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 #include <asm/prom.h>
 117
 118 /*
 119  * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
 120  * The direct mapping extends to max_pfn_mapped, so that we can directly access
 121  * apertures, ACPI and other tables without having to play with fixmaps.
 122  */
 123 unsigned long max_low_pfn_mapped;
 124 unsigned long max_pfn_mapped;
 125
 126 #ifdef CONFIG_DMI
 127 RESERVE_BRK(dmi_alloc, 65536);
 128 #endif
 129
 130
 131 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
 132 unsigned long _brk_end = (unsigned long)__brk_base;
 133
 134 #ifdef CONFIG_X86_64
 135 int default_cpu_present_to_apicid(int mps_cpu)
 136 {
 137         return __default_cpu_present_to_apicid(mps_cpu);
 138 }
 139
 140 int default_check_phys_apicid_present(int phys_apicid)
 141 {
 142         return __default_check_phys_apicid_present(phys_apicid);
 143 }
 144 #endif
 145
 146 #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
 298 static void __init cleanup_highmap_brk_end(void)
 299 {
 300         pud_t *pud;
 301         pmd_t *pmd;
 302
 303         mmu_cr4_features = read_cr4();
 304
 305         /*
 306          * _brk_end cannot change anymore, but it and _end may be
 307          * located on different 2M pages. cleanup_highmap(), however,
 308          * can only consider _end when it runs, so destroy any
 309          * mappings beyond _brk_end here.
 310          */
 311         pud = pud_offset(pgd_offset_k(_brk_end), _brk_end);
 312         pmd = pmd_offset(pud, _brk_end - 1);
 313         while (++pmd <= pmd_offset(pud, (unsigned long)_end - 1))
 314                 pmd_clear(pmd);
 315 }
 316 #else
 317 static inline void init_gbpages(void)
 318 {
 319 }
 320 static inline void cleanup_highmap_brk_end(void)
 321 {
 322 }
 323 #endif
 324
 325 static void __init reserve_brk(void)
 326 {
 327         if (_brk_end > _brk_start)
 328                 memblock_x86_reserve_range(__pa(_brk_start), __pa(_brk_end), "BRK");
 329
 330         /* Mark brk area as locked down and no longer taking any
 331            new allocations */
 332         _brk_start = 0;
 333
 334         cleanup_highmap_brk_end();
 335 }
 336
 337 #ifdef CONFIG_BLK_DEV_INITRD
 338
 339 #define MAX_MAP_CHUNK   (NR_FIX_BTMAPS << PAGE_SHIFT)
 340 static void __init relocate_initrd(void)
 341 {
 342         /* Assume only end is not page aligned */
 343         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 344         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 345         u64 area_size     = PAGE_ALIGN(ramdisk_size);
 346         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 347         u64 ramdisk_here;
 348         unsigned long slop, clen, mapaddr;
 349         char *p, *q;
 350
 351         /* We need to move the initrd down into lowmem */
 352         ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
 353                                          PAGE_SIZE);
 354
 355         if (ramdisk_here == MEMBLOCK_ERROR)
 356                 panic("Cannot find place for new RAMDISK of size %lld\n",
 357                          ramdisk_size);
 358
 359         /* Note: this includes all the lowmem currently occupied by
 360            the initrd, we rely on that fact to keep the data intact. */
 361         memblock_x86_reserve_range(ramdisk_here, ramdisk_here + area_size, "NEW RAMDISK");
 362         initrd_start = ramdisk_here + PAGE_OFFSET;
 363         initrd_end   = initrd_start + ramdisk_size;
 364         printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
 365                          ramdisk_here, ramdisk_here + ramdisk_size);
 366
 367         q = (char *)initrd_start;
 368
 369         /* Copy any lowmem portion of the initrd */
 370         if (ramdisk_image < end_of_lowmem) {
 371                 clen = end_of_lowmem - ramdisk_image;
 372                 p = (char *)__va(ramdisk_image);
 373                 memcpy(q, p, clen);
 374                 q += clen;
 375                 ramdisk_image += clen;
 376                 ramdisk_size  -= clen;
 377         }
 378
 379         /* Copy the highmem portion of the initrd */
 380         while (ramdisk_size) {
 381                 slop = ramdisk_image & ~PAGE_MASK;
 382                 clen = ramdisk_size;
 383                 if (clen > MAX_MAP_CHUNK-slop)
 384                         clen = MAX_MAP_CHUNK-slop;
