1 THE LINUX/x86 BOOT PROTOCOL 2 --------------------------- 3 4On the x86 platform, the Linux kernel uses a rather complicated boot 5convention. This has evolved partially due to historical aspects, as 6well as the desire in the early days to have the kernel itself be a 7bootable image, the complicated PC memory model and due to changed 8expectations in the PC industry caused by the effective demise of 9real-mode DOS as a mainstream operating system. 10 11Currently, the following versions of the Linux/x86 boot protocol exist. 12 13Old kernels: zImage/Image support only. Some very early kernels 14 may not even support a command line. 15 16Protocol 2.00: (Kernel 1.3.73) Added bzImage and initrd support, as 17 well as a formalized way to communicate between the 18 boot loader and the kernel. setup.S made relocatable, 19 although the traditional setup area still assumed 20 writable. 21 22Protocol 2.01: (Kernel 1.3.76) Added a heap overrun warning. 23 24Protocol 2.02: (Kernel 2.4.0-test3-pre3) New command line protocol. 25 Lower the conventional memory ceiling. No overwrite 26 of the traditional setup area, thus making booting 27 safe for systems which use the EBDA from SMM or 32-bit 28 BIOS entry points. zImage deprecated but still 29 supported. 30 31Protocol 2.03: (Kernel 2.4.18-pre1) Explicitly makes the highest possible 32 initrd address available to the bootloader. 33 34Protocol 2.04: (Kernel 2.6.14) Extend the syssize field to four bytes. 35 36Protocol 2.05: (Kernel 2.6.20) Make protected mode kernel relocatable. 37 Introduce relocatable_kernel and kernel_alignment fields. 38 39Protocol 2.06: (Kernel 2.6.22) Added a field that contains the size of 40 the boot command line. 41 42Protocol 2.07: (Kernel 2.6.24) Added paravirtualised boot protocol. 43 Introduced hardware_subarch and hardware_subarch_data 44 and KEEP_SEGMENTS flag in load_flags. 45 46Protocol 2.08: (Kernel 2.6.26) Added crc32 checksum and ELF format 47 payload. Introduced payload_offset and payload_length 48 fields to aid in locating the payload. 49 50Protocol 2.09: (Kernel 2.6.26) Added a field of 64-bit physical 51 pointer to single linked list of struct setup_data. 52 53Protocol 2.10: (Kernel 2.6.31) Added a protocol for relaxed alignment 54 beyond the kernel_alignment added, new init_size and 55 pref_address fields. Added extended boot loader IDs. 56 57Protocol 2.11: (Kernel 3.6) Added a field for offset of EFI handover 58 protocol entry point. 59 60Protocol 2.12: (Kernel 3.8) Added the xloadflags field and extension fields 61 to struct boot_params for loading bzImage and ramdisk 62 above 4G in 64bit. 63 64**** MEMORY LAYOUT 65 66The traditional memory map for the kernel loader, used for Image or 67zImage kernels, typically looks like: 68 69 | | 700A0000 +------------------------+ 71 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA. 7209A000 +------------------------+ 73 | Command line | 74 | Stack/heap | For use by the kernel real-mode code. 75098000 +------------------------+ 76 | Kernel setup | The kernel real-mode code. 77090200 +------------------------+ 78 | Kernel boot sector | The kernel legacy boot sector. 79090000 +------------------------+ 80 | Protected-mode kernel | The bulk of the kernel image. 81010000 +------------------------+ 82 | Boot loader | <- Boot sector entry point 0000:7C00 83001000 +------------------------+ 84 | Reserved for MBR/BIOS | 85000800 +------------------------+ 86 | Typically used by MBR | 87000600 +------------------------+ 88 | BIOS use only | 89000000 +------------------------+ 90 91 92When using bzImage, the protected-mode kernel was relocated to 930x100000 ("high memory"), and the kernel real-mode block (boot sector, 94setup, and stack/heap) was made relocatable to any address between 950x10000 and end of low memory. Unfortunately, in protocols 2.00 and 962.01 the 0x90000+ memory range is still used internally by the kernel; 97the 2.02 protocol resolves that problem. 98 99It is desirable to keep the "memory ceiling" -- the highest point in 100low memory touched by the boot loader -- as low as possible, since 101some newer BIOSes have begun to allocate some rather large amounts of 102memory, called the Extended BIOS Data Area, near the top of low 103memory. The boot loader should use the "INT 12h" BIOS call to verify 104how much low memory is available. 