arch/x86/kernel/setup.c

   1 /*
   2  *  Copyright (C) 1995  Linus Torvalds
   3  *
   4  *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
   5  *
   6  *  Memory region support
   7  *      David Parsons <orc@pell.chi.il.us>, July-August 1999
   8  *
   9  *  Added E820 sanitization routine (removes overlapping memory regions);
  10  *  Brian Moyle <bmoyle@mvista.com>, February 2001
  11  *
  12  * Moved CPU detection code to cpu/${cpu}.c
  13  *    Patrick Mochel <mochel@osdl.org>, March 2002
  14  *
  15  *  Provisions for empty E820 memory regions (reported by certain BIOSes).
  16  *  Alex Achenbach <xela@slit.de>, December 2002.
  17  *
  18  */
  19
  20 /*
  21  * This file handles the architecture-dependent parts of initialization
  22  */
  23
  24 #include <linux/sched.h>
  25 #include <linux/mm.h>
  26 #include <linux/mmzone.h>
  27 #include <linux/screen_info.h>
  28 #include <linux/ioport.h>
  29 #include <linux/acpi.h>
  30 #include <linux/sfi.h>
  31 #include <linux/apm_bios.h>
  32 #include <linux/initrd.h>
  33 #include <linux/bootmem.h>
  34 #include <linux/memblock.h>
  35 #include <linux/seq_file.h>
  36 #include <linux/console.h>
  37 #include <linux/root_dev.h>
  38 #include <linux/highmem.h>
  39 #include <linux/module.h>
  40 #include <linux/efi.h>
  41 #include <linux/init.h>
  42 #include <linux/edd.h>
  43 #include <linux/iscsi_ibft.h>
  44 #include <linux/nodemask.h>
  45 #include <linux/kexec.h>
  46 #include <linux/dmi.h>
  47 #include <linux/pfn.h>
  48 #include <linux/pci.h>
  49 #include <asm/pci-direct.h>
  50 #include <linux/init_ohci1394_dma.h>
  51 #include <linux/kvm_para.h>
  52 #include <linux/dma-contiguous.h>
  53
  54 #include <linux/errno.h>
  55 #include <linux/kernel.h>
  56 #include <linux/stddef.h>
  57 #include <linux/unistd.h>
  58 #include <linux/ptrace.h>
  59 #include <linux/user.h>
  60 #include <linux/delay.h>
  61
  62 #include <linux/kallsyms.h>
  63 #include <linux/cpufreq.h>
  64 #include <linux/dma-mapping.h>
  65 #include <linux/ctype.h>
  66 #include <linux/uaccess.h>
  67
  68 #include <linux/percpu.h>
  69 #include <linux/crash_dump.h>
  70 #include <linux/tboot.h>
  71 #include <linux/jiffies.h>
  72
  73 #include <video/edid.h>
  74
  75 #include <asm/mtrr.h>
  76 #include <asm/apic.h>
  77 #include <asm/realmode.h>
  78 #include <asm/e820.h>
  79 #include <asm/mpspec.h>
  80 #include <asm/setup.h>
  81 #include <asm/efi.h>
  82 #include <asm/timer.h>
  83 #include <asm/i8259.h>
  84 #include <asm/sections.h>
  85 #include <asm/io_apic.h>
  86 #include <asm/ist.h>
  87 #include <asm/setup_arch.h>
  88 #include <asm/bios_ebda.h>
  89 #include <asm/cacheflush.h>
  90 #include <asm/processor.h>
  91 #include <asm/bugs.h>
  92 #include <asm/kasan.h>
  93
  94 #include <asm/vsyscall.h>
  95 #include <asm/cpu.h>
  96 #include <asm/desc.h>
  97 #include <asm/dma.h>
  98 #include <asm/iommu.h>
  99 #include <asm/gart.h>
 100 #include <asm/mmu_context.h>
 101 #include <asm/proto.h>
 102
 103 #include <asm/paravirt.h>
 104 #include <asm/hypervisor.h>
 105 #include <asm/olpc_ofw.h>
 106
 107 #include <asm/percpu.h>
 108 #include <asm/topology.h>
 109 #include <asm/apicdef.h>
 110 #include <asm/amd_nb.h>
 111 #include <asm/mce.h>
 112 #include <asm/alternative.h>
 113 #include <asm/prom.h>
 114
 115 /*
 116  * max_low_pfn_mapped: highest direct mapped pfn under 4GB
 117  * max_pfn_mapped:     highest direct mapped pfn over 4GB
 118  *
 119  * The direct mapping only covers E820_RAM regions, so the ranges and gaps are
 120  * represented by pfn_mapped
 121  */
 122 unsigned long max_low_pfn_mapped;
 123 unsigned long max_pfn_mapped;
 124
 125 #ifdef CONFIG_DMI
 126 RESERVE_BRK(dmi_alloc, 65536);
 127 #endif
 128
 129
 130 static __initdata unsigned long _brk_start = (unsigned long)__brk_base;
 131 unsigned long _brk_end = (unsigned long)__brk_base;
 132
 133 #ifdef CONFIG_X86_64
 134 int default_cpu_present_to_apicid(int mps_cpu)
 135 {
 136         return __default_cpu_present_to_apicid(mps_cpu);
 137 }
 138
 139 int default_check_phys_apicid_present(int phys_apicid)
 140 {
 141         return __default_check_phys_apicid_present(phys_apicid);
 142 }
 143 #endif
 144
 145 struct boot_params boot_params;
 146
 147 /*
 148  * Machine setup..
