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
  71 /* This value is set up by the early boot code to point to the value
  72    immediately after the boot time page tables.  It contains a *physical*
  73    address, and must not be in the .bss segment! */
  74 unsigned long init_pg_tables_end __initdata = ~0UL;
  75
  76 /*
  77  * Machine setup..
  78  */
  79 static struct resource data_resource = {
  80         .name   = "Kernel data",
  81         .start  = 0,
  82         .end    = 0,
  83         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  84 };
  85
  86 static struct resource code_resource = {
  87         .name   = "Kernel code",
  88         .start  = 0,
  89         .end    = 0,
  90         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  91 };
  92
  93 static struct resource bss_resource = {
  94         .name   = "Kernel bss",
  95         .start  = 0,
  96         .end    = 0,
  97         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  98 };
  99
 100 static struct resource video_ram_resource = {
 101         .name   = "Video RAM area",
 102         .start  = 0xa0000,
 103         .end    = 0xbffff,
 104         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 105 };
 106
 107 static struct resource standard_io_resources[] = { {
 108         .name   = "dma1",
 109         .start  = 0x0000,
 110         .end    = 0x001f,
 111         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 112 }, {
 113         .name   = "pic1",
 114         .start  = 0x0020,
 115         .end    = 0x0021,
 116         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 117 }, {
 118         .name   = "timer0",
 119         .start  = 0x0040,
 120         .end    = 0x0043,
 121         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 122 }, {
 123         .name   = "timer1",
 124         .start  = 0x0050,
 125         .end    = 0x0053,
 126         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 127 }, {
 128         .name   = "keyboard",
 129         .start  = 0x0060,
 130         .end    = 0x0060,
 131         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 132 }, {
 133         .name   = "keyboard",
 134         .start  = 0x0064,
 135         .end    = 0x0064,
 136         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 137 }, {
 138         .name   = "dma page reg",
 139         .start  = 0x0080,
 140         .end    = 0x008f,
 141         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 142 }, {
 143         .name   = "pic2",
 144         .start  = 0x00a0,
 145         .end    = 0x00a1,
 146         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 147 }, {
 148         .name   = "dma2",
 149         .start  = 0x00c0,
 150         .end    = 0x00df,
 151         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 152 }, {
 153         .name   = "fpu",
 154         .start  = 0x00f0,
 155         .end    = 0x00ff,
 156         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 157 } };
 158
 159 /* cpu data as detected by the assembly code in head.S */
 160 struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
 161 /* common cpu data for all cpus */
 162 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
 163 EXPORT_SYMBOL(boot_cpu_data);
 164
 165 unsigned int def_to_bigsmp;
 166
 167 #ifndef CONFIG_X86_PAE
 168 unsigned long mmu_cr4_features;
 169 #else
 170 unsigned long mmu_cr4_features = X86_CR4_PAE;
 171 #endif
 172
 173 /* for MCA, but anyone else can use it if they want */
 174 unsigned int machine_id;
 175 unsigned int machine_submodel_id;
 176 unsigned int BIOS_revision;
 177
 178 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
 179 int bootloader_type;
 180
 181 /* user-defined highmem size */
 182 static unsigned int highmem_pages = -1;
 183
 184 /*
 185  * Setup options
 186  */
 187 struct screen_info screen_info;
 188 EXPORT_SYMBOL(screen_info);
 189 struct apm_info apm_info;
 190 EXPORT_SYMBOL(apm_info);
 191 struct edid_info edid_info;
 192 EXPORT_SYMBOL_GPL(edid_info);
 193 struct ist_info ist_info;
 194 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 195         defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 196 EXPORT_SYMBOL(ist_info);
 197 #endif
 198
 199 extern void early_cpu_init(void);
 200 extern int root_mountflags;
 201
 202 unsigned long saved_video_mode;
 203
 204 #define RAMDISK_IMAGE_START_MASK        0x07FF
 205 #define RAMDISK_PROMPT_FLAG             0x8000
 206 #define RAMDISK_LOAD_FLAG               0x4000
 207
 208 static char __initdata command_line[COMMAND_LINE_SIZE];
 209
 210 #ifndef CONFIG_DEBUG_BOOT_PARAMS
 211 struct boot_params __initdata boot_params;
 212 #else
 213 struct boot_params boot_params;
 214 #endif
 215
 216 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 217 struct edd edd;
 218 #ifdef CONFIG_EDD_MODULE
 219 EXPORT_SYMBOL(edd);
 220 #endif
 221 /**
 222  * copy_edd() - Copy the BIOS EDD information
 223  *              from boot_params into a safe place.
