| 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/dmi.h> |
| 86 | #include <asm/io_apic.h> |
| 87 | #include <asm/ist.h> |
| 88 | #include <asm/setup_arch.h> |
| 89 | #include <asm/bios_ebda.h> |
| 90 | #include <asm/cacheflush.h> |
| 91 | #include <asm/processor.h> |
| 92 | #include <asm/bugs.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 | #ifdef CONFIG_X86_64 |
| 112 | #include <asm/numa_64.h> |
| 113 | #endif |
| 114 | #include <asm/mce.h> |
| 115 | #include <asm/alternative.h> |
| 116 | #include <asm/prom.h> |
| 117 | |
| 118 | /* |
| 119 | * max_low_pfn_mapped: highest direct mapped pfn under 4GB |
| 120 | * max_pfn_mapped: highest direct mapped pfn over 4GB |
| 121 | * |
| 122 | * The direct mapping only covers E820_RAM regions, so the ranges and gaps are |
| 123 | * represented by pfn_mapped |
| 124 | */ |
| 125 | unsigned long max_low_pfn_mapped; |
| 126 | unsigned long max_pfn_mapped; |
| 127 | |
| 128 | #ifdef CONFIG_DMI |
| 129 | RESERVE_BRK(dmi_alloc, 65536); |
| 130 | #endif |
| 131 | |
| 132 | |
| 133 | static __initdata unsigned long _brk_start = (unsigned long)__brk_base; |
| 134 | unsigned long _brk_end = (unsigned long)__brk_base; |
| 135 | |
| 136 | #ifdef CONFIG_X86_64 |
| 137 | int default_cpu_present_to_apicid(int mps_cpu) |
| 138 | { |
| 139 | return __default_cpu_present_to_apicid(mps_cpu); |
| 140 | } |
| 141 | |
| 142 | int default_check_phys_apicid_present(int phys_apicid) |
| 143 | { |
| 144 | return __default_check_phys_apicid_present(phys_apicid); |
| 145 | } |
| 146 | #endif |
| 147 | |
| 148 | #ifndef CONFIG_DEBUG_BOOT_PARAMS |
| 149 | struct boot_params __initdata boot_params; |
| 150 | #else |
| 151 | struct boot_params boot_params; |
| 152 | #endif |
| 153 | |
| 154 | /* |
| 155 | * Machine setup.. |
| 156 | */ |
| 157 | static struct resource data_resource = { |
| 158 | .name = "Kernel data", |
| 159 | .start = 0, |
| 160 | .end = 0, |
| 161 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM |
| 162 | }; |
| 163 | |
| 164 | static struct resource code_resource = { |
| 165 | .name = "Kernel code", |
| 166 | .start = 0, |
| 167 | .end = 0, |
| 168 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM |
| 169 | }; |
| 170 | |
| 171 | static struct resource bss_resource = { |
| 172 | .name = "Kernel bss", |
| 173 | .start = 0, |
| 174 | .end = 0, |
| 175 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM |
| 176 | }; |
| 177 | |
| 178 | |
| 179 | #ifdef CONFIG_X86_32 |
| 180 | /* cpu data as detected by the assembly code in head.S */ |
| 181 | struct cpuinfo_x86 new_cpu_data __cpuinitdata = {0, 0, 0, 0, -1, 1, 0, 0, -1}; |
| 182 | /* common cpu data for all cpus */ |
| 183 | struct cpuinfo_x86 boot_cpu_data __read_mostly = {0, 0, 0, 0, -1, 1, 0, 0, -1}; |
| 184 | EXPORT_SYMBOL(boot_cpu_data); |
| 185 | |
| 186 | unsigned int def_to_bigsmp; |
| 187 | |
| 188 | /* for MCA, but anyone else can use it if they want */ |
| 189 | unsigned int machine_id; |
| 190 | unsigned int machine_submodel_id; |
| 191 | unsigned int BIOS_revision; |
| 192 | |
| 193 | struct apm_info apm_info; |
| 194 | EXPORT_SYMBOL(apm_info); |
| 195 | |
| 196 | #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \ |
| 197 | defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE) |
| 198 | struct ist_info ist_info; |
| 199 | EXPORT_SYMBOL(ist_info); |
| 200 | #else |
| 201 | struct ist_info ist_info; |
| 202 | #endif |
| 203 | |
| 204 | #else |
| 205 | struct cpuinfo_x86 boot_cpu_data __read_mostly = { |
| 206 | .x86_phys_bits = MAX_PHYSMEM_BITS, |
| 207 | }; |
| 208 | EXPORT_SYMBOL(boot_cpu_data); |
| 209 | #endif |
| 210 | |
| 211 | |
| 212 | #if !defined(CONFIG_X86_PAE) || defined(CONFIG_X86_64) |
| 213 | unsigned long mmu_cr4_features; |
| 214 | #else |
| 215 | unsigned long mmu_cr4_features = X86_CR4_PAE; |
| 216 | #endif |
| 217 | |
| 218 | /* Boot loader ID and version as integers, for the benefit of proc_dointvec */ |
| 219 | int bootloader_type, bootloader_version; |
| 220 | |
| 221 | /* |
| 222 | * Setup options |
| 223 | */ |
| 224 | struct screen_info screen_info; |
| 225 | EXPORT_SYMBOL(screen_info); |
| 226 | struct edid_info edid_info; |