 385                 mapaddr = ramdisk_image & PAGE_MASK;
 386                 p = early_memremap(mapaddr, clen+slop);
 387                 memcpy(q, p+slop, clen);
 388                 early_iounmap(p, clen+slop);
 389                 q += clen;
 390                 ramdisk_image += clen;
 391                 ramdisk_size  -= clen;
 392         }
 393         /* high pages is not converted by early_res_to_bootmem */
 394         ramdisk_image = boot_params.hdr.ramdisk_image;
 395         ramdisk_size  = boot_params.hdr.ramdisk_size;
 396         printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
 397                 " %08llx - %08llx\n",
 398                 ramdisk_image, ramdisk_image + ramdisk_size - 1,
 399                 ramdisk_here, ramdisk_here + ramdisk_size - 1);
 400 }
 401
 402 static void __init reserve_initrd(void)
 403 {
 404         /* Assume only end is not page aligned */
 405         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 406         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 407         u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 408         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 409
 410         if (!boot_params.hdr.type_of_loader ||
 411             !ramdisk_image || !ramdisk_size)
 412                 return;         /* No initrd provided by bootloader */
 413
 414         initrd_start = 0;
 415
 416         if (ramdisk_size >= (end_of_lowmem>>1)) {
 417                 memblock_x86_free_range(ramdisk_image, ramdisk_end);
 418                 printk(KERN_ERR "initrd too large to handle, "
 419                        "disabling initrd\n");
 420                 return;
 421         }
 422
 423         printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
 424                         ramdisk_end);
 425
 426
 427         if (ramdisk_end <= end_of_lowmem) {
 428                 /* All in lowmem, easy case */
 429                 /*
 430                  * don't need to reserve again, already reserved early
 431                  * in i386_start_kernel
 432                  */
 433                 initrd_start = ramdisk_image + PAGE_OFFSET;
 434                 initrd_end = initrd_start + ramdisk_size;
 435                 return;
 436         }
 437
 438         relocate_initrd();
 439
 440         memblock_x86_free_range(ramdisk_image, ramdisk_end);
 441 }
 442 #else
 443 static void __init reserve_initrd(void)
 444 {
 445 }
 446 #endif /* CONFIG_BLK_DEV_INITRD */
 447
 448 static void __init parse_setup_data(void)
 449 {
 450         struct setup_data *data;
 451         u64 pa_data;
 452
 453         if (boot_params.hdr.version < 0x0209)
 454                 return;
 455         pa_data = boot_params.hdr.setup_data;
 456         while (pa_data) {
 457                 u32 data_len, map_len;
 458
 459                 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
 460                               (u64)sizeof(struct setup_data));
 461                 data = early_memremap(pa_data, map_len);
 462                 data_len = data->len + sizeof(struct setup_data);
 463                 if (data_len > map_len) {
 464                         early_iounmap(data, map_len);
 465                         data = early_memremap(pa_data, data_len);
 466                         map_len = data_len;
 467                 }
 468
 469                 switch (data->type) {
 470                 case SETUP_E820_EXT:
 471                         parse_e820_ext(data);
 472                         break;
 473                 case SETUP_DTB:
 474                         add_dtb(pa_data);
 475                         break;
 476                 default:
 477                         break;
 478                 }
 479                 pa_data = data->next;
 480                 early_iounmap(data, map_len);
 481         }
 482 }
 483
 484 static void __init e820_reserve_setup_data(void)
 485 {
 486         struct setup_data *data;
 487         u64 pa_data;
 488         int found = 0;
 489
 490         if (boot_params.hdr.version < 0x0209)
 491                 return;
 492         pa_data = boot_params.hdr.setup_data;
 493         while (pa_data) {
 494                 data = early_memremap(pa_data, sizeof(*data));
 495                 e820_update_range(pa_data, sizeof(*data)+data->len,
 496                          E820_RAM, E820_RESERVED_KERN);
