1 /* 2 * Kexec bzImage loader 3 * 4 * Copyright (C) 2014 Red Hat Inc. 5 * Authors: 6 * Vivek Goyal <vgoyal@redhat.com> 7 * 8 * This source code is licensed under the GNU General Public License, 9 * Version 2. See the file COPYING for more details. 10 */ 11 12 #define pr_fmt(fmt) "kexec-bzImage64: " fmt 13 14 #include <linux/string.h> 15 #include <linux/printk.h> 16 #include <linux/errno.h> 17 #include <linux/slab.h> 18 #include <linux/kexec.h> 19 #include <linux/kernel.h> 20 #include <linux/mm.h> 21 #include <linux/efi.h> 22 #include <linux/verify_pefile.h> 23 #include <keys/system_keyring.h> 24 25 #include <asm/bootparam.h> 26 #include <asm/setup.h> 27 #include <asm/crash.h> 28 #include <asm/efi.h> 29 30 #define MAX_ELFCOREHDR_STR_LEN 30 /* elfcorehdr=0x<64bit-value> */ 31 32 /* 33 * Defines lowest physical address for various segments. Not sure where 34 * exactly these limits came from. Current bzimage64 loader in kexec-tools 35 * uses these so I am retaining it. It can be changed over time as we gain 36 * more insight. 37 */ 38 #define MIN_PURGATORY_ADDR 0x3000 39 #define MIN_BOOTPARAM_ADDR 0x3000 40 #define MIN_KERNEL_LOAD_ADDR 0x100000 41 #define MIN_INITRD_LOAD_ADDR 0x1000000 42 43 /* 44 * This is a place holder for all boot loader specific data structure which 45 * gets allocated in one call but gets freed much later during cleanup 46 * time. Right now there is only one field but it can grow as need be. 47 */ 48 struct bzimage64_data { 49 /* 50 * Temporary buffer to hold bootparams buffer. This should be 51 * freed once the bootparam segment has been loaded. 52 */ 53 void *bootparams_buf; 54 }; 55 56 static int setup_initrd(struct boot_params *params, 57 unsigned long initrd_load_addr, unsigned long initrd_len) 58 { 59 params->hdr.ramdisk_image = initrd_load_addr & 0xffffffffUL; 60 params->hdr.ramdisk_size = initrd_len & 0xffffffffUL; 61 62 params->ext_ramdisk_image = initrd_load_addr >> 32; 63 params->ext_ramdisk_size = initrd_len >> 32; 64 65 return 0; 66 } 67 68 static int setup_cmdline(struct kimage *image, struct boot_params *params, 69 unsigned long bootparams_load_addr, 70 unsigned long cmdline_offset, char *cmdline, 71 unsigned long cmdline_len) 72 { 73 char *cmdline_ptr = ((char *)params) + cmdline_offset; 74 unsigned long cmdline_ptr_phys, len; 75 uint32_t cmdline_low_32, cmdline_ext_32; 76 77 memcpy(cmdline_ptr, cmdline, cmdline_len); 78 if (image->type == KEXEC_TYPE_CRASH) { 79 len = sprintf(cmdline_ptr + cmdline_len - 1, 80 " elfcorehdr=0x%lx", image->arch.elf_load_addr); 81 cmdline_len += len; 82 } 83 cmdline_ptr[cmdline_len - 1] = '\0'; 84 85 pr_debug("Final command line is: %s\n", cmdline_ptr); 86 cmdline_ptr_phys = bootparams_load_addr + cmdline_offset; 87 cmdline_low_32 = cmdline_ptr_phys & 0xffffffffUL; 88 cmdline_ext_32 = cmdline_ptr_phys >> 32; 89 90 params->hdr.cmd_line_ptr = cmdline_low_32; 91 if (cmdline_ext_32) 92 params->ext_cmd_line_ptr = cmdline_ext_32; 93 94 return 0; 95 } 96 97 static int setup_e820_entries(struct boot_params *params) 98 { 99 unsigned int nr_e820_entries; 100 101 nr_e820_entries = e820_saved.nr_map; 102 103 /* TODO: Pass entries more than E820MAX in bootparams setup data */ 104 if (nr_e820_entries > E820MAX) 105 nr_e820_entries = E820MAX; 106 107 params->e820_entries = nr_e820_entries; 108 memcpy(¶ms->e820_map, &e820_saved.map, 109 