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/nodemask.h>
  43 #include <linux/kexec.h>
  44 #include <linux/crash_dump.h>
  45 #include <linux/dmi.h>
  46 #include <linux/pfn.h>
  47 #include <linux/pci.h>
  48 #include <linux/init_ohci1394_dma.h>
  49
  50 #include <video/edid.h>
  51
  52 #include <asm/mtrr.h>
  53 #include <asm/apic.h>
  54 #include <asm/e820.h>
  55 #include <asm/mpspec.h>
  56 #include <asm/mmzone.h>
  57 #include <asm/setup.h>
  58 #include <asm/arch_hooks.h>
  59 #include <asm/sections.h>
  60 #include <asm/io_apic.h>
  61 #include <asm/ist.h>
  62 #include <asm/io.h>
  63 #include <asm/vmi.h>
  64 #include <setup_arch.h>
  65 #include <bios_ebda.h>
  66 #include <asm/cacheflush.h>
  67
  68 /* This value is set up by the early boot code to point to the value
  69    immediately after the boot time page tables.  It contains a *physical*
  70    address, and must not be in the .bss segment! */
  71 unsigned long init_pg_tables_end __initdata = ~0UL;
  72
  73 /*
  74  * Machine setup..
  75  */
  76 static struct resource data_resource = {
  77         .name   = "Kernel data",
  78         .start  = 0,
  79         .end    = 0,
  80         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  81 };
  82
  83 static struct resource code_resource = {
  84         .name   = "Kernel code",
  85         .start  = 0,
  86         .end    = 0,
  87         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  88 };
  89
  90 static struct resource bss_resource = {
  91         .name   = "Kernel bss",
  92         .start  = 0,
  93         .end    = 0,
  94         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
  95 };
  96
  97 static struct resource video_ram_resource = {
  98         .name   = "Video RAM area",
  99         .start  = 0xa0000,
 100         .end    = 0xbffff,
 101         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 102 };
 103
 104 static struct resource standard_io_resources[] = { {
 105         .name   = "dma1",
 106         .start  = 0x0000,
 107         .end    = 0x001f,
 108         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 109 }, {
 110         .name   = "pic1",
 111         .start  = 0x0020,
 112         .end    = 0x0021,
 113         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 114 }, {
 115         .name   = "timer0",
 116         .start  = 0x0040,
 117         .end    = 0x0043,
 118         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 119 }, {
 120         .name   = "timer1",
 121         .start  = 0x0050,
 122         .end    = 0x0053,
 123         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 124 }, {
 125         .name   = "keyboard",
 126         .start  = 0x0060,
 127         .end    = 0x006f,
 128         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 129 }, {
 130         .name   = "dma page reg",
 131         .start  = 0x0080,
 132         .end    = 0x008f,
 133         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 134 }, {
 135         .name   = "pic2",
 136         .start  = 0x00a0,
 137         .end    = 0x00a1,
 138         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 139 }, {
 140         .name   = "dma2",
 141         .start  = 0x00c0,
 142         .end    = 0x00df,
 143         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 144 }, {
 145         .name   = "fpu",
 146         .start  = 0x00f0,
 147         .end    = 0x00ff,
 148         .flags  = IORESOURCE_BUSY | IORESOURCE_IO
 149 } };
 150
 151 /* cpu data as detected by the assembly code in head.S */
 152 struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
 153 /* common cpu data for all cpus */
 154 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
 155 EXPORT_SYMBOL(boot_cpu_data);
 156
 157 #ifndef CONFIG_X86_PAE
 158 unsigned long mmu_cr4_features;
 159 #else
 160 unsigned long mmu_cr4_features = X86_CR4_PAE;
 161 #endif
 162
 163 /* for MCA, but anyone else can use it if they want */
 164 unsigned int machine_id;
 165 unsigned int machine_submodel_id;
 166 unsigned int BIOS_revision;
 167 unsigned int mca_pentium_flag;
 168
 169 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
 170 int bootloader_type;
 171
 172 /* user-defined highmem size */
 173 static unsigned int highmem_pages = -1;
 174
 175 /*
 176  * Setup options
 177  */
 178 struct screen_info screen_info;
 179 EXPORT_SYMBOL(screen_info);
 180 struct apm_info apm_info;
 181 EXPORT_SYMBOL(apm_info);
 182 struct edid_info edid_info;
 183 EXPORT_SYMBOL_GPL(edid_info);
 184 struct ist_info ist_info;
 185 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 186         defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 187 EXPORT_SYMBOL(ist_info);
 188 #endif
 189
 190 extern void early_cpu_init(void);
 191 extern int root_mountflags;
 192
 193 unsigned long saved_videomode;
 194
 195 #define RAMDISK_IMAGE_START_MASK        0x07FF
 196 #define RAMDISK_PROMPT_FLAG             0x8000
 197 #define RAMDISK_LOAD_FLAG               0x4000
 198
 199 static char __initdata command_line[COMMAND_LINE_SIZE];
 200
 201 #ifndef CONFIG_DEBUG_BOOT_PARAMS
 202 struct boot_params __initdata boot_params;
 203 #else
 204 struct boot_params boot_params;
 205 #endif
 206
 207 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 208 struct edd edd;
 209 #ifdef CONFIG_EDD_MODULE
 210 EXPORT_SYMBOL(edd);
 211 #endif
 212 /**
 213  * copy_edd() - Copy the BIOS EDD information
 214  *              from boot_params into a safe place.