 149  */
 150 static struct resource data_resource = {
 151         .name   = "Kernel data",
 152         .start  = 0,
 153         .end    = 0,
 154         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 155 };
 156
 157 static struct resource code_resource = {
 158         .name   = "Kernel code",
 159         .start  = 0,
 160         .end    = 0,
 161         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 162 };
 163
 164 static struct resource bss_resource = {
 165         .name   = "Kernel bss",
 166         .start  = 0,
 167         .end    = 0,
 168         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
 169 };
 170
 171
 172 #ifdef CONFIG_X86_32
 173 /* cpu data as detected by the assembly code in head.S */
 174 struct cpuinfo_x86 new_cpu_data = {
 175         .wp_works_ok = -1,
 176 };
 177 /* common cpu data for all cpus */
 178 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
 179         .wp_works_ok = -1,
 180 };
 181 EXPORT_SYMBOL(boot_cpu_data);
 182
 183 unsigned int def_to_bigsmp;
 184
 185 /* for MCA, but anyone else can use it if they want */
 186 unsigned int machine_id;
 187 unsigned int machine_submodel_id;
 188 unsigned int BIOS_revision;
 189
 190 struct apm_info apm_info;
 191 EXPORT_SYMBOL(apm_info);
 192
 193 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
 194         defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
 195 struct ist_info ist_info;
 196 EXPORT_SYMBOL(ist_info);
 197 #else
 198 struct ist_info ist_info;
 199 #endif
 200
 201 #else
 202 struct cpuinfo_x86 boot_cpu_data __read_mostly = {
 203         .x86_phys_bits = MAX_PHYSMEM_BITS,
 204 };
 205 EXPORT_SYMBOL(boot_cpu_data);
 206 #endif
 207
 208
 209 #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64)
 210 __visible unsigned long mmu_cr4_features;
 211 #else
 212 __visible unsigned long mmu_cr4_features = X86_CR4_PAE;
 213 #endif
 214
 215 /* Boot loader ID and version as integers, for the benefit of proc_dointvec */
 216 int bootloader_type, bootloader_version;
 217
 218 /*
 219  * Setup options
 220  */
 221 struct screen_info screen_info;
 222 EXPORT_SYMBOL(screen_info);
 223 struct edid_info edid_info;
 224 EXPORT_SYMBOL_GPL(edid_info);
 225
 226 extern int root_mountflags;
 227
 228 unsigned long saved_video_mode;
 229
 230 #define RAMDISK_IMAGE_START_MASK        0x07FF
 231 #define RAMDISK_PROMPT_FLAG             0x8000
 232 #define RAMDISK_LOAD_FLAG               0x4000
 233
 234 static char __initdata command_line[COMMAND_LINE_SIZE];
 235 #ifdef CONFIG_CMDLINE_BOOL
 236 static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE;
 237 #endif
 238
 239 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
 240 struct edd edd;
 241 #ifdef CONFIG_EDD_MODULE
 242 EXPORT_SYMBOL(edd);
 243 #endif
 244 /**
 245  * copy_edd() - Copy the BIOS EDD information
 246  *              from boot_params into a safe place.
 247  *
 248  */
 249 static inline void __init copy_edd(void)
 250 {
 251      memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
 252             sizeof(edd.mbr_signature));
 253      memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
 254      edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
 255      edd.edd_info_nr = boot_params.eddbuf_entries;
 256 }
 257 #else
 258 static inline void __init copy_edd(void)
 259 {
 260 }
 261 #endif
 262
 263 void * __init extend_brk(size_t size, size_t align)
 264 {
 265         size_t mask = align - 1;
 266         void *ret;
 267
 268         BUG_ON(_brk_start == 0);
 269         BUG_ON(align & mask);
 270
 271         _brk_end = (_brk_end + mask) & ~mask;
 272         BUG_ON((char *)(_brk_end + size) > __brk_limit);
 273
 274         ret = (void *)_brk_end;
 275         _brk_end += size;
 276
 277         memset(ret, 0, size);
 278
 279         return ret;
 280 }
 281
 282 #ifdef CONFIG_X86_32
 283 static void __init cleanup_highmap(void)
 284 {
 285 }
 286 #endif
 287
 288 static void __init reserve_brk(void)
 289 {
 290         if (_brk_end > _brk_start)
 291                 memblock_reserve(__pa_symbol(_brk_start),
 292                                  _brk_end - _brk_start);
 293
 294         /* Mark brk area as locked down and no longer taking any
 295            new allocations */
 296         _brk_start = 0;
 297 }
 298
 299 u64 relocated_ramdisk;
 300
 301 #ifdef CONFIG_BLK_DEV_INITRD
 302
 303 static u64 __init get_ramdisk_image(void)
 304 {
 305         u64 ramdisk_image = boot_params.hdr.ramdisk_image;
 306
 307         ramdisk_image |= (u64)boot_params.ext_ramdisk_image << 32;
 308
 309         return ramdisk_image;
 310 }
 311 static u64 __init get_ramdisk_size(void)
 312 {
 313         u64 ramdisk_size = boot_params.hdr.ramdisk_size;
 314
 315         ramdisk_size |= (u64)boot_params.ext_ramdisk_size << 32;
 316
 317         return ramdisk_size;
 318 }
 319
 320 static void __init relocate_initrd(void)
 321 {
 322         /* Assume only end is not page aligned */
 323         u64 ramdisk_image = get_ramdisk_image();
 324         u64 ramdisk_size  = get_ramdisk_size();
 325         u64 area_size     = PAGE_ALIGN(ramdisk_size);
 326
 327         /* We need to move the initrd down into directly mapped mem */
 328         relocated_ramdisk = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
 329                                                    area_size, PAGE_SIZE);
 330
 331         if (!relocated_ramdisk)
 332                 panic("Cannot find place for new RAMDISK of size %lld\n",
 333                       ramdisk_size);
 334
 335         /* Note: this includes all the mem currently occupied by
 336            the initrd, we rely on that fact to keep the data intact. */