nr_e820_entries * sizeof(struct e820entry)); 110 111 return 0; 112 } 113 114 #ifdef CONFIG_EFI 115 static int setup_efi_info_memmap(struct boot_params *params, 116 unsigned long params_load_addr, 117 unsigned int efi_map_offset, 118 unsigned int efi_map_sz) 119 { 120 void *efi_map = (void *)params + efi_map_offset; 121 unsigned long efi_map_phys_addr = params_load_addr + efi_map_offset; 122 struct efi_info *ei = ¶ms->efi_info; 123 124 if (!efi_map_sz) 125 return 0; 126 127 efi_runtime_map_copy(efi_map, efi_map_sz); 128 129 ei->efi_memmap = efi_map_phys_addr & 0xffffffff; 130 ei->efi_memmap_hi = efi_map_phys_addr >> 32; 131 ei->efi_memmap_size = efi_map_sz; 132 133 return 0; 134 } 135 136 static int 137 prepare_add_efi_setup_data(struct boot_params *params, 138 unsigned long params_load_addr, 139 unsigned int efi_setup_data_offset) 140 { 141 unsigned long setup_data_phys; 142 struct setup_data *sd = (void *)params + efi_setup_data_offset; 143 struct efi_setup_data *esd = (void *)sd + sizeof(struct setup_data); 144 145 esd->fw_vendor = efi.fw_vendor; 146 esd->runtime = efi.runtime; 147 esd->tables = efi.config_table; 148 esd->smbios = efi.smbios; 149 150 sd->type = SETUP_EFI; 151 sd->len = sizeof(struct efi_setup_data); 152 153 /* Add setup data */ 154 setup_data_phys = params_load_addr + efi_setup_data_offset; 155 sd->next = params->hdr.setup_data; 156 params->hdr.setup_data = setup_data_phys; 157 158 return 0; 159 } 160 161 static int 162 setup_efi_state(struct boot_params *params, unsigned long params_load_addr, 163 unsigned int efi_map_offset, unsigned int efi_map_sz, 164 unsigned int efi_setup_data_offset) 165 { 166 struct efi_info *current_ei = &boot_params.efi_info; 167 struct efi_info *ei = ¶ms->efi_info; 168 169 if (!current_ei->efi_memmap_size) 170 return 0; 171 172 /* 173 * If 1:1 mapping is not enabled, second kernel can not setup EFI 174 * and use EFI run time services. User space will have to pass 175 * acpi_rsdp=<addr> on kernel command line to make second kernel boot 176 * without efi. 177 */ 178 if (efi_enabled(EFI_OLD_MEMMAP)) 179 return 0; 180 181 ei->efi_loader_signature = current_ei->efi_loader_signature; 182 ei->efi_systab = current_ei->efi_systab; 183 ei->efi_systab_hi = current_ei->efi_systab_hi; 184 185 ei->efi_memdesc_version = current_ei->efi_memdesc_version; 186 ei->efi_memdesc_size = efi_get_runtime_map_desc_size(); 187 188 setup_efi_info_memmap(params, params_load_addr, efi_map_offset, 189 efi_map_sz); 190 prepare_add_efi_setup_data(params, params_load_addr, 191 efi_setup_data_offset); 192 return 0; 193 } 194 #endif /* CONFIG_EFI */ 195 196 static int 197 setup_boot_parameters(struct kimage *image, struct boot_params *params, 198 unsigned long params_load_addr, 199 unsigned int efi_map_offset, unsigned int efi_map_sz, 200 unsigned int efi_setup_data_offset) 201 { 202 unsigned int nr_e820_entries; 203 unsigned long long mem_k, start, end; 204 int i, ret = 0; 205 206 /* Get subarch from existing bootparams */ 207 params->hdr.hardware_subarch = boot_params.hdr.hardware_subarch; 208 209 /* Copying screen_info will do? */ 210 memcpy(¶ms->screen_info, &boot_params.screen_info, 211 sizeof(struct screen_info)); 212 213 /* Fill in memsize later */ 214 params->screen_info.ext_mem_k = 0; 215 params->alt_mem_k = 0; 216 217 /* Default APM info */ 218 memset(¶ms->apm_bios_info, 