 497                 found = 1;
 498                 pa_data = data->next;
 499                 early_iounmap(data, sizeof(*data));
 500         }
 501         if (!found)
 502                 return;
 503
 504         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 505         memcpy(&e820_saved, &e820, sizeof(struct e820map));
 506         printk(KERN_INFO "extended physical RAM map:\n");
 507         e820_print_map("reserve setup_data");
 508 }
 509
 510 static void __init memblock_x86_reserve_range_setup_data(void)
 511 {
 512         struct setup_data *data;
 513         u64 pa_data;
 514         char buf[32];
 515
 516         if (boot_params.hdr.version < 0x0209)
 517                 return;
 518         pa_data = boot_params.hdr.setup_data;
 519         while (pa_data) {
 520                 data = early_memremap(pa_data, sizeof(*data));
 521                 sprintf(buf, "setup data %x", data->type);
 522                 memblock_x86_reserve_range(pa_data, pa_data+sizeof(*data)+data->len, buf);
 523                 pa_data = data->next;
 524                 early_iounmap(data, sizeof(*data));
 525         }
 526 }
 527
 528 /*
 529  * --------- Crashkernel reservation ------------------------------
 530  */
 531
 532 #ifdef CONFIG_KEXEC
 533
 534 static inline unsigned long long get_total_mem(void)
 535 {
 536         unsigned long long total;
 537
 538         total = max_pfn - min_low_pfn;
 539
 540         return total << PAGE_SHIFT;
 541 }
 542
 543 /*
 544  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
 545  * would limit the kernel to the low 512 MiB due to mapping restrictions.
 546  * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
 547  * limit once kexec-tools are fixed.
 548  */
 549 #ifdef CONFIG_X86_32
 550 # define CRASH_KERNEL_ADDR_MAX  (512 << 20)
 551 #else
 552 # define CRASH_KERNEL_ADDR_MAX  (896 << 20)
 553 #endif
 554
 555 static void __init reserve_crashkernel(void)
 556 {
 557         unsigned long long total_mem;
 558         unsigned long long crash_size, crash_base;
 559         int ret;
 560
 561         total_mem = get_total_mem();
 562
 563         ret = parse_crashkernel(boot_command_line, total_mem,
 564                         &crash_size, &crash_base);
 565         if (ret != 0 || crash_size <= 0)
 566                 return;
 567
 568         /* 0 means: find the address automatically */
 569         if (crash_base <= 0) {
 570                 const unsigned long long alignment = 16<<20;    /* 16M */
 571
 572                 /*
 573                  *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
 574                  */
 575                 crash_base = memblock_find_in_range(alignment,
 576                                CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
 577
 578                 if (crash_base == MEMBLOCK_ERROR) {
 579                         pr_info("crashkernel reservation failed - No suitable area found.\n");
 580                         return;
 581                 }
 582         } else {
 583                 unsigned long long start;
 584
 585                 start = memblock_find_in_range(crash_base,
 586                                  crash_base + crash_size, crash_size, 1<<20);
 587                 if (start != crash_base) {
 588                         pr_info("crashkernel reservation failed - memory is in use.\n");
 589                         return;
 590                 }
 591         }
 592         memblock_x86_reserve_range(crash_base, crash_base + crash_size, "CRASH KERNEL");
 593
 594         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 595                         "for crashkernel (System RAM: %ldMB)\n",
 596                         (unsigned long)(crash_size >> 20),
 597                         (unsigned long)(crash_base >> 20),
 598                         (unsigned long)(total_mem >> 20));
 599
 600         crashk_res.start = crash_base;
 601         crashk_res.end   = crash_base + crash_size - 1;
 602         insert_resource(&iomem_resource, &crashk_res);
 603 }
 604 #else
 605 static void __init reserve_crashkernel(void)
 606 {
 607 }
 608 #endif
 609
 610 static struct resource standard_io_resources[] = {
 611         { .name = "dma1", .start = 0x00, .end = 0x1f,
 612                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 613         { .name = "pic1", .start = 0x20, .end = 0x21,