 215  *
 216  */
 217 static inline void copy_edd(void)
 218 {
 219      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 220             sizeof(edd.mbr_signature));
 221      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 222      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 223      edd.edd_info_nr = boot_params.eddbuf_entries;
 224 }
 225 #else
 226 static inline void copy_edd(void)
 227 {
 228 }
 229 #endif
 230
 231 int __initdata user_defined_memmap = 0;
 232
 233 /*
 234  * "mem=nopentium" disables the 4MB page tables.
 235  * "mem=XXX[kKmM]" defines a memory region from HIGH_MEM
 236  * to <mem>, overriding the bios size.
 237  * "memmap=XXX[KkmM]@XXX[KkmM]" defines a memory region from
 238  * <start> to <start>+<mem>, overriding the bios size.
 239  *
 240  * HPA tells me bootloaders need to parse mem=, so no new
 241  * option should be mem=  [also see Documentation/i386/boot.txt]
 242  */
 243 static int __init parse_mem(char *arg)
 244 {
 245         if (!arg)
 246                 return -EINVAL;
 247
 248         if (strcmp(arg, "nopentium") == 0) {
 249                 setup_clear_cpu_cap(X86_FEATURE_PSE);
 250         } else {
 251                 /* If the user specifies memory size, we
 252                  * limit the BIOS-provided memory map to
 253                  * that size. exactmap can be used to specify
 254                  * the exact map. mem=number can be used to
 255                  * trim the existing memory map.
 256                  */
 257                 unsigned long long mem_size;
 258
 259                 mem_size = memparse(arg, &arg);
 260                 limit_regions(mem_size);
 261                 user_defined_memmap = 1;
 262         }
 263         return 0;
 264 }
 265 early_param("mem", parse_mem);
 266
 267 #ifdef CONFIG_PROC_VMCORE
 268 /* elfcorehdr= specifies the location of elf core header
 269  * stored by the crashed kernel.
 270  */
 271 static int __init parse_elfcorehdr(char *arg)
 272 {
 273         if (!arg)
 274                 return -EINVAL;
 275
 276         elfcorehdr_addr = memparse(arg, &arg);
 277         return 0;
 278 }
 279 early_param("elfcorehdr", parse_elfcorehdr);
 280 #endif /* CONFIG_PROC_VMCORE */
 281
 282 /*
 283  * highmem=size forces highmem to be exactly 'size' bytes.
 284  * This works even on boxes that have no highmem otherwise.
 285  * This also works to reduce highmem size on bigger boxes.
 286  */
 287 static int __init parse_highmem(char *arg)
 288 {
 289         if (!arg)
 290                 return -EINVAL;
 291
 292         highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
 293         return 0;
 294 }
 295 early_param("highmem", parse_highmem);
 296
 297 /*
 298  * vmalloc=size forces the vmalloc area to be exactly 'size'
 299  * bytes. This can be used to increase (or decrease) the
 300  * vmalloc area - the default is 128m.
