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 #else
 298 static inline void init_gbpages(void)
 299 {
 300 }
 301 static void __init cleanup_highmap(void)
 302 {
 303 }
 304 #endif
 305
 306 static void __init reserve_brk(void)
 307 {
 308         if (_brk_end > _brk_start)
 309                 memblock_reserve(__pa(_brk_start),
 310                                  __pa(_brk_end) - __pa(_brk_start));
 311
 312         /* Mark brk area as locked down and no longer taking any
 313            new allocations */
 314         _brk_start = 0;
 315 }
 316
 317 #ifdef CONFIG_BLK_DEV_INITRD
 318
 319 #define MAX_MAP_CHUNK   (NR_FIX_BTMAPS << PAGE_SHIFT)
 320 static void __init relocate_initrd(void)
 321 {
 322         /* Assume only end is not page aligned */
 323         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 324         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 325         u64 area_size     = PAGE_ALIGN(ramdisk_size);
 326         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 327         u64 ramdisk_here;
 328         unsigned long slop, clen, mapaddr;
 329         char *p, *q;
 330
 331         /* We need to move the initrd down into lowmem */
 332         ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
 333                                          PAGE_SIZE);
 334
 335         if (!ramdisk_here)
 336                 panic("Cannot find place for new RAMDISK of size %lld\n",
 337                          ramdisk_size);
 338
 339         /* Note: this includes all the lowmem currently occupied by
 340            the initrd, we rely on that fact to keep the data intact. */
 341         memblock_reserve(ramdisk_here, area_size);
 342         initrd_start = ramdisk_here + PAGE_OFFSET;
 343         initrd_end   = initrd_start + ramdisk_size;
 344         printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
 345                          ramdisk_here, ramdisk_here + ramdisk_size);
 346
 347         q = (char *)initrd_start;
 348
 349         /* Copy any lowmem portion of the initrd */
 350         if (ramdisk_image < end_of_lowmem) {
 351                 clen = end_of_lowmem - ramdisk_image;
 352                 p = (char *)__va(ramdisk_image);
 353                 memcpy(q, p, clen);
 354                 q += clen;
 355                 ramdisk_image += clen;
 356                 ramdisk_size  -= clen;
 357         }
 358
 359         /* Copy the highmem portion of the initrd */
 360         while (ramdisk_size) {
 361                 slop = ramdisk_image & ~PAGE_MASK;
 362                 clen = ramdisk_size;
 363                 if (clen > MAX_MAP_CHUNK-slop)
 364                         clen = MAX_MAP_CHUNK-slop;
 365                 mapaddr = ramdisk_image & PAGE_MASK;
 366                 p = early_memremap(mapaddr, clen+slop);
 367                 memcpy(q, p+slop, clen);
 368                 early_iounmap(p, clen+slop);
 369                 q += clen;
 370                 ramdisk_image += clen;
 371                 ramdisk_size  -= clen;
 372         }
 373         /* high pages is not converted by early_res_to_bootmem */
 374         ramdisk_image = boot_params.hdr.ramdisk_image;
 375         ramdisk_size  = boot_params.hdr.ramdisk_size;
 376         printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
 377                 " %08llx - %08llx\n",
 378                 ramdisk_image, ramdisk_image + ramdisk_size - 1,
 379                 ramdisk_here, ramdisk_here + ramdisk_size - 1);
 380 }
 381
 382 static void __init reserve_initrd(void)
 383 {
 384         /* Assume only end is not page aligned */
 385         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 386         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 387         u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 388         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 389
 390         if (!boot_params.hdr.type_of_loader ||
 391             !ramdisk_image || !ramdisk_size)
 392                 return;         /* No initrd provided by bootloader */
 393
 394         initrd_start = 0;
 395
 396         if (ramdisk_size >= (end_of_lowmem>>1)) {
 397                 memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
 398                 printk(KERN_ERR "initrd too large to handle, "
 399                        "disabling initrd\n");