 614                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 615         { .name = "timer0", .start = 0x40, .end = 0x43,
 616                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 617         { .name = "timer1", .start = 0x50, .end = 0x53,
 618                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 619         { .name = "keyboard", .start = 0x60, .end = 0x60,
 620                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 621         { .name = "keyboard", .start = 0x64, .end = 0x64,
 622                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 623         { .name = "dma page reg", .start = 0x80, .end = 0x8f,
 624                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 625         { .name = "pic2", .start = 0xa0, .end = 0xa1,
 626                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 627         { .name = "dma2", .start = 0xc0, .end = 0xdf,
 628                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 629         { .name = "fpu", .start = 0xf0, .end = 0xff,
 630                 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
 631 };
 632
 633 void __init reserve_standard_io_resources(void)
 634 {
 635         int i;
 636
 637         /* request I/O space for devices used on all i[345]86 PCs */
 638         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 639                 request_resource(&ioport_resource, &standard_io_resources[i]);
 640
 641 }
 642
 643 /*
 644  * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
 645  * is_kdump_kernel() to determine if we are booting after a panic. Hence
 646  * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
 647  */
 648
 649 #ifdef CONFIG_CRASH_DUMP
 650 /* elfcorehdr= specifies the location of elf core header
 651  * stored by the crashed kernel. This option will be passed
 652  * by kexec loader to the capture kernel.
 653  */
 654 static int __init setup_elfcorehdr(char *arg)
 655 {
 656         char *end;
 657         if (!arg)
 658                 return -EINVAL;
 659         elfcorehdr_addr = memparse(arg, &end);
 660         return end > arg ? 0 : -EINVAL;
 661 }
 662 early_param("elfcorehdr", setup_elfcorehdr);
 663 #endif
 664
 665 static __init void reserve_ibft_region(void)
 666 {
 667         unsigned long addr, size = 0;
 668
 669         addr = find_ibft_region(&size);
 670
 671         if (size)
 672                 memblock_x86_reserve_range(addr, addr + size, "* ibft");
 673 }
 674
 675 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 676
 677 static void __init trim_bios_range(void)
 678 {
 679         /*
 680          * A special case is the first 4Kb of memory;
 681          * This is a BIOS owned area, not kernel ram, but generally
 682          * not listed as such in the E820 table.
 683          *
 684          * This typically reserves additional memory (64KiB by default)
 685          * since some BIOSes are known to corrupt low memory.  See the
 686          * Kconfig help text for X86_RESERVE_LOW.
 687          */
 688         e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
 689                           E820_RAM, E820_RESERVED);
 690
 691         /*
 692          * special case: Some BIOSen report the PC BIOS
 693          * area (640->1Mb) as ram even though it is not.
 694          * take them out.
 695          */
 696         e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 697         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 698 }
 699
 700 static int __init parse_reservelow(char *p)
 701 {
 702         unsigned long long size;
 703
 704         if (!p)
 705                 return -EINVAL;
 706
 707         size = memparse(p, &p);
 708
 709         if (size < 4096)
 710                 size = 4096;
 711
 712         if (size > 640*1024)
 713                 size = 640*1024;
 714
 715         reserve_low = size;
 716
 717         return 0;
 718 }
 719
 720 early_param("reservelow", parse_reservelow);
 721
 722 /*
 723  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 724  * passed the efi memmap, systab, etc., so we should use these data structures
 725  * for initialization.  Note, the efi init code path is determined by the
 726  * global efi_enabled. This allows the same kernel image to be used on existing
 727  * systems (with a traditional BIOS) as well as on EFI systems.
