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
 297 static void __init cleanup_highmap_brk_end(void)
 298 {
 299         pud_t *pud;
 300         pmd_t *pmd;
 301
 302         mmu_cr4_features = read_cr4();
 303
 304         /*
 305          * _brk_end cannot change anymore, but it and _end may be
 306          * located on different 2M pages. cleanup_highmap(), however,
 307          * can only consider _end when it runs, so destroy any
 308          * mappings beyond _brk_end here.
 309          */
 310         pud = pud_offset(pgd_offset_k(_brk_end), _brk_end);
 311         pmd = pmd_offset(pud, _brk_end - 1);
 312         while (++pmd <= pmd_offset(pud, (unsigned long)_end - 1))
 313                 pmd_clear(pmd);
 314 }
 315 #else
 316 static inline void init_gbpages(void)
 317 {
 318 }
 319 static inline void cleanup_highmap_brk_end(void)
 320 {
 321 }
 322 #endif
 323
 324 static void __init reserve_brk(void)
 325 {
 326         if (_brk_end > _brk_start)
 327                 memblock_x86_reserve_range(__pa(_brk_start), __pa(_brk_end), "BRK");
 328
 329         /* Mark brk area as locked down and no longer taking any
 330            new allocations */
 331         _brk_start = 0;
 332
 333         cleanup_highmap_brk_end();
 334 }
 335
 336 #ifdef CONFIG_BLK_DEV_INITRD
 337
 338 #define MAX_MAP_CHUNK   (NR_FIX_BTMAPS << PAGE_SHIFT)
 339 static void __init relocate_initrd(void)
 340 {
 341         /* Assume only end is not page aligned */
 342         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 343         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 344         u64 area_size     = PAGE_ALIGN(ramdisk_size);
 345         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 346         u64 ramdisk_here;
 347         unsigned long slop, clen, mapaddr;
 348         char *p, *q;
 349
 350         /* We need to move the initrd down into lowmem */
 351         ramdisk_here = memblock_find_in_range(0, end_of_lowmem, area_size,
 352                                          PAGE_SIZE);
 353
 354         if (ramdisk_here == MEMBLOCK_ERROR)
 355                 panic("Cannot find place for new RAMDISK of size %lld\n",
 356                          ramdisk_size);
 357
 358         /* Note: this includes all the lowmem currently occupied by
 359            the initrd, we rely on that fact to keep the data intact. */
 360         memblock_x86_reserve_range(ramdisk_here, ramdisk_here + area_size, "NEW RAMDISK");
 361         initrd_start = ramdisk_here + PAGE_OFFSET;
 362         initrd_end   = initrd_start + ramdisk_size;
 363         printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
 364                          ramdisk_here, ramdisk_here + ramdisk_size);
 365
 366         q = (char *)initrd_start;
 367
 368         /* Copy any lowmem portion of the initrd */
 369         if (ramdisk_image < end_of_lowmem) {
 370                 clen = end_of_lowmem - ramdisk_image;
 371                 p = (char *)__va(ramdisk_image);
 372                 memcpy(q, p, clen);
 373                 q += clen;
 374                 ramdisk_image += clen;
 375                 ramdisk_size  -= clen;
 376         }
 377
 378         /* Copy the highmem portion of the initrd */
 379         while (ramdisk_size) {
 380                 slop = ramdisk_image & ~PAGE_MASK;
 381                 clen = ramdisk_size;
 382                 if (clen > MAX_MAP_CHUNK-slop)
 383                         clen = MAX_MAP_CHUNK-slop;
 384                 mapaddr = ramdisk_image & PAGE_MASK;
 385                 p = early_memremap(mapaddr, clen+slop);
 386                 memcpy(q, p+slop, clen);
 387                 early_iounmap(p, clen+slop);
 388                 q += clen;
 389                 ramdisk_image += clen;
 390                 ramdisk_size  -= clen;
 391         }
 392         /* high pages is not converted by early_res_to_bootmem */
 393         ramdisk_image = boot_params.hdr.ramdisk_image;
 394         ramdisk_size  = boot_params.hdr.ramdisk_size;
 395         printk(KERN_INFO "Move RAMDISK from %016llx - %016llx to"
 396                 " %08llx - %08llx\n",
 397                 ramdisk_image, ramdisk_image + ramdisk_size - 1,
 398                 ramdisk_here, ramdisk_here + ramdisk_size - 1);
 399 }
 400
 401 static void __init reserve_initrd(void)
 402 {
 403         /* Assume only end is not page aligned */