0, sizeof(params->apm_bios_info)); 219 220 /* Default drive info */ 221 memset(¶ms->hd0_info, 0, sizeof(params->hd0_info)); 222 memset(¶ms->hd1_info, 0, sizeof(params->hd1_info)); 223 224 /* Default sysdesc table */ 225 params->sys_desc_table.length = 0; 226 227 if (image->type == KEXEC_TYPE_CRASH) { 228 ret = crash_setup_memmap_entries(image, params); 229 if (ret) 230 return ret; 231 } else 232 setup_e820_entries(params); 233 234 nr_e820_entries = params->e820_entries; 235 236 for (i = 0; i < nr_e820_entries; i++) { 237 if (params->e820_map[i].type != E820_RAM) 238 continue; 239 start = params->e820_map[i].addr; 240 end = params->e820_map[i].addr + params->e820_map[i].size - 1; 241 242 if ((start <= 0x100000) && end > 0x100000) { 243 mem_k = (end >> 10) - (0x100000 >> 10); 244 params->screen_info.ext_mem_k = mem_k; 245 params->alt_mem_k = mem_k; 246 if (mem_k > 0xfc00) 247 params->screen_info.ext_mem_k = 0xfc00; /* 64M*/ 248 if (mem_k > 0xffffffff) 249 params->alt_mem_k = 0xffffffff; 250 } 251 } 252 253 #ifdef CONFIG_EFI 254 /* Setup EFI state */ 255 setup_efi_state(params, params_load_addr, efi_map_offset, efi_map_sz, 256 efi_setup_data_offset); 257 #endif 258 259 /* Setup EDD info */ 260 memcpy(params->eddbuf, boot_params.eddbuf, 261 EDDMAXNR * sizeof(struct edd_info)); 262 params->eddbuf_entries = boot_params.eddbuf_entries; 263 264 memcpy(params->edd_mbr_sig_buffer, boot_params.edd_mbr_sig_buffer, 265 EDD_MBR_SIG_MAX * sizeof(unsigned int)); 266 267 return ret; 268 } 269 270 int bzImage64_probe(const char *buf, unsigned long len) 271 { 272 int ret = -ENOEXEC; 273 struct setup_header *header; 274 275 /* kernel should be atleast two sectors long */ 276 if (len < 2 * 512) { 277 pr_err("File is too short to be a bzImage\n"); 278 return ret; 279 } 280 281 header = (struct setup_header *)(buf + offsetof(struct boot_params, hdr)); 282 if (memcmp((char *)&header->header, "HdrS", 4) != 0) { 283 pr_err("Not a bzImage\n"); 284 return ret; 285 } 286 287 if (header->boot_flag != 0xAA55) { 288 pr_err("No x86 boot sector present\n"); 289 return ret; 290 } 291 292 if (header->version < 0x020C) { 293 pr_err("Must be at least protocol version 2.12\n"); 294 return ret; 295 } 296 297 if (!(header->loadflags & LOADED_HIGH)) { 298 pr_err("zImage not a bzImage\n"); 299 return ret; 300 } 301 302 if (!(header->xloadflags & XLF_KERNEL_64)) { 303 pr_err("Not a bzImage64. XLF_KERNEL_64 is not set.\n"); 304 return ret; 305 } 306 307 if (!(header->xloadflags & XLF_CAN_BE_LOADED_ABOVE_4G)) { 308 pr_err("XLF_CAN_BE_LOADED_ABOVE_4G is not set.\n"); 309 return ret; 310 } 311 312 /* 313 * Can't handle 32bit EFI as it does not allow loading kernel 314 * above 4G. This should be handled by 32bit bzImage loader 315 */ 316 if (efi_enabled(EFI_RUNTIME_SERVICES) && !efi_enabled(EFI_64BIT)) { 317 pr_debug("EFI is 32 bit. Can't load kernel above 4G.