 301  */
 302 static int __init parse_vmalloc(char *arg)
 303 {
 304         if (!arg)
 305                 return -EINVAL;
 306
 307         __VMALLOC_RESERVE = memparse(arg, &arg);
 308         return 0;
 309 }
 310 early_param("vmalloc", parse_vmalloc);
 311
 312 /*
 313  * reservetop=size reserves a hole at the top of the kernel address space which
 314  * a hypervisor can load into later.  Needed for dynamically loaded hypervisors,
 315  * so relocating the fixmap can be done before paging initialization.
 316  */
 317 static int __init parse_reservetop(char *arg)
 318 {
 319         unsigned long address;
 320
 321         if (!arg)
 322                 return -EINVAL;
 323
 324         address = memparse(arg, &arg);
 325         reserve_top_address(address);
 326         return 0;
 327 }
 328 early_param("reservetop", parse_reservetop);
 329
 330 /*
 331  * Determine low and high memory ranges:
 332  */
 333 unsigned long __init find_max_low_pfn(void)
 334 {
 335         unsigned long max_low_pfn;
 336
 337         max_low_pfn = max_pfn;
 338         if (max_low_pfn > MAXMEM_PFN) {
 339                 if (highmem_pages == -1)
 340                         highmem_pages = max_pfn - MAXMEM_PFN;
 341                 if (highmem_pages + MAXMEM_PFN < max_pfn)
 342                         max_pfn = MAXMEM_PFN + highmem_pages;
 343                 if (highmem_pages + MAXMEM_PFN > max_pfn) {
 344                         printk("only %luMB highmem pages available, ignoring highmem size of %uMB.\n", pages_to_mb(max_pfn - MAXMEM_PFN), pages_to_mb(highmem_pages));
 345                         highmem_pages = 0;
 346                 }
 347                 max_low_pfn = MAXMEM_PFN;
 348 #ifndef CONFIG_HIGHMEM
 349                 /* Maximum memory usable is what is directly addressable */
 350                 printk(KERN_WARNING "Warning only %ldMB will be used.\n",
 351                                         MAXMEM>>20);
 352                 if (max_pfn > MAX_NONPAE_PFN)
 353                         printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
 354                 else
 355                         printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
 356                 max_pfn = MAXMEM_PFN;
 357 #else /* !CONFIG_HIGHMEM */
 358 #ifndef CONFIG_HIGHMEM64G
 359                 if (max_pfn > MAX_NONPAE_PFN) {
 360                         max_pfn = MAX_NONPAE_PFN;
 361                         printk(KERN_WARNING "Warning only 4GB will be used.\n");
 362                         printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
 363                 }
 364 #endif /* !CONFIG_HIGHMEM64G */
 365 #endif /* !CONFIG_HIGHMEM */
 366         } else {
 367                 if (highmem_pages == -1)
 368                         highmem_pages = 0;
 369 #ifdef CONFIG_HIGHMEM
 370                 if (highmem_pages >= max_pfn) {
 371                         printk(KERN_ERR "highmem size specified (%uMB) is bigger than pages available (%luMB)!.\n", pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
 372                         highmem_pages = 0;
 373                 }
 374                 if (highmem_pages) {
 375                         if (max_low_pfn-highmem_pages < 64*1024*1024/PAGE_SIZE){
 376                                 printk(KERN_ERR "highmem size %uMB results in smaller than 64MB lowmem, ignoring it.\n", pages_to_mb(highmem_pages));
 377                                 highmem_pages = 0;
 378                         }
 379                         max_low_pfn -= highmem_pages;
 380                 }
 381 #else
 382                 if (highmem_pages)
 383                         printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
 384 #endif
 385         }
 386         return max_low_pfn;
 387 }
 388
 389 /*
 390  * workaround for Dell systems that neglect to reserve EBDA
 391  */
 392 static void __init reserve_ebda_region(void)
 393 {
 394         unsigned int addr;
 395         addr = get_bios_ebda();
 396         if (addr)
 397                 reserve_bootmem(addr, PAGE_SIZE, BOOTMEM_DEFAULT);
 398 }
 399
 400 #ifndef CONFIG_NEED_MULTIPLE_NODES
 401 void __init setup_bootmem_allocator(void);
 402 static unsigned long __init setup_memory(void)
 403 {