 404         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 405         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 406         u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 407         u64 end_of_lowmem = max_low_pfn_mapped << PAGE_SHIFT;
 408
 409         if (!boot_params.hdr.type_of_loader ||
 410             !ramdisk_image || !ramdisk_size)
 411                 return;         /* No initrd provided by bootloader */
 412
 413         initrd_start = 0;
 414
 415         if (ramdisk_size >= (end_of_lowmem>>1)) {
 416                 memblock_x86_free_range(ramdisk_image, ramdisk_end);
 417                 printk(KERN_ERR "initrd too large to handle, "
 418                        "disabling initrd\n");
 419                 return;
 420         }
 421
 422         printk(KERN_INFO "RAMDISK: %08llx - %08llx\n", ramdisk_image,
 423                         ramdisk_end);
 424
 425
 426         if (ramdisk_end <= end_of_lowmem) {
 427                 /* All in lowmem, easy case */
 428                 /*
 429                  * don't need to reserve again, already reserved early
 430                  * in i386_start_kernel
 431                  */
 432                 initrd_start = ramdisk_image + PAGE_OFFSET;
 433                 initrd_end = initrd_start + ramdisk_size;
 434                 return;
 435         }
 436
 437         relocate_initrd();
 438
 439         memblock_x86_free_range(ramdisk_image, ramdisk_end);
 440 }
 441 #else
 442 static void __init reserve_initrd(void)
 443 {
 444 }
 445 #endif /* CONFIG_BLK_DEV_INITRD */
 446
 447 static void __init parse_setup_data(void)
 448 {
 449         struct setup_data *data;
 450         u64 pa_data;
 451
 452         if (boot_params.hdr.version < 0x0209)
 453                 return;
 454         pa_data = boot_params.hdr.setup_data;
 455         while (pa_data) {
 456                 data = early_memremap(pa_data, PAGE_SIZE);
 457                 switch (data->type) {
 458                 case SETUP_E820_EXT:
 459                         parse_e820_ext(data, pa_data);
 460                         break;
 461                 default:
 462                         break;
 463                 }
 464                 pa_data = data->next;
 465                 early_iounmap(data, PAGE_SIZE);
 466         }
 467 }
 468
 469 static void __init e820_reserve_setup_data(void)
 470 {
 471         struct setup_data *data;
 472         u64 pa_data;
 473         int found = 0;
 474
 475         if (boot_params.hdr.version < 0x0209)
 476                 return;
 477         pa_data = boot_params.hdr.setup_data;
 478         while (pa_data) {
 479                 data = early_memremap(pa_data, sizeof(*data));
 480                 e820_update_range(pa_data, sizeof(*data)+data->len,
 481                          E820_RAM, E820_RESERVED_KERN);
 482                 found = 1;
 483                 pa_data = data->next;
 484                 early_iounmap(data, sizeof(*data));
 485         }
 486         if (!found)
 487                 return;
 488
 489         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 490         memcpy(&e820_saved, &e820, sizeof(struct e820map));
 491         printk(KERN_INFO "extended physical RAM map:\n");
 492         e820_print_map("reserve setup_data");
 493 }
 494
 495 static void __init memblock_x86_reserve_range_setup_data(void)
 496 {
 497         struct setup_data *data;
 498         u64 pa_data;
 499         char buf[32];
 500
 501         if (boot_params.hdr.version < 0x0209)
 502                 return;
 503         pa_data = boot_params.hdr.setup_data;
 504         while (pa_data) {
 505                 data = early_memremap(pa_data, sizeof(*data));
 506                 sprintf(buf, "setup data %x", data->type);
 507                 memblock_x86_reserve_range(pa_data, pa_data+sizeof(*data)+data->len, buf);
 508                 pa_data = data->next;
 509                 early_iounmap(data, sizeof(*data));
 510         }
 511 }
 512
 513 /*
 514  * --------- Crashkernel reservation ------------------------------
 515  */
 516
 517 #ifdef CONFIG_KEXEC
 518
 519 static inline unsigned long long get_total_mem(void)
 520 {
 521         unsigned long long total;
 522
 523         total = max_pfn - min_low_pfn;
 524
 525         return total << PAGE_SHIFT;
 526 }
 527
 528 /*
 529  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
 530  * would limit the kernel to the low 512 MiB due to mapping restrictions.
