arch/x86/kernel/setup_32.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/apm_bios.h>
  31 #include <linux/initrd.h>
  32 #include <linux/bootmem.h>
  33 #include <linux/seq_file.h>
  34 #include <linux/console.h>
  35 #include <linux/mca.h>
  36 #include <linux/root_dev.h>
  37 #include <linux/highmem.h>
  38 #include <linux/module.h>
  39 #include <linux/efi.h>
  40 #include <linux/init.h>
  41 #include <linux/edd.h>
  42 #include <linux/iscsi_ibft.h>
  43 #include <linux/nodemask.h>
  44 #include <linux/kexec.h>
  45 #include <linux/crash_dump.h>
  46 #include <linux/dmi.h>
  47 #include <linux/pfn.h>
  48 #include <linux/pci.h>
  49 #include <linux/init_ohci1394_dma.h>
  50 #include <linux/kvm_para.h>
  51
  52 #include <video/edid.h>
  53
  54 #include <asm/mtrr.h>
  55 #include <asm/apic.h>
  56 #include <asm/e820.h>
  57 #include <asm/mpspec.h>
  58 #include <asm/mmzone.h>
  59 #include <asm/setup.h>
  60 #include <asm/arch_hooks.h>
  61 #include <asm/sections.h>
  62 #include <asm/io_apic.h>
  63 #include <asm/ist.h>
  64 #include <asm/io.h>
  65 #include <asm/vmi.h>
  66 #include <setup_arch.h>
  67 #include <asm/bios_ebda.h>
  68 #include <asm/cacheflush.h>
  69 #include <asm/processor.h>
  70 #include <asm/efi.h>
  71
  72 /* This value is set up by the early boot code to point to the value
  73    immediately after the boot time page tables.  It contains a *physical*
  74    address, and must not be in the .bss segment! */
  75 unsigned long init_pg_tables_start __initdata = ~0UL;
  76 unsigned long init_pg_tables_end __initdata = ~0UL;
  77
  78 /*
  79  * Machine setup..
  80  */
  81 static struct resource data_resource = {
  82         .name   = "Kernel data",
  83         .start  = 0,
  84         .end    = 0,
  85         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  86 };
  87
  88 static struct resource code_resource = {
  89         .name   = "Kernel code",
  90         .start  = 0,
  91         .end    = 0,
  92         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  93 };
  94
  95 static struct resource bss_resource = {
  96         .name   = "Kernel bss",
  97         .start  = 0,
  98         .end    = 0,
  99         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 100 };
 101
 102 static struct resource video_ram_resource = {
 103         .name   = "Video RAM area",
 104         .start  = 0xa0000,
 105         .end    = 0xbffff,
 106         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 107 };
 108
 109 static struct resource standard_io_resources[] = { {
 110         .name   = "dma1",
 111         .start  = 0x0000,
 112         .end    = 0x001f,
 113         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 114 }, {
 115         .name   = "pic1",
 116         .start  = 0x0020,
 117         .end    = 0x0021,
 118         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 119 }, {
 120         .name   = "timer0",
 121         .start  = 0x0040,
 122         .end    = 0x0043,
 123         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 124 }, {
 125         .name   = "timer1",
 126         .start  = 0x0050,
 127         .end    = 0x0053,
 128         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 129 }, {
 130         .name   = "keyboard",
 131         .start  = 0x0060,
 132         .end    = 0x0060,
 133         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 134 }, {
 135         .name   = "keyboard",
 136         .start  = 0x0064,
 137         .end    = 0x0064,
 138         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 139 }, {
 140         .name   = "dma page reg",
 141         .start  = 0x0080,
 142         .end    = 0x008f,
 143         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 144 }, {
 145         .name   = "pic2",
 146         .start  = 0x00a0,
 147         .end    = 0x00a1,
 148         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 149 }, {
 150         .name   = "dma2",
 151         .start  = 0x00c0,
 152         .end    = 0x00df,
 153         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 154 }, {
 155         .name   = "fpu",
 156         .start  = 0x00f0,