 400                 return;
 401         }
 402
 403         printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
 404                         ramdisk_end);
 405
 406
 407         if (ramdisk_end <= end_of_lowmem) {
 408                 /* All in lowmem, easy case */
 409                 /*
 410                  * don't need to reserve again, already reserved early
 411                  * in i386_start_kernel
 412                  */
 413                 initrd_start = ramdisk_image + PAGE_OFFSET;
 414                 initrd_end = initrd_start + ramdisk_size;
 415                 return;
 416         }
 417
 418         relocate_initrd();
 419
 420         memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
 421 }
 422 #else
 423 static void __init reserve_initrd(void)
 424 {
 425 }
 426 #endif /* CONFIG_BLK_DEV_INITRD */
 427
 428 static void __init parse_setup_data(void)
 429 {
 430         struct setup_data *data;
 431         u64 pa_data;
 432
 433         if (boot_params.hdr.version < 0x0209)
 434                 return;
 435         pa_data = boot_params.hdr.setup_data;
 436         while (pa_data) {
 437                 u32 data_len, map_len;
 438
 439                 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
 440                               (u64)sizeof(struct setup_data));
 441                 data = early_memremap(pa_data, map_len);
 442                 data_len = data->len + sizeof(struct setup_data);
 443                 if (data_len > map_len) {
 444                         early_iounmap(data, map_len);
 445                         data = early_memremap(pa_data, data_len);
 446                         map_len = data_len;
 447                 }
 448
 449                 switch (data->type) {
 450                 case SETUP_E820_EXT:
 451                         parse_e820_ext(data);
 452                         break;
 453                 case SETUP_DTB:
 454                         add_dtb(pa_data);
 455                         break;
 456                 default:
 457                         break;
 458                 }
 459                 pa_data = data->next;
 460                 early_iounmap(data, map_len);
 461         }
 462 }
 463
 464 static void __init e820_reserve_setup_data(void)
 465 {
 466         struct setup_data *data;
 467         u64 pa_data;
 468         int found = 0;
 469
 470         if (boot_params.hdr.version < 0x0209)
 471                 return;
 472         pa_data = boot_params.hdr.setup_data;
 473         while (pa_data) {
 474                 data = early_memremap(pa_data, sizeof(*data));
 475                 e820_update_range(pa_data, sizeof(*data)+data->len,
 476                          E820_RAM, E820_RESERVED_KERN);
 477                 found = 1;
 478                 pa_data = data->next;
 479                 early_iounmap(data, sizeof(*data));
 480         }
 481         if (!found)
 482                 return;
 483
 484         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 485         memcpy(&e820_saved, &e820, sizeof(struct e820map));
 486         printk(KERN_INFO "extended physical RAM map:\n");
 487         e820_print_map("reserve setup_data");
 488 }
 489
 490 static void __init memblock_x86_reserve_range_setup_data(void)
 491 {
 492         struct setup_data *data;
 493         u64 pa_data;
 494
 495         if (boot_params.hdr.version < 0x0209)
 496                 return;
 497         pa_data = boot_params.hdr.setup_data;
 498         while (pa_data) {
 499                 data = early_memremap(pa_data, sizeof(*data));
 500                 memblock_reserve(pa_data, sizeof(*data) + data->len);
 501                 pa_data = data->next;
 502                 early_iounmap(data, sizeof(*data));
 503         }
 504 }
 505
 506 /*
 507  * --------- Crashkernel reservation ------------------------------
 508  */
 509
 510 #ifdef CONFIG_KEXEC
 511
 512 static inline unsigned long long get_total_mem(void)
 513 {
 514         unsigned long long total;
 515
 516         total = max_pfn - min_low_pfn;
 517
 518         return total << PAGE_SHIFT;
 519 }
 520
 521 /*
 522  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
 523  * would limit the kernel to the low 512 MiB due to mapping restrictions.
 524  * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
 525  * limit once kexec-tools are fixed.