 224  *
 225  */
 226 static inline void copy_edd(void)
 227 {
 228      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 229             sizeof(edd.mbr_signature));
 230      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 231      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 232      edd.edd_info_nr = boot_params.eddbuf_entries;
 233 }
 234 #else
 235 static inline void copy_edd(void)
 236 {
 237 }
 238 #endif
 239
 240 #ifdef CONFIG_PROC_VMCORE
 241 /* elfcorehdr= specifies the location of elf core header
 242  * stored by the crashed kernel.
 243  */
 244 static int __init parse_elfcorehdr(char *arg)
 245 {
 246         if (!arg)
 247                 return -EINVAL;
 248
 249         elfcorehdr_addr = memparse(arg, &arg);
 250         return 0;
 251 }
 252 early_param("elfcorehdr", parse_elfcorehdr);
 253 #endif /* CONFIG_PROC_VMCORE */
 254
 255 /*
 256  * highmem=size forces highmem to be exactly 'size' bytes.
 257  * This works even on boxes that have no highmem otherwise.
 258  * This also works to reduce highmem size on bigger boxes.
 259  */
 260 static int __init parse_highmem(char *arg)
 261 {
 262         if (!arg)
 263                 return -EINVAL;
 264
 265         highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
 266         return 0;
 267 }
 268 early_param("highmem", parse_highmem);
 269
 270 /*
 271  * vmalloc=size forces the vmalloc area to be exactly 'size'
 272  * bytes. This can be used to increase (or decrease) the
 273  * vmalloc area - the default is 128m.
 274  */
 275 static int __init parse_vmalloc(char *arg)
 276 {
 277         if (!arg)
 278                 return -EINVAL;
 279
 280         __VMALLOC_RESERVE = memparse(arg, &arg);
 281         return 0;
 282 }
 283 early_param("vmalloc", parse_vmalloc);
 284
 285 /*
 286  * reservetop=size reserves a hole at the top of the kernel address space which
 287  * a hypervisor can load into later.  Needed for dynamically loaded hypervisors,
 288  * so relocating the fixmap can be done before paging initialization.
 289  */
 290 static int __init parse_reservetop(char *arg)
 291 {
 292         unsigned long address;
 293
 294         if (!arg)
 295                 return -EINVAL;
 296
 297         address = memparse(arg, &arg);
 298         reserve_top_address(address);
 299         return 0;
 300 }
 301 early_param("reservetop", parse_reservetop);
 302
 303 /*
 304  * Determine low and high memory ranges:
 305  */
 306 unsigned long __init find_max_low_pfn(void)
 307 {
 308         unsigned long max_low_pfn;
 309
 310         max_low_pfn = max_pfn;
 311         if (max_low_pfn > MAXMEM_PFN) {
 312                 if (highmem_pages == -1)
 313                         highmem_pages = max_pfn - MAXMEM_PFN;
 314                 if (highmem_pages + MAXMEM_PFN < max_pfn)
 315                         max_pfn = MAXMEM_PFN + highmem_pages;
 316                 if (highmem_pages + MAXMEM_PFN > max_pfn) {
 317                         printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
 318                         highmem_pages = 0;
 319                 }
 320                 max_low_pfn = MAXMEM_PFN;
 321 #ifndef CONFIG_HIGHMEM
 322                 /* Maximum memory usable is what is directly addressable */
 323                 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
 324                                         MAXMEM>>20);
 325                 if (max_pfn > MAX_NONPAE_PFN)
 326                         printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
 327                 else
 328                         printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
 329                 max_pfn = MAXMEM_PFN;
 330 #else /* !CONFIG_HIGHMEM */
 331 #ifndef CONFIG_HIGHMEM64G
 332                 if (max_pfn > MAX_NONPAE_PFN) {
 333                         max_pfn = MAX_NONPAE_PFN;
 334                         printk(KERN_WARNING "Warning only 4GB will be used.\n");
 335                         printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
 336                 }
 337 #endif /* !CONFIG_HIGHMEM64G */
 338 #endif /* !CONFIG_HIGHMEM */
 339         } else {
 340                 if (highmem_pages == -1)
 341                         highmem_pages = 0;
 342 #ifdef CONFIG_HIGHMEM
 343                 if (highmem_pages >= max_pfn) {
 344                         printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
 345                         highmem_pages = 0;
 346                 }
 347                 if (highmem_pages) {
 348                         if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
 349                                 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
 350                                 highmem_pages = 0;
 351                         }
 352                         max_low_pfn -= highmem_pages;
 353                 }
 354 #else
 355                 if (highmem_pages)