105 106Unfortunately, if INT 12h reports that the amount of memory is too 107low, there is usually nothing the boot loader can do but to report an 108error to the user. The boot loader should therefore be designed to 109take up as little space in low memory as it reasonably can. For 110zImage or old bzImage kernels, which need data written into the 1110x90000 segment, the boot loader should make sure not to use memory 112above the 0x9A000 point; too many BIOSes will break above that point. 113 114For a modern bzImage kernel with boot protocol version >= 2.02, a 115memory layout like the following is suggested: 116 117 ~ ~ 118 | Protected-mode kernel | 119100000 +------------------------+ 120 | I/O memory hole | 1210A0000 +------------------------+ 122 | Reserved for BIOS | Leave as much as possible unused 123 ~ ~ 124 | Command line | (Can also be below the X+10000 mark) 125X+10000 +------------------------+ 126 | Stack/heap | For use by the kernel real-mode code. 127X+08000 +------------------------+ 128 | Kernel setup | The kernel real-mode code. 129 | Kernel boot sector | The kernel legacy boot sector. 130X +------------------------+ 131 | Boot loader | <- Boot sector entry point 0000:7C00 132001000 +------------------------+ 133 | Reserved for MBR/BIOS | 134000800 +------------------------+ 135 | Typically used by MBR | 136000600 +------------------------+ 137 | BIOS use only | 138000000 +------------------------+ 139 140... where the address X is as low as the design of the boot loader 141permits. 142 143 144**** THE REAL-MODE KERNEL HEADER 145 146In the following text, and anywhere in the kernel boot sequence, "a 147sector" refers to 512 bytes. It is independent of the actual sector 148size of the underlying medium. 149 150The first step in loading a Linux kernel should be to load the 151real-mode code (boot sector and setup code) and then examine the 152following header at offset 0x01f1. The real-mode code can total up to 15332K, although the boot loader may choose to load only the first two 154sectors (1K) and then examine the bootup sector size. 155 156The header looks like: 157 158Offset Proto Name Meaning 159/Size 160 16101F1/1 ALL(1 setup_sects The size of the setup in sectors 16201F2/2 ALL root_flags If set, the root is mounted readonly 16301F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras 16401F8/2 ALL ram_size DO NOT USE - for bootsect.S use only 16501FA/2 ALL vid_mode Video mode control 16601FC/2 ALL root_dev Default root device number 16701FE/2 ALL boot_flag 0xAA55 magic number 1680200/2 2.00+ jump Jump instruction 1690202/4 2.00+ header Magic signature "HdrS" 1700206/2 2.00+ version Boot protocol version supported 1710208/4 2.00+ realmode_swtch Boot loader hook (see below) 172020C/2 2.00+ start_sys_seg The load-low segment (0x1000) (obsolete) 173020E/2 2.00+ kernel_version Pointer to kernel version string 1740210/1 2.00+ type_of_loader Boot loader identifier 1750211/1 2.00+ loadflags Boot protocol option flags 1760212/2 2.00+ setup_move_size Move to high memory size (used with hooks) 1770214/4 2.00+ code32_start Boot loader hook (see below) 1780218/4 2.00+ ramdisk_image initrd load address (set by boot loader) 179021C/4 2.00+ ramdisk_size initrd size (set by boot loader) 1800220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only 1810224/2 2.01+ heap_end_ptr Free memory after setup end 1820226/1 2.02+(3 ext_loader_ver Extended boot loader version 1830227/1 2.02+(3 ext_loader_type Extended boot loader ID 1840228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line 185022C/4 2.03+ initrd_addr_max Highest legal initrd address 1860230/4 2.05+ kernel_alignment Physical addr alignment required for kernel 1870234/1 2.05+ relocatable_kernel Whether kernel is relocatable or not 1880235/1 2.10+ min_alignment Minimum alignment, as a power of two 1890236/2 2.12+ xloadflags Boot protocol option flags 1900238/4 2.06+ cmdline_size Maximum size of the kernel command line 191023C/4 2.07+ hardware_subarch Hardware subarchitecture 1920240/8 2.07+ hardware_subarch_data Subarchitecture-specific data 1930248/4 2.08+ payload_offset Offset of kernel payload 194024C/4 2.08+ payload_length Length of kernel payload 1950250/8 2.09+ setup_data 64-bit physical pointer to linked list 196 of struct setup_data 1970258/8 2.10+ pref_address Preferred loading address 1980260/4 2.10+ init_size Linear memory required during initialization 1990264/4 2.11+ handover_offset Offset of handover entry point 200 201(1) For backwards compatibility, if the setup_sects field contains 0, the 202 real value is 4. 