| 227 | EXPORT_SYMBOL_GPL(edid_info); |
| 228 | |
| 229 | extern int root_mountflags; |
| 230 | |
| 231 | unsigned long saved_video_mode; |
| 232 | |
| 233 | #define RAMDISK_IMAGE_START_MASK 0x07FF |
| 234 | #define RAMDISK_PROMPT_FLAG 0x8000 |
| 235 | #define RAMDISK_LOAD_FLAG 0x4000 |
| 236 | |
| 237 | static char __initdata command_line[COMMAND_LINE_SIZE]; |
| 238 | #ifdef CONFIG_CMDLINE_BOOL |
| 239 | static char __initdata builtin_cmdline[COMMAND_LINE_SIZE] = CONFIG_CMDLINE; |
| 240 | #endif |
| 241 | |
| 242 | #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE) |
| 243 | struct edd edd; |
| 244 | #ifdef CONFIG_EDD_MODULE |
| 245 | EXPORT_SYMBOL(edd); |
| 246 | #endif |
| 247 | /** |
| 248 | * copy_edd() - Copy the BIOS EDD information |
| 249 | * from boot_params into a safe place. |
| 250 | * |
| 251 | */ |
| 252 | static inline void __init copy_edd(void) |
| 253 | { |
| 254 | memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer, |
| 255 | sizeof(edd.mbr_signature)); |
| 256 | memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info)); |
| 257 | edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries; |
| 258 | edd.edd_info_nr = boot_params.eddbuf_entries; |
| 259 | } |
| 260 | #else |
| 261 | static inline void __init copy_edd(void) |
| 262 | { |
| 263 | } |
| 264 | #endif |
| 265 | |
| 266 | void * __init extend_brk(size_t size, size_t align) |
| 267 | { |
| 268 | size_t mask = align - 1; |
| 269 | void *ret; |
| 270 | |
| 271 | BUG_ON(_brk_start == 0); |
| 272 | BUG_ON(align & mask); |
| 273 | |
| 274 | _brk_end = (_brk_end + mask) & ~mask; |
| 275 | BUG_ON((char *)(_brk_end + size) > __brk_limit); |
| 276 | |
| 277 | ret = (void *)_brk_end; |
| 278 | _brk_end += size; |
| 279 | |
| 280 | memset(ret, 0, size); |
| 281 | |
| 282 | return ret; |
| 283 | } |
| 284 | |
| 285 | #ifdef CONFIG_X86_64 |
| 286 | static void __init init_gbpages(void) |
| 287 | { |
| 288 | if (direct_gbpages && cpu_has_gbpages) |
| 289 | printk(KERN_INFO "Using GB pages for direct mapping\n"); |
| 290 | else |
| 291 | direct_gbpages = 0; |
| 292 | } |
| 293 | #else |
| 294 | static inline void init_gbpages(void) |
| 295 | { |
| 296 | } |
| 297 | static void __init cleanup_highmap(void) |
| 298 | { |
| 299 | } |
| 300 | #endif |
| 301 | |
| 302 | static void __init reserve_brk(void) |
| 303 | { |
| 304 | if (_brk_end > _brk_start) |
| 305 | memblock_reserve(__pa(_brk_start), |
| 306 | __pa(_brk_end) - __pa(_brk_start)); |
| 307 | |
| 308 | /* Mark brk area as locked down and no longer taking any |
| 309 | new allocations */ |
| 310 | _brk_start = 0; |
| 311 | } |
| 312 | |
| 313 | #ifdef CONFIG_BLK_DEV_INITRD |
| 314 | |
| 315 | #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT) |
| 316 | static void __init relocate_initrd(void) |
| 317 | { |
| 318 | /* Assume only end is not page aligned */ |
| 319 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; |
| 320 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; |
| 321 | u64 area_size = PAGE_ALIGN(ramdisk_size); |
| 322 | u64 ramdisk_here; |
| 323 | unsigned long slop, clen, mapaddr; |
| 324 | char *p, *q; |
| 325 | |
| 326 | /* We need to move the initrd down into directly mapped mem */ |
| 327 | ramdisk_here = memblock_find_in_range(0, PFN_PHYS(max_low_pfn_mapped), |
| 328 | area_size, PAGE_SIZE); |
| 329 | |
| 330 | if (!ramdisk_here) |
| 331 | panic("Cannot find place for new RAMDISK of size %lld\n", |
| 332 | ramdisk_size); |
| 333 | |
| 334 | /* Note: this includes all the mem currently occupied by |
| 335 | the initrd, we rely on that fact to keep the data intact. */ |
| 336 | memblock_reserve(ramdisk_here, area_size); |
| 337 | initrd_start = ramdisk_here + PAGE_OFFSET; |
| 338 | initrd_end = initrd_start + ramdisk_size; |
| 339 | printk(KERN_INFO "Allocated new RAMDISK: [mem %#010llx-%#010llx]\n", |
| 340 | ramdisk_here, ramdisk_here + ramdisk_size - 1); |
| 341 | |
| 342 | q = (char *)initrd_start; |
| 343 | |
| 344 | /* Copy the initrd */ |
| 345 | while (ramdisk_size) { |
| 346 | slop = ramdisk_image & ~PAGE_MASK; |
| 347 | clen = ramdisk_size; |
| 348 | if (clen > MAX_MAP_CHUNK-slop) |
| 349 | clen = MAX_MAP_CHUNK-slop; |