 337         memblock_reserve(relocated_ramdisk, area_size);
 338         initrd_start = relocated_ramdisk + PAGE_OFFSET;
 339         initrd_end   = initrd_start + ramdisk_size;
 340         printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n",
 341                relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 342
 343         copy_from_early_mem((void *)initrd_start, ramdisk_image, ramdisk_size);
 344
 345         printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to"
 346                 " [mem %#010llx-%#010llx]\n",
 347                 ramdisk_image, ramdisk_image + ramdisk_size - 1,
 348                 relocated_ramdisk, relocated_ramdisk + ramdisk_size - 1);
 349 }
 350
 351 static void __init early_reserve_initrd(void)
 352 {
 353         /* Assume only end is not page aligned */
 354         u64 ramdisk_image = get_ramdisk_image();
 355         u64 ramdisk_size  = get_ramdisk_size();
 356         u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 357
 358         if (!boot_params.hdr.type_of_loader ||
 359             !ramdisk_image || !ramdisk_size)
 360                 return;         /* No initrd provided by bootloader */
 361
 362         memblock_reserve(ramdisk_image, ramdisk_end - ramdisk_image);
 363 }
 364 static void __init reserve_initrd(void)
 365 {
 366         /* Assume only end is not page aligned */
 367         u64 ramdisk_image = get_ramdisk_image();
 368         u64 ramdisk_size  = get_ramdisk_size();
 369         u64 ramdisk_end   = PAGE_ALIGN(ramdisk_image + ramdisk_size);
 370         u64 mapped_size;
 371
 372         if (!boot_params.hdr.type_of_loader ||
 373             !ramdisk_image || !ramdisk_size)
 374                 return;         /* No initrd provided by bootloader */
 375
 376         initrd_start = 0;
 377
 378         mapped_size = memblock_mem_size(max_pfn_mapped);
 379         if (ramdisk_size >= (mapped_size>>1))
 380                 panic("initrd too large to handle, "
 381                        "disabling initrd (%lld needed, %lld available)\n",
 382                        ramdisk_size, mapped_size>>1);
 383
 384         printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
 385                         ramdisk_end - 1);
 386
 387         if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
 388                                 PFN_DOWN(ramdisk_end))) {
 389                 /* All are mapped, easy case */
 390                 initrd_start = ramdisk_image + PAGE_OFFSET;
 391                 initrd_end = initrd_start + ramdisk_size;
 392                 return;
 393         }
 394
 395         relocate_initrd();
 396
 397         memblock_free(ramdisk_image, ramdisk_end - ramdisk_image);
 398 }
 399 #else
 400 static void __init early_reserve_initrd(void)
 401 {
 402 }
 403 static void __init reserve_initrd(void)
 404 {
 405 }
 406 #endif /* CONFIG_BLK_DEV_INITRD */
 407
 408 static void __init parse_setup_data(void)
 409 {
 410         struct setup_data *data;
 411         u64 pa_data, pa_next;
 412
 413         pa_data = boot_params.hdr.setup_data;
 414         while (pa_data) {
 415                 u32 data_len, data_type;
 416
 417                 data = early_memremap(pa_data, sizeof(*data));
 418                 data_len = data->len + sizeof(struct setup_data);
 419                 data_type = data->type;
 420                 pa_next = data->next;
 421                 early_memunmap(data, sizeof(*data));
 422
 423                 switch (data_type) {
 424                 case SETUP_E820_EXT:
 425                         parse_e820_ext(pa_data, data_len);
 426                         break;
 427                 case SETUP_DTB:
 428                         add_dtb(pa_data);
 429                         break;
 430                 case SETUP_EFI:
 431                         parse_efi_setup(pa_data, data_len);
 432                         break;
 433                 default:
 434                         break;
 435                 }
 436                 pa_data = pa_next;
 437         }
 438 }
 439
 440 static void __init e820_reserve_setup_data(void)
 441 {
 442         struct setup_data *data;
 443         u64 pa_data;
 444
 445         pa_data = boot_params.hdr.setup_data;
 446         if (!pa_data)
 447                 return;
 448
 449         while (pa_data) {
 450                 data = early_memremap(pa_data, sizeof(*data));
 451                 e820_update_range(pa_data, sizeof(*data)+data->len,
 452                          E820_RAM, E820_RESERVED_KERN);
 453                 pa_data = data->next;
 454                 early_memunmap(data, sizeof(*data));
 455         }
 456
 457         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 458         memcpy(&e820_saved, &e820, sizeof(struct e820map));
 459         printk(KERN_INFO "extended physical RAM map:\n");
 460         e820_print_map("reserve setup_data");
 461 }
 462
 463 static void __init memblock_x86_reserve_range_setup_data(void)
 464 {
 465         struct setup_data *data;
 466         u64 pa_data;
 467
 468         pa_data = boot_params.hdr.setup_data;
 469         while (pa_data) {
 470                 data = early_memremap(pa_data, sizeof(*data));
 471                 memblock_reserve(pa_data, sizeof(*data) + data->len);
 472                 pa_data = data->next;
 473                 early_memunmap(data, sizeof(*data));
 474         }
 475 }
 476
 477 /*
 478  * --------- Crashkernel reservation ------------------------------
 479  */
 480
 481 #ifdef CONFIG_KEXEC_CORE
 482
 483 /*
 484  * Keep the crash kernel below this limit.  On 32 bits earlier kernels
 485  * would limit the kernel to the low 512 MiB due to mapping restrictions.
