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 #include <asm/kexec-bzimage64.h> 30 31 #define MAX_ELFCOREHDR_STR_LEN 30 /* elfcorehdr=0x<64bit-value> */ 32 33 /* 34 * Defines lowest physical address for various segments. Not sure where 35 * exactly these limits came from. Current bzimage64 loader in kexec-tools 36 * uses these so I am retaining it. It can be changed over time as we gain 37 * more insight. 38 */ 39 #define MIN_PURGATORY_ADDR 0x3000 40 #define MIN_BOOTPARAM_ADDR 0x3000 41 #define MIN_KERNEL_LOAD_ADDR 0x100000 42 #define MIN_INITRD_LOAD_ADDR 0x1000000 43 44 /* 45 * This is a place holder for all boot loader specific data structure which 46 * gets allocated in one call but gets freed much later during cleanup 47 * time. Right now there is only one field but it can grow as need be. 48 */ 49 struct bzimage64_data { 50 /* 51 * Temporary buffer to hold bootparams buffer. This should be 52 * freed once the bootparam segment has been loaded. 53 */ 54 void *bootparams_buf; 55 }; 56 57 static int setup_initrd(struct boot_params *params, 58 unsigned long initrd_load_addr, unsigned long initrd_len) 59 { 60 params->hdr.ramdisk_image = initrd_load_addr & 0xffffffffUL; 61 params->hdr.ramdisk_size = initrd_len & 0xffffffffUL; 62 63 params->ext_ramdisk_image = initrd_load_addr >> 32; 64 params->ext_ramdisk_size = initrd_len >> 32; 65 66 return 0; 67 } 68 69 static int setup_cmdline(struct kimage *image, struct boot_params *params, 70 unsigned long bootparams_load_addr, 71 unsigned long cmdline_offset, char *cmdline, 72 unsigned long cmdline_len) 73 { 74 char *cmdline_ptr = ((char *)params) + cmdline_offset; 75 unsigned long cmdline_ptr_phys, len = 0; 76 uint32_t cmdline_low_32, cmdline_ext_32; 77 78 if (image->type == KEXEC_TYPE_CRASH) { 79 len = sprintf(cmdline_ptr, 80 "elfcorehdr=0x%lx ", image->arch.elf_load_addr); 81 } 82 memcpy(cmdline_ptr + len, cmdline, cmdline_len); 83 cmdline_len += len; 84 85 cmdline_ptr[cmdline_len - 1] = '\0'; 86 87 pr_debug("Final command line is: %s\n", cmdline_ptr); 88 cmdline_ptr_phys = bootparams_load_addr + cmdline_offset; 89 cmdline_low_32 = cmdline_ptr_phys & 0xffffffffUL; 90 cmdline_ext_32 = cmdline_ptr_phys >> 32; 91 92 params->hdr.cmd_line_ptr = cmdline_low_32; 93 if (cmdline_ext_32) 94 params->ext_cmd_line_ptr = cmdline_ext_32; 95 96 return 0; 97 } 98 99 static int setup_e820_entries(struct boot_params *params) 100 { 101 unsigned int nr_e820_entries; 102 103 nr_e820_entries = e820_saved.nr_map; 104 105 /* TODO: Pass entries more than E820MAX in bootparams setup data */ 106 if (nr_e820_entries > E820MAX) 107 nr_e820_entries = E820MAX; 108 109 params->e820_entries = nr_e820_entries; 110 memcpy(¶ms->e820_map, &e820_saved.map, 111 nr_e820_entries * sizeof(struct e820entry)); 112 113 return 0; 114 } 115 116 #ifdef CONFIG_EFI 117 static int setup_efi_info_memmap(struct boot_params *params, 118 unsigned long params_load_addr, 119 unsigned int efi_map_offset, 120 unsigned int efi_map_sz) 121 { 122 void *efi_map = (void *)params + efi_map_offset; 123 unsigned long efi_map_phys_addr = params_load_addr + efi_map_offset; 124 struct efi_info *ei = ¶ms->efi_info; 125 126 if (!efi_map_sz) 127 return 0; 128 129 efi_runtime_map_copy(efi_map, efi_map_sz); 