 531  * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this
 532  * limit once kexec-tools are fixed.
 533  */
 534 #ifdef CONFIG_X86_32
 535 # define CRASH_KERNEL_ADDR_MAX  (512 << 20)
 536 #else
 537 # define CRASH_KERNEL_ADDR_MAX  (896 << 20)
 538 #endif
 539
 540 static void __init reserve_crashkernel(void)
 541 {
 542         unsigned long long total_mem;
 543         unsigned long long crash_size, crash_base;
 544         int ret;
 545
 546         total_mem = get_total_mem();
 547
 548         ret = parse_crashkernel(boot_command_line, total_mem,
 549                         &crash_size, &crash_base);
 550         if (ret != 0 || crash_size <= 0)
 551                 return;
 552
 553         /* 0 means: find the address automatically */
 554         if (crash_base <= 0) {
 555                 const unsigned long long alignment = 16<<20;    /* 16M */
 556
 557                 /*
 558                  *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
 559                  */
 560                 crash_base = memblock_find_in_range(alignment,
 561                                CRASH_KERNEL_ADDR_MAX, crash_size, alignment);
 562
 563                 if (crash_base == MEMBLOCK_ERROR) {
 564                         pr_info("crashkernel reservation failed - No suitable area found.\n");
 565                         return;
 566                 }
 567         } else {
 568                 unsigned long long start;
 569
 570                 start = memblock_find_in_range(crash_base,
 571                                  crash_base + crash_size, crash_size, 1<<20);
 572                 if (start != crash_base) {
 573                         pr_info("crashkernel reservation failed - memory is in use.\n");
 574                         return;
 575                 }
 576         }
 577         memblock_x86_reserve_range(crash_base, crash_base + crash_size, "CRASH KERNEL");
 578
 579         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 580                         "for crashkernel (System RAM: %ldMB)\n",
 581                         (unsigned long)(crash_size >> 20),
 582                         (unsigned long)(crash_base >> 20),
 583                         (unsigned long)(total_mem >> 20));
 584
 585         crashk_res.start = crash_base;
 586         crashk_res.end   = crash_base + crash_size - 1;
 587         insert_resource(&iomem_resource, &crashk_res);
 588 }
 589 #else
 590 static void __init reserve_crashkernel(void)
 591 {
 592 }
 593 #endif
 594
 595 static struct resource standard_io_resources[] = {
 596         { .name = "dma1", .start = 0x00, .end = 0x1f,
 597                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 598         { .name = "pic1", .start = 0x20, .end = 0x21,
 599                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 600         { .name = "timer0", .start = 0x40, .end = 0x43,
 601                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 602         { .name = "timer1", .start = 0x50, .end = 0x53,
 603                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 604         { .name = "keyboard", .start = 0x60, .end = 0x60,
 605                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 606         { .name = "keyboard", .start = 0x64, .end = 0x64,
 607                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 608         { .name = "dma page reg", .start = 0x80, .end = 0x8f,
 609                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 610         { .name = "pic2", .start = 0xa0, .end = 0xa1,
 611                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 612         { .name = "dma2", .start = 0xc0, .end = 0xdf,
 613                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 614         { .name = "fpu", .start = 0xf0, .end = 0xff,
 615                 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
 616 };
 617
 618 void __init reserve_standard_io_resources(void)
 619 {
 620         int i;
 621
 622         /* request I/O space for devices used on all i[345]86 PCs */
 623         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 624                 request_resource(&ioport_resource, &standard_io_resources[i]);
 625
 626 }
 627
 628 /*
 629  * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
 630  * is_kdump_kernel() to determine if we are booting after a panic. Hence
 631  * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
 632  */
 633
 634 #ifdef CONFIG_CRASH_DUMP
 635 /* elfcorehdr= specifies the location of elf core header
 636  * stored by the crashed kernel. This option will be passed
 637  * by kexec loader to the capture kernel.
 638  */
 639 static int __init setup_elfcorehdr(char *arg)
 640 {
 641         char *end;
 642         if (!arg)
 643                 return -EINVAL;
 644         elfcorehdr_addr = memparse(arg, &end);
 645         return end > arg ? 0 : -EINVAL;
 646 }
 647 early_param("elfcorehdr", setup_elfcorehdr);
 648 #endif
 649
 650 static __init void reserve_ibft_region(void)
 651 {
 652         unsigned long addr, size = 0;
 653
 654         addr = find_ibft_region(&size);
 655
 656         if (size)
 657                 memblock_x86_reserve_range(addr, addr + size, "* ibft");
 658 }
 659
 660 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 661
 662 static void __init trim_bios_range(void)
 663 {
 664         /*
 665          * A special case is the first 4Kb of memory;
 666          * This is a BIOS owned area, not kernel ram, but generally
 667          * not listed as such in the E820 table.