 486  * On 64bit, old kexec-tools need to under 896MiB.
 487  */
 488 #ifdef CONFIG_X86_32
 489 # define CRASH_KERNEL_ADDR_LOW_MAX      (512 << 20)
 490 # define CRASH_KERNEL_ADDR_HIGH_MAX     (512 << 20)
 491 #else
 492 # define CRASH_KERNEL_ADDR_LOW_MAX      (896UL<<20)
 493 # define CRASH_KERNEL_ADDR_HIGH_MAX     MAXMEM
 494 #endif
 495
 496 static void __init reserve_crashkernel_low(void)
 497 {
 498 #ifdef CONFIG_X86_64
 499         const unsigned long long alignment = 16<<20;    /* 16M */
 500         unsigned long long low_base = 0, low_size = 0;
 501         unsigned long total_low_mem;
 502         unsigned long long base;
 503         bool auto_set = false;
 504         int ret;
 505
 506         total_low_mem = memblock_mem_size(1UL<<(32-PAGE_SHIFT));
 507         /* crashkernel=Y,low */
 508         ret = parse_crashkernel_low(boot_command_line, total_low_mem,
 509                                                 &low_size, &base);
 510         if (ret != 0) {
 511                 /*
 512                  * two parts from lib/swiotlb.c:
 513                  * -swiotlb size: user-specified with swiotlb= or default.
 514                  *
 515                  * -swiotlb overflow buffer: now hardcoded to 32k. We round it
 516                  * to 8M for other buffers that may need to stay low too. Also
 517                  * make sure we allocate enough extra low memory so that we
 518                  * don't run out of DMA buffers for 32-bit devices.
 519                  */
 520                 low_size = max(swiotlb_size_or_default() + (8UL<<20), 256UL<<20);
 521                 auto_set = true;
 522         } else {
 523                 /* passed with crashkernel=0,low ? */
 524                 if (!low_size)
 525                         return;
 526         }
 527
 528         low_base = memblock_find_in_range(low_size, (1ULL<<32),
 529                                         low_size, alignment);
 530
 531         if (!low_base) {
 532                 if (!auto_set)
 533                         pr_info("crashkernel low reservation failed - No suitable area found.\n");
 534
 535                 return;
 536         }
 537
 538         memblock_reserve(low_base, low_size);
 539         pr_info("Reserving %ldMB of low memory at %ldMB for crashkernel (System low RAM: %ldMB)\n",
 540                         (unsigned long)(low_size >> 20),
 541                         (unsigned long)(low_base >> 20),
 542                         (unsigned long)(total_low_mem >> 20));
 543         crashk_low_res.start = low_base;
 544         crashk_low_res.end   = low_base + low_size - 1;
 545         insert_resource(&iomem_resource, &crashk_low_res);
 546 #endif
 547 }
 548
 549 static void __init reserve_crashkernel(void)
 550 {
 551         const unsigned long long alignment = 16<<20;    /* 16M */
 552         unsigned long long total_mem;
 553         unsigned long long crash_size, crash_base;
 554         bool high = false;
 555         int ret;
 556
 557         total_mem = memblock_phys_mem_size();
 558
 559         /* crashkernel=XM */
 560         ret = parse_crashkernel(boot_command_line, total_mem,
 561                         &crash_size, &crash_base);
 562         if (ret != 0 || crash_size <= 0) {
 563                 /* crashkernel=X,high */
 564                 ret = parse_crashkernel_high(boot_command_line, total_mem,
 565                                 &crash_size, &crash_base);
 566                 if (ret != 0 || crash_size <= 0)
 567                         return;
 568                 high = true;
 569         }
 570
 571         /* 0 means: find the address automatically */
 572         if (crash_base <= 0) {
 573                 /*
 574                  *  kexec want bzImage is below CRASH_KERNEL_ADDR_MAX
 575                  */
 576                 crash_base = memblock_find_in_range(alignment,
 577                                         high ? CRASH_KERNEL_ADDR_HIGH_MAX :
 578                                                CRASH_KERNEL_ADDR_LOW_MAX,
 579                                         crash_size, alignment);
 580
 581                 if (!crash_base) {
 582                         pr_info("crashkernel reservation failed - No suitable area found.\n");
 583                         return;
 584                 }
 585
 586         } else {
 587                 unsigned long long start;
 588
 589                 start = memblock_find_in_range(crash_base,
 590                                  crash_base + crash_size, crash_size, 1<<20);
 591                 if (start != crash_base) {
 592                         pr_info("crashkernel reservation failed - memory is in use.\n");
 593                         return;
 594                 }
 595         }
 596         memblock_reserve(crash_base, crash_size);
 597
 598         printk(KERN_INFO "Reserving %ldMB of memory at %ldMB "
 599                         "for crashkernel (System RAM: %ldMB)\n",
 600                         (unsigned long)(crash_size >> 20),
 601                         (unsigned long)(crash_base >> 20),
 602                         (unsigned long)(total_mem >> 20));
 603
 604         crashk_res.start = crash_base;
 605         crashk_res.end   = crash_base + crash_size - 1;
 606         insert_resource(&iomem_resource, &crashk_res);
 607
 608         if (crash_base >= (1ULL<<32))
 609                 reserve_crashkernel_low();
 610 }
 611 #else
 612 static void __init reserve_crashkernel(void)
 613 {
 614 }
 615 #endif
 616
 617 static struct resource standard_io_resources[] = {
 618         { .name = "dma1", .start = 0x00, .end = 0x1f,
 619                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 620         { .name = "pic1", .start = 0x20, .end = 0x21,