| 350 | mapaddr = ramdisk_image & PAGE_MASK; |
| 351 | p = early_memremap(mapaddr, clen+slop); |
| 352 | memcpy(q, p+slop, clen); |
| 353 | early_iounmap(p, clen+slop); |
| 354 | q += clen; |
| 355 | ramdisk_image += clen; |
| 356 | ramdisk_size -= clen; |
| 357 | } |
| 358 | |
| 359 | ramdisk_image = boot_params.hdr.ramdisk_image; |
| 360 | ramdisk_size = boot_params.hdr.ramdisk_size; |
| 361 | printk(KERN_INFO "Move RAMDISK from [mem %#010llx-%#010llx] to" |
| 362 | " [mem %#010llx-%#010llx]\n", |
| 363 | ramdisk_image, ramdisk_image + ramdisk_size - 1, |
| 364 | ramdisk_here, ramdisk_here + ramdisk_size - 1); |
| 365 | } |
| 366 | |
| 367 | static u64 __init get_mem_size(unsigned long limit_pfn) |
| 368 | { |
| 369 | int i; |
| 370 | u64 mapped_pages = 0; |
| 371 | unsigned long start_pfn, end_pfn; |
| 372 | |
| 373 | for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, NULL) { |
| 374 | start_pfn = min_t(unsigned long, start_pfn, limit_pfn); |
| 375 | end_pfn = min_t(unsigned long, end_pfn, limit_pfn); |
| 376 | mapped_pages += end_pfn - start_pfn; |
| 377 | } |
| 378 | |
| 379 | return mapped_pages << PAGE_SHIFT; |
| 380 | } |
| 381 | static void __init reserve_initrd(void) |
| 382 | { |
| 383 | /* Assume only end is not page aligned */ |
| 384 | u64 ramdisk_image = boot_params.hdr.ramdisk_image; |
| 385 | u64 ramdisk_size = boot_params.hdr.ramdisk_size; |
| 386 | u64 ramdisk_end = PAGE_ALIGN(ramdisk_image + ramdisk_size); |
| 387 | u64 mapped_size; |
| 388 | |
| 389 | if (!boot_params.hdr.type_of_loader || |
| 390 | !ramdisk_image || !ramdisk_size) |
| 391 | return; /* No initrd provided by bootloader */ |
| 392 | |
| 393 | initrd_start = 0; |
| 394 | |
| 395 | mapped_size = get_mem_size(max_low_pfn_mapped); |
| 396 | if (ramdisk_size >= (mapped_size>>1)) |
| 397 | panic("initrd too large to handle, " |
| 398 | "disabling initrd (%lld needed, %lld available)\n", |
| 399 | ramdisk_size, mapped_size>>1); |
| 400 | |
| 401 | printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image, |
| 402 | ramdisk_end - 1); |
| 403 | |
| 404 | if (ramdisk_end <= (max_low_pfn_mapped<<PAGE_SHIFT) && |
| 405 | pfn_range_is_mapped(PFN_DOWN(ramdisk_image), |
| 406 | PFN_DOWN(ramdisk_end))) { |
| 407 | /* All are mapped, easy case */ |
| 408 | /* |
| 409 | * don't need to reserve again, already reserved early |
| 410 | * in i386_start_kernel |
| 411 | */ |
| 412 | initrd_start = ramdisk_image + PAGE_OFFSET; |
| 413 | initrd_end = initrd_start + ramdisk_size; |
| 414 | return; |
| 415 | } |
| 416 | |
| 417 | relocate_initrd(); |
| 418 | |
| 419 | memblock_free(ramdisk_image, ramdisk_end - ramdisk_image); |
| 420 | } |
| 421 | #else |
| 422 | static void __init reserve_initrd(void) |
| 423 | { |
| 424 | } |
| 425 | #endif /* CONFIG_BLK_DEV_INITRD */ |
| 426 | |
| 427 | static void __init parse_setup_data(void) |
| 428 | { |
| 429 | struct setup_data *data; |
| 430 | u64 pa_data; |
| 431 | |
| 432 | if (boot_params.hdr.version < 0x0209) |
| 433 | return; |
| 434 | pa_data = boot_params.hdr.setup_data; |
| 435 | while (pa_data) { |
| 436 | u32 data_len, map_len; |
| 437 | |
| 438 | map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK), |
| 439 | (u64)sizeof(struct setup_data)); |
| 440 | data = early_memremap(pa_data, map_len); |
| 441 | data_len = data->len + sizeof(struct setup_data); |
| 442 | if (data_len > map_len) { |
| 443 | early_iounmap(data, map_len); |
| 444 | data = early_memremap(pa_data, data_len); |
| 445 | map_len = data_len; |
| 446 | } |
| 447 | |
| 448 | switch (data->type) { |
| 449 | case SETUP_E820_EXT: |
| 450 | parse_e820_ext(data); |
| 451 | break; |
| 452 | case SETUP_DTB: |
| 453 | add_dtb(pa_data); |
| 454 | break; |
| 455 | default: |
| 456 | break; |
| 457 | } |
| 458 | pa_data = data->next; |
| 459 | early_iounmap(data, map_len); |
| 460 | } |
| 461 | } |
| 462 | |
| 463 | static void __init e820_reserve_setup_data(void) |
| 464 | { |
| 465 | struct setup_data *data; |
| 466 | u64 pa_data; |
| 467 | int found = 0; |
| 468 | |
| 469 | if (boot_params.hdr.version < 0x0209) |
| 470 | return; |
| 471 | pa_data = boot_params.hdr.setup_data; |
| 472 | while (pa_data) { |