 668          *
 669          * This typically reserves additional memory (64KiB by default)
 670          * since some BIOSes are known to corrupt low memory.  See the
 671          * Kconfig help text for X86_RESERVE_LOW.
 672          */
 673         e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE),
 674                           E820_RAM, E820_RESERVED);
 675
 676         /*
 677          * special case: Some BIOSen report the PC BIOS
 678          * area (640->1Mb) as ram even though it is not.
 679          * take them out.
 680          */
 681         e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 682         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 683 }
 684
 685 static int __init parse_reservelow(char *p)
 686 {
 687         unsigned long long size;
 688
 689         if (!p)
 690                 return -EINVAL;
 691
 692         size = memparse(p, &p);
 693
 694         if (size < 4096)
 695                 size = 4096;
 696
 697         if (size > 640*1024)
 698                 size = 640*1024;
 699
 700         reserve_low = size;
 701
 702         return 0;
 703 }
 704
 705 early_param("reservelow", parse_reservelow);
 706
 707 /*
 708  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 709  * passed the efi memmap, systab, etc., so we should use these data structures
 710  * for initialization.  Note, the efi init code path is determined by the
 711  * global efi_enabled. This allows the same kernel image to be used on existing
 712  * systems (with a traditional BIOS) as well as on EFI systems.
 713  */
 714 /*
 715  * setup_arch - architecture-specific boot-time initializations
 716  *
 717  * Note: On x86_64, fixmaps are ready for use even before this is called.
 718  */
 719
 720 void __init setup_arch(char **cmdline_p)
 721 {
 722         int acpi = 0;
 723         int amd = 0;
 724         unsigned long flags;
 725
 726 #ifdef CONFIG_X86_32
 727         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 728         visws_early_detect();
 729
 730         /*
 731          * copy kernel address range established so far and switch
 732          * to the proper swapper page table
 733          */
 734         clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 735                         initial_page_table + KERNEL_PGD_BOUNDARY,
 736                         KERNEL_PGD_PTRS);
 737
 738         load_cr3(swapper_pg_dir);
 739         __flush_tlb_all();
 740 #else
 741         printk(KERN_INFO "Command line: %s\n", boot_command_line);
 742 #endif
 743
 744         /*
 745          * If we have OLPC OFW, we might end up relocating the fixmap due to
 746          * reserve_top(), so do this before touching the ioremap area.
 747          */
 748         olpc_ofw_detect();
 749
 750         early_trap_init();
 751         early_cpu_init();
 752         early_ioremap_init();
 753
 754         setup_olpc_ofw_pgd();
 755
 756         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 757         screen_info = boot_params.screen_info;
 758         edid_info = boot_params.edid_info;
 759 #ifdef CONFIG_X86_32
 760         apm_info.bios = boot_params.apm_bios_info;
 761         ist_info = boot_params.ist_info;
 762         if (boot_params.sys_desc_table.length != 0) {
 763                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 764                 machine_id = boot_params.sys_desc_table.table[0];
 765                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 766                 BIOS_revision = boot_params.sys_desc_table.table[2];
 767         }
 768 #endif
 769         saved_video_mode = boot_params.hdr.vid_mode;
 770         bootloader_type = boot_params.hdr.type_of_loader;
 771         if ((bootloader_type >> 4) == 0xe) {
 772                 bootloader_type &= 0xf;
 773                 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 774         }
 775         bootloader_version  = bootloader_type & 0xf;
 776         bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 777
 778 #ifdef CONFIG_BLK_DEV_RAM
 779         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 780         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 781         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 782 #endif
 783 #ifdef CONFIG_EFI
 784         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 785 #ifdef CONFIG_X86_32
 786                      "EL32",
 787 #else
 788                      "EL64",
 789 #endif
 790          4)) {
 791                 efi_enabled = 1;
 792                 efi_memblock_x86_reserve_range();
 793         }
 794 #endif
 795
 796         x86_init.oem.arch_setup();
 797
 798         iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 799         setup_memory_map();
 800         parse_setup_data();
 801         /* update the e820_saved too */
 802         e820_reserve_setup_data();