203 204(2) For boot protocol prior to 2.04, the upper two bytes of the syssize 205 field are unusable, which means the size of a bzImage kernel 206 cannot be determined. 207 208(3) Ignored, but safe to set, for boot protocols 2.02-2.09. 209 210If the "HdrS" (0x53726448) magic number is not found at offset 0x202, 211the boot protocol version is "old". Loading an old kernel, the 212following parameters should be assumed: 213 214 Image type = zImage 215 initrd not supported 216 Real-mode kernel must be located at 0x90000. 217 218Otherwise, the "version" field contains the protocol version, 219e.g. protocol version 2.01 will contain 0x0201 in this field. When 220setting fields in the header, you must make sure only to set fields 221supported by the protocol version in use. 222 223 224**** DETAILS OF HEADER FIELDS 225 226For each field, some are information from the kernel to the bootloader 227("read"), some are expected to be filled out by the bootloader 228("write"), and some are expected to be read and modified by the 229bootloader ("modify"). 230 231All general purpose boot loaders should write the fields marked 232(obligatory). Boot loaders who want to load the kernel at a 233nonstandard address should fill in the fields marked (reloc); other 234boot loaders can ignore those fields. 235 236The byte order of all fields is littleendian (this is x86, after all.) 237 238Field name: setup_sects 239Type: read 240Offset/size: 0x1f1/1 241Protocol: ALL 242 243 The size of the setup code in 512-byte sectors. If this field is 244 0, the real value is 4. The real-mode code consists of the boot 245 sector (always one 512-byte sector) plus the setup code. 246 247Field name: root_flags 248Type: modify (optional) 249Offset/size: 0x1f2/2 250Protocol: ALL 251 252 If this field is nonzero, the root defaults to readonly. The use of 253 this field is deprecated; use the "ro" or "rw" options on the 254 command line instead. 255 256Field name: syssize 257Type: read 258Offset/size: 0x1f4/4 (protocol 2.04+) 0x1f4/2 (protocol ALL) 259Protocol: 2.04+ 260 261 The size of the protected-mode code in units of 16-byte paragraphs. 262 For protocol versions older than 2.04 this field is only two bytes 263 wide, and therefore cannot be trusted for the size of a kernel if 264 the LOAD_HIGH flag is set. 265 266Field name: ram_size 267Type: kernel internal 268Offset/size: 0x1f8/2 269Protocol: ALL 270 271 This field is obsolete. 272 273Field name: vid_mode 274Type: modify (obligatory) 275Offset/size: 0x1fa/2 276 277 Please see the section on SPECIAL COMMAND LINE OPTIONS. 278 279Field name: root_dev 280Type: modify (optional) 281Offset/size: 0x1fc/2 282Protocol: ALL 283 284 The default root device device number. The use of this field is 285 deprecated, use the "root=" option on the command line instead. 286 287Field name: boot_flag 288Type: read 289Offset/size: 0x1fe/2 290Protocol: ALL 291 292 Contains 0xAA55. This is the closest thing old Linux kernels have 293 to a magic number. 294 295Field name: jump 296Type: read 297Offset/size: 0x200/2 298Protocol: 2.00+ 299 300 Contains an x86 jump instruction, 0xEB followed by a signed offset 301 relative to byte 0x202. This can be used to determine the size of 302 the header. 303 304Field name: header 305Type: read 306Offset/size: 0x202/4 307Protocol: 2.00+ 308 309 Contains the magic number "HdrS" (0x53726448). 310 311Field name: version 312Type: read 313Offset/size: 0x206/2 314Protocol: 2.00+ 315 316 Contains the boot protocol version, in (major << 8)+minor format, 317 e.g. 0x0204 for version 2.04, and 0x0a11 for a hypothetical version 318 10.17. 