 621                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 622         { .name = "timer0", .start = 0x40, .end = 0x43,
 623                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 624         { .name = "timer1", .start = 0x50, .end = 0x53,
 625                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 626         { .name = "keyboard", .start = 0x60, .end = 0x60,
 627                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 628         { .name = "keyboard", .start = 0x64, .end = 0x64,
 629                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 630         { .name = "dma page reg", .start = 0x80, .end = 0x8f,
 631                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 632         { .name = "pic2", .start = 0xa0, .end = 0xa1,
 633                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 634         { .name = "dma2", .start = 0xc0, .end = 0xdf,
 635                 .flags = IORESOURCE_BUSY | IORESOURCE_IO },
 636         { .name = "fpu", .start = 0xf0, .end = 0xff,
 637                 .flags = IORESOURCE_BUSY | IORESOURCE_IO }
 638 };
 639
 640 void __init reserve_standard_io_resources(void)
 641 {
 642         int i;
 643
 644         /* request I/O space for devices used on all i[345]86 PCs */
 645         for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++)
 646                 request_resource(&ioport_resource, &standard_io_resources[i]);
 647
 648 }
 649
 650 static __init void reserve_ibft_region(void)
 651 {
 652         unsigned long addr, size = 0;
 653
 654         addr = find_ibft_region(&size);
 655
 656         if (size)
 657                 memblock_reserve(addr, size);
 658 }
 659
 660 static bool __init snb_gfx_workaround_needed(void)
 661 {
 662 #ifdef CONFIG_PCI
 663         int i;
 664         u16 vendor, devid;
 665         static const __initconst u16 snb_ids[] = {
 666                 0x0102,
 667                 0x0112,
 668                 0x0122,
 669                 0x0106,
 670                 0x0116,
 671                 0x0126,
 672                 0x010a,
 673         };
 674
 675         /* Assume no if something weird is going on with PCI */
 676         if (!early_pci_allowed())
 677                 return false;
 678
 679         vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
 680         if (vendor != 0x8086)
 681                 return false;
 682
 683         devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
 684         for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
 685                 if (devid == snb_ids[i])
 686                         return true;
 687 #endif
 688
 689         return false;
 690 }
 691
 692 /*
 693  * Sandy Bridge graphics has trouble with certain ranges, exclude
 694  * them from allocation.
 695  */
 696 static void __init trim_snb_memory(void)
 697 {
 698         static const __initconst unsigned long bad_pages[] = {
 699                 0x20050000,
 700                 0x20110000,
 701                 0x20130000,
 702                 0x20138000,
 703                 0x40004000,
 704         };
 705         int i;
 706
 707         if (!snb_gfx_workaround_needed())
 708                 return;
 709
 710         printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
 711
 712         /*
 713          * Reserve all memory below the 1 MB mark that has not
 714          * already been reserved.
 715          */
 716         memblock_reserve(0, 1<<20);
 717
 718         for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
 719                 if (memblock_reserve(bad_pages[i], PAGE_SIZE))
 720                         printk(KERN_WARNING "failed to reserve 0x%08lx\n",
 721                                bad_pages[i]);
 722         }
 723 }
 724
 725 /*
 726  * Here we put platform-specific memory range workarounds, i.e.
 727  * memory known to be corrupt or otherwise in need to be reserved on
 728  * specific platforms.
 729  *
 730  * If this gets used more widely it could use a real dispatch mechanism.
 731  */
 732 static void __init trim_platform_memory_ranges(void)
 733 {
 734         trim_snb_memory();
 735 }
 736
 737 static void __init trim_bios_range(void)
 738 {
 739         /*
 740          * A special case is the first 4Kb of memory;
 741          * This is a BIOS owned area, not kernel ram, but generally
 742          * not listed as such in the E820 table.
 743          *
 744          * This typically reserves additional memory (64KiB by default)
 745          * since some BIOSes are known to corrupt low memory.  See the
 746          * Kconfig help text for X86_RESERVE_LOW.
 747          */
 748         e820_update_range(0, PAGE_SIZE, E820_RAM, E820_RESERVED);
 749
 750         /*
 751          * special case: Some BIOSen report the PC BIOS
 752          * area (640->1Mb) as ram even though it is not.
 753          * take them out.
 754          */
 755         e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
 756
 757         sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 758 }
 759
 760 /* called before trim_bios_range() to spare extra sanitize */
 761 static void __init e820_add_kernel_range(void)
 762 {
 763         u64 start = __pa_symbol(_text);
 764         u64 size = __pa_symbol(_end) - start;
 765
 766         /*
 767          * Complain if .text .data and .bss are not marked as E820_RAM and
 768          * attempt to fix it by adding the range. We may have a confused BIOS,
 769          * or the user may have used memmap=exactmap or memmap=xxM$yyM to
 770          * exclude kernel range. If we really are running on top non-RAM,
 771          * we will crash later anyways.