| 473 | data = early_memremap(pa_data, sizeof(*data)); |
| 474 | e820_update_range(pa_data, sizeof(*data)+data->len, |
| 475 | E820_RAM, E820_RESERVED_KERN); |
| 476 | found = 1; |
| 477 | pa_data = data->next; |
| 478 | early_iounmap(data, sizeof(*data)); |
| 479 | } |
| 480 | if (!found) |
| 481 | return; |
| 482 | |
| 483 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); |
| 484 | memcpy(&e820_saved, &e820, sizeof(struct e820map)); |
| 485 | printk(KERN_INFO "extended physical RAM map:\n"); |
| 486 | e820_print_map("reserve setup_data"); |
| 487 | } |
| 488 | |
| 489 | static void __init memblock_x86_reserve_range_setup_data(void) |
| 490 | { |
| 491 | struct setup_data *data; |
| 492 | u64 pa_data; |
| 493 | |
| 494 | if (boot_params.hdr.version < 0x0209) |
| 495 | return; |
| 496 | pa_data = boot_params.hdr.setup_data; |
| 497 | while (pa_data) { |
| 498 | data = early_memremap(pa_data, sizeof(*data)); |
| 499 | memblock_reserve(pa_data, sizeof(*data) + data->len); |
| 500 | pa_data = data->next; |
| 501 | early_iounmap(data, sizeof(*data)); |
| 502 | } |
| 503 | } |
| 504 | |
| 505 | /* |
| 506 | * --------- Crashkernel reservation ------------------------------ |
| 507 | */ |
| 508 | |
| 509 | #ifdef CONFIG_KEXEC |
| 510 | |
| 511 | /* |
| 512 | * Keep the crash kernel below this limit. On 32 bits earlier kernels |
| 513 | * would limit the kernel to the low 512 MiB due to mapping restrictions. |
| 514 | * On 64 bits, kexec-tools currently limits us to 896 MiB; increase this |
| 515 | * limit once kexec-tools are fixed. |
| 516 | */ |
| 517 | #ifdef CONFIG_X86_32 |
| 518 | # define CRASH_KERNEL_ADDR_MAX (512 << 20) |
| 519 | #else |
| 520 | # define CRASH_KERNEL_ADDR_MAX (896 << 20) |
| 521 | #endif |
| 522 | |
| 523 | static void __init reserve_crashkernel(void) |
| 524 | { |
| 525 | unsigned long long total_mem; |
| 526 | unsigned long long crash_size, crash_base; |
| 527 | int ret; |
| 528 | |
| 529 | total_mem = memblock_phys_mem_size(); |
| 530 | |
| 531 | ret = parse_crashkernel(boot_command_line, total_mem, |
| 532 | &crash_size, &crash_base); |
| 533 | if (ret != 0 || crash_size <= 0) |
| 534 | return; |
| 535 | |
| 536 | /* 0 means: find the address automatically */ |
| 537 | if (crash_base <= 0) { |
| 538 | const unsigned long long alignment = 16<<20; /* 16M */ |
| 539 | |
| 540 | /* |
| 541 | * kexec want bzImage is below CRASH_KERNEL_ADDR_MAX |
| 542 | */ |
| 543 | crash_base = memblock_find_in_range(alignment, |
| 544 | CRASH_KERNEL_ADDR_MAX, crash_size, alignment); |
| 545 | |
| 546 | if (!crash_base) { |
| 547 | pr_info("crashkernel reservation failed - No suitable area found.\n"); |
| 548 | return; |
| 549 | } |
| 550 | } else { |
| 551 | unsigned long long start; |
| 552 | |
| 553 | start = memblock_find_in_range(crash_base, |
| 554 | crash_base + crash_size, crash_size, 1<<20); |
| 555 | if (start != crash_base) { |
| 556 | pr_info("crashkernel reservation failed - memory is in use.\n"); |
| 557 | return; |
| 558 | } |
| 559 | } |
| 560 | memblock_reserve(crash_base, crash_size); |
| 561 | |
| 562 | printk(KERN_INFO "Reserving %ldMB of memory at %ldMB " |
| 563 | "for crashkernel (System RAM: %ldMB)\n", |
| 564 | (unsigned long)(crash_size >> 20), |
| 565 | (unsigned long)(crash_base >> 20), |
| 566 | (unsigned long)(total_mem >> 20)); |
| 567 | |
| 568 | crashk_res.start = crash_base; |
| 569 | crashk_res.end = crash_base + crash_size - 1; |
| 570 | insert_resource(&iomem_resource, &crashk_res); |
| 571 | } |
| 572 | #else |
| 573 | static void __init reserve_crashkernel(void) |
| 574 | { |
| 575 | } |
| 576 | #endif |
| 577 | |
| 578 | static struct resource standard_io_resources[] = { |
| 579 | { .name = "dma1", .start = 0x00, .end = 0x1f, |
| 580 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 581 | { .name = "pic1", .start = 0x20, .end = 0x21, |
| 582 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 583 | { .name = "timer0", .start = 0x40, .end = 0x43, |
| 584 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 585 | { .name = "timer1", .start = 0x50, .end = 0x53, |
| 586 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 587 | { .name = "keyboard", .start = 0x60, .end = 0x60, |
