined(CONFIG_VGA_CONSOLE)
1054         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1055                 conswitchp = &vga_con;
1056 #elif defined(CONFIG_DUMMY_CONSOLE)
1057         conswitchp = &dummy_con;
1058 #endif
1059 #endif
1060         x86_init.oem.banner();
1061
1062         mcheck_init();
1063
1064         local_irq_save(flags);
1065         arch_init_ideal_nop5();
1066         local_irq_restore(flags);
1067 }
1068
1069 #ifdef CONFIG_X86_32
1070
1071 static struct resource video_ram_resource = {
1072         .name   = "Video RAM area",
1073         .start  = 0xa0000,
1074         .end    = 0xbffff,
1075         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
1076 };
1077
1078 void __init i386_reserve_resources(void)
1079 {
1080         request_resource(&iomem_resource, &video_ram_resource);
1081         reserve_standard_io_resources();
1082 }
1083
1084 #endif /* CONFIG_X86_32 */