 404         /*
 405          * partially used pages are not usable - thus
 406          * we are rounding upwards:
 407          */
 408         min_low_pfn = PFN_UP(init_pg_tables_end);
 409
 410         find_max_pfn();
 411
 412         max_low_pfn = find_max_low_pfn();
 413
 414 #ifdef CONFIG_HIGHMEM
 415         highstart_pfn = highend_pfn = max_pfn;
 416         if (max_pfn > max_low_pfn) {
 417                 highstart_pfn = max_low_pfn;
 418         }
 419         printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
 420                 pages_to_mb(highend_pfn - highstart_pfn));
 421         num_physpages = highend_pfn;
 422         high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
 423 #else
 424         num_physpages = max_low_pfn;
 425         high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
 426 #endif
 427 #ifdef CONFIG_FLATMEM
 428         max_mapnr = num_physpages;
 429 #endif
 430         printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
 431                         pages_to_mb(max_low_pfn));
 432
 433         setup_bootmem_allocator();
 434
 435         return max_low_pfn;
 436 }
 437
 438 void __init zone_sizes_init(void)
 439 {
 440         unsigned long max_zone_pfns[MAX_NR_ZONES];
 441         memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
 442         max_zone_pfns[ZONE_DMA] =
 443                 virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
 444         max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
 445 #ifdef CONFIG_HIGHMEM
 446         max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
 447         add_active_range(0, 0, highend_pfn);
 448 #else
 449         add_active_range(0, 0, max_low_pfn);
 450 #endif
 451
 452         free_area_init_nodes(max_zone_pfns);
 453 }
 454 #else
 455 extern unsigned long __init setup_memory(void);
 456 extern void zone_sizes_init(void);
 457 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
 458
 459 static inline unsigned long long get_total_mem(void)
 460 {
 461         unsigned long long total;
 462
 463         total = max_low_pfn - min_low_pfn;
 464 #ifdef CONFIG_HIGHMEM
 465         total += highend_pfn - highstart_pfn;
 466 #endif
 467
 468         return total << PAGE_SHIFT;
 469 }
 470
 471 #ifdef CONFIG_KEXEC
 472 static void __init reserve_crashkernel(void)
 473 {
 474         unsigned long long total_mem;
 475         unsigned long long crash_size, crash_base;
 476         int ret;
 477
 478         total_mem = get_total_mem();
 479
 480         ret = parse_crashkernel(boot_command_line, total_mem,
 481                         &crash_size, &crash_base);
 482         if (ret == 0 && crash_size > 0) {
 483                 if (crash_base > 0) {
 484                         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 485                                         "for crashkernel (System RAM: %ldMB)\n",
 486                                         (unsigned long)(crash_size >> 20),
 487                                         (unsigned long)(crash_base >> 20),
 488                                         (unsigned long)(total_mem >> 20));
 489                         crashk_res.start = crash_base;
 490                         crashk_res.end   = crash_base + crash_size - 1;
 491                         reserve_bootmem(crash_base, crash_size,
 492                                         BOOTMEM_DEFAULT);
 493                 } else
 494                         printk(KERN_INFO "crashkernel reservation failed - "
 495                                         "you have to specify a base address\n");
 496         }
 497 }
 498 #else
 499 static inline void __init reserve_crashkernel(void)
 500 {}
 501 #endif
 502
 503 #ifdef CONFIG_BLK_DEV_INITRD
 504
 505 static bool do_relocate_initrd = false;
 506
 507 static void __init reserve_initrd(void)
 508 {
 509         unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
 510         unsigned long ramdisk_size  = boot_params.hdr.ramdisk_size;
 511         unsigned long ramdisk_end   = ramdisk_image + ramdisk_size;
 512         unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT;
 513         unsigned long ramdisk_here;
 514
 515         initrd_start = 0;
 516