| 588 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 589 | { .name = "keyboard", .start = 0x64, .end = 0x64, |
| 590 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 591 | { .name = "dma page reg", .start = 0x80, .end = 0x8f, |
| 592 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 593 | { .name = "pic2", .start = 0xa0, .end = 0xa1, |
| 594 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 595 | { .name = "dma2", .start = 0xc0, .end = 0xdf, |
| 596 | .flags = IORESOURCE_BUSY | IORESOURCE_IO }, |
| 597 | { .name = "fpu", .start = 0xf0, .end = 0xff, |
| 598 | .flags = IORESOURCE_BUSY | IORESOURCE_IO } |
| 599 | }; |
| 600 | |
| 601 | void __init reserve_standard_io_resources(void) |
| 602 | { |
| 603 | int i; |
| 604 | |
| 605 | /* request I/O space for devices used on all i[345]86 PCs */ |
| 606 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) |
| 607 | request_resource(&ioport_resource, &standard_io_resources[i]); |
| 608 | |
| 609 | } |
| 610 | |
| 611 | static __init void reserve_ibft_region(void) |
| 612 | { |
| 613 | unsigned long addr, size = 0; |
| 614 | |
| 615 | addr = find_ibft_region(&size); |
| 616 | |
| 617 | if (size) |
| 618 | memblock_reserve(addr, size); |
| 619 | } |
| 620 | |
| 621 | static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10; |
| 622 | |
| 623 | static void __init trim_bios_range(void) |
| 624 | { |
| 625 | /* |
| 626 | * A special case is the first 4Kb of memory; |
| 627 | * This is a BIOS owned area, not kernel ram, but generally |
| 628 | * not listed as such in the E820 table. |
| 629 | * |
| 630 | * This typically reserves additional memory (64KiB by default) |
| 631 | * since some BIOSes are known to corrupt low memory. See the |
| 632 | * Kconfig help text for X86_RESERVE_LOW. |
| 633 | */ |
| 634 | e820_update_range(0, ALIGN(reserve_low, PAGE_SIZE), |
| 635 | E820_RAM, E820_RESERVED); |
| 636 | |
| 637 | /* |
| 638 | * special case: Some BIOSen report the PC BIOS |
| 639 | * area (640->1Mb) as ram even though it is not. |
| 640 | * take them out. |
| 641 | */ |
| 642 | e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1); |
| 643 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); |
| 644 | } |
| 645 | |
| 646 | static int __init parse_reservelow(char *p) |
| 647 | { |
| 648 | unsigned long long size; |
| 649 | |
| 650 | if (!p) |
| 651 | return -EINVAL; |
| 652 | |
| 653 | size = memparse(p, &p); |
| 654 | |
| 655 | if (size < 4096) |
| 656 | size = 4096; |
| 657 | |
| 658 | if (size > 640*1024) |
| 659 | size = 640*1024; |
| 660 | |
| 661 | reserve_low = size; |
| 662 | |
| 663 | return 0; |
| 664 | } |
| 665 | |
| 666 | early_param("reservelow", parse_reservelow); |
| 667 | |
| 668 | /* |
| 669 | * Determine if we were loaded by an EFI loader. If so, then we have also been |
| 670 | * passed the efi memmap, systab, etc., so we should use these data structures |
| 671 | * for initialization. Note, the efi init code path is determined by the |
| 672 | * global efi_enabled. This allows the same kernel image to be used on existing |
| 673 | * systems (with a traditional BIOS) as well as on EFI systems. |
| 674 | */ |
| 675 | /* |
| 676 | * setup_arch - architecture-specific boot-time initializations |
| 677 | * |
| 678 | * Note: On x86_64, fixmaps are ready for use even before this is called. |
| 679 | */ |
| 680 | |
| 681 | void __init setup_arch(char **cmdline_p) |
| 682 | { |
| 683 | #ifdef CONFIG_X86_32 |
| 684 | memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data)); |
| 685 | visws_early_detect(); |
| 686 | |
| 687 | /* |
| 688 | * copy kernel address range established so far and switch |
| 689 | * to the proper swapper page table |
| 690 | */ |
| 691 | clone_pgd_range(swapper_pg_dir + KERNEL_PGD_BOUNDARY, |
| 692 | initial_page_table + KERNEL_PGD_BOUNDARY, |
| 693 | KERNEL_PGD_PTRS); |
| 694 | |
| 695 | load_cr3(swapper_pg_dir); |
| 696 | __flush_tlb_all(); |
| 697 | #else |
| 698 | printk(KERN_INFO "Command line: %s\n", boot_command_line); |
| 699 | #endif |
| 700 | |
| 701 | /* |
| 702 | * If we have OLPC OFW, we might end up relocating the fixmap due to |