\n"); 318 return ret; 319 } 320 321 /* I've got a bzImage */ 322 pr_debug("It's a relocatable bzImage64\n"); 323 ret = 0; 324 325 return ret; 326 } 327 328 void *bzImage64_load(struct kimage *image, char *kernel, 329 unsigned long kernel_len, char *initrd, 330 unsigned long initrd_len, char *cmdline, 331 unsigned long cmdline_len) 332 { 333 334 struct setup_header *header; 335 int setup_sects, kern16_size, ret = 0; 336 unsigned long setup_header_size, params_cmdline_sz, params_misc_sz; 337 struct boot_params *params; 338 unsigned long bootparam_load_addr, kernel_load_addr, initrd_load_addr; 339 unsigned long purgatory_load_addr; 340 unsigned long kernel_bufsz, kernel_memsz, kernel_align; 341 char *kernel_buf; 342 struct bzimage64_data *ldata; 343 struct kexec_entry64_regs regs64; 344 void *stack; 345 unsigned int setup_hdr_offset = offsetof(struct boot_params, hdr); 346 unsigned int efi_map_offset, efi_map_sz, efi_setup_data_offset; 347 348 header = (struct setup_header *)(kernel + setup_hdr_offset); 349 setup_sects = header->setup_sects; 350 if (setup_sects == 0) 351 setup_sects = 4; 352 353 kern16_size = (setup_sects + 1) * 512; 354 if (kernel_len < kern16_size) { 355 pr_err("bzImage truncated\n"); 356 return ERR_PTR(-ENOEXEC); 357 } 358 359 if (cmdline_len > header->cmdline_size) { 360 pr_err("Kernel command line too long\n"); 361 return ERR_PTR(-EINVAL); 362 } 363 364 /* 365 * In case of crash dump, we will append elfcorehdr=<addr> to 366 * command line. Make sure it does not overflow 367 */ 368 if (cmdline_len + MAX_ELFCOREHDR_STR_LEN > header->cmdline_size) { 369 pr_debug("Appending elfcorehdr=<addr> to command line exceeds maximum allowed length\n"); 370 return ERR_PTR(-EINVAL); 371 } 372 373 /* Allocate and load backup region */ 374 if (image->type == KEXEC_TYPE_CRASH) { 375 ret = crash_load_segments(image); 376 if (ret) 377 return ERR_PTR(ret); 378 } 379 380 /* 381 * Load purgatory. For 64bit entry point, purgatory code can be 382 * anywhere. 383 */ 384 ret = kexec_load_purgatory(image, MIN_PURGATORY_ADDR, ULONG_MAX, 1, 385 &purgatory_load_addr); 386 if (ret) { 387 pr_err("Loading purgatory failed\n"); 388 return ERR_PTR(ret); 389 } 390 391 pr_debug("Loaded purgatory at 0x%lx\n", purgatory_load_addr); 392 393 394 /* 395 * Load Bootparams and cmdline and space for efi stuff. 396 * 397 * Allocate memory together for multiple data structures so 398 * that they all can go in single area/segment and we don't 399 * have to create separate segment for each. Keeps things 400 * little bit simple 401 */ 402 efi_map_sz = efi_get_runtime_map_size(); 403 efi_map_sz = ALIGN(efi_map_sz, 16); 404 params_cmdline_sz = sizeof(struct boot_params) + cmdline_len + 405 MAX_ELFCOREHDR_STR_LEN; 406 params_cmdline_sz = ALIGN(params_cmdline_sz, 16); 407 params_misc_sz = params_cmdline_sz + efi_map_sz + 408 sizeof(struct setup_data) + 409 sizeof(struct efi_setup_data); 410 411 params = kzalloc(params_misc_sz, GFP_KERNEL); 412 if (!params) 413 return ERR_PTR(-ENOMEM); 414 efi_map_offset = params_cmdline_sz; 415 efi_setup_data_offset = efi_map_offset + efi_map_sz; 416 417 /* Copy setup header onto bootparams. Documentation/x86/boot.txt */ 418 setup_header_size = 0x0202 + kernel[0x0201] - setup_hdr_offset; 419 420 /* Is there a limit on setup header size? */ 421 memcpy(¶ms->hdr, (kernel + setup_hdr_offset), setup_header_size); 422 423 ret = kexec_add_buffer(image, (char *)params, params_misc_sz, 424 params_misc_sz, 16, MIN_BOOTPARAM_ADDR, 425 ULONG_MAX, 1, &bootparam_load_addr); 426 if (ret) 427 goto out_free_params; 428 pr_debug("Loaded boot_param, command line and misc at 0x%lx bufsz=0x%lx memsz=0x%lx\n", 429 bootparam_load_addr, params_misc_sz, params_misc_sz); 430 431 /* Load kernel */ 432 kernel_buf = kernel + kern16_size; 433 kernel_bufsz = kernel_len - kern16_size; 434 kernel_memsz = PAGE_ALIGN(header->init_size); 435 kernel_align = header->kernel_alignment; 436 437 ret = kexec_add_buffer(image, kernel_buf, 438 kernel_bufsz, kernel_memsz, kernel_align, 439 MIN_KERNEL_LOAD_ADDR, ULONG_MAX, 1, 440 &kernel_load_addr); 441 if (ret) 442 goto out_free_params; 443 444 pr_debug("Loaded 64bit kernel at 0x%lx bufsz=0x%lx memsz=0x%lx\n", 445 kernel_load_addr, kernel_memsz, kernel_memsz); 446 447 /* Load initrd high */ 448 if (initrd) { 449 ret = kexec_add_buffer(image, initrd, initrd_len, initrd_len, 450 PAGE_SIZE, MIN_INITRD_LOAD_ADDR, 451 ULONG_MAX, 1, &initrd_load_addr); 452 if (ret) 453 goto out_free_params; 454 455 pr_debug("Loaded initrd at 0x%lx bufsz=0x%lx memsz=0x%lx\n", 456 initrd_load_addr, initrd_len, initrd_len); 457 458 setup_initrd(params, initrd_load_addr, initrd_len); 459 } 460 461 setup_cmdline(image, params, bootparam_load_addr, 462 sizeof(struct boot_params), cmdline, cmdline_len); 463 464 /* bootloader info. Do we need a separate ID for kexec kernel loader? */ 465 params->hdr.type_of_loader = 0x0D << 4; 466 params->hdr.loadflags = 0; 467 468 /* Setup purgatory regs for entry */ 469 ret = kexec_purgatory_get_set_symbol(image, "entry64_regs", ®s64, 470 sizeof(regs64), 1); 471 if (ret) 472 goto out_free_params; 473 474 regs64.rbx = 0; /* Bootstrap Processor */ 475 regs64.rsi = bootparam_load_addr; 476 regs64.rip = kernel_load_addr + 0x200; 477 stack = kexec_purgatory_get_symbol_addr(image, "stack_end"); 478 if (IS_ERR(stack)) { 479 pr_err("Could not find address of symbol stack_end\n"); 480 ret = -EINVAL; 481 goto out_free_params; 482 } 483 484 regs64.rsp = (unsigned long)stack; 485 ret = kexec_purgatory_get_set_symbol(image, "entry64_regs", ®s64, 486 sizeof(regs64), 0); 487 if (ret) 488 goto out_free_params; 489 490 ret = setup_boot_parameters(image, params, bootparam_load_addr, 491 efi_map_offset, efi_map_sz, 492 efi_setup_data_offset); 493 if (ret) 494 goto out_free_params; 495 496 /* Allocate loader specific data */ 497 ldata = kzalloc(sizeof(struct bzimage64_data), GFP_KERNEL); 498 if (!ldata) { 499 ret = -ENOMEM; 500 goto out_free_params; 501 } 502 503 /* 504 * Store pointer to params so that it could be freed after loading 505 * params segment has been loaded and contents have been copied 506 * somewhere else. 507 */ 508 ldata->bootparams_buf = params; 509 return ldata; 510 511 out_free_params: 512 kfree(params); 513 return ERR_PTR(ret); 514 } 515 516 /* This cleanup function is called after various segments have been loaded */ 517 int bzImage64_cleanup(void *loader_data) 518 { 519 struct bzimage64_data *ldata = loader_data; 520 521 if (!ldata) 522 return 0; 523 524 kfree(ldata->bootparams_buf); 525 ldata->bootparams_buf = NULL; 526 527 return 0; 528 } 529 530 #ifdef CONFIG_KEXEC_BZIMAGE_VERIFY_SIG 531 int bzImage64_verify_sig(const char *kernel, unsigned long kernel_len) 532 { 533 bool trusted; 534 int ret; 535 536 ret = verify_pefile_signature(kernel, kernel_len, 537 system_trusted_keyring, &trusted); 538 if (ret < 0) 539 return ret; 540 if (!trusted) 541 return -EKEYREJECTED; 542 return 0; 543 } 544 #endif 545 546 struct kexec_file_ops kexec_bzImage64_ops = { 547 .probe = bzImage64_probe, 548 .load = bzImage64_load, 549 .cleanup = bzImage64_cleanup, 550 #ifdef CONFIG_KEXEC_BZIMAGE_VERIFY_SIG 551 .verify_sig = bzImage64_verify_sig, 552 #endif 553 }; 554