319 320Field name: realmode_swtch 321Type: modify (optional) 322Offset/size: 0x208/4 323Protocol: 2.00+ 324 325 Boot loader hook (see ADVANCED BOOT LOADER HOOKS below.) 326 327Field name: start_sys_seg 328Type: read 329Offset/size: 0x20c/2 330Protocol: 2.00+ 331 332 The load low segment (0x1000). Obsolete. 333 334Field name: kernel_version 335Type: read 336Offset/size: 0x20e/2 337Protocol: 2.00+ 338 339 If set to a nonzero value, contains a pointer to a NUL-terminated 340 human-readable kernel version number string, less 0x200. This can 341 be used to display the kernel version to the user. This value 342 should be less than (0x200*setup_sects). 343 344 For example, if this value is set to 0x1c00, the kernel version 345 number string can be found at offset 0x1e00 in the kernel file. 346 This is a valid value if and only if the "setup_sects" field 347 contains the value 15 or higher, as: 348 349 0x1c00 < 15*0x200 (= 0x1e00) but 350 0x1c00 >= 14*0x200 (= 0x1c00) 351 352 0x1c00 >> 9 = 14, so the minimum value for setup_secs is 15. 353 354Field name: type_of_loader 355Type: write (obligatory) 356Offset/size: 0x210/1 357Protocol: 2.00+ 358 359 If your boot loader has an assigned id (see table below), enter 360 0xTV here, where T is an identifier for the boot loader and V is 361 a version number. Otherwise, enter 0xFF here. 362 363 For boot loader IDs above T = 0xD, write T = 0xE to this field and 364 write the extended ID minus 0x10 to the ext_loader_type field. 365 Similarly, the ext_loader_ver field can be used to provide more than 366 four bits for the bootloader version. 367 368 For example, for T = 0x15, V = 0x234, write: 369 370 type_of_loader <- 0xE4 371 ext_loader_type <- 0x05 372 ext_loader_ver <- 0x23 373 374 Assigned boot loader ids (hexadecimal): 375 376 0 LILO (0x00 reserved for pre-2.00 bootloader) 377 1 Loadlin 378 2 bootsect-loader (0x20, all other values reserved) 379 3 Syslinux 380 4 Etherboot/gPXE/iPXE 381 5 ELILO 382 7 GRUB 383 8 U-Boot 384 9 Xen 385 A Gujin 386 B Qemu 387 C Arcturus Networks uCbootloader 388 D kexec-tools 389 E Extended (see ext_loader_type) 390 F Special (0xFF = undefined) 391 10 Reserved 392 11 Minimal Linux Bootloader <http://sebastian-plotz.blogspot.de> 393 12 OVMF UEFI virtualization stack 394 395 Please contact <hpa@zytor.com> if you need a bootloader ID 396 value assigned. 397 398Field name: loadflags 399Type: modify (obligatory) 400Offset/size: 0x211/1 401Protocol: 2.00+ 402 403 This field is a bitmask. 404 405 Bit 0 (read): LOADED_HIGH 406 - If 0, the protected-mode code is loaded at 0x10000. 407 - If 1, the protected-mode code is loaded at 0x100000. 408 409 Bit 1 (kernel internal): ALSR_FLAG 410 - Used internally by the compressed kernel to communicate 411 KASLR status to kernel proper. 412 If 1, KASLR enabled. 413 If 0, KASLR disabled. 414 415 Bit 5 (write): QUIET_FLAG 416 - If 0, print early messages. 417 - If 1, suppress early messages. 418 This requests to the kernel (decompressor and early 419 kernel) to not write early messages that require 420 accessing the display hardware directly. 421 422 Bit 6 (write): KEEP_SEGMENTS 423 Protocol: 2.07+ 424 - If 0, reload the segment registers in the 32bit entry point. 425 - If 1, do not reload the segment registers in the 32bit entry point. 426 Assume that %cs %ds %ss %es are all set to flat segments with 427 a base of 0 (or the equivalent for their environment). 428 429 Bit 7 (write): CAN_USE_HEAP 430 Set this bit to 1 to indicate that the value entered in the 431 heap_end_ptr is valid. If this field is clear, some setup code 432 functionality will be disabled. 433 434Field name: setup_move_size 435Type: modify (obligatory) 436Offset/size: 0x212/2 437Protocol: 2.00-2.01 438 439 When using protocol 2.00 or 2.01, if the real mode kernel is not 440 loaded at 0x90000, it gets moved there later in the loading 441 sequence. Fill in this field if you want additional data (such as 442 the kernel command line) moved in addition to the real-mode kernel 443 itself. 444 445 The unit is bytes starting with the beginning of the boot sector. 446 447 This field is can be ignored when the protocol is 2.02 or higher, or 448 if the real-mode code is loaded at 0x90000. 