130 131 ei->efi_memmap = efi_map_phys_addr & 0xffffffff; 132 ei->efi_memmap_hi = efi_map_phys_addr >> 32; 133 ei->efi_memmap_size = efi_map_sz; 134 135 return 0; 136 } 137 138 static int 139 prepare_add_efi_setup_data(struct boot_params *params, 140 unsigned long params_load_addr, 141 unsigned int efi_setup_data_offset) 142 { 143 unsigned long setup_data_phys; 144 struct setup_data *sd = (void *)params + efi_setup_data_offset; 145 struct efi_setup_data *esd = (void *)sd + sizeof(struct setup_data); 146 147 esd->fw_vendor = efi.fw_vendor; 148 esd->runtime = efi.runtime; 149 esd->tables = efi.config_table; 150 esd->smbios = efi.smbios; 151 152 sd->type = SETUP_EFI; 153 sd->len = sizeof(struct efi_setup_data); 154 155 /* Add setup data */ 156 setup_data_phys = params_load_addr + efi_setup_data_offset; 157 sd->next = params->hdr.setup_data; 158 params->hdr.setup_data = setup_data_phys; 159 160 return 0; 161 } 162 163 static int 164 setup_efi_state(struct boot_params *params, unsigned long params_load_addr, 165 unsigned int efi_map_offset, unsigned int efi_map_sz, 166 unsigned int efi_setup_data_offset) 167 { 168 struct efi_info *current_ei = &boot_params.efi_info; 169 struct efi_info *ei = ¶ms->efi_info; 170 171 if (!current_ei->efi_memmap_size) 172 return 0; 173 174 /* 175 * If 1:1 mapping is not enabled, second kernel can not setup EFI 176 * and use EFI run time services. User space will have to pass 177 * acpi_rsdp=<addr> on kernel command line to make second kernel boot 178 * without efi. 179 */ 180 if (efi_enabled(EFI_OLD_MEMMAP)) 181 return 0; 182 183 ei->efi_loader_signature = current_ei->efi_loader_signature; 184 ei->efi_systab = current_ei->efi_systab; 185 ei->efi_systab_hi = current_ei->efi_systab_hi; 186 187 ei->efi_memdesc_version = current_ei->efi_memdesc_version; 188 ei->efi_memdesc_size = efi_get_runtime_map_desc_size(); 189 190 setup_efi_info_memmap(params, params_load_addr, efi_map_offset, 191 efi_map_sz); 192 prepare_add_efi_setup_data(params, params_load_addr, 193 efi_setup_data_offset); 194 return 0; 195 } 196 #endif /* CONFIG_EFI */ 197 198 static int 199 setup_boot_parameters(struct kimage *image, struct boot_params *params, 200 unsigned long params_load_addr, 201 unsigned int efi_map_offset, unsigned int efi_map_sz, 202 unsigned int efi_setup_data_offset) 203 { 204 unsigned int nr_e820_entries; 205 unsigned long long mem_k, start, end; 206 int i, ret = 0; 207 208 /* Get subarch from existing bootparams */ 209 params->hdr.hardware_subarch = boot_params.hdr.hardware_subarch; 210 211 /* Copying screen_info will do? */ 212 memcpy(¶ms->screen_info, &boot_params.screen_info, 213 sizeof(struct screen_info)); 214 215 /* Fill in memsize later */ 216 params->screen_info.ext_mem_k = 0; 217 params->alt_mem_k = 0; 218 219 /* Default APM info */ 220 memset(¶ms->apm_bios_info, 0, sizeof(params->apm_bios_info)); 221 222 /* Default drive info */ 223 memset(¶ms->hd0_info, 0, sizeof(params->hd0_info)); 224 memset(¶ms->hd1_info, 0, sizeof(params->hd1_info)); 225 226 if (image->type == KEXEC_TYPE_CRASH) { 227 ret = crash_setup_memmap_entries(image, params); 228 if (ret) 229 return ret; 230 } else 231 setup_e820_entries(params); 232 233 nr_e820_entries = params->e820_entries; 234 235 for (i = 0; i < nr_e820_entries; i++) { 236 if (params->e820_map[i].type != E820_RAM) 237 continue; 238 start = params->e820_map[i].addr; 239 end = params->e820_map[i].addr + params->e820_map[i].size - 1; 240 241 if ((start <= 0x100000) && end > 0x100000) { 242 mem_k = (end >> 10) - (0x100000 >> 10); 243 params->screen_info.ext_mem_k = mem_k; 244 params->alt_mem_k = mem_k; 245 if (mem_k > 0xfc00) 246 params->screen_info.ext_mem_k = 0xfc00; /* 64M*/ 247 if (mem_k > 0xffffffff) 248 params->alt_mem_k = 0xffffffff; 249 } 250 } 251 252 #ifdef CONFIG_EFI 253 /* Setup EFI state */ 254 setup_efi_state(params, params_load_addr, efi_map_offset, efi_map_sz, 255 efi_setup_data_offset); 256 #endif 257 258 /* Setup EDD info */ 259 memcpy(params->eddbuf, boot_params.eddbuf, 260 EDDMAXNR * sizeof(struct edd_info)); 261 params->eddbuf_entries = boot_params.eddbuf_entries; 262 263 memcpy(params->edd_mbr_sig_buffer, boot_params.edd_mbr_sig_buffer, 264 EDD_MBR_SIG_MAX * sizeof(unsigned int)); 265 266 return ret; 267 } 268 269 static int bzImage64_probe(const char *buf, unsigned long len) 270 { 271 int ret = -ENOEXEC; 272 struct setup_header *header; 273 274 /* kernel should be atleast two sectors long */ 275 if (len < 2 * 512) { 276 pr_err("File is too short to be a bzImage\n"); 277 return ret; 278 } 279 280 header = (struct setup_header *)(buf + offsetof(struct boot_params, hdr)); 281 if (memcmp((char *)&header->header, "HdrS", 4) != 0) { 282 pr_err("Not a bzImage\n"); 283 return ret; 284 } 285 286 if (header->boot_flag != 0xAA55) { 287 pr_err("No x86 boot sector present\n"); 288 return ret; 289 } 290 291 if (header->version < 0x020C) { 292 pr_err("Must be at least protocol version 2.12\n"); 293 return ret; 294 } 295 296 if (!(header->loadflags & LOADED_HIGH)) { 297 pr_err("zImage not a bzImage\n"); 298 return ret; 299 } 300 301 if (!(header->xloadflags & XLF_KERNEL_64)) { 302 pr_err("Not a bzImage64. XLF_KERNEL_64 is not set.\n"); 303 return ret; 304 } 305 306 if (!(header->xloadflags & XLF_CAN_BE_LOADED_ABOVE_4G)) { 307 pr_err("XLF_CAN_BE_LOADED_ABOVE_4G is not set.\n"); 308 return ret; 309 } 310 311 /* 312 * Can't handle 32bit EFI as it does not allow loading kernel 313 * above 4G. This should be handled by 32bit bzImage loader 314 */ 315 if (efi_enabled(EFI_RUNTIME_SERVICES) && !efi_enabled(EFI_64BIT)) { 316 pr_debug("EFI is 32 bit. Can't load kernel above 4G.\n"); 317 return ret; 318 } 319 320 /* I've got a bzImage */ 321 pr_debug("It's a relocatable bzImage64\n"); 322 ret = 0; 323 324 return ret; 325 } 326 327 static void *bzImage64_load(struct kimage *image, char *kernel, 328 unsigned long kernel_len, char *initrd, 329 unsigned long initrd_len, char *cmdline, 330 unsigned long cmdline_len) 331 { 332 333 struct setup_header *header; 334 int setup_sects, kern16_size, ret = 0; 335 unsigned long setup_header_size, params_cmdline_sz, params_misc_sz; 336 struct boot_params *params; 337 unsigned long bootparam_load_addr, kernel_load_addr, initrd_load_addr; 338 unsigned long purgatory_load_addr; 339 unsigned long kernel_bufsz, kernel_memsz, kernel_align; 340 char *kernel_buf; 341 struct bzimage64_data *ldata; 342 struct kexec_entry64_regs regs64; 343 