 157         .end    = 0x00ff,
 158         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 159 } };
 160
 161 /* cpu data as detected by the assembly code in head.S */
 162 struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
 163 /* common cpu data for all cpus */
 164 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
 165 EXPORT_SYMBOL(boot_cpu_data);
 166
 167 unsigned int def_to_bigsmp;
 168
 169 #ifndef CONFIG_X86_PAE
 170 unsigned long mmu_cr4_features;
 171 #else
 172 unsigned long mmu_cr4_features = X86_CR4_PAE;
 173 #endif
 174
 175 /* for MCA, but anyone else can use it if they want */
 176 unsigned int machine_id;
 177 unsigned int machine_submodel_id;
 178 unsigned int BIOS_revision;
 179
 180 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
 181 int bootloader_type;
 182
 183 /* user-defined highmem size */
 184 static unsigned int highmem_pages = -1;
 185
 186 /*
 187  * Setup options
 188  */
 189 struct screen_info screen_info;
 190 EXPORT_SYMBOL(screen_info);
 191 struct apm_info apm_info;
 192 EXPORT_SYMBOL(apm_info);
 193 struct edid_info edid_info;
 194 EXPORT_SYMBOL_GPL(edid_info);
 195 struct ist_info ist_info;
 196 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 197         defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 198 EXPORT_SYMBOL(ist_info);
 199 #endif
 200
 201 extern void early_cpu_init(void);
 202 extern int root_mountflags;
 203
 204 unsigned long saved_video_mode;
 205
 206 #define RAMDISK_IMAGE_START_MASK        0x07FF
 207 #define RAMDISK_PROMPT_FLAG             0x8000
 208 #define RAMDISK_LOAD_FLAG               0x4000
 209
 210 static char __initdata command_line[COMMAND_LINE_SIZE];
 211
 212 #ifndef CONFIG_DEBUG_BOOT_PARAMS
 213 struct boot_params __initdata boot_params;
 214 #else
 215 struct boot_params boot_params;
 216 #endif
 217
 218 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 219 struct edd edd;
 220 #ifdef CONFIG_EDD_MODULE
 221 EXPORT_SYMBOL(edd);
 222 #endif
 223 /**
 224  * copy_edd() - Copy the BIOS EDD information
 225  *              from boot_params into a safe place.
 226  *
 227  */
 228 static inline void copy_edd(void)
 229 {
 230      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 231             sizeof(edd.mbr_signature));
 232      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 233      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 234      edd.edd_info_nr = boot_params.eddbuf_entries;
 235 }
 236 #else
 237 static inline void copy_edd(void)
 238 {
 239 }
 240 #endif
 241
 242 #ifdef CONFIG_PROC_VMCORE
 243 /* elfcorehdr= specifies the location of elf core header
 244  * stored by the crashed kernel.
 245  */
 246 static int __init parse_elfcorehdr(char *arg)
 247 {
 248         if (!arg)
 249                 return -EINVAL;
 250
 251         elfcorehdr_addr = memparse(arg, &arg);
 252         return 0;
 253 }
 254 early_param("elfcorehdr", parse_elfcorehdr);
 255 #endif /* CONFIG_PROC_VMCORE */
 256
 257 /*
 258  * highmem=size forces highmem to be exactly 'size' bytes.
 259  * This works even on boxes that have no highmem otherwise.
 260  * This also works to reduce highmem size on bigger boxes.
 261  */
 262 static int __init parse_highmem(char *arg)
 263 {
 264         if (!arg)
 265                 return -EINVAL;
 266
 267         highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
 268         return 0;
 269 }
 270 early_param("highmem", parse_highmem);
 271
 272 /*
 273  * vmalloc=size forces the vmalloc area to be exactly 'size'
 274  * bytes. This can be used to increase (or decrease) the
 275  * vmalloc area - the default is 128m.
 276  */
 277 static int __init parse_vmalloc(char *arg)
 278 {
 279         if (!arg)
 280                 return -EINVAL;
 281
 282         __VMALLOC_RESERVE = memparse(arg, &arg);
 283         return 0;
 284 }
 285 early_param("vmalloc", parse_vmalloc);
 286
 287 /*
 288  * reservetop=size reserves a hole at the top of the kernel address space which
 289  * a hypervisor can load into later.  Needed for dynamically loaded hypervisors,
 290  * so relocating the fixmap can be done before paging initialization.