 356                         printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
 357 #endif
 358         }
 359         return max_low_pfn;
 360 }
 361
 362 #ifndef CONFIG_NEED_MULTIPLE_NODES
 363 static void __init setup_bootmem_allocator(void);
 364 static unsigned long __init setup_memory(void)
 365 {
 366         /*
 367          * partially used pages are not usable - thus
 368          * we are rounding upwards:
 369          */
 370         min_low_pfn = PFN_UP(init_pg_tables_end);
 371
 372         max_low_pfn = find_max_low_pfn();
 373
 374 #ifdef CONFIG_HIGHMEM
 375         highstart_pfn = highend_pfn = max_pfn;
 376         if (max_pfn > max_low_pfn) {
 377                 highstart_pfn = max_low_pfn;
 378         }
 379         printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
 380                 pages_to_mb(highend_pfn - highstart_pfn));
 381         num_physpages = highend_pfn;
 382         high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
 383 #else
 384         num_physpages = max_low_pfn;
 385         high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
 386 #endif
 387 #ifdef CONFIG_FLATMEM
 388         max_mapnr = num_physpages;
 389 #endif
 390         printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
 391                         pages_to_mb(max_low_pfn));
 392
 393         setup_bootmem_allocator();
 394
 395         return max_low_pfn;
 396 }
 397
 398 static void __init zone_sizes_init(void)
 399 {
 400         unsigned long max_zone_pfns[MAX_NR_ZONES];
 401         memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
 402         max_zone_pfns[ZONE_DMA] =
 403                 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
 404         max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
 405 #ifdef CONFIG_HIGHMEM
 406         max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
 407         add_active_range(0, 0, highend_pfn);
 408 #else
 409         add_active_range(0, 0, max_low_pfn);
 410 #endif
 411
 412         free_area_init_nodes(max_zone_pfns);
 413 }
 414 #else
 415 extern unsigned long __init setup_memory(void);
 416 extern void zone_sizes_init(void);
 417 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
 418
 419 static inline unsigned long long get_total_mem(void)
 420 {
 421         unsigned long long total;
 422
 423         total = max_low_pfn - min_low_pfn;
 424 #ifdef CONFIG_HIGHMEM
 425         total += highend_pfn - highstart_pfn;
 426 #endif
 427
 428         return total << PAGE_SHIFT;
 429 }
 430
 431 #ifdef CONFIG_KEXEC
 432 static void __init reserve_crashkernel(void)
 433 {
 434         unsigned long long total_mem;
 435         unsigned long long crash_size, crash_base;
 436         int ret;
 437
 438         total_mem = get_total_mem();
 439
 440         ret = parse_crashkernel(boot_command_line, total_mem,
 441                         &crash_size, &crash_base);
 442         if (ret == 0 && crash_size > 0) {
 443                 if (crash_base > 0) {
 444                         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 445                                         "for crashkernel (System RAM: %ldMB)\n",
 446                                         (unsigned long)(crash_size >> 20),
 447                                         (unsigned long)(crash_base >> 20),
 448                                         (unsigned long)(total_mem >> 20));
 449                         crashk_res.start = crash_base;
 450                         crashk_res.end   = crash_base + crash_size - 1;
 451                         reserve_bootmem(crash_base, crash_size,
 452                                         BOOTMEM_DEFAULT);
 453                 } else
 454                         printk(KERN_INFO "crashkernel reservation failed - "
 455                                         "you have to specify a base address\n");
 456         }
 457 }
 458 #else
 459 static inline void __init reserve_crashkernel(void)
 460 {}
 461 #endif
 462
 463 #ifdef CONFIG_BLK_DEV_INITRD
 464
 465 static bool do_relocate_initrd = false;
 466
 467 static void __init reserve_initrd(void)
 468 {
 469         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 470         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 471         u64 ramdisk_end   = ramdisk_image + ramdisk_size;
 472         u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
 473         u64 ramdisk_here;
 474
 475         if (!boot_params.hdr.type_of_loader ||
 476             !ramdisk_image || !ramdisk_size)
 477                 return;         /* No initrd provided by bootloader */
 478
 479         initrd_start = 0;
 480
 481         if (ramdisk_size >= end_of_lowmem/2) {
 482                 free_early(ramdisk_image, ramdisk_image + ramdisk_size - 1);
 483                 printk(KERN_ERR "initrd too large to handle, "
 484                        "disabling initrd\n");