449 450Field name: code32_start 451Type: modify (optional, reloc) 452Offset/size: 0x214/4 453Protocol: 2.00+ 454 455 The address to jump to in protected mode. This defaults to the load 456 address of the kernel, and can be used by the boot loader to 457 determine the proper load address. 458 459 This field can be modified for two purposes: 460 461 1. as a boot loader hook (see ADVANCED BOOT LOADER HOOKS below.) 462 463 2. if a bootloader which does not install a hook loads a 464 relocatable kernel at a nonstandard address it will have to modify 465 this field to point to the load address. 466 467Field name: ramdisk_image 468Type: write (obligatory) 469Offset/size: 0x218/4 470Protocol: 2.00+ 471 472 The 32-bit linear address of the initial ramdisk or ramfs. Leave at 473 zero if there is no initial ramdisk/ramfs. 474 475Field name: ramdisk_size 476Type: write (obligatory) 477Offset/size: 0x21c/4 478Protocol: 2.00+ 479 480 Size of the initial ramdisk or ramfs. Leave at zero if there is no 481 initial ramdisk/ramfs. 482 483Field name: bootsect_kludge 484Type: kernel internal 485Offset/size: 0x220/4 486Protocol: 2.00+ 487 488 This field is obsolete. 489 490Field name: heap_end_ptr 491Type: write (obligatory) 492Offset/size: 0x224/2 493Protocol: 2.01+ 494 495 Set this field to the offset (from the beginning of the real-mode 496 code) of the end of the setup stack/heap, minus 0x0200. 497 498Field name: ext_loader_ver 499Type: write (optional) 500Offset/size: 0x226/1 501Protocol: 2.02+ 502 503 This field is used as an extension of the version number in the 504 type_of_loader field. The total version number is considered to be 505 (type_of_loader & 0x0f) + (ext_loader_ver << 4). 506 507 The use of this field is boot loader specific. If not written, it 508 is zero. 509 510 Kernels prior to 2.6.31 did not recognize this field, but it is safe 511 to write for protocol version 2.02 or higher. 512 513Field name: ext_loader_type 514Type: write (obligatory if (type_of_loader & 0xf0) == 0xe0) 515Offset/size: 0x227/1 516Protocol: 2.02+ 517 518 This field is used as an extension of the type number in 519 type_of_loader field. If the type in type_of_loader is 0xE, then 520 the actual type is (ext_loader_type + 0x10). 521 522 This field is ignored if the type in type_of_loader is not 0xE. 523 524 Kernels prior to 2.6.31 did not recognize this field, but it is safe 525 to write for protocol version 2.02 or higher. 526 527Field name: cmd_line_ptr 528Type: write (obligatory) 529Offset/size: 0x228/4 530Protocol: 2.02+ 531 532 Set this field to the linear address of the kernel command line. 533 The kernel command line can be located anywhere between the end of 534 the setup heap and 0xA0000; it does not have to be located in the 535 same 64K segment as the real-mode code itself. 536 537 Fill in this field even if your boot loader does not support a 538 command line, in which case you can point this to an empty string 539 (or better yet, to the string "auto".) If this field is left at 540 zero, the kernel will assume that your boot loader does not support 541 the 2.02+ protocol. 542 543Field name: initrd_addr_max 544Type: read 545Offset/size: 0x22c/4 546Protocol: 2.03+ 547 548 The maximum address that may be occupied by the initial 549 ramdisk/ramfs contents. For boot protocols 2.02 or earlier, this 550 field is not present, and the maximum address is 0x37FFFFFF. (This 551 address is defined as the address of the highest safe byte, so if 552 your ramdisk is exactly 131072 bytes long and this field is 553 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.) 554 555Field name: kernel_alignment 556Type: read/modify (reloc) 557Offset/size: 0x230/4 558Protocol: 2.05+ (read), 2.10+ (modify) 559 560 Alignment unit required by the kernel (if relocatable_kernel is 561 true.) A relocatable kernel that is loaded at an alignment 562 incompatible with the value in this field will be realigned during 563 kernel initialization. 564 565 Starting with protocol version 2.10, this reflects the kernel 566 alignment preferred for optimal performance; it is possible for the 567 loader to modify this field to permit a lesser alignment. See the 568 min_alignment and pref_address field below. 