 291  */
 292 static int __init parse_reservetop(char *arg)
 293 {
 294         unsigned long address;
 295
 296         if (!arg)
 297                 return -EINVAL;
 298
 299         address = memparse(arg, &arg);
 300         reserve_top_address(address);
 301         return 0;
 302 }
 303 early_param("reservetop", parse_reservetop);
 304
 305 /*
 306  * Determine low and high memory ranges:
 307  */
 308 unsigned long __init find_max_low_pfn(void)
 309 {
 310         unsigned long max_low_pfn;
 311
 312         max_low_pfn = max_pfn;
 313         if (max_low_pfn > MAXMEM_PFN) {
 314                 if (highmem_pages == -1)
 315                         highmem_pages = max_pfn - MAXMEM_PFN;
 316                 if (highmem_pages + MAXMEM_PFN < max_pfn)
 317                         max_pfn = MAXMEM_PFN + highmem_pages;
 318                 if (highmem_pages + MAXMEM_PFN > max_pfn) {
 319                         printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
 320                         highmem_pages = 0;
 321                 }
 322                 max_low_pfn = MAXMEM_PFN;
 323 #ifndef CONFIG_HIGHMEM
 324                 /* Maximum memory usable is what is directly addressable */
 325                 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
 326                                         MAXMEM>>20);
 327                 if (max_pfn > MAX_NONPAE_PFN)
 328                         printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
 329                 else
 330                         printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
 331                 max_pfn = MAXMEM_PFN;
 332 #else /* !CONFIG_HIGHMEM */
 333 #ifndef CONFIG_HIGHMEM64G
 334                 if (max_pfn > MAX_NONPAE_PFN) {
 335                         max_pfn = MAX_NONPAE_PFN;
 336                         printk(KERN_WARNING "Warning only 4GB will be used.\n");
 337                         printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
 338                 }
 339 #endif /* !CONFIG_HIGHMEM64G */
 340 #endif /* !CONFIG_HIGHMEM */
 341         } else {
 342                 if (highmem_pages == -1)
 343                         highmem_pages = 0;
 344 #ifdef CONFIG_HIGHMEM
 345                 if (highmem_pages >= max_pfn) {
 346                         printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
 347                         highmem_pages = 0;
 348                 }
 349                 if (highmem_pages) {
 350                         if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
 351                                 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
 352                                 highmem_pages = 0;
 353                         }
 354                         max_low_pfn -= highmem_pages;
 355                 }
 356 #else
 357                 if (highmem_pages)
 358                         printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
 359 #endif
 360         }
 361         return max_low_pfn;
 362 }
 363
 364 #ifndef CONFIG_NEED_MULTIPLE_NODES
 365 static void __init setup_bootmem_allocator(void);
 366 static unsigned long __init setup_memory(void)
 367 {
 368         /*
 369          * partially used pages are not usable - thus
 370          * we are rounding upwards:
 371          */
 372         min_low_pfn = PFN_UP(init_pg_tables_end);
 373
 374         max_low_pfn = find_max_low_pfn();
 375
 376 #ifdef CONFIG_HIGHMEM
 377         highstart_pfn = highend_pfn = max_pfn;
 378         if (max_pfn > max_low_pfn) {
 379                 highstart_pfn = max_low_pfn;
 380         }
 381         memory_present(0, 0, highend_pfn);
 382         printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
 383                 pages_to_mb(highend_pfn - highstart_pfn));
 384         num_physpages = highend_pfn;
 385         high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
 386 #else
 387         memory_present(0, 0, max_low_pfn);
 388         num_physpages = max_low_pfn;
 389         high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
 390 #endif
 391 #ifdef CONFIG_FLATMEM
 392         max_mapnr = num_physpages;
 393 #endif
 394         printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
 395                         pages_to_mb(max_low_pfn));
 396
 397         setup_bootmem_allocator();
 398
 399         return max_low_pfn;
 400 }
 401
 402 static void __init zone_sizes_init(void)
 403 {
 404         unsigned long max_zone_pfns[MAX_NR_ZONES];
 405         memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