 517         if (!boot_params.hdr.type_of_loader ||
 518             !ramdisk_image || !ramdisk_size)
 519                 return;         /* No initrd provided by bootloader */
 520
 521         if (ramdisk_end < ramdisk_image) {
 522                 printk(KERN_ERR "initrd wraps around end of memory, "
 523                        "disabling initrd\n");
 524                 return;
 525         }
 526         if (ramdisk_size >= end_of_lowmem/2) {
 527                 printk(KERN_ERR "initrd too large to handle, "
 528                        "disabling initrd\n");
 529                 return;
 530         }
 531         if (ramdisk_end <= end_of_lowmem) {
 532                 /* All in lowmem, easy case */
 533                 reserve_bootmem(ramdisk_image, ramdisk_size, BOOTMEM_DEFAULT);
 534                 initrd_start = ramdisk_image + PAGE_OFFSET;
 535                 initrd_end = initrd_start+ramdisk_size;
 536                 return;
 537         }
 538
 539         /* We need to move the initrd down into lowmem */
 540         ramdisk_here = (end_of_lowmem - ramdisk_size) & PAGE_MASK;
 541
 542         /* Note: this includes all the lowmem currently occupied by
 543            the initrd, we rely on that fact to keep the data intact. */
 544         reserve_bootmem(ramdisk_here, ramdisk_size, BOOTMEM_DEFAULT);
 545         initrd_start = ramdisk_here + PAGE_OFFSET;
 546         initrd_end   = initrd_start + ramdisk_size;
 547
 548         do_relocate_initrd = true;
 549 }
 550
 551 #define MAX_MAP_CHUNK   (NR_FIX_BTMAPS << PAGE_SHIFT)
 552
 553 static void __init relocate_initrd(void)
 554 {
 555         unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
 556         unsigned long ramdisk_size  = boot_params.hdr.ramdisk_size;
 557         unsigned long end_of_lowmem = max_low_pfn << PAGE_SHIFT;
 558         unsigned long ramdisk_here;
 559         unsigned long slop, clen, mapaddr;
 560         char *p, *q;
 561
 562         if (!do_relocate_initrd)
 563                 return;
 564
 565         ramdisk_here = initrd_start - PAGE_OFFSET;
 566
 567         q = (char *)initrd_start;
 568
 569         /* Copy any lowmem portion of the initrd */
 570         if (ramdisk_image < end_of_lowmem) {
 571                 clen = end_of_lowmem - ramdisk_image;
 572                 p = (char *)__va(ramdisk_image);
 573                 memcpy(q, p, clen);
 574                 q += clen;
 575                 ramdisk_image += clen;
 576                 ramdisk_size  -= clen;
 577         }
 578
 579         /* Copy the highmem portion of the initrd */
 580         while (ramdisk_size) {
 581                 slop = ramdisk_image & ~PAGE_MASK;
 582                 clen = ramdisk_size;
 583                 if (clen > MAX_MAP_CHUNK-slop)
 584                         clen = MAX_MAP_CHUNK-slop;
 585                 mapaddr = ramdisk_image & PAGE_MASK;
 586                 p = early_ioremap(mapaddr, clen+slop);
 587                 memcpy(q, p+slop, clen);
 588                 early_iounmap(p, clen+slop);
 589                 q += clen;
 590                 ramdisk_image += clen;
 591                 ramdisk_size  -= clen;
 592         }
 593 }
 594
 595 #endif /* CONFIG_BLK_DEV_INITRD */
 596
 597 void __init setup_bootmem_allocator(void)
 598 {
 599         unsigned long bootmap_size;
 600         /*
 601          * Initialize the boot-time allocator (with low memory only):
 602          */
 603         bootmap_size = init_bootmem(min_low_pfn, max_low_pfn);
 604
 605         register_bootmem_low_pages(max_low_pfn);
 606
 607         /*
 608          * Reserve the bootmem bitmap itself as well. We do this in two
 609          * steps (first step was init_bootmem()) because this catches
 610          * the (very unlikely) case of us accidentally initializing the
 611          * bootmem allocator with an invalid RAM area.
 612          */
 613         reserve_bootmem(__pa_symbol(_text), (PFN_PHYS(min_low_pfn) +
 614                          bootmap_size + PAGE_SIZE-1) - __pa_symbol(_text),
 615                          BOOTMEM_DEFAULT);
 616
 617         /*
 618          * reserve physical page 0 - it's a special BIOS page on many boxes,
 619          * enabling clean reboots, SMP operation, laptop functions.