| 703 | * reserve_top(), so do this before touching the ioremap area. |
| 704 | */ |
| 705 | olpc_ofw_detect(); |
| 706 | |
| 707 | early_trap_init(); |
| 708 | early_cpu_init(); |
| 709 | early_ioremap_init(); |
| 710 | |
| 711 | setup_olpc_ofw_pgd(); |
| 712 | |
| 713 | ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev); |
| 714 | screen_info = boot_params.screen_info; |
| 715 | edid_info = boot_params.edid_info; |
| 716 | #ifdef CONFIG_X86_32 |
| 717 | apm_info.bios = boot_params.apm_bios_info; |
| 718 | ist_info = boot_params.ist_info; |
| 719 | if (boot_params.sys_desc_table.length != 0) { |
| 720 | machine_id = boot_params.sys_desc_table.table[0]; |
| 721 | machine_submodel_id = boot_params.sys_desc_table.table[1]; |
| 722 | BIOS_revision = boot_params.sys_desc_table.table[2]; |
| 723 | } |
| 724 | #endif |
| 725 | saved_video_mode = boot_params.hdr.vid_mode; |
| 726 | bootloader_type = boot_params.hdr.type_of_loader; |
| 727 | if ((bootloader_type >> 4) == 0xe) { |
| 728 | bootloader_type &= 0xf; |
| 729 | bootloader_type |= (boot_params.hdr.ext_loader_type+0x10) << 4; |
| 730 | } |
| 731 | bootloader_version = bootloader_type & 0xf; |
| 732 | bootloader_version |= boot_params.hdr.ext_loader_ver << 4; |
| 733 | |
| 734 | #ifdef CONFIG_BLK_DEV_RAM |
| 735 | rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK; |
| 736 | rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0); |
| 737 | rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0); |
| 738 | #endif |
| 739 | #ifdef CONFIG_EFI |
| 740 | if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, |
| 741 | "EL32", 4)) { |
| 742 | efi_enabled = 1; |
| 743 | efi_64bit = false; |
| 744 | } else if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature, |
| 745 | "EL64", 4)) { |
| 746 | efi_enabled = 1; |
| 747 | efi_64bit = true; |
| 748 | } |
| 749 | if (efi_enabled && efi_memblock_x86_reserve_range()) |
| 750 | efi_enabled = 0; |
| 751 | #endif |
| 752 | |
| 753 | x86_init.oem.arch_setup(); |
| 754 | |
| 755 | iomem_resource.end = (1ULL << boot_cpu_data.x86_phys_bits) - 1; |
| 756 | setup_memory_map(); |
| 757 | parse_setup_data(); |
| 758 | /* update the e820_saved too */ |
| 759 | e820_reserve_setup_data(); |
| 760 | |
| 761 | copy_edd(); |
| 762 | |
| 763 | if (!boot_params.hdr.root_flags) |
| 764 | root_mountflags &= ~MS_RDONLY; |
| 765 | init_mm.start_code = (unsigned long) _text; |
| 766 | init_mm.end_code = (unsigned long) _etext; |
| 767 | init_mm.end_data = (unsigned long) _edata; |
| 768 | init_mm.brk = _brk_end; |
| 769 | |
| 770 | code_resource.start = virt_to_phys(_text); |
| 771 | code_resource.end = virt_to_phys(_etext)-1; |
| 772 | data_resource.start = virt_to_phys(_etext); |
| 773 | data_resource.end = virt_to_phys(_edata)-1; |
| 774 | bss_resource.start = virt_to_phys(&__bss_start); |
| 775 | bss_resource.end = virt_to_phys(&__bss_stop)-1; |
| 776 | |
| 777 | #ifdef CONFIG_CMDLINE_BOOL |
| 778 | #ifdef CONFIG_CMDLINE_OVERRIDE |
| 779 | strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); |
| 780 | #else |
| 781 | if (builtin_cmdline[0]) { |
| 782 | /* append boot loader cmdline to builtin */ |
| 783 | strlcat(builtin_cmdline, " ", COMMAND_LINE_SIZE); |
| 784 | strlcat(builtin_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
| 785 | strlcpy(boot_command_line, builtin_cmdline, COMMAND_LINE_SIZE); |
| 786 | } |
| 787 | #endif |
| 788 | #endif |
| 789 | |
| 790 | strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE); |
| 791 | *cmdline_p = command_line; |
| 792 | |
| 793 | /* |
| 794 | * x86_configure_nx() is called before parse_early_param() to detect |
| 795 | * whether hardware doesn't support NX (so that the early EHCI debug |
| 796 | * console setup can safely call set_fixmap()). It may then be called |
| 797 | * again from within noexec_setup() during parsing early parameters |
| 798 | * to honor the respective command line option. |
| 799 | */ |
| 800 | x86_configure_nx(); |
| 801 | |
| 802 | parse_early_param(); |
| 803 | |
| 804 | x86_report_nx(); |
| 805 | |
| 806 | /* after early param, so could get panic from serial */ |