 772          */
 773         if (e820_all_mapped(start, start + size, E820_RAM))
 774                 return;
 775
 776         pr_warn(".text .data .bss are not marked as E820_RAM!\n");
 777         e820_remove_range(start, size, E820_RAM, 0);
 778         e820_add_region(start, size, E820_RAM);
 779 }
 780
 781 static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
 782
 783 static int __init parse_reservelow(char *p)
 784 {
 785         unsigned long long size;
 786
 787         if (!p)
 788                 return -EINVAL;
 789
 790         size = memparse(p, &p);
 791
 792         if (size < 4096)
 793                 size = 4096;
 794
 795         if (size > 640*1024)
 796                 size = 640*1024;
 797
 798         reserve_low = size;
 799
 800         return 0;
 801 }
 802
 803 early_param("reservelow", parse_reservelow);
 804
 805 static void __init trim_low_memory_range(void)
 806 {
 807         memblock_reserve(0, ALIGN(reserve_low, PAGE_SIZE));
 808 }
 809
 810 /*
 811  * Dump out kernel offset information on panic.
 812  */
 813 static int
 814 dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
 815 {
 816         if (kaslr_enabled()) {
 817                 pr_emerg("Kernel Offset: 0x%lx from 0x%lx (relocation range: 0x%lx-0x%lx)\n",
 818                          kaslr_offset(),
 819                          __START_KERNEL,
 820                          __START_KERNEL_map,
 821                          MODULES_VADDR-1);
 822         } else {
 823                 pr_emerg("Kernel Offset: disabled\n");
 824         }
 825
 826         return 0;
 827 }
 828
 829 /*
 830  * Determine if we were loaded by an EFI loader.  If so, then we have also been
 831  * passed the efi memmap, systab, etc., so we should use these data structures
 832  * for initialization.  Note, the efi init code path is determined by the
 833  * global efi_enabled. This allows the same kernel image to be used on existing
 834  * systems (with a traditional BIOS) as well as on EFI systems.
 835  */
 836 /*
 837  * setup_arch - architecture-specific boot-time initializations
 838  *
 839  * Note: On x86_64, fixmaps are ready for use even before this is called.
 840  */
 841
 842 void __init setup_arch(char **cmdline_p)
 843 {
 844         memblock_reserve(__pa_symbol(_text),
 845                          (unsigned long)__bss_stop - (unsigned long)_text);
 846
 847         early_reserve_initrd();
 848
 849         /*
 850          * At this point everything still needed from the boot loader
 851          * or BIOS or kernel text should be early reserved or marked not
 852          * RAM in e820. All other memory is free game.
 853          */
 854
 855 #ifdef CONFIG_X86_32
 856         memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
 857
 858         /*
 859          * copy kernel address range established so far and switch
 860          * to the proper swapper page table
 861          */
 862         clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 863                         initial_page_table + KERNEL_PGD_BOUNDARY,
 864                         KERNEL_PGD_PTRS);
 865
 866         load_cr3(swapper_pg_dir);
 867         /*
 868          * Note: Quark X1000 CPUs advertise PGE incorrectly and require
 869          * a cr3 based tlb flush, so the following __flush_tlb_all()
 870          * will not flush anything because the cpu quirk which clears
 871          * X86_FEATURE_PGE has not been invoked yet. Though due to the
 872          * load_cr3() above the TLB has been flushed already. The
 873          * quirk is invoked before subsequent calls to __flush_tlb_all()
 874          * so proper operation is guaranteed.
 875          */
 876         __flush_tlb_all();
 877 #else
 878         printk(KERN_INFO "Command line: %s\n", boot_command_line);
 879 #endif
 880
 881         /*
 882          * If we have OLPC OFW, we might end up relocating the fixmap due to
 883          * reserve_top(), so do this before touching the ioremap area.