 526  */
 527 #ifdef CONFIG_X86_32
 528 # define CRASH_KERNEL_ADDR_MAX  (512 << 20)
 529 #else
 530 # define CRASH_KERNEL_ADDR_MAX  (896 << 20)
 531 #endif
 532
 533 static void __init reserve_crashkernel(void)
 534 {
 535         unsigned long long total_mem;
 536         unsigned long long crash_size, crash_base;
 537         int ret;
 538
 539         total_mem = get_total_mem();
 540
 541         ret = parse_crashkernel(boot_command_line, total_mem,
 542                         &crash_size, &crash_base);
 543         if (ret != 0 || crash_size <= 0)
 544                 return;
 545
 546         /* 0 means: find the address automatically */
 547         if (crash_base <= 0) {
 548                 const unsigned long long alignment = 16<<20;    /* 16M */
 549
 550                 /*
 551                  *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
 552                  */
 553                 crash_base = memblock_find_in_range(alignment,
 554                                CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
 555
 556                 if (!crash_base) {
 557                         pr_info("crashkernel reservation failed - No suitable area found.\n");
 558                         return;
 559                 }
 560         } else {
 561                 unsigned long long start;
 562
 563                 start = memblock_find_in_range(crash_base,
 564                                  crash_base + crash_size, crash_size, 1<<20);
 565                 if (start != crash_base) {
 566                         pr_info("crashkernel reservation failed - memory is in use.\n");
 567                         return;
 568                 }
 569         }
 570         memblock_reserve(crash_base, crash_size);
 571
 572         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 573                         "for crashkernel (System RAM: %ldMB)\n",
 574                         (unsigned long)(crash_size >> 20),
 575                         (unsigned long)(crash_base >> 20),
 576                         (unsigned long)(total_mem >> 20));
 577
 578         crashk_res.start = crash_base;
 579         crashk_res.end   = crash_base + crash_size - 1;
 580         insert_resource(&iomem_resource, &crashk_res);
 581 }
 582 #else
 583 static void __init reserve_crashkernel(void)
 584 {
 585 }
 586 #endif
 587
 588 static struct resource standard_io_resources[] = {
 589         { .name = "dma1", .start = 0x00, .end = 0x1f,
 590                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 591         { .name = "pic1", .start = 0x20, .end = 0x21,
 592                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 593         { .name = "timer0", .start = 0x40, .end = 0x43,
 594                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 595         { .name = "timer1", .start = 0x50, .end = 0x53,
 596                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 597         { .name = "keyboard", .start = 0x60, .end = 0x60,
 598                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 599         { .name = "keyboard", .start = 0x64, .end = 0x64,
 600                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 601         { .name = "dma page reg", .start = 0x80, .end = 0x8f,
 602                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 603         { .name = "pic2", .start = 0xa0, .end = 0xa1,
 604                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 605         { .name = "dma2", .start = 0xc0, .end = 0xdf,
 606                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 607         { .name = "fpu", .start = 0xf0, .end = 0xff,
 608                 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
 609 };
 610
 611 void __init reserve_standard_io_resources(void)
 612 {
 613         int i;
 614
 615         /* request I/O space for devices used on all i[345]86 PCs */
 616         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 617                 request_resource(&ioport_resource, &standard_io_resources[i]);
 618
 619 }
 620
 621 static __init void reserve_ibft_region(void)
 622 {
 623         unsigned long addr, size = 0;
 624
 625         addr = find_ibft_region(&size);
 626
 627         if (size)
 628                 memblock_reserve(addr, size);
 629 }
 630
 631 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 632
 633 static void __init trim_bios_range(void)
 634 {
 635         /*
 636          * A special case is the first 4Kb of memory;
 637          * This is a BIOS owned area, not kernel ram, but generally
 638          * not listed as such in the E820 table.
 639          *
 640          * This typically reserves additional memory (64KiB by default)
 641          * since some BIOSes are known to corrupt low memory.  See the
 642          * Kconfig help text for X86_RESERVE_LOW.
 643          */
 644         e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
 645                           E820_RAM, E820_RESERVED);
 646
 647         /*
 648          * special case: Some BIOSen report the PC BIOS
 649          * area (640->1Mb) as ram even though it is not.
 650          * take them out.