 803
 804         copy_edd();
 805
 806         if (!boot_params.hdr.root_flags)
 807                 root_mountflags &= ~MS_RDONLY;
 808         init_mm.start_code = (unsigned long) _text;
 809         init_mm.end_code = (unsigned long) _etext;
 810         init_mm.end_data = (unsigned long) _edata;
 811         init_mm.brk = _brk_end;
 812
 813         code_resource.start = virt_to_phys(_text);
 814         code_resource.end = virt_to_phys(_etext)-1;
 815         data_resource.start = virt_to_phys(_etext);
 816         data_resource.end = virt_to_phys(_edata)-1;
 817         bss_resource.start = virt_to_phys(&__bss_start);
 818         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 819
 820 #ifdef CONFIG_CMDLINE_BOOL
 821 #ifdef CONFIG_CMDLINE_OVERRIDE
 822         strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 823 #else
 824         if (builtin_cmdline[0]) {
 825                 /* append boot loader cmdline to builtin */
 826                 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 827                 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 828                 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 829         }
 830 #endif
 831 #endif
 832
 833         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 834         *cmdline_p = command_line;
 835
 836         /*
 837          * x86_configure_nx() is called before parse_early_param() to detect
 838          * whether hardware doesn't support NX (so that the early EHCI debug
 839          * console setup can safely call set_fixmap()). It may then be called
 840          * again from within noexec_setup() during parsing early parameters
 841          * to honor the respective command line option.
 842          */
 843         x86_configure_nx();
 844
 845         parse_early_param();
 846
 847         x86_report_nx();
 848
 849         /* after early param, so could get panic from serial */
 850         memblock_x86_reserve_range_setup_data();
 851
 852         if (acpi_mps_check()) {
 853 #ifdef CONFIG_X86_LOCAL_APIC
 854                 disable_apic = 1;
 855 #endif
 856                 setup_clear_cpu_cap(X86_FEATURE_APIC);
 857         }
 858
 859 #ifdef CONFIG_PCI
 860         if (pci_early_dump_regs)
 861                 early_dump_pci_devices();
 862 #endif
 863
 864         finish_e820_parsing();
 865
 866         if (efi_enabled)
 867                 efi_init();
 868
 869         dmi_scan_machine();
 870
 871         /*
 872          * VMware detection requires dmi to be available, so this
 873          * needs to be done after dmi_scan_machine, for the BP.
 874          */
 875         init_hypervisor_platform();
 876
 877         x86_init.resources.probe_roms();
 878
 879         /* after parse_early_param, so could debug it */
 880         insert_resource(&iomem_resource, &code_resource);
 881         insert_resource(&iomem_resource, &data_resource);
 882         insert_resource(&iomem_resource, &bss_resource);
 883
 884         trim_bios_range();
 885 #ifdef CONFIG_X86_32
 886         if (ppro_with_ram_bug()) {
 887                 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
 888                                   E820_RESERVED);
 889                 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 890                 printk(KERN_INFO "fixed physical RAM map:\n");
 891                 e820_print_map("bad_ppro");
 892         }
 893 #else
 894         early_gart_iommu_check();
 895 #endif
 896
 897         /*
 898          * partially used pages are not usable - thus
 899          * we are rounding upwards:
 900          */
 901         max_pfn = e820_end_of_ram_pfn();
 902
 903         /* update e820 for memory not covered by WB MTRRs */
 904         mtrr_bp_init();
 905         if (mtrr_trim_uncached_memory(max_pfn))
 906                 max_pfn = e820_end_of_ram_pfn();
 907
 908 #ifdef CONFIG_X86_32
 909         /* max_low_pfn get updated here */
 910         find_low_pfn_range();
 911 #else
 912         num_physpages = max_pfn;
 913
 914         check_x2apic();
 915
 916         /* How many end-of-memory variables you have, grandma! */
 917         /* need this before calling reserve_initrd */
 918         if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
 919                 max_low_pfn = e820_end_of_low_ram_pfn();
 920         else
 921                 max_low_pfn = max_pfn;
 922
 923         high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
 924 #endif
 925
 926         /*
 927          * Find and reserve possible boot-time SMP configuration:
 928          */
 929         find_smp_config();
 930
 931         reserve_ibft_region();
 932
 933         /*
 934          * Need to conclude brk, before memblock_x86_fill()
 935          *  it could use memblock_find_in_range, could overlap with
 936          *  brk area.