 406         max_zone_pfns[ZONE_DMA] =
 407                 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
 408         max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
 409         remove_all_active_ranges();
 410 #ifdef CONFIG_HIGHMEM
 411         max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
 412         e820_register_active_regions(0, 0, highend_pfn);
 413 #else
 414         e820_register_active_regions(0, 0, max_low_pfn);
 415 #endif
 416
 417         free_area_init_nodes(max_zone_pfns);
 418 }
 419 #else
 420 extern unsigned long __init setup_memory(void);
 421 extern void zone_sizes_init(void);
 422 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
 423
 424 static inline unsigned long long get_total_mem(void)
 425 {
 426         unsigned long long total;
 427
 428         total = max_low_pfn - min_low_pfn;
 429 #ifdef CONFIG_HIGHMEM
 430         total += highend_pfn - highstart_pfn;
 431 #endif
 432
 433         return total << PAGE_SHIFT;
 434 }
 435
 436 #ifdef CONFIG_KEXEC
 437 static void __init reserve_crashkernel(void)
 438 {
 439         unsigned long long total_mem;
 440         unsigned long long crash_size, crash_base;
 441         int ret;
 442
 443         total_mem = get_total_mem();
 444
 445         ret = parse_crashkernel(boot_command_line, total_mem,
 446                         &crash_size, &crash_base);
 447         if (ret == 0 && crash_size > 0) {
 448                 if (crash_base > 0) {
 449                         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 450                                         "for crashkernel (System RAM: %ldMB)\n",
 451                                         (unsigned long)(crash_size >> 20),
 452                                         (unsigned long)(crash_base >> 20),
 453                                         (unsigned long)(total_mem >> 20));
 454
 455                         if (reserve_bootmem(crash_base, crash_size,
 456                                         BOOTMEM_EXCLUSIVE) < 0) {
 457                                 printk(KERN_INFO "crashkernel reservation "
 458                                         "failed - memory is in use\n");
 459                                 return;
 460                         }
 461
 462                         crashk_res.start = crash_base;
 463                         crashk_res.end   = crash_base + crash_size - 1;
 464                 } else
 465                         printk(KERN_INFO "crashkernel reservation failed - "
 466                                         "you have to specify a base address\n");
 467         }
 468 }
 469 #else
 470 static inline void __init reserve_crashkernel(void)
 471 {}
 472 #endif
 473
 474 #ifdef CONFIG_BLK_DEV_INITRD
 475
 476 static bool do_relocate_initrd = false;
 477
 478 static void __init reserve_initrd(void)
 479 {
 480         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 481         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 482         u64 ramdisk_end   = ramdisk_image + ramdisk_size;
 483         u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
 484         u64 ramdisk_here;
 485
 486         if (!boot_params.hdr.type_of_loader ||
 487             !ramdisk_image || !ramdisk_size)
 488                 return;         /* No initrd provided by bootloader */
 489
 490         initrd_start = 0;
 491
 492         if (ramdisk_size >= end_of_lowmem/2) {
 493                 free_early(ramdisk_image, ramdisk_end);
 494                 printk(KERN_ERR "initrd too large to handle, "
 495                        "disabling initrd\n");
 496                 return;
 497         }
 498
 499         printk(KERN_INFO "old RAMDISK: %08llx - %08llx\n", ramdisk_image,
 500                         ramdisk_end);
 501
 502
 503         if (ramdisk_end <= end_of_lowmem) {
 504                 /* All in lowmem, easy case */
 505                 /*
 506                  * don't need to reserve again, already reserved early
 507                  * in i386_start_kernel
 508                  */
 509                 initrd_start = ramdisk_image + PAGE_OFFSET;
 510                 initrd_end = initrd_start+ramdisk_size;
 511                 return;
 512         }
 513
 514         /* We need to move the initrd down into lowmem */
 515         ramdisk_here = find_e820_area(min_low_pfn<<PAGE_SHIFT,
 516                                  end_of_lowmem, ramdisk_size,
 517                                  PAGE_SIZE);
 518
 519         if (ramdisk_here == -1ULL)