 485                 return;
 486         }
 487
 488         if (ramdisk_end <= end_of_lowmem) {
 489                 /* All in lowmem, easy case */
 490                 /*
 491                  * don't need to reserve again, already reserved early
 492                  * in i386_start_kernel
 493                  */
 494                 initrd_start = ramdisk_image + PAGE_OFFSET;
 495                 initrd_end = initrd_start+ramdisk_size;
 496                 return;
 497         }
 498
 499         /* We need to move the initrd down into lowmem */
 500         ramdisk_here = find_e820_area(min_low_pfn<<PAGE_SHIFT,
 501                                  end_of_lowmem, ramdisk_size,
 502                                  PAGE_SIZE);
 503
 504         /* Note: this includes all the lowmem currently occupied by
 505            the initrd, we rely on that fact to keep the data intact. */
 506         reserve_early(ramdisk_here, ramdisk_here + ramdisk_size - 1,
 507                          "NEW RAMDISK");
 508         initrd_start = ramdisk_here + PAGE_OFFSET;
 509         initrd_end   = initrd_start + ramdisk_size;
 510
 511         do_relocate_initrd = true;
 512 }
 513
 514 #define MAX_MAP_CHUNK   (NR_FIX_BTMAPS << PAGE_SHIFT)
 515
 516 static void __init relocate_initrd(void)
 517 {
 518         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 519         u64 ramdisk_size  = boot_params.hdr.ramdisk_size;
 520         u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
 521         u64 ramdisk_here;
 522         unsigned long slop, clen, mapaddr;
 523         char *p, *q;
 524
 525         if (!do_relocate_initrd)
 526                 return;
 527
 528         ramdisk_here = initrd_start - PAGE_OFFSET;
 529
 530         q = (char *)initrd_start;
 531
 532         /* Copy any lowmem portion of the initrd */
 533         if (ramdisk_image < end_of_lowmem) {
 534                 clen = end_of_lowmem - ramdisk_image;
 535                 p = (char *)__va(ramdisk_image);
 536                 memcpy(q, p, clen);
 537                 q += clen;
 538                 /* need to free these low pages...*/
 539                 printk(KERN_INFO "Freeing old partial RAMDISK %08llx-%08llx\n",
 540                          ramdisk_image, ramdisk_image + clen - 1);
 541                 free_bootmem(ramdisk_image, clen);
 542                 ramdisk_image += clen;
 543                 ramdisk_size  -= clen;
 544         }
 545
 546         /* Copy the highmem portion of the initrd */
 547         while (ramdisk_size) {
 548                 slop = ramdisk_image & ~PAGE_MASK;
 549                 clen = ramdisk_size;
 550                 if (clen > MAX_MAP_CHUNK-slop)
 551                         clen = MAX_MAP_CHUNK-slop;
 552                 mapaddr = ramdisk_image & PAGE_MASK;
 553                 p = early_ioremap(mapaddr, clen+slop);
 554                 memcpy(q, p+slop, clen);
 555                 early_iounmap(p, clen+slop);
 556                 q += clen;
 557                 ramdisk_image += clen;
 558                 ramdisk_size  -= clen;
 559         }
 560         /* high pages is not converted by early_res_to_bootmem */
 561         ramdisk_image = boot_params.hdr.ramdisk_image;
 562         ramdisk_size  = boot_params.hdr.ramdisk_size;
 563         printk(KERN_INFO "Copied RAMDISK from %016llx - %016llx to %08llx - %08llx\n",
 564                 ramdisk_image, ramdisk_image + ramdisk_size - 1,
 565                 ramdisk_here, ramdisk_here + ramdisk_size - 1);
 566 }
 567
 568 #endif /* CONFIG_BLK_DEV_INITRD */
 569
 570 void __init setup_bootmem_allocator(void)
 571 {
 572         unsigned long bootmap_size, bootmap;
 573         /*
 574          * Initialize the boot-time allocator (with low memory only):
 575          */
 576         bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT;
 577         bootmap = find_e820_area(min_low_pfn<<PAGE_SHIFT,
 578                                  max_low_pfn<<PAGE_SHIFT, bootmap_size,
 579                                  PAGE_SIZE);
 580         if (bootmap == -1L)
 581                 panic("Cannot find bootmem map of size %ld\n", bootmap_size);
 582         reserve_early(bootmap, bootmap + bootmap_size - 1, "BOOTMAP");
 583 #ifdef CONFIG_BLK_DEV_INITRD
 584         reserve_initrd();
 585 #endif
 586         bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, max_low_pfn);
 587         printk(KERN_INFO "  low ram: %08lx - %08lx\n",
 588                  min_low_pfn<<PAGE_SHIFT, max_low_pfn<<PAGE_SHIFT);
 589         printk(KERN_INFO "  bootmap [%08lx -  %08lx]\n",
 590                  bootmap, bootmap + bootmap_size - 1);
 591         register_bootmem_low_pages(max_low_pfn);
 592         early_res_to_bootmem(0, max_low_pfn<<PAGE_SHIFT);
 593
 594 #ifdef CONFIG_ACPI_SLEEP
 595         /*
 596          * Reserve low memory region for sleep support.