 728  */
 729 /*
 730  * setup_arch - architecture-specific boot-time initializations
 731  *
 732  * Note: On x86_64, fixmaps are ready for use even before this is called.
 733  */
 734
 735 void __init setup_arch(char **cmdline_p)
 736 {
 737         unsigned long flags;
 738
 739 #ifdef CONFIG_X86_32
 740         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 741         visws_early_detect();
 742
 743         /*
 744          * copy kernel address range established so far and switch
 745          * to the proper swapper page table
 746          */
 747         clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 748                         initial_page_table + KERNEL_PGD_BOUNDARY,
 749                         KERNEL_PGD_PTRS);
 750
 751         load_cr3(swapper_pg_dir);
 752         __flush_tlb_all();
 753 #else
 754         printk(KERN_INFO "Command line: %s\n", boot_command_line);
 755 #endif
 756
 757         /*
 758          * If we have OLPC OFW, we might end up relocating the fixmap due to
 759          * reserve_top(), so do this before touching the ioremap area.
 760          */
 761         olpc_ofw_detect();
 762
 763         early_trap_init();
 764         early_cpu_init();
 765         early_ioremap_init();
 766
 767         setup_olpc_ofw_pgd();
 768
 769         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 770         screen_info = boot_params.screen_info;
 771         edid_info = boot_params.edid_info;
 772 #ifdef CONFIG_X86_32
 773         apm_info.bios = boot_params.apm_bios_info;
 774         ist_info = boot_params.ist_info;
 775         if (boot_params.sys_desc_table.length != 0) {
 776                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 777                 machine_id = boot_params.sys_desc_table.table[0];
 778                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 779                 BIOS_revision = boot_params.sys_desc_table.table[2];
 780         }
 781 #endif
 782         saved_video_mode = boot_params.hdr.vid_mode;
 783         bootloader_type = boot_params.hdr.type_of_loader;
 784         if ((bootloader_type >> 4) == 0xe) {
 785                 bootloader_type &= 0xf;
 786                 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 787         }
 788         bootloader_version  = bootloader_type & 0xf;
 789         bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 790
 791 #ifdef CONFIG_BLK_DEV_RAM
 792         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 793         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 794         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 795 #endif
 796 #ifdef CONFIG_EFI
 797         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 798 #ifdef CONFIG_X86_32
 799                      "EL32",
 800 #else
 801                      "EL64",
 802 #endif
 803          4)) {
 804                 efi_enabled = 1;
 805                 efi_memblock_x86_reserve_range();
 806         }
 807 #endif
 808
 809         x86_init.oem.arch_setup();
 810
 811         iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 812         setup_memory_map();
 813         parse_setup_data();
 814         /* update the e820_saved too */
 815         e820_reserve_setup_data();
 816
 817         copy_edd();
 818
 819         if (!boot_params.hdr.root_flags)
 820                 root_mountflags &= ~MS_RDONLY;
 821         init_mm.start_code = (unsigned long) _text;
 822         init_mm.end_code = (unsigned long) _etext;
 823         init_mm.end_data = (unsigned long) _edata;
 824         init_mm.brk = _brk_end;
 825
 826         code_resource.start = virt_to_phys(_text);
 827         code_resource.end = virt_to_phys(_etext)-1;
 828         data_resource.start = virt_to_phys(_etext);
 829         data_resource.end = virt_to_phys(_edata)-1;
 830         bss_resource.start = virt_to_phys(&__bss_start);
 831         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 832
 833 #ifdef CONFIG_CMDLINE_BOOL
 834 #ifdef CONFIG_CMDLINE_OVERRIDE
 835         strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 836 #else
 837         if (builtin_cmdline[0]) {
 838                 /* append boot loader cmdline to builtin */
 839                 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 840                 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 841                 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 842         }
 843 #endif
 844 #endif
 845
 846         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 847         *cmdline_p = command_line;
 848
 849         /*
 850          * x86_configure_nx() is called before parse_early_param() to detect
 851          * whether hardware doesn't support NX (so that the early EHCI debug
 852          * console setup can safely call set_fixmap()). It may then be called
 853          * again from within noexec_setup() during parsing early parameters
 854          * to honor the respective command line option.