 620          */
 621         reserve_bootmem(0, PAGE_SIZE, BOOTMEM_DEFAULT);
 622
 623         /* reserve EBDA region, it's a 4K region */
 624         reserve_ebda_region();
 625
 626     /* could be an AMD 768MPX chipset. Reserve a page  before VGA to prevent
 627        PCI prefetch into it (errata #56). Usually the page is reserved anyways,
 628        unless you have no PS/2 mouse plugged in. */
 629         if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
 630             boot_cpu_data.x86 == 6)
 631              reserve_bootmem(0xa0000 - 4096, 4096, BOOTMEM_DEFAULT);
 632
 633 #ifdef CONFIG_SMP
 634         /*
 635          * But first pinch a few for the stack/trampoline stuff
 636          * FIXME: Don't need the extra page at 4K, but need to fix
 637          * trampoline before removing it. (see the GDT stuff)
 638          */
 639         reserve_bootmem(PAGE_SIZE, PAGE_SIZE, BOOTMEM_DEFAULT);
 640 #endif
 641 #ifdef CONFIG_ACPI_SLEEP
 642         /*
 643          * Reserve low memory region for sleep support.
 644          */
 645         acpi_reserve_bootmem();
 646 #endif
 647 #ifdef CONFIG_X86_FIND_SMP_CONFIG
 648         /*
 649          * Find and reserve possible boot-time SMP configuration:
 650          */
 651         find_smp_config();
 652 #endif
 653 #ifdef CONFIG_BLK_DEV_INITRD
 654         reserve_initrd();
 655 #endif
 656         numa_kva_reserve();
 657         reserve_crashkernel();
 658 }
 659
 660 /*
 661  * The node 0 pgdat is initialized before all of these because
 662  * it's needed for bootmem.  node>0 pgdats have their virtual
 663  * space allocated before the pagetables are in place to access
 664  * them, so they can't be cleared then.
 665  *
 666  * This should all compile down to nothing when NUMA is off.
 667  */
 668 static void __init remapped_pgdat_init(void)
 669 {
 670         int nid;
 671
 672         for_each_online_node(nid) {
 673                 if (nid != 0)
 674                         memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
 675         }
 676 }
 677
 678 #ifdef CONFIG_MCA
 679 static void set_mca_bus(int x)
 680 {
 681         MCA_bus = x;
 682 }
 683 #else
 684 static void set_mca_bus(int x) { }
 685 #endif
 686
 687 /* Overridden in paravirt.c if CONFIG_PARAVIRT */
 688 char * __init __attribute__((weak)) memory_setup(void)
 689 {
 690         return machine_specific_memory_setup();
 691 }
 692
 693 /*
 694  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 695  * passed the efi memmap, systab, etc., so we should use these data structures
 696  * for initialization.  Note, the efi init code path is determined by the
 697  * global efi_enabled. This allows the same kernel image to be used on existing
 698  * systems (with a traditional BIOS) as well as on EFI systems.
 699  */
 700 void __init setup_arch(char **cmdline_p)
 701 {
 702         unsigned long max_low_pfn;
 703
 704         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 705         pre_setup_arch_hook();
 706         early_cpu_init();
 707         early_ioremap_init();
 708
 709 #ifdef CONFIG_EFI
 710         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 711                      "EL32", 4))
 712                 efi_enabled = 1;
 713 #endif
 714
 715         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 716         screen_info = boot_params.screen_info;
 717         edid_info = boot_params.edid_info;
 718         apm_info.bios = boot_params.apm_bios_info;
 719         ist_info = boot_params.ist_info;
 720         saved_videomode = boot_params.hdr.vid_mode;
 721         if( boot_params.sys_desc_table.length != 0 ) {
 722                 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
 723                 machine_id = boot_params.sys_desc_table.table[0];
 724                 machine_submodel_id = boot_params.sys_desc_table.table[1];
 725                 BIOS_revision = boot_params.sys_desc_table.table[2];
 726         }
 727         bootloader_type = boot_params.hdr.type_of_loader;
 728
 729 #ifdef CONFIG_BLK_DEV_RAM
 730         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 731         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 732         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 733 #endif
 734         ARCH_SETUP
 735
 736         printk(KERN_INFO "BIOS-provided physical RAM map:\n");
 737         print_memory_map(memory_setup());
 738
 739         copy_edd();
 740
 741         if (!boot_params.hdr.root_flags)