 884          */
 885         olpc_ofw_detect();
 886
 887         early_trap_init();
 888         early_cpu_init();
 889         early_ioremap_init();
 890
 891         setup_olpc_ofw_pgd();
 892
 893         ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
 894         screen_info = boot_params.screen_info;
 895         edid_info = boot_params.edid_info;
 896 #ifdef CONFIG_X86_32
 897         apm_info.bios = boot_params.apm_bios_info;
 898         ist_info = boot_params.ist_info;
 899 #endif
 900         saved_video_mode = boot_params.hdr.vid_mode;
 901         bootloader_type = boot_params.hdr.type_of_loader;
 902         if ((bootloader_type >> 4) == 0xe) {
 903                 bootloader_type &= 0xf;
 904                 bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4;
 905         }
 906         bootloader_version  = bootloader_type & 0xf;
 907         bootloader_version |= boot_params.hdr.ext_loader_ver << 4;
 908
 909 #ifdef CONFIG_BLK_DEV_RAM
 910         rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
 911         rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 912         rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 913 #endif
 914 #ifdef CONFIG_EFI
 915         if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 916                      EFI32_LOADER_SIGNATURE, 4)) {
 917                 set_bit(EFI_BOOT, &efi.flags);
 918         } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
 919                      EFI64_LOADER_SIGNATURE, 4)) {
 920                 set_bit(EFI_BOOT, &efi.flags);
 921                 set_bit(EFI_64BIT, &efi.flags);
 922         }
 923
 924         if (efi_enabled(EFI_BOOT))
 925                 efi_memblock_x86_reserve_range();
 926 #endif
 927
 928         x86_init.oem.arch_setup();
 929
 930         iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1;
 931         setup_memory_map();
 932         parse_setup_data();
 933
 934         copy_edd();
 935
 936         if (!boot_params.hdr.root_flags)
 937                 root_mountflags &= ~MS_RDONLY;
 938         init_mm.start_code = (unsigned long) _text;
 939         init_mm.end_code = (unsigned long) _etext;
 940         init_mm.end_data = (unsigned long) _edata;
 941         init_mm.brk = _brk_end;
 942
 943         mpx_mm_init(&init_mm);
 944
 945         code_resource.start = __pa_symbol(_text);
 946         code_resource.end = __pa_symbol(_etext)-1;
 947         data_resource.start = __pa_symbol(_etext);
 948         data_resource.end = __pa_symbol(_edata)-1;
 949         bss_resource.start = __pa_symbol(__bss_start);
 950         bss_resource.end = __pa_symbol(__bss_stop)-1;
 951
 952 #ifdef CONFIG_CMDLINE_BOOL
 953 #ifdef CONFIG_CMDLINE_OVERRIDE
 954         strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 955 #else
 956         if (builtin_cmdline[0]) {
 957                 /* append boot loader cmdline to builtin */
 958                 strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE);
 959                 strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE);
 960                 strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE);
 961         }
 962 #endif
 963 #endif
 964
 965         strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 966         *cmdline_p = command_line;
 967
 968         /*
 969          * x86_configure_nx() is called before parse_early_param() to detect
 970          * whether hardware doesn't support NX (so that the early EHCI debug
 971          * console setup can safely call set_fixmap()). It may then be called
 972          * again from within noexec_setup() during parsing early parameters
 973          * to honor the respective command line option.
 974          */
 975         x86_configure_nx();
 976
 977         parse_early_param();
 978
 979         x86_report_nx();
 980
 981         /* after early param, so could get panic from serial */
 982         memblock_x86_reserve_range_setup_data();
 983
 984         if (acpi_mps_check()) {
 985 #ifdef CONFIG_X86_LOCAL_APIC
 986                 disable_apic = 1;
 987 #endif
 988                 setup_clear_cpu_cap(X86_FEATURE_APIC);
 989         }
 990
 991 #ifdef CONFIG_PCI
 992         if (pci_early_dump_regs)
 993                 early_dump_pci_devices();
 994 #endif
 995
 996         /* update the e820_saved too */
 997         e820_reserve_setup_data();
 998         finish_e820_parsing();
 999
1000         if (efi_enabled(EFI_BOOT))
1001                 efi_init();
1002
1003         dmi_scan_machine();
1004         dmi_memdev_walk();
1005         dmi_set_dump_stack_arch_desc();
1006
1007         /*
1008          * VMware detection requires dmi to be available, so this
1009          * needs to be done after dmi_scan_machine, for the BP.
1010          */
1011         init_hypervisor_platform();
1012
1013         x86_init.resources.probe_roms();
1014
1015         /* after parse_early_param, so could debug it */
1016         insert_resource(&iomem_resource, &code_resource);
1017         insert_resource(&iomem_resource, &data_resource);
1018         insert_resource(&iomem_resource, &bss_resource);
1019
1020         e820_add_kernel_range();
1021         trim_bios_range();
1022 #ifdef CONFIG_X86_32
1023         if (ppro_with_ram_bug()) {
1024                 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
1025                                   E820_RESERVED);
1026                 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
1027                 printk(KERN_INFO "fixed physical RAM map:\n");
1028                 e820_print_map("bad_ppro");
1029         }
1030 #else
1031         early_gart_iommu_check();
1032 #endif
1033
1034         /*
1035          * partially used pages are not usable - thus
1036          * we are rounding upwards:
1037          */
1038         max_pfn = e820_end_of_ram_pfn();
1039
1040         /* update e820 for memory not covered by WB MTRRs */
1041         mtrr_bp_init();
1042         if (mtrr_trim_uncached_memory(max_pfn))
1043                 max_pfn = e820_end_of_ram_pfn();
1044
1045 #ifdef CONFIG_X86_32
1046         /* max_low_pfn get updated here */
1047         find_low_pfn_range();
1048 #else
1049         check_x2apic();
1050
1051         /* How many end-of-memory variables you have, grandma! */
1052         /* need this before calling reserve_initrd */
1053         if (max_pfn > (1UL<<(32 - PAGE_SHIFT)))
1054                 max_low_pfn = e820_end_of_low_ram_pfn();
1055         else
1056                 max_low_pfn = max_pfn;
1057
1058         high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1;
1059 #endif
1060
1061         /*
1062          * Find and reserve possible boot-time SMP configuration:
1063          */
1064         find_smp_config();
1065
1066         reserve_ibft_region();
1067
1068         early_alloc_pgt_buf();
1069
1070         /*
1071          * Need to conclude brk, before memblock_x86_fill()
1072          *  it could use memblock_find_in_range, could overlap with
1073          *  brk area.
1074          */
1075         reserve_brk();
1076
1077         cleanup_highmap();
1078
1079         memblock_set_current_limit(ISA_END_ADDRESS);
1080         memblock_x86_fill();
1081
1082         if (efi_enabled(EFI_BOOT))
1083                 efi_find_mirror();
1084
1085         /*
1086          * The EFI specification says that boot service code won't be called
1087          * after ExitBootServices(). This is, in fact, a lie.