 855          */
 856         x86_configure_nx();
 857
 858         parse_early_param();
 859
 860         x86_report_nx();
 861
 862         /* after early param, so could get panic from serial */
 863         memblock_x86_reserve_range_setup_data();
 864
 865         if (acpi_mps_check()) {
 866 #ifdef CONFIG_X86_LOCAL_APIC
 867                 disable_apic = 1;
 868 #endif
 869                 setup_clear_cpu_cap(X86_FEATURE_APIC);
 870         }
 871
 872 #ifdef CONFIG_PCI
 873         if (pci_early_dump_regs)
 874                 early_dump_pci_devices();
 875 #endif
 876
 877         finish_e820_parsing();
 878
 879         if (efi_enabled)
 880                 efi_init();
 881
 882         dmi_scan_machine();
 883
 884         /*
 885          * VMware detection requires dmi to be available, so this
 886          * needs to be done after dmi_scan_machine, for the BP.
 887          */
 888         init_hypervisor_platform();
 889
 890         x86_init.resources.probe_roms();
 891
 892         /* after parse_early_param, so could debug it */
 893         insert_resource(&iomem_resource, &code_resource);
 894         insert_resource(&iomem_resource, &data_resource);
 895         insert_resource(&iomem_resource, &bss_resource);
 896
 897         trim_bios_range();
 898 #ifdef CONFIG_X86_32
 899         if (ppro_with_ram_bug()) {
 900                 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
 901                                   E820_RESERVED);
 902                 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 903                 printk(KERN_INFO "fixed physical RAM map:\n");
 904                 e820_print_map("bad_ppro");
 905         }
 906 #else
 907         early_gart_iommu_check();
 908 #endif
 909
 910         /*
 911          * partially used pages are not usable - thus
 912          * we are rounding upwards:
 913          */
 914         max_pfn = e820_end_of_ram_pfn();
 915
 916         /* update e820 for memory not covered by WB MTRRs */
 917         mtrr_bp_init();
 918         if (mtrr_trim_uncached_memory(max_pfn))
 919                 max_pfn = e820_end_of_ram_pfn();
 920
 921 #ifdef CONFIG_X86_32
 922         /* max_low_pfn get updated here */
 923         find_low_pfn_range();
 924 #else
 925         num_physpages = max_pfn;
 926
 927         check_x2apic();
 928
 929         /* How many end-of-memory variables you have, grandma! */
 930         /* need this before calling reserve_initrd */
 931         if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
 932                 max_low_pfn = e820_end_of_low_ram_pfn();
 933         else
 934                 max_low_pfn = max_pfn;
 935
 936         high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
 937 #endif
 938
 939         /*
 940          * Find and reserve possible boot-time SMP configuration:
 941          */
 942         find_smp_config();
 943
 944         reserve_ibft_region();
 945
 946         /*
 947          * Need to conclude brk, before memblock_x86_fill()
 948          *  it could use memblock_find_in_range, could overlap with
 949          *  brk area.
 950          */
 951         reserve_brk();
 952
 953         memblock.current_limit = get_max_mapped();
 954         memblock_x86_fill();
 955
 956         /* preallocate 4k for mptable mpc */
 957         early_reserve_e820_mpc_new();
 958
 959 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
 960         setup_bios_corruption_check();
 961 #endif
 962
 963         printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
 964                         max_pfn_mapped<<PAGE_SHIFT);
 965
 966         reserve_trampoline_memory();
 967
 968 #ifdef CONFIG_ACPI_SLEEP
 969         /*
 970          * Reserve low memory region for sleep support.