 520                 panic("Cannot find place for new RAMDISK of size %lld\n",
 521                          ramdisk_size);
 522
 523         /* Note: this includes all the lowmem currently occupied by
 524            the initrd, we rely on that fact to keep the data intact. */
 525         reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
 526                          "NEW RAMDISK");
 527         initrd_start = ramdisk_here + PAGE_OFFSET;
 528         initrd_end   = initrd_start + ramdisk_size;
 529         printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
 530                          ramdisk_here, ramdisk_here + ramdisk_size);
 531
 532         do_relocate_initrd = true;
 533 }
 534
 535 #define MAX_MAP_CHUNK   (NR_FIX_BTMAPS << PAGE_SHIFT)
 536
 537 static void __init relocate_initrd(void)
 538 {
 539         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 540         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 541         u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
 542         u64 ramdisk_here;
 543         unsigned long slop, clen, mapaddr;
 544         char *p, *q;
 545
 546         if (!do_relocate_initrd)
 547                 return;
 548
 549         ramdisk_here = initrd_start - PAGE_OFFSET;
 550
 551         q = (char *)initrd_start;
 552
 553         /* Copy any lowmem portion of the initrd */
 554         if (ramdisk_image < end_of_lowmem) {
 555                 clen = end_of_lowmem - ramdisk_image;
 556                 p = (char *)__va(ramdisk_image);
 557                 memcpy(q, p, clen);
 558                 q += clen;
 559                 /* need to free these low pages...*/
 560                 printk(KERN_INFO "Freeing old partial RAMDISK %08llx-%08llx\n",
 561                          ramdisk_image, ramdisk_image + clen - 1);
 562                 free_bootmem(ramdisk_image, clen);
 563                 ramdisk_image += clen;
 564                 ramdisk_size  -= clen;
 565         }
 566
 567         /* Copy the highmem portion of the initrd */
 568         while (ramdisk_size) {
 569                 slop = ramdisk_image & ~PAGE_MASK;
 570                 clen = ramdisk_size;
 571                 if (clen > MAX_MAP_CHUNK-slop)
 572                         clen = MAX_MAP_CHUNK-slop;
 573                 mapaddr = ramdisk_image & PAGE_MASK;
 574                 p = early_ioremap(mapaddr, clen+slop);
 575                 memcpy(q, p+slop, clen);
 576                 early_iounmap(p, clen+slop);
 577                 q += clen;
 578                 ramdisk_image += clen;
 579                 ramdisk_size  -= clen;
 580         }
 581         /* high pages is not converted by early_res_to_bootmem */
 582         ramdisk_image = boot_params.hdr.ramdisk_image;
 583         ramdisk_size  = boot_params.hdr.ramdisk_size;
 584         printk(KERN_INFO "Copied RAMDISK from %016llx - %016llx to %08llx - %08llx\n",
 585                 ramdisk_image, ramdisk_image + ramdisk_size - 1,
 586                 ramdisk_here, ramdisk_here + ramdisk_size - 1);
 587 }
 588
 589 #endif /* CONFIG_BLK_DEV_INITRD */
 590
 591 void __init setup_bootmem_allocator(void)
 592 {
 593         int i;
 594         unsigned long bootmap_size, bootmap;
 595         /*
 596          * Initialize the boot-time allocator (with low memory only):
 597          */
 598         bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT;
 599         bootmap = find_e820_area(min_low_pfn<<PAGE_SHIFT,
 600                                  max_pfn_mapped<<PAGE_SHIFT, bootmap_size,
 601                                  PAGE_SIZE);
 602         if (bootmap == -1L)
 603                 panic("Cannot find bootmem map of size %ld\n", bootmap_size);
 604         reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP");
 605 #ifdef CONFIG_BLK_DEV_INITRD
 606         reserve_initrd();
 607 #endif
 608         bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, max_low_pfn);
 609         printk(KERN_INFO "  mapped low ram: 0 - %08lx\n",
 610                  max_pfn_mapped<<PAGE_SHIFT);
 611         printk(KERN_INFO "  low ram: %08lx - %08lx\n",
 612                  min_low_pfn<<PAGE_SHIFT, max_low_pfn<<PAGE_SHIFT);
 613         printk(KERN_INFO "  bootmap %08lx - %08lx\n",
 614                  bootmap, bootmap + bootmap_size);
 615         for_each_online_node(i)
 616                 free_bootmem_with_active_regions(i, max_low_pfn);
 617         early_res_to_bootmem(0, max_low_pfn<<PAGE_SHIFT);
 618
 619 #ifdef CONFIG_ACPI_SLEEP
 620         /*
 621          * Reserve low memory region for sleep support.