 651          */
 652         e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 653         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 654 }
 655
 656 static int __init parse_reservelow(char *p)
 657 {
 658         unsigned long long size;
 659
 660         if (!p)
 661                 return -EINVAL;
 662
 663         size = memparse(p, &p);
 664
 665         if (size < 4096)
 666                 size = 4096;
 667
 668         if (size > 640*1024)
 669                 size = 640*1024;
 670
 671         reserve_low = size;
 672
 673         return 0;
 674 }
 675
 676 early_param("reservelow", parse_reservelow);
 677
 678 /*
 679  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 680  * passed the efi memmap, systab, etc., so we should use these data structures
 681  * for initialization.  Note, the efi init code path is determined by the
 682  * global efi_enabled. This allows the same kernel image to be used on existing
 683  * systems (with a traditional BIOS) as well as on EFI systems.
 684  */
 685 /*
 686  * setup_arch - architecture-specific boot-time initializations
 687  *
 688  * Note: On x86_64, fixmaps are ready for use even before this is called.
 689  */
 690
 691 void __init setup_arch(char **cmdline_p)
 692 {
 693 #ifdef CONFIG_X86_32
 694         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 695         visws_early_detect();
 696
 697         /*
 698          * copy kernel address range established so far and switch
 699          * to the proper swapper page table
 700          */
 701         clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 702                         initial_page_table + KERNEL_PGD_BOUNDARY,
 703                         KERNEL_PGD_PTRS);
 704
 705         load_cr3(swapper_pg_dir);
 706         __flush_tlb_all();
 707 #else
 708         printk(KERN_INFO "Command line: %s\n", boot_command_line);
 709 #endif
 710
 711         /*
 712          * If we have OLPC OFW, we might end up relocating the fixmap due to
 713          * reserve_top(), so do this before touching the ioremap area.
 714          */
 715         olpc_ofw_detect();
 716
 717         early_trap_init();
 718         early_cpu_init();
 719         early_ioremap_init();
 720
 721         setup_olpc_ofw_pgd();
 722
 723         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 724         screen_info = boot_params.screen_info;
 725         edid_info = boot_params.edid_info;
 726 #ifdef CONFIG_X86_32
 727         apm_info.bios = boot_params.apm_bios_info;
 728         ist_info = boot_params.ist_info;
 729         if (boot_params.sys_desc_table.length != 0) {
 730                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 731                 machine_id = boot_params.sys_desc_table.table[0];
 732                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 733                 BIOS_revision = boot_params.sys_desc_table.table[2];
 734         }
 735 #endif
 736         saved_video_mode = boot_params.hdr.vid_mode;
 737         bootloader_type = boot_params.hdr.type_of_loader;
 738         if ((bootloader_type >> 4) == 0xe) {
 739                 bootloader_type &= 0xf;
 740                 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 741         }
 742         bootloader_version  = bootloader_type & 0xf;
 743         bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 744
 745 #ifdef CONFIG_BLK_DEV_RAM
 746         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 747         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 748         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 749 #endif
 750 #ifdef CONFIG_EFI
 751         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 752                      EFI_LOADER_SIGNATURE, 4)) {
 753                 efi_enabled = 1;
 754                 efi_memblock_x86_reserve_range();
 755         }
 756 #endif
 757
 758         x86_init.oem.arch_setup();
 759
 760         iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 761         setup_memory_map();
 762         parse_setup_data();
 763         /* update the e820_saved too */
 764         e820_reserve_setup_data();
 765
 766         copy_edd();
 767
 768         if (!boot_params.hdr.root_flags)
 769                 root_mountflags &= ~MS_RDONLY;
 770         init_mm.start_code = (unsigned long) _text;
 771         init_mm.end_code = (unsigned long) _etext;
 772         init_mm.end_data = (unsigned long) _edata;
 773         init_mm.brk = _brk_end;
 774
 775         code_resource.start = virt_to_phys(_text);
 776         code_resource.end = virt_to_phys(_etext)-1;
 777         data_resource.start = virt_to_phys(_etext);
 778         data_resource.end = virt_to_phys(_edata)-1;
 779         bss_resource.start = virt_to_phys(&__bss_start);
 780         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 781
 782 #ifdef CONFIG_CMDLINE_BOOL
 783 #ifdef CONFIG_CMDLINE_OVERRIDE
 784         strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 785 #else
 786         if (builtin_cmdline[0]) {
 787                 /* append boot loader cmdline to builtin */
 788                 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 789                 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 790                 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 791         }
 792 #endif
 793 #endif
 794
 795         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 796         *cmdline_p = command_line;
 797
 798         /*
 799          * x86_configure_nx() is called before parse_early_param() to detect
 800          * whether hardware doesn't support NX (so that the early EHCI debug
 801          * console setup can safely call set_fixmap()). It may then be called
 802          * again from within noexec_setup() during parsing early parameters
 803          * to honor the respective command line option.