 742                 root_mountflags &= ~MS_RDONLY;
 743         init_mm.start_code = (unsigned long) _text;
 744         init_mm.end_code = (unsigned long) _etext;
 745         init_mm.end_data = (unsigned long) _edata;
 746         init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
 747
 748         code_resource.start = virt_to_phys(_text);
 749         code_resource.end = virt_to_phys(_etext)-1;
 750         data_resource.start = virt_to_phys(_etext);
 751         data_resource.end = virt_to_phys(_edata)-1;
 752         bss_resource.start = virt_to_phys(&__bss_start);
 753         bss_resource.end = virt_to_phys(&__bss_stop)-1;
 754
 755         parse_early_param();
 756
 757         if (user_defined_memmap) {
 758                 printk(KERN_INFO "user-defined physical RAM map:\n");
 759                 print_memory_map("user");
 760         }
 761
 762         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 763         *cmdline_p = command_line;
 764
 765         if (efi_enabled)
 766                 efi_init();
 767
 768         max_low_pfn = setup_memory();
 769
 770         /* update e820 for memory not covered by WB MTRRs */
 771         mtrr_bp_init();
 772         if (mtrr_trim_uncached_memory(max_pfn))
 773                 max_low_pfn = setup_memory();
 774
 775 #ifdef CONFIG_VMI
 776         /*
 777          * Must be after max_low_pfn is determined, and before kernel
 778          * pagetables are setup.
 779          */
 780         vmi_init();
 781 #endif
 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 #ifdef CONFIG_SMP
 794         smp_alloc_memory(); /* AP processor realmode stacks in low memory*/
 795 #endif
 796         paging_init();
 797
 798         /*
 799          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
 800          */
 801
 802 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
 803         if (init_ohci1394_dma_early)
 804                 init_ohci1394_dma_on_all_controllers();
 805 #endif
 806
 807         remapped_pgdat_init();
 808         sparse_init();
 809         zone_sizes_init();
 810
 811         /*
 812          * NOTE: at this point the bootmem allocator is fully available.
 813          */
 814
 815 #ifdef CONFIG_BLK_DEV_INITRD
 816         relocate_initrd();
 817 #endif
 818
 819         paravirt_post_allocator_init();
 820
 821         dmi_scan_machine();
 822
 823         io_delay_init();
 824
 825 #ifdef CONFIG_X86_GENERICARCH
 826         generic_apic_probe();
 827 #endif
 828
 829 #ifdef CONFIG_ACPI
 830         /*
 831          * Parse the ACPI tables for possible boot-time SMP configuration.
 832          */
 833         acpi_boot_table_init();
 834 #endif
 835
 836         early_quirks();
 837
 838 #ifdef CONFIG_ACPI
 839         acpi_boot_init();
 840
 841 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
 842         if (def_to_bigsmp)
 843                 printk(KERN_WARNING "More than 8 CPUs detected and "
 844                         "CONFIG_X86_PC cannot handle it.\nUse "
 845                         "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
 846 #endif
 847 #endif
 848 #ifdef CONFIG_X86_LOCAL_APIC
 849         if (smp_found_config)
 850                 get_smp_config();
 851 #endif
 852
 853         e820_register_memory();
 854         e820_mark_nosave_regions();
 855
 856 #ifdef CONFIG_VT
 857 #if defined(CONFIG_VGA_CONSOLE)
 858         if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
 859                 conswitchp = &vga_con;
 860 #elif defined(CONFIG_DUMMY_CONSOLE)
 861         conswitchp = &dummy_con;
 862 #endif
 863 #endif
 864 }
 865
 866 /*
 867  * Request address space for all standard resources
 868  *
 869  * This is called just before pcibios_init(), which is also a
 870  * subsys_initcall, but is linked in later (in arch/i386/pci/common.c).
 871  */
 872 static int __init request_standard_resources(void)
 873 {
 874         int i;
 875
 876         printk(KERN_INFO "Setting up standard PCI resources\n");
 877         init_iomem_resources(&code_resource, &data_resource, &bss_resource);
 878
 879         request_resource(&iomem_resource, &video_ram_resource);
 880
 881         /* request I/O space for devices used on all i[345]86 PCs */
 882         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 883                 request_resource(&ioport_resource, &standard_io_resources[i]);
 884         return 0;
 885 }
 886
 887 subsys_initcall(request_standard_resources);