569 570Field name: relocatable_kernel 571Type: read (reloc) 572Offset/size: 0x234/1 573Protocol: 2.05+ 574 575 If this field is nonzero, the protected-mode part of the kernel can 576 be loaded at any address that satisfies the kernel_alignment field. 577 After loading, the boot loader must set the code32_start field to 578 point to the loaded code, or to a boot loader hook. 579 580Field name: min_alignment 581Type: read (reloc) 582Offset/size: 0x235/1 583Protocol: 2.10+ 584 585 This field, if nonzero, indicates as a power of two the minimum 586 alignment required, as opposed to preferred, by the kernel to boot. 587 If a boot loader makes use of this field, it should update the 588 kernel_alignment field with the alignment unit desired; typically: 589 590 kernel_alignment = 1 << min_alignment 591 592 There may be a considerable performance cost with an excessively 593 misaligned kernel. Therefore, a loader should typically try each 594 power-of-two alignment from kernel_alignment down to this alignment. 595 596Field name: xloadflags 597Type: read 598Offset/size: 0x236/2 599Protocol: 2.12+ 600 601 This field is a bitmask. 602 603 Bit 0 (read): XLF_KERNEL_64 604 - If 1, this kernel has the legacy 64-bit entry point at 0x200. 605 606 Bit 1 (read): XLF_CAN_BE_LOADED_ABOVE_4G 607 - If 1, kernel/boot_params/cmdline/ramdisk can be above 4G. 608 609 Bit 2 (read): XLF_EFI_HANDOVER_32 610 - If 1, the kernel supports the 32-bit EFI handoff entry point 611 given at handover_offset. 612 613 Bit 3 (read): XLF_EFI_HANDOVER_64 614 - If 1, the kernel supports the 64-bit EFI handoff entry point 615 given at handover_offset + 0x200. 616 617 Bit 4 (read): XLF_EFI_KEXEC 618 - If 1, the kernel supports kexec EFI boot with EFI runtime support. 619 620Field name: cmdline_size 621Type: read 622Offset/size: 0x238/4 623Protocol: 2.06+ 624 625 The maximum size of the command line without the terminating 626 zero. This means that the command line can contain at most 627 cmdline_size characters. With protocol version 2.05 and earlier, the 628 maximum size was 255. 629 630Field name: hardware_subarch 631Type: write (optional, defaults to x86/PC) 632Offset/size: 0x23c/4 633Protocol: 2.07+ 634 635 In a paravirtualized environment the hardware low level architectural 636 pieces such as interrupt handling, page table handling, and 637 accessing process control registers needs to be done differently. 638 639 This field allows the bootloader to inform the kernel we are in one 640 one of those environments. 641 642 0x00000000 The default x86/PC environment 643 0x00000001 lguest 644 0x00000002 Xen 645 0x00000003 Moorestown MID 646 0x00000004 CE4100 TV Platform 647 648Field name: hardware_subarch_data 649Type: write (subarch-dependent) 650Offset/size: 0x240/8 651Protocol: 2.07+ 652 653 A pointer to data that is specific to hardware subarch 654 This field is currently unused for the default x86/PC environment, 655 do not modify. 656 657Field name: payload_offset 658Type: read 659Offset/size: 0x248/4 660Protocol: 2.08+ 661 662 If non-zero then this field contains the offset from the beginning 663 of the protected-mode code to the payload. 664 665 The payload may be compressed. The format of both the compressed and 666 uncompressed data should be determined using the standard magic 667 numbers. The currently supported compression formats are gzip 668 (magic numbers 1F 8B or 1F 9E), bzip2 (magic number 42 5A), LZMA 669 (magic number 5D 00), XZ (magic number FD 37), and LZ4 (magic number 670 02 21). The uncompressed payload is currently always ELF (magic 671 number 7F 45 4C 46). 672 673Field name: payload_length 674Type: read 675Offset/size: 0x24c/4 676Protocol: 2.08+ 677 678 The length of the payload. 679 680Field name: setup_data 681Type: write (special) 682Offset/size: 0x250/8 683Protocol: 2.09+ 684 685 The 64-bit physical pointer to NULL terminated single linked list of 686 struct setup_data. This is used to define a more extensible boot 687 parameters passing mechanism. The definition of struct setup_data is 688 as follow: 689 690 struct setup_data { 691 u64 next; 692 u32 type; 693 u32 len; 694 u8 data[0]; 695 }; 696 697 Where, the next is a 64-bit physical pointer to the next node of 698 linked list, the next field of the last node is 0; the type is used 699 to identify the contents of data; the len is the length of data 700 field; the data holds the real payload. 