1088          */
1089         if (efi_enabled(EFI_MEMMAP))
1090                 efi_reserve_boot_services();
1091
1092         /* preallocate 4k for mptable mpc */
1093         early_reserve_e820_mpc_new();
1094
1095 #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION
1096         setup_bios_corruption_check();
1097 #endif
1098
1099 #ifdef CONFIG_X86_32
1100         printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n",
1101                         (max_pfn_mapped<<PAGE_SHIFT) - 1);
1102 #endif
1103
1104         reserve_real_mode();
1105
1106         trim_platform_memory_ranges();
1107         trim_low_memory_range();
1108
1109         init_mem_mapping();
1110
1111         early_trap_pf_init();
1112
1113         setup_real_mode();
1114
1115         memblock_set_current_limit(get_max_mapped());
1116
1117         /*
1118          * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
1119          */
1120
1121 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
1122         if (init_ohci1394_dma_early)
1123                 init_ohci1394_dma_on_all_controllers();
1124 #endif
1125         /* Allocate bigger log buffer */
1126         setup_log_buf(1);
1127
1128         reserve_initrd();
1129
1130 #if defined(CONFIG_ACPI) && defined(CONFIG_BLK_DEV_INITRD)
1131         acpi_initrd_override((void *)initrd_start, initrd_end - initrd_start);
1132 #endif
1133
1134         vsmp_init();
1135
1136         io_delay_init();
1137
1138         /*
1139          * Parse the ACPI tables for possible boot-time SMP configuration.
1140          */
1141         acpi_boot_table_init();
1142
1143         early_acpi_boot_init();
1144
1145         initmem_init();
1146         dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
1147
1148         /*
1149          * Reserve memory for crash kernel after SRAT is parsed so that it
1150          * won't consume hotpluggable memory.
1151          */
1152         reserve_crashkernel();
1153
1154         memblock_find_dma_reserve();
1155
1156 #ifdef CONFIG_KVM_GUEST
1157         kvmclock_init();
1158 #endif
1159
1160         x86_init.paging.pagetable_init();
1161
1162         kasan_init();
1163
1164         if (boot_cpu_data.cpuid_level >= 0) {
1165                 /* A CPU has %cr4 if and only if it has CPUID */
1166                 mmu_cr4_features = __read_cr4();
1167                 if (trampoline_cr4_features)
1168                         *trampoline_cr4_features = mmu_cr4_features;
1169         }
1170
1171 #ifdef CONFIG_X86_32
1172         /* sync back kernel address range */
1173         clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
1174                         swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
1175                         KERNEL_PGD_PTRS);
1176 #endif
1177
1178         tboot_probe();
1179
1180         map_vsyscall();
1181
1182         generic_apic_probe();
1183
1184         early_quirks();
1185
1186         /*
1187          * Read APIC and some other early information from ACPI tables.
1188          */
1189         acpi_boot_init();
1190         sfi_init();
1191         x86_dtb_init();
1192
1193         /*
1194          * get boot-time SMP configuration:
1195          */
1196         if (smp_found_config)
1197                 get_smp_config();
1198
1199         prefill_possible_map();
1200
1201         init_cpu_to_node();
1202
1203         init_apic_mappings();
1204         io_apic_init_mappings();
1205
1206         kvm_guest_init();
1207
1208         e820_reserve_resources();
1209         e820_mark_nosave_regions(max_low_pfn);
1210
1211         x86_init.resources.reserve_resources();
1212
1213         e820_setup_gap();
1214
1215 #ifdef CONFIG_VT
1216 #if defined(CONFIG_VGA_CONSOLE)
1217         if (!efi_enabled(EFI_BOOT) || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
1218                 conswitchp = &vga_con;
1219 #elif defined(CONFIG_DUMMY_CONSOLE)
1220         conswitchp = &dummy_con;
1221 #endif
1222 #endif
1223         x86_init.oem.banner();
1224
1225         x86_init.timers.wallclock_init();
1226
1227         mcheck_init();
1228
1229         arch_init_ideal_nops();
1230
1231         register_refined_jiffies(CLOCK_TICK_RATE);
1232
1233 #ifdef CONFIG_EFI
1234         if (efi_enabled(EFI_BOOT))
1235                 efi_apply_memmap_quirks();
1236 #endif
1237 }
1238
1239 #ifdef CONFIG_X86_32
1240
1241 static struct resource video_ram_resource = {
1242         .name   = "Video RAM area",
1243         .start  = 0xa0000,
1244         .end    = 0xbffff,
1245         .flags  = IORESOURCE_BUSY | IORESOURCE_MEM
1246 };
1247
1248 void __init i386_reserve_resources(void)
1249 {
1250         request_resource(&iomem_resource, &video_ram_resource);
1251         reserve_standard_io_resources();
1252 }
1253
1254 #endif /* CONFIG_X86_32 */
1255
1256 static struct notifier_block kernel_offset_notifier = {
1257         .notifier_call = dump_kernel_offset
1258 };
1259
1260 static int __init register_kernel_offset_dumper(void)
1261 {
1262         atomic_notifier_chain_register(&panic_notifier_list,
1263                                         &kernel_offset_notifier);
1264         return 0;
1265 }
1266 __initcall(register_kernel_offset_dumper);