 597          */
 598         acpi_reserve_bootmem();
 599 #endif
 600 #ifdef CONFIG_X86_FIND_SMP_CONFIG
 601         /*
 602          * Find and reserve possible boot-time SMP configuration:
 603          */
 604         find_smp_config();
 605 #endif
 606         numa_kva_reserve();
 607         reserve_crashkernel();
 608
 609         reserve_ibft_region();
 610 }
 611
 612 /*
 613  * The node 0 pgdat is initialized before all of these because
 614  * it's needed for bootmem.  node>0 pgdats have their virtual
 615  * space allocated before the pagetables are in place to access
 616  * them, so they can't be cleared then.
 617  *
 618  * This should all compile down to nothing when NUMA is off.
 619  */
 620 static void __init remapped_pgdat_init(void)
 621 {
 622         int nid;
 623
 624         for_each_online_node(nid) {
 625                 if (nid != 0)
 626                         memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
 627         }
 628 }
 629
 630 #ifdef CONFIG_MCA
 631 static void set_mca_bus(int x)
 632 {
 633         MCA_bus = x;
 634 }
 635 #else
 636 static void set_mca_bus(int x) { }
 637 #endif
 638
 639 #ifdef CONFIG_NUMA
 640 /*
 641  * In the golden day, when everything among i386 and x86_64 will be
 642  * integrated, this will not live here
 643  */
 644 void *x86_cpu_to_node_map_early_ptr;
 645 int x86_cpu_to_node_map_init[NR_CPUS] = {
 646         [0 ... NR_CPUS-1] = NUMA_NO_NODE
 647 };
 648 DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
 649 #endif
 650
 651 /*
 652  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 653  * passed the efi memmap, systab, etc., so we should use these data structures
 654  * for initialization.  Note, the efi init code path is determined by the
 655  * global efi_enabled. This allows the same kernel image to be used on existing
 656  * systems (with a traditional BIOS) as well as on EFI systems.
 657  */
 658 void __init setup_arch(char **cmdline_p)
 659 {
 660         unsigned long max_low_pfn;
 661
 662         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 663         pre_setup_arch_hook();
 664         early_cpu_init();
 665         early_ioremap_init();
 666
 667 #ifdef CONFIG_EFI
 668         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 669                      "EL32", 4))
 670                 efi_enabled = 1;
 671 #endif
 672
 673         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 674         screen_info = boot_params.screen_info;
 675         edid_info = boot_params.edid_info;
 676         apm_info.bios = boot_params.apm_bios_info;
 677         ist_info = boot_params.ist_info;
 678         saved_video_mode = boot_params.hdr.vid_mode;
 679         if( boot_params.sys_desc_table.length != 0 ) {
 680                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 681                 machine_id = boot_params.sys_desc_table.table[0];
 682                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 683                 BIOS_revision = boot_params.sys_desc_table.table[2];
 684         }
 685         bootloader_type = boot_params.hdr.type_of_loader;
 686
 687 #ifdef CONFIG_BLK_DEV_RAM
 688         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 689         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 690         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 691 #endif
 692         ARCH_SETUP
 693
 694         setup_memory_map();
 695
 696         copy_edd();
 697
 698         if (!boot_params.hdr.root_flags)
 699                 root_mountflags &= ~MS_RDONLY;
 700         init_mm.start_code = (unsigned long) _text;
 701         init_mm.end_code = (unsigned long) _etext;
 702         init_mm.end_data = (unsigned long) _edata;
 703         init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
 704
 705         code_resource.start = virt_to_phys(_text);
 706         code_resource.end = virt_to_phys(_etext)-1;
 707         data_resource.start = virt_to_phys(_etext);
 708         data_resource.end = virt_to_phys(_edata)-1;
 709         bss_resource.start = virt_to_phys(&__bss_start);
 710         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 711
 712         parse_early_param();
 713
 714         finish_e820_parsing();
 715
 716         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 717         *cmdline_p = command_line;
 718
 719         if (efi_enabled)
 720                 efi_init();
 721
 722         /* update e820 for memory not covered by WB MTRRs */
 723         propagate_e820_map();
 724         mtrr_bp_init();
 725         if (mtrr_trim_uncached_memory(max_pfn))
 726                 propagate_e820_map();
 727
 728         max_low_pfn = setup_memory();
 729
 730 #ifdef CONFIG_KVM_CLOCK
 731         kvmclock_init();
 732 #endif
 733
 734 #ifdef CONFIG_VMI
 735         /*
 736          * Must be after max_low_pfn is determined, and before kernel
 737          * pagetables are setup.