void *stack; 344 unsigned int setup_hdr_offset = offsetof(struct boot_params, hdr); 345 unsigned int efi_map_offset, efi_map_sz, efi_setup_data_offset; 346 347 header = (struct setup_header *)(kernel + setup_hdr_offset); 348 setup_sects = header->setup_sects; 349 if (setup_sects == 0) 350 setup_sects = 4; 351 352 kern16_size = (setup_sects + 1) * 512; 353 if (kernel_len < kern16_size) { 354 pr_err("bzImage truncated\n"); 355 return ERR_PTR(-ENOEXEC); 356 } 357 358 if (cmdline_len > header->cmdline_size) { 359 pr_err("Kernel command line too long\n"); 360 return ERR_PTR(-EINVAL); 361 } 362 363 /* 364 * In case of crash dump, we will append elfcorehdr=<addr> to 365 * command line. Make sure it does not overflow 366 */ 367 if (cmdline_len + MAX_ELFCOREHDR_STR_LEN > header->cmdline_size) { 368 pr_debug("Appending elfcorehdr=<addr> to command line exceeds maximum allowed length\n"); 369 return ERR_PTR(-EINVAL); 370 } 371 372 /* Allocate and load backup region */ 373 if (image->type == KEXEC_TYPE_CRASH) { 374 ret = crash_load_segments(image); 375 if (ret) 376 return ERR_PTR(ret); 377 } 378 379 /* 380 * Load purgatory. For 64bit entry point, purgatory code can be 381 * anywhere. 382 */ 383 ret = kexec_load_purgatory(image, MIN_PURGATORY_ADDR, ULONG_MAX, 1, 384 &purgatory_load_addr); 385 if (ret) { 386 pr_err("Loading purgatory failed\n"); 387 return ERR_PTR(ret); 388 } 389 390 pr_debug("Loaded purgatory at 0x%lx\n", purgatory_load_addr); 391 392 393 /* 394 * Load Bootparams and cmdline and space for efi stuff. 395 * 396 * Allocate memory together for multiple data structures so 397 * that they all can go in single area/segment and we don't 398 * have to create separate segment for each. Keeps things 399 * little bit simple 400 */ 401 efi_map_sz = efi_get_runtime_map_size(); 402 efi_map_sz = ALIGN(efi_map_sz, 16); 403 params_cmdline_sz = sizeof(struct boot_params) + cmdline_len + 404 MAX_ELFCOREHDR_STR_LEN; 405 params_cmdline_sz = ALIGN(params_cmdline_sz, 16); 406 params_misc_sz = params_cmdline_sz + efi_map_sz + 407 sizeof(struct setup_data) + 408 sizeof(struct efi_setup_data); 409 410 params = kzalloc(params_misc_sz, GFP_KERNEL); 411 if (!params) 412 return ERR_PTR(-ENOMEM); 413 efi_map_offset = params_cmdline_sz; 414 efi_setup_data_offset = efi_map_offset + efi_map_sz; 415 416 /* Copy setup header onto bootparams. Documentation/x86/boot.txt */ 417 setup_header_size = 0x0202 + kernel[0x0201] - setup_hdr_offset; 418 419 /* Is there a limit on setup header size? */ 420 memcpy(¶ms->hdr, (kernel + setup_hdr_offset), setup_header_size); 421 422 ret = kexec_add_buffer(image, (char *)params, params_misc_sz, 423 params_misc_sz, 16, MIN_BOOTPARAM_ADDR, 424 ULONG_MAX, 1, &bootparam_load_addr); 425 if (ret) 426 goto out_free_params; 427 pr_debug("Loaded boot_param, command line and misc at 0x%lx bufsz=0x%lx memsz=0x%lx\n", 428 bootparam_load_addr, params_misc_sz, params_misc_sz); 429 430 /* Load kernel */ 431 kernel_buf = kernel + kern16_size; 432 kernel_bufsz = kernel_len - kern16_size; 433 kernel_memsz = PAGE_ALIGN(header->init_size); 434 kernel_align = header->kernel_alignment; 435 436 ret = kexec_add_buffer(image, kernel_buf, 437 kernel_bufsz, kernel_memsz, kernel_align, 438 