 937          */
 938         reserve_brk();
 939
 940         memblock.current_limit = get_max_mapped();
 941         memblock_x86_fill();
 942
 943         /* preallocate 4k for mptable mpc */
 944         early_reserve_e820_mpc_new();
 945
 946 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
 947         setup_bios_corruption_check();
 948 #endif
 949
 950         printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
 951                         max_pfn_mapped<<PAGE_SHIFT);
 952
 953         reserve_trampoline_memory();
 954
 955 #ifdef CONFIG_ACPI_SLEEP
 956         /*
 957          * Reserve low memory region for sleep support.
 958          * even before init_memory_mapping
 959          */
 960         acpi_reserve_wakeup_memory();
 961 #endif
 962         init_gbpages();
 963
 964         /* max_pfn_mapped is updated here */
 965         max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
 966         max_pfn_mapped = max_low_pfn_mapped;
 967
 968         memblock.current_limit = get_max_mapped();
 969
 970         /*
 971          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 972          */
 973
 974 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 975         if (init_ohci1394_dma_early)
 976                 init_ohci1394_dma_on_all_controllers();
 977 #endif
 978
 979         reserve_initrd();
 980
 981         reserve_crashkernel();
 982
 983         vsmp_init();
 984
 985         io_delay_init();
 986
 987         /*
 988          * Parse the ACPI tables for possible boot-time SMP configuration.
 989          */
 990         acpi_boot_table_init();
 991
 992         early_acpi_boot_init();
 993
 994 #ifdef CONFIG_ACPI_NUMA
 995         /*
 996          * Parse SRAT to discover nodes.
 997          */
 998         acpi = !x86_acpi_numa_init();
 999 #endif
1000
1001 #ifdef CONFIG_AMD_NUMA
1002         if (!acpi)
1003                 amd = !amd_numa_init();
1004 #endif
1005
1006         initmem_init(acpi, amd);
1007         memblock_find_dma_reserve();
1008         dma32_reserve_bootmem();
1009
1010 #ifdef CONFIG_KVM_CLOCK
1011         kvmclock_init();
1012 #endif
1013
1014         x86_init.paging.pagetable_setup_start(swapper_pg_dir);
1015         paging_init();
1016         x86_init.paging.pagetable_setup_done(swapper_pg_dir);
1017
1018 #ifdef CONFIG_X86_32
1019         /* sync back kernel address range */
1020         clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1021                         swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1022                         KERNEL_PGD_PTRS);
1023 #endif
1024
1025         tboot_probe();
1026
1027 #ifdef CONFIG_X86_64
1028         map_vsyscall();
1029 #endif
1030
1031         generic_apic_probe();
1032
1033         early_quirks();
1034
1035         /*
1036          * Read APIC and some other early information from ACPI tables.
1037          */
1038         acpi_boot_init();
1039
1040         sfi_init();
1041
1042         /*
1043          * get boot-time SMP configuration:
1044          */
1045         if (smp_found_config)
1046                 get_smp_config();
1047
1048         prefill_possible_map();
1049
1050         init_cpu_to_node();
1051
1052         init_apic_mappings();
1053         ioapic_and_gsi_init();
1054
1055         kvm_guest_init();
1056
1057         e820_reserve_resources();
1058         e820_mark_nosave_regions(max_low_pfn);
1059
1060         x86_init.resources.reserve_resources();
1061
1062         e820_setup_gap();
1063
1064 #ifdef CONFIG_VT
1065 #if defined(CONFIG_VGA_CONSOLE)
1066         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1067                 conswitchp = &vga_con;
1068 #elif defined(CONFIG_DUMMY_CONSOLE)
1069         conswitchp = &dummy_con;
1070 #endif
1071 #endif
1072         x86_init.oem.banner();
1073
1074         mcheck_init();
1075
1076         local_irq_save(flags);
1077         arch_init_ideal_nop5();
1078         local_irq_restore(flags);
1079 }
1080
1081 #ifdef CONFIG_X86_32
1082
1083 static struct resource video_ram_resource = {
1084         .name   = "Video RAM area",
1085         .start  = 0xa0000,
1086         .end    = 0xbffff,
1087         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
1088 };
1089
1090 void __init i386_reserve_resources(void)
1091 {
1092         request_resource(&iomem_resource, &video_ram_resource);
1093         reserve_standard_io_resources();
1094 }
1095
1096 #endif /* CONFIG_X86_32 */