 971          * even before init_memory_mapping
 972          */
 973         acpi_reserve_wakeup_memory();
 974 #endif
 975         init_gbpages();
 976
 977         /* max_pfn_mapped is updated here */
 978         max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
 979         max_pfn_mapped = max_low_pfn_mapped;
 980
 981 #ifdef CONFIG_X86_64
 982         if (max_pfn > max_low_pfn) {
 983                 max_pfn_mapped = init_memory_mapping(1UL<<32,
 984                                                      max_pfn<<PAGE_SHIFT);
 985                 /* can we preseve max_low_pfn ?*/
 986                 max_low_pfn = max_pfn;
 987         }
 988 #endif
 989         memblock.current_limit = get_max_mapped();
 990
 991         /*
 992          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 993          */
 994
 995 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 996         if (init_ohci1394_dma_early)
 997                 init_ohci1394_dma_on_all_controllers();
 998 #endif
 999
1000         reserve_initrd();
1001
1002         reserve_crashkernel();
1003
1004         vsmp_init();
1005
1006         io_delay_init();
1007
1008         /*
1009          * Parse the ACPI tables for possible boot-time SMP configuration.
1010          */
1011         acpi_boot_table_init();
1012
1013         early_acpi_boot_init();
1014
1015         initmem_init();
1016         memblock_find_dma_reserve();
1017         dma32_reserve_bootmem();
1018
1019 #ifdef CONFIG_KVM_CLOCK
1020         kvmclock_init();
1021 #endif
1022
1023         x86_init.paging.pagetable_setup_start(swapper_pg_dir);
1024         paging_init();
1025         x86_init.paging.pagetable_setup_done(swapper_pg_dir);
1026
1027 #ifdef CONFIG_X86_32
1028         /* sync back kernel address range */
1029         clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1030                         swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1031                         KERNEL_PGD_PTRS);
1032 #endif
1033
1034         tboot_probe();
1035
1036 #ifdef CONFIG_X86_64
1037         map_vsyscall();
1038 #endif
1039
1040         generic_apic_probe();
1041
1042         early_quirks();
1043
1044         /*
1045          * Read APIC and some other early information from ACPI tables.
1046          */
1047         acpi_boot_init();
1048         sfi_init();
1049         x86_dtb_init();
1050
1051         /*
1052          * get boot-time SMP configuration:
1053          */
1054         if (smp_found_config)
1055                 get_smp_config();
1056
1057         prefill_possible_map();
1058
1059         init_cpu_to_node();
1060
1061         init_apic_mappings();
1062         ioapic_and_gsi_init();
1063
1064         kvm_guest_init();
1065
1066         e820_reserve_resources();
1067         e820_mark_nosave_regions(max_low_pfn);
1068
1069         x86_init.resources.reserve_resources();
1070
1071         e820_setup_gap();
1072
1073 #ifdef CONFIG_VT
1074 #if defined(CONFIG_VGA_CONSOLE)
1075         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1076                 conswitchp = &vga_con;
1077 #elif defined(CONFIG_DUMMY_CONSOLE)
1078         conswitchp = &dummy_con;
1079 #endif
1080 #endif
1081         x86_init.oem.banner();
1082
1083         x86_init.timers.wallclock_init();
1084
1085         mcheck_init();
1086
1087         local_irq_save(flags);
1088         arch_init_ideal_nop5();
1089         local_irq_restore(flags);
1090 }
1091
1092 #ifdef CONFIG_X86_32
1093
1094 static struct resource video_ram_resource = {
1095         .name   = "Video RAM area",
1096         .start  = 0xa0000,
1097         .end    = 0xbffff,
1098         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
1099 };
1100
1101 void __init i386_reserve_resources(void)
1102 {
1103         request_resource(&iomem_resource, &video_ram_resource);
1104         reserve_standard_io_resources();
1105 }
1106
1107 #endif /* CONFIG_X86_32 */