701 702 This list may be modified at a number of points during the bootup 703 process. Therefore, when modifying this list one should always make 704 sure to consider the case where the linked list already contains 705 entries. 706 707Field name: pref_address 708Type: read (reloc) 709Offset/size: 0x258/8 710Protocol: 2.10+ 711 712 This field, if nonzero, represents a preferred load address for the 713 kernel. A relocating bootloader should attempt to load at this 714 address if possible. 715 716 A non-relocatable kernel will unconditionally move itself and to run 717 at this address. 718 719Field name: init_size 720Type: read 721Offset/size: 0x260/4 722 723 This field indicates the amount of linear contiguous memory starting 724 at the kernel runtime start address that the kernel needs before it 725 is capable of examining its memory map. This is not the same thing 726 as the total amount of memory the kernel needs to boot, but it can 727 be used by a relocating boot loader to help select a safe load 728 address for the kernel. 729 730 The kernel runtime start address is determined by the following algorithm: 731 732 if (relocatable_kernel) 733 runtime_start = align_up(load_address, kernel_alignment) 734 else 735 runtime_start = pref_address 736 737Field name: handover_offset 738Type: read 739Offset/size: 0x264/4 740 741 This field is the offset from the beginning of the kernel image to 742 the EFI handover protocol entry point. Boot loaders using the EFI 743 handover protocol to boot the kernel should jump to this offset. 744 745 See EFI HANDOVER PROTOCOL below for more details. 746 747 748**** THE IMAGE CHECKSUM 749 750From boot protocol version 2.08 onwards the CRC-32 is calculated over 751the entire file using the characteristic polynomial 0x04C11DB7 and an 752initial remainder of 0xffffffff. The checksum is appended to the 753file; therefore the CRC of the file up to the limit specified in the 754syssize field of the header is always 0. 755 756 757**** THE KERNEL COMMAND LINE 758 759The kernel command line has become an important way for the boot 760loader to communicate with the kernel. Some of its options are also 761relevant to the boot loader itself, see "special command line options" 762below. 763 764The kernel command line is a null-terminated string. The maximum 765length can be retrieved from the field cmdline_size. Before protocol 766version 2.06, the maximum was 255 characters. A string that is too 767long will be automatically truncated by the kernel. 768 769If the boot protocol version is 2.02 or later, the address of the 770kernel command line is given by the header field cmd_line_ptr (see 771above.) This address can be anywhere between the end of the setup 772heap and 0xA0000. 773 774If the protocol version is *not* 2.02 or higher, the kernel 775command line is entered using the following protocol: 776 777 At offset 0x0020 (word), "cmd_line_magic", enter the magic 778 number 0xA33F. 779 780 At offset 0x0022 (word), "cmd_line_offset", enter the offset 781 of the kernel command line (relative to the start of the 782 real-mode kernel). 783 784 The kernel command line *must* be within the memory region 785 covered by setup_move_size, so you may need to adjust this 786 field. 787 788 789**** MEMORY LAYOUT OF THE REAL-MODE CODE 790 791The real-mode code requires a stack/heap to be set up, as well as 792memory allocated for the kernel command line. This needs to be done 793in the real-mode accessible memory in bottom megabyte. 794 795It should be noted that modern machines often have a sizable Extended 796BIOS Data Area (EBDA). As a result, it is advisable to use as little 797of the low megabyte as possible. 798 799Unfortunately, under the following circumstances the 0x90000 memory 800segment has to be used: 801 802 - When loading a zImage kernel ((loadflags & 0x01) == 0). 803 - When loading a 2.01 or earlier boot protocol kernel. 