 804          */
 805         x86_configure_nx();
 806
 807         parse_early_param();
 808
 809         x86_report_nx();
 810
 811         /* after early param, so could get panic from serial */
 812         memblock_x86_reserve_range_setup_data();
 813
 814         if (acpi_mps_check()) {
 815 #ifdef CONFIG_X86_LOCAL_APIC
 816                 disable_apic = 1;
 817 #endif
 818                 setup_clear_cpu_cap(X86_FEATURE_APIC);
 819         }
 820
 821 #ifdef CONFIG_PCI
 822         if (pci_early_dump_regs)
 823                 early_dump_pci_devices();
 824 #endif
 825
 826         finish_e820_parsing();
 827
 828         if (efi_enabled)
 829                 efi_init();
 830
 831         dmi_scan_machine();
 832
 833         /*
 834          * VMware detection requires dmi to be available, so this
 835          * needs to be done after dmi_scan_machine, for the BP.
 836          */
 837         init_hypervisor_platform();
 838
 839         x86_init.resources.probe_roms();
 840
 841         /* after parse_early_param, so could debug it */
 842         insert_resource(&iomem_resource, &code_resource);
 843         insert_resource(&iomem_resource, &data_resource);
 844         insert_resource(&iomem_resource, &bss_resource);
 845
 846         trim_bios_range();
 847 #ifdef CONFIG_X86_32
 848         if (ppro_with_ram_bug()) {
 849                 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
 850                                   E820_RESERVED);
 851                 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 852                 printk(KERN_INFO "fixed physical RAM map:\n");
 853                 e820_print_map("bad_ppro");
 854         }
 855 #else
 856         early_gart_iommu_check();
 857 #endif
 858
 859         /*
 860          * partially used pages are not usable - thus
 861          * we are rounding upwards:
 862          */
 863         max_pfn = e820_end_of_ram_pfn();
 864
 865         /* update e820 for memory not covered by WB MTRRs */
 866         mtrr_bp_init();
 867         if (mtrr_trim_uncached_memory(max_pfn))
 868                 max_pfn = e820_end_of_ram_pfn();
 869
 870 #ifdef CONFIG_X86_32
 871         /* max_low_pfn get updated here */
 872         find_low_pfn_range();
 873 #else
 874         num_physpages = max_pfn;
 875
 876         check_x2apic();
 877
 878         /* How many end-of-memory variables you have, grandma! */
 879         /* need this before calling reserve_initrd */
 880         if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
 881                 max_low_pfn = e820_end_of_low_ram_pfn();
 882         else
 883                 max_low_pfn = max_pfn;
 884
 885         high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
 886 #endif
 887
 888         /*
 889          * Find and reserve possible boot-time SMP configuration:
 890          */
 891         find_smp_config();
 892
 893         reserve_ibft_region();
 894
 895         /*
 896          * Need to conclude brk, before memblock_x86_fill()
 897          *  it could use memblock_find_in_range, could overlap with
 898          *  brk area.
 899          */
 900         reserve_brk();
 901
 902         cleanup_highmap();
 903
 904         memblock.current_limit = get_max_mapped();
 905         memblock_x86_fill();
 906
 907         /*
 908          * The EFI specification says that boot service code won't be called
 909          * after ExitBootServices(). This is, in fact, a lie.