| 807 | memblock_x86_reserve_range_setup_data(); |
| 808 | |
| 809 | if (acpi_mps_check()) { |
| 810 | #ifdef CONFIG_X86_LOCAL_APIC |
| 811 | disable_apic = 1; |
| 812 | #endif |
| 813 | setup_clear_cpu_cap(X86_FEATURE_APIC); |
| 814 | } |
| 815 | |
| 816 | #ifdef CONFIG_PCI |
| 817 | if (pci_early_dump_regs) |
| 818 | early_dump_pci_devices(); |
| 819 | #endif |
| 820 | |
| 821 | finish_e820_parsing(); |
| 822 | |
| 823 | if (efi_enabled) |
| 824 | efi_init(); |
| 825 | |
| 826 | dmi_scan_machine(); |
| 827 | |
| 828 | /* |
| 829 | * VMware detection requires dmi to be available, so this |
| 830 | * needs to be done after dmi_scan_machine, for the BP. |
| 831 | */ |
| 832 | init_hypervisor_platform(); |
| 833 | |
| 834 | x86_init.resources.probe_roms(); |
| 835 | |
| 836 | /* after parse_early_param, so could debug it */ |
| 837 | insert_resource(&iomem_resource, &code_resource); |
| 838 | insert_resource(&iomem_resource, &data_resource); |
| 839 | insert_resource(&iomem_resource, &bss_resource); |
| 840 | |
| 841 | /* |
| 842 | * Complain if .text .data and .bss are not marked as E820_RAM and |
| 843 | * attempt to fix it by adding the range. We may have a confused BIOS, |
| 844 | * or the user may have incorrectly supplied it via memmap=exactmap. If |
| 845 | * we really are running on top non-RAM, we will crash later anyways. |
| 846 | */ |
| 847 | if (!e820_all_mapped(code_resource.start, __pa(__brk_limit), E820_RAM)) { |
| 848 | pr_warn(".text .data .bss are not marked as E820_RAM!\n"); |
| 849 | |
| 850 | e820_add_region(code_resource.start, |
| 851 | __pa(__brk_limit) - code_resource.start + 1, |
| 852 | E820_RAM); |
| 853 | } |
| 854 | |
| 855 | trim_bios_range(); |
| 856 | #ifdef CONFIG_X86_32 |
| 857 | if (ppro_with_ram_bug()) { |
| 858 | e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM, |
| 859 | E820_RESERVED); |
| 860 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); |
| 861 | printk(KERN_INFO "fixed physical RAM map:\n"); |
| 862 | e820_print_map("bad_ppro"); |
| 863 | } |
| 864 | #else |
| 865 | early_gart_iommu_check(); |
| 866 | #endif |
| 867 | |
| 868 | /* |
| 869 | * partially used pages are not usable - thus |
| 870 | * we are rounding upwards: |
| 871 | */ |
| 872 | max_pfn = e820_end_of_ram_pfn(); |
| 873 | |
| 874 | /* update e820 for memory not covered by WB MTRRs */ |
| 875 | mtrr_bp_init(); |
| 876 | if (mtrr_trim_uncached_memory(max_pfn)) |
| 877 | max_pfn = e820_end_of_ram_pfn(); |
| 878 | |
| 879 | #ifdef CONFIG_X86_32 |
| 880 | /* max_low_pfn get updated here */ |
| 881 | find_low_pfn_range(); |
| 882 | #else |
| 883 | num_physpages = max_pfn; |
| 884 | |
| 885 | check_x2apic(); |
| 886 | |
| 887 | /* How many end-of-memory variables you have, grandma! */ |
| 888 | /* need this before calling reserve_initrd */ |
| 889 | if (max_pfn > (1UL<<(32 - PAGE_SHIFT))) |
| 890 | max_low_pfn = e820_end_of_low_ram_pfn(); |
| 891 | else |
| 892 | max_low_pfn = max_pfn; |
| 893 | |
| 894 | high_memory = (void *)__va(max_pfn * PAGE_SIZE - 1) + 1; |
| 895 | #endif |
| 896 | |
| 897 | /* |
| 898 | * Find and reserve possible boot-time SMP configuration: |
| 899 | */ |
| 900 | find_smp_config(); |
| 901 | |
| 902 | reserve_ibft_region(); |
| 903 | |
| 904 | /* |
| 905 | * Need to conclude brk, before memblock_x86_fill() |
| 906 | * it could use memblock_find_in_range, could overlap with |
| 907 | * brk area. |
| 908 | */ |
| 909 | reserve_brk(); |
| 910 | |
| 911 | cleanup_highmap(); |
| 912 | |
| 913 | memblock.current_limit = ISA_END_ADDRESS; |
| 914 | memblock_x86_fill(); |
| 915 | |
| 916 | /* |
| 917 | * The EFI specification says that boot service code won't be called |
| 918 | * after ExitBootServices(). This is, in fact, a lie. |
| 919 | */ |
| 920 | if (efi_enabled) |
| 921 | efi_reserve_boot_services(); |
| 922 | |
| 923 | /* preallocate 4k for mptable mpc */ |
| 924 | early_reserve_e820_mpc_new(); |
| 925 | |
| 926 | #ifdef CONFIG_X86_CHECK_BIOS_CORRUPTION |
| 927 | setup_bios_corruption_check(); |
| 928 | #endif |
| 929 | |
| 930 | printk(KERN_DEBUG "initial memory mapped: [mem 0x00000000-%#010lx]\n", |
| 931 | (max_pfn_mapped<<PAGE_SHIFT) - 1); |
| 932 | |