 622          */
 623         acpi_reserve_bootmem();
 624 #endif
 625 #ifdef CONFIG_X86_FIND_SMP_CONFIG
 626         /*
 627          * Find and reserve possible boot-time SMP configuration:
 628          */
 629         find_smp_config();
 630 #endif
 631         reserve_crashkernel();
 632
 633         reserve_ibft_region();
 634 }
 635
 636 /*
 637  * The node 0 pgdat is initialized before all of these because
 638  * it's needed for bootmem.  node>0 pgdats have their virtual
 639  * space allocated before the pagetables are in place to access
 640  * them, so they can't be cleared then.
 641  *
 642  * This should all compile down to nothing when NUMA is off.
 643  */
 644 static void __init remapped_pgdat_init(void)
 645 {
 646         int nid;
 647
 648         for_each_online_node(nid) {
 649                 if (nid != 0)
 650                         memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
 651         }
 652 }
 653
 654 #ifdef CONFIG_MCA
 655 static void set_mca_bus(int x)
 656 {
 657         MCA_bus = x;
 658 }
 659 #else
 660 static void set_mca_bus(int x) { }
 661 #endif
 662
 663 #ifdef CONFIG_NUMA
 664 /*
 665  * In the golden day, when everything among i386 and x86_64 will be
 666  * integrated, this will not live here
 667  */
 668 void *x86_cpu_to_node_map_early_ptr;
 669 int x86_cpu_to_node_map_init[NR_CPUS] = {
 670         [0 ... NR_CPUS-1] = NUMA_NO_NODE
 671 };
 672 DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
 673 #endif
 674
 675 /*
 676  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 677  * passed the efi memmap, systab, etc., so we should use these data structures
 678  * for initialization.  Note, the efi init code path is determined by the
 679  * global efi_enabled. This allows the same kernel image to be used on existing
 680  * systems (with a traditional BIOS) as well as on EFI systems.
 681  */
 682 void __init setup_arch(char **cmdline_p)
 683 {
 684         unsigned long max_low_pfn;
 685
 686         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 687         pre_setup_arch_hook();
 688         early_cpu_init();
 689         early_ioremap_init();
 690         reserve_setup_data();
 691
 692 #ifdef CONFIG_EFI
 693         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 694                      "EL32", 4)) {
 695                 efi_enabled = 1;
 696                 efi_reserve_early();
 697         }
 698 #endif
 699
 700         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 701         screen_info = boot_params.screen_info;
 702         edid_info = boot_params.edid_info;
 703         apm_info.bios = boot_params.apm_bios_info;
 704         ist_info = boot_params.ist_info;
 705         saved_video_mode = boot_params.hdr.vid_mode;
 706         if( boot_params.sys_desc_table.length != 0 ) {
 707                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 708                 machine_id = boot_params.sys_desc_table.table[0];
 709                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 710                 BIOS_revision = boot_params.sys_desc_table.table[2];
 711         }
 712         bootloader_type = boot_params.hdr.type_of_loader;
 713
 714 #ifdef CONFIG_BLK_DEV_RAM
 715         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 716         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 717         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 718 #endif
 719         ARCH_SETUP
 720
 721         setup_memory_map();
 722
 723         copy_edd();
 724
 725         if (!boot_params.hdr.root_flags)
 726                 root_mountflags &= ~MS_RDONLY;
 727         init_mm.start_code = (unsigned long) _text;
 728         init_mm.end_code = (unsigned long) _etext;
 729         init_mm.end_data = (unsigned long) _edata;
 730         init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
 731
 732         code_resource.start = virt_to_phys(_text);
 733         code_resource.end = virt_to_phys(_etext)-1;
 734         data_resource.start = virt_to_phys(_etext);
 735         data_resource.end = virt_to_phys(_edata)-1;
 736         bss_resource.start = virt_to_phys(&__bss_start);
 737         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 738
 739         parse_setup_data();
 740
 741         parse_early_param();
 742
 743         finish_e820_parsing();
 744
 745         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 746         *cmdline_p = command_line;
 747
 748         if (efi_enabled)
 749                 efi_init();
 750
 751         e820_register_active_regions(0, 0, -1UL);
 752         /*
 753          * partially used pages are not usable - thus
 754          * we are rounding upwards:
 755          */
 756         max_pfn = e820_end_of_ram();
 757
 758         /* preallocate 4k for mptable mpc */
 759         early_reserve_e820_mpc_new();
 760         /* update e820 for memory not covered by WB MTRRs */
 761         mtrr_bp_init();
 762         if (mtrr_trim_uncached_memory(max_pfn)) {
 763                 remove_all_active_ranges();
 764                 e820_register_active_regions(0, 0, -1UL);
 765                 max_pfn = e820_end_of_ram();
 766         }
 767
 768         max_low_pfn = setup_memory();
 769
 770 #ifdef CONFIG_KVM_CLOCK
 771         kvmclock_init();
 772 #endif
 773
 774 #ifdef CONFIG_VMI
 775         /*
 776          * Must be after max_low_pfn is determined, and before kernel
 777          * pagetables are setup.