MIN_KERNEL_LOAD_ADDR, ULONG_MAX, 1, 439 &kernel_load_addr); 440 if (ret) 441 goto out_free_params; 442 443 pr_debug("Loaded 64bit kernel at 0x%lx bufsz=0x%lx memsz=0x%lx\n", 444 kernel_load_addr, kernel_memsz, kernel_memsz); 445 446 /* Load initrd high */ 447 if (initrd) { 448 ret = kexec_add_buffer(image, initrd, initrd_len, initrd_len, 449 PAGE_SIZE, MIN_INITRD_LOAD_ADDR, 450 ULONG_MAX, 1, &initrd_load_addr); 451 if (ret) 452 goto out_free_params; 453 454 pr_debug("Loaded initrd at 0x%lx bufsz=0x%lx memsz=0x%lx\n", 455 initrd_load_addr, initrd_len, initrd_len); 456 457 setup_initrd(params, initrd_load_addr, initrd_len); 458 } 459 460 setup_cmdline(image, params, bootparam_load_addr, 461 sizeof(struct boot_params), cmdline, cmdline_len); 462 463 /* bootloader info. Do we need a separate ID for kexec kernel loader? */ 464 params->hdr.type_of_loader = 0x0D << 4; 465 params->hdr.loadflags = 0; 466 467 /* Setup purgatory regs for entry */ 468 ret = kexec_purgatory_get_set_symbol(image, "entry64_regs", ®s64, 469 sizeof(regs64), 1); 470 if (ret) 471 goto out_free_params; 472 473 regs64.rbx = 0; /* Bootstrap Processor */ 474 regs64.rsi = bootparam_load_addr; 475 regs64.rip = kernel_load_addr + 0x200; 476 stack = kexec_purgatory_get_symbol_addr(image, "stack_end"); 477 if (IS_ERR(stack)) { 478 pr_err("Could not find address of symbol stack_end\n"); 479 ret = -EINVAL; 480 goto out_free_params; 481 } 482 483 regs64.rsp = (unsigned long)stack; 484 ret = kexec_purgatory_get_set_symbol(image, "entry64_regs", ®s64, 485 sizeof(regs64), 0); 486 if (ret) 487 goto out_free_params; 488 489 ret = setup_boot_parameters(image, params, bootparam_load_addr, 490 efi_map_offset, efi_map_sz, 491 efi_setup_data_offset); 492 if (ret) 493 goto out_free_params; 494 495 /* Allocate loader specific data */ 496 ldata = kzalloc(sizeof(struct bzimage64_data), GFP_KERNEL); 497 if (!ldata) { 498 ret = -ENOMEM; 499 goto out_free_params; 500 } 501 502 /* 503 * Store pointer to params so that it could be freed after loading 504 * params segment has been loaded and contents have been copied 505 * somewhere else. 506 */ 507 ldata->bootparams_buf = params; 508 return ldata; 509 510 out_free_params: 511 kfree(params); 512 return ERR_PTR(ret); 513 } 514 515 /* This cleanup function is called after various segments have been loaded */ 516 static int bzImage64_cleanup(void *loader_data) 517 { 518 struct bzimage64_data *ldata = loader_data; 519 520 if (!ldata) 521 return 0; 522 523 kfree(ldata->bootparams_buf); 524 ldata->bootparams_buf = NULL; 525 526 return 0; 527 } 528 529 #ifdef CONFIG_KEXEC_BZIMAGE_VERIFY_SIG 530 static int bzImage64_verify_sig(const char *kernel, unsigned long kernel_len) 531 { 532 bool trusted; 533 int ret; 534 535 ret = verify_pefile_signature(kernel, kernel_len, 536 system_trusted_keyring, 537 VERIFYING_KEXEC_PE_SIGNATURE, 538 &trusted); 539 if (ret < 0) 540 return ret; 541 if (!trusted) 542 return -EKEYREJECTED; 543 return 0; 544 } 545 #endif 546 547 struct kexec_file_ops kexec_bzImage64_ops = { 548 .probe = bzImage64_probe, 549 .load = bzImage64_load, 550 .cleanup = bzImage64_cleanup, 551 #ifdef CONFIG_KEXEC_BZIMAGE_VERIFY_SIG 552 .verify_sig = bzImage64_verify_sig, 553 #endif 554 }; 555