| 933 | setup_real_mode(); |
| 934 | |
| 935 | init_gbpages(); |
| 936 | |
| 937 | init_mem_mapping(); |
| 938 | |
| 939 | memblock.current_limit = get_max_mapped(); |
| 940 | dma_contiguous_reserve(0); |
| 941 | |
| 942 | /* |
| 943 | * NOTE: On x86-32, only from this point on, fixmaps are ready for use. |
| 944 | */ |
| 945 | |
| 946 | #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT |
| 947 | if (init_ohci1394_dma_early) |
| 948 | init_ohci1394_dma_on_all_controllers(); |
| 949 | #endif |
| 950 | /* Allocate bigger log buffer */ |
| 951 | setup_log_buf(1); |
| 952 | |
| 953 | reserve_initrd(); |
| 954 | |
| 955 | reserve_crashkernel(); |
| 956 | |
| 957 | vsmp_init(); |
| 958 | |
| 959 | io_delay_init(); |
| 960 | |
| 961 | /* |
| 962 | * Parse the ACPI tables for possible boot-time SMP configuration. |
| 963 | */ |
| 964 | acpi_boot_table_init(); |
| 965 | |
| 966 | early_acpi_boot_init(); |
| 967 | |
| 968 | initmem_init(); |
| 969 | memblock_find_dma_reserve(); |
| 970 | |
| 971 | #ifdef CONFIG_KVM_GUEST |
| 972 | kvmclock_init(); |
| 973 | #endif |
| 974 | |
| 975 | x86_init.paging.pagetable_init(); |
| 976 | |
| 977 | if (boot_cpu_data.cpuid_level >= 0) { |
| 978 | /* A CPU has %cr4 if and only if it has CPUID */ |
| 979 | mmu_cr4_features = read_cr4(); |
| 980 | if (trampoline_cr4_features) |
| 981 | *trampoline_cr4_features = mmu_cr4_features; |
| 982 | } |
| 983 | |
| 984 | #ifdef CONFIG_X86_32 |
| 985 | /* sync back kernel address range */ |
| 986 | clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY, |
| 987 | swapper_pg_dir + KERNEL_PGD_BOUNDARY, |
| 988 | KERNEL_PGD_PTRS); |
| 989 | #endif |
| 990 | |
| 991 | tboot_probe(); |
| 992 | |
| 993 | #ifdef CONFIG_X86_64 |
| 994 | map_vsyscall(); |
| 995 | #endif |
| 996 | |
| 997 | generic_apic_probe(); |
| 998 | |
| 999 | early_quirks(); |
| 1000 | |
| 1001 | /* |
| 1002 | * Read APIC and some other early information from ACPI tables. |
| 1003 | */ |
| 1004 | acpi_boot_init(); |
| 1005 | sfi_init(); |
| 1006 | x86_dtb_init(); |
| 1007 | |
| 1008 | /* |
| 1009 | * get boot-time SMP configuration: |
| 1010 | */ |
| 1011 | if (smp_found_config) |
| 1012 | get_smp_config(); |
| 1013 | |
| 1014 | prefill_possible_map(); |
| 1015 | |
| 1016 | init_cpu_to_node(); |
| 1017 | |
| 1018 | init_apic_mappings(); |
| 1019 | if (x86_io_apic_ops.init) |
| 1020 | x86_io_apic_ops.init(); |
| 1021 | |
| 1022 | kvm_guest_init(); |
| 1023 | |
| 1024 | e820_reserve_resources(); |
| 1025 | e820_mark_nosave_regions(max_low_pfn); |
| 1026 | |
| 1027 | x86_init.resources.reserve_resources(); |
| 1028 | |
| 1029 | e820_setup_gap(); |
| 1030 | |
| 1031 | #ifdef CONFIG_VT |
| 1032 | #if defined(CONFIG_VGA_CONSOLE) |
| 1033 | if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY)) |
| 1034 | conswitchp = &vga_con; |
| 1035 | #elif defined(CONFIG_DUMMY_CONSOLE) |
| 1036 | conswitchp = &dummy_con; |
| 1037 | #endif |
| 1038 | #endif |
| 1039 | x86_init.oem.banner(); |
| 1040 | |
| 1041 | x86_init.timers.wallclock_init(); |
| 1042 | |
| 1043 | mcheck_init(); |
| 1044 | |
| 1045 | arch_init_ideal_nops(); |
| 1046 | |
| 1047 | register_refined_jiffies(CLOCK_TICK_RATE); |
| 1048 | |
| 1049 | #ifdef CONFIG_EFI |
| 1050 | /* Once setup is done above, disable efi_enabled on mismatched |
| 1051 | * firmware/kernel archtectures since there is no support for |
| 1052 | * runtime services. |
| 1053 | */ |
| 1054 | if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) { |
| 1055 | pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n"); |
| 1056 | efi_unmap_memmap(); |
| 1057 | efi_enabled = 0; |
| 1058 | } |
| 1059 | #endif |
| 1060 | } |
| 1061 | |
| 1062 | #ifdef CONFIG_X86_32 |
| 1063 | |
| 1064 | static struct resource video_ram_resource = { |
| 1065 | .name = "Video RAM area", |
| 1066 | .start = 0xa0000, |
| 1067 | .end = 0xbffff, |
| 1068 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM |
| 1069 | }; |
| 1070 | |
| 1071 | void __init i386_reserve_resources(void) |
| 1072 | { |
| 1073 | request_resource(&iomem_resource, &video_ram_resource); |
| 1074 | reserve_standard_io_resources(); |
| 1075 | } |
| 1076 | |
| 1077 | #endif /* CONFIG_X86_32 */ |