 910          */
 911         if (efi_enabled)
 912                 efi_reserve_boot_services();
 913
 914         /* preallocate 4k for mptable mpc */
 915         early_reserve_e820_mpc_new();
 916
 917 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
 918         setup_bios_corruption_check();
 919 #endif
 920
 921         printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
 922                         max_pfn_mapped<<PAGE_SHIFT);
 923
 924         setup_trampolines();
 925
 926         init_gbpages();
 927
 928         /* max_pfn_mapped is updated here */
 929         max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
 930         max_pfn_mapped = max_low_pfn_mapped;
 931
 932 #ifdef CONFIG_X86_64
 933         if (max_pfn > max_low_pfn) {
 934                 max_pfn_mapped = init_memory_mapping(1UL<<32,
 935                                                      max_pfn<<PAGE_SHIFT);
 936                 /* can we preseve max_low_pfn ?*/
 937                 max_low_pfn = max_pfn;
 938         }
 939 #endif
 940         memblock.current_limit = get_max_mapped();
 941
 942         /*
 943          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 944          */
 945
 946 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 947         if (init_ohci1394_dma_early)
 948                 init_ohci1394_dma_on_all_controllers();
 949 #endif
 950         /* Allocate bigger log buffer */
 951         setup_log_buf(1);
 952
 953         reserve_initrd();
 954
 955         reserve_crashkernel();
 956
 957         vsmp_init();
 958
 959         io_delay_init();
 960
 961         /*
 962          * Parse the ACPI tables for possible boot-time SMP configuration.
 963          */
 964         acpi_boot_table_init();
 965
 966         early_acpi_boot_init();
 967
 968         initmem_init();
 969         memblock_find_dma_reserve();
 970
 971 #ifdef CONFIG_KVM_CLOCK
 972         kvmclock_init();
 973 #endif
 974
 975         x86_init.paging.pagetable_setup_start(swapper_pg_dir);
 976         paging_init();
 977         x86_init.paging.pagetable_setup_done(swapper_pg_dir);
 978
 979         if (boot_cpu_data.cpuid_level >= 0) {
 980                 /* A CPU has %cr4 if and only if it has CPUID */
 981                 mmu_cr4_features = read_cr4();
 982         }
 983
 984 #ifdef CONFIG_X86_32
 985         /* sync back kernel address range */
 986         clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
 987                         swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 988                         KERNEL_PGD_PTRS);
 989 #endif
 990
 991         tboot_probe();
 992
 993 #ifdef CONFIG_X86_64
 994         map_vsyscall();
 995 #endif
 996
 997         generic_apic_probe();
 998
 999         early_quirks();
1000
1001         /*
1002          * Read APIC and some other early information from ACPI tables.
1003          */
1004         acpi_boot_init();
1005         sfi_init();
1006         x86_dtb_init();
1007
1008         /*
1009          * get boot-time SMP configuration:
1010          */
1011         if (smp_found_config)
1012                 get_smp_config();
1013
1014         prefill_possible_map();
1015
1016         init_cpu_to_node();
1017
1018         init_apic_mappings();
1019         ioapic_and_gsi_init();
1020
1021         kvm_guest_init();
1022
1023         e820_reserve_resources();
1024         e820_mark_nosave_regions(max_low_pfn);
1025
1026         x86_init.resources.reserve_resources();
1027
1028         e820_setup_gap();
1029
1030 #ifdef CONFIG_VT
1031 #if defined(CONFIG_VGA_CONSOLE)
1032         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1033                 conswitchp = &vga_con;
1034 #elif defined(CONFIG_DUMMY_CONSOLE)
1035         conswitchp = &dummy_con;
1036 #endif
1037 #endif
1038         x86_init.oem.banner();
1039
1040         x86_init.timers.wallclock_init();
1041
1042         x86_platform.wallclock_init();
1043
1044         mcheck_init();
1045
1046         arch_init_ideal_nops();
1047 }
1048
1049 #ifdef CONFIG_X86_32
1050
1051 static struct resource video_ram_resource = {
1052         .name   = "Video RAM area",
1053         .start  = 0xa0000,
1054         .end    = 0xbffff,
1055         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
1056 };
1057
1058 void __init i386_reserve_resources(void)
1059 {
1060         request_resource(&iomem_resource, &video_ram_resource);
1061         reserve_standard_io_resources();
1062 }
1063
1064 #endif /* CONFIG_X86_32 */