 738          */
 739         vmi_init();
 740 #endif
 741         kvm_guest_init();
 742
 743         /*
 744          * NOTE: before this point _nobody_ is allowed to allocate
 745          * any memory using the bootmem allocator.  Although the
 746          * allocator is now initialised only the first 8Mb of the kernel
 747          * virtual address space has been mapped.  All allocations before
 748          * paging_init() has completed must use the alloc_bootmem_low_pages()
 749          * variant (which allocates DMA'able memory) and care must be taken
 750          * not to exceed the 8Mb limit.
 751          */
 752
 753         paging_init();
 754
 755         /*
 756          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 757          */
 758
 759 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 760         if (init_ohci1394_dma_early)
 761                 init_ohci1394_dma_on_all_controllers();
 762 #endif
 763
 764         remapped_pgdat_init();
 765         sparse_init();
 766         zone_sizes_init();
 767
 768         /*
 769          * NOTE: at this point the bootmem allocator is fully available.
 770          */
 771
 772 #ifdef CONFIG_BLK_DEV_INITRD
 773         relocate_initrd();
 774 #endif
 775
 776         paravirt_post_allocator_init();
 777
 778         dmi_scan_machine();
 779
 780         io_delay_init();
 781
 782 #ifdef CONFIG_X86_SMP
 783         /*
 784          * setup to use the early static init tables during kernel startup
 785          * X86_SMP will exclude sub-arches that don't deal well with it.
 786          */
 787         x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init;
 788         x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init;
 789 #ifdef CONFIG_NUMA
 790         x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init;
 791 #endif
 792 #endif
 793
 794 #ifdef CONFIG_X86_GENERICARCH
 795         generic_apic_probe();
 796 #endif
 797
 798 #ifdef CONFIG_ACPI
 799         /*
 800          * Parse the ACPI tables for possible boot-time SMP configuration.
 801          */
 802         acpi_boot_table_init();
 803 #endif
 804
 805         early_quirks();
 806
 807 #ifdef CONFIG_ACPI
 808         acpi_boot_init();
 809
 810 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
 811         if (def_to_bigsmp)
 812                 printk(KERN_WARNING "More than 8 CPUs detected and "
 813                         "CONFIG_X86_PC cannot handle it.\nUse "
 814                         "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
 815 #endif
 816 #endif
 817 #ifdef CONFIG_X86_LOCAL_APIC
 818         if (smp_found_config)
 819                 get_smp_config();
 820 #endif
 821
 822         e820_setup_gap();
 823         e820_mark_nosave_regions(max_low_pfn);
 824
 825 #ifdef CONFIG_VT
 826 #if defined(CONFIG_VGA_CONSOLE)
 827         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
 828                 conswitchp = &vga_con;
 829 #elif defined(CONFIG_DUMMY_CONSOLE)
 830         conswitchp = &dummy_con;
 831 #endif
 832 #endif
 833 }
 834
 835 /*
 836  * Request address space for all standard resources
 837  *
 838  * This is called just before pcibios_init(), which is also a
 839  * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
 840  */
 841 static int __init request_standard_resources(void)
 842 {
 843         int i;
 844
 845         printk(KERN_INFO "Setting up standard PCI resources\n");
 846         init_iomem_resources(&code_resource, &data_resource, &bss_resource);
 847
 848         request_resource(&iomem_resource, &video_ram_resource);
 849
 850         /* request I/O space for devices used on all i[345]86 PCs */
 851         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 852                 request_resource(&ioport_resource, &standard_io_resources[i]);
 853         return 0;
 854 }
 855
 856 subsys_initcall(request_standard_resources);