 778          */
 779         vmi_init();
 780 #endif
 781         kvm_guest_init();
 782
 783         /*
 784          * NOTE: before this point _nobody_ is allowed to allocate
 785          * any memory using the bootmem allocator.  Although the
 786          * allocator is now initialised only the first 8Mb of the kernel
 787          * virtual address space has been mapped.  All allocations before
 788          * paging_init() has completed must use the alloc_bootmem_low_pages()
 789          * variant (which allocates DMA'able memory) and care must be taken
 790          * not to exceed the 8Mb limit.
 791          */
 792
 793         paging_init();
 794
 795         /*
 796          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 797          */
 798
 799 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 800         if (init_ohci1394_dma_early)
 801                 init_ohci1394_dma_on_all_controllers();
 802 #endif
 803
 804         remapped_pgdat_init();
 805         sparse_init();
 806         zone_sizes_init();
 807
 808         /*
 809          * NOTE: at this point the bootmem allocator is fully available.
 810          */
 811
 812 #ifdef CONFIG_BLK_DEV_INITRD
 813         relocate_initrd();
 814 #endif
 815
 816         paravirt_post_allocator_init();
 817
 818         dmi_scan_machine();
 819
 820         io_delay_init();
 821
 822 #ifdef CONFIG_X86_SMP
 823         /*
 824          * setup to use the early static init tables during kernel startup
 825          * X86_SMP will exclude sub-arches that don't deal well with it.
 826          */
 827         x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init;
 828         x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init;
 829 #ifdef CONFIG_NUMA
 830         x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init;
 831 #endif
 832 #endif
 833
 834 #ifdef CONFIG_X86_GENERICARCH
 835         generic_apic_probe();
 836 #endif
 837
 838 #ifdef CONFIG_ACPI
 839         /*
 840          * Parse the ACPI tables for possible boot-time SMP configuration.
 841          */
 842         acpi_boot_table_init();
 843 #endif
 844
 845         early_quirks();
 846
 847 #ifdef CONFIG_ACPI
 848         acpi_boot_init();
 849 #endif
 850 #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
 851         if (smp_found_config)
 852                 get_smp_config();
 853 #endif
 854 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
 855         if (def_to_bigsmp)
 856                 printk(KERN_WARNING "More than 8 CPUs detected and "
 857                         "CONFIG_X86_PC cannot handle it.\nUse "
 858                         "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
 859 #endif
 860
 861         e820_setup_gap();
 862         e820_mark_nosave_regions(max_low_pfn);
 863
 864 #ifdef CONFIG_VT
 865 #if defined(CONFIG_VGA_CONSOLE)
 866         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
 867                 conswitchp = &vga_con;
 868 #elif defined(CONFIG_DUMMY_CONSOLE)
 869         conswitchp = &dummy_con;
 870 #endif
 871 #endif
 872 }
 873
 874 /*
 875  * Request address space for all standard resources
 876  *
 877  * This is called just before pcibios_init(), which is also a
 878  * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
 879  */
 880 static int __init request_standard_resources(void)
 881 {
 882         int i;
 883
 884         printk(KERN_INFO "Setting up standard PCI resources\n");
 885         init_iomem_resources(&code_resource, &data_resource, &bss_resource);
 886
 887         request_resource(&iomem_resource, &video_ram_resource);
 888
 889         /* request I/O space for devices used on all i[345]86 PCs */
 890         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 891                 request_resource(&ioport_resource, &standard_io_resources[i]);
 892         return 0;
 893 }
 894
 895 subsys_initcall(request_standard_resources);