804 805 -> For the 2.00 and 2.01 boot protocols, the real-mode code 806 can be loaded at another address, but it is internally 807 relocated to 0x90000. For the "old" protocol, the 808 real-mode code must be loaded at 0x90000. 809 810When loading at 0x90000, avoid using memory above 0x9a000. 811 812For boot protocol 2.02 or higher, the command line does not have to be 813located in the same 64K segment as the real-mode setup code; it is 814thus permitted to give the stack/heap the full 64K segment and locate 815the command line above it. 816 817The kernel command line should not be located below the real-mode 818code, nor should it be located in high memory. 819 820 821**** SAMPLE BOOT CONFIGURATION 822 823As a sample configuration, assume the following layout of the real 824mode segment: 825 826 When loading below 0x90000, use the entire segment: 827 828 0x0000-0x7fff Real mode kernel 829 0x8000-0xdfff Stack and heap 830 0xe000-0xffff Kernel command line 831 832 When loading at 0x90000 OR the protocol version is 2.01 or earlier: 833 834 0x0000-0x7fff Real mode kernel 835 0x8000-0x97ff Stack and heap 836 0x9800-0x9fff Kernel command line 837 838Such a boot loader should enter the following fields in the header: 839 840 unsigned long base_ptr; /* base address for real-mode segment */ 841 842 if ( setup_sects == 0 ) { 843 setup_sects = 4; 844 } 845 846 if ( protocol >= 0x0200 ) { 847 type_of_loader = <type code>; 848 if ( loading_initrd ) { 849 ramdisk_image = <initrd_address>; 850 ramdisk_size = <initrd_size>; 851 } 852 853 if ( protocol >= 0x0202 && loadflags & 0x01 ) 854 heap_end = 0xe000; 855 else 856 heap_end = 0x9800; 857 858 if ( protocol >= 0x0201 ) { 859 heap_end_ptr = heap_end - 0x200; 860 loadflags |= 0x80; /* CAN_USE_HEAP */ 861 } 862 863 if ( protocol >= 0x0202 ) { 864 cmd_line_ptr = base_ptr + heap_end; 865 strcpy(cmd_line_ptr, cmdline); 866 } else { 867 cmd_line_magic = 0xA33F; 868 cmd_line_offset = heap_end; 869 setup_move_size = heap_end + strlen(cmdline)+1; 870 strcpy(base_ptr+cmd_line_offset, cmdline); 871 } 872 } else { 873 /* Very old kernel */ 874 875 heap_end = 0x9800; 876 877 cmd_line_magic = 0xA33F; 878 cmd_line_offset = heap_end; 879 880 /* A very old kernel MUST have its real-mode code 881 loaded at 0x90000 */ 882 883 if ( base_ptr != 0x90000 ) { 884 /* Copy the real-mode kernel */ 885 memcpy(0x90000, base_ptr, (setup_sects+1)*512); 886 base_ptr = 0x90000; /* Relocated */ 887 } 888 889 strcpy(0x90000+cmd_line_offset, cmdline); 890 891 /* It is recommended to clear memory up to the 32K mark */ 892 memset(0x90000 + (setup_sects+1)*512, 0, 893 (64-(setup_sects+1))*512); 894 } 895 896 897**** LOADING THE REST OF THE KERNEL 898 899The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512 900in the kernel file (again, if setup_sects == 0 the real value is 4.) 901It should be loaded at address 0x10000 for Image/zImage kernels and 9020x100000 for bzImage kernels. 903 904The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01 905bit (LOAD_HIGH) in the loadflags field is set: 906 907 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01); 908 load_address = is_bzImage ? 0x100000 : 0x10000; 909 910Note that Image/zImage kernels can be up to 512K in size, and thus use 911the entire 0x10000-0x90000 range of memory. This means it is pretty 912much a requirement for these kernels to load the real-mode part at 9130x90000. bzImage kernels allow much more flexibility. 914 915 916**** SPECIAL COMMAND LINE OPTIONS 917 918If the command line provided by the boot loader is entered by the 919user, the user may expect the following command line options to work. 920They should normally not be deleted from the kernel command line even 921though not all of them are actually meaningful to the kernel. Boot 922loader authors who need additional command line options for the boot 923loader itself should get them registered in 924Documentation/kernel-parameters.txt to make sure they will not 925conflict with actual kernel options now or in the future. 926 927 vga=<mode> 928 <mode> here is either an integer (in C notation, either 929 decimal, octal, or hexadecimal) or one of the strings 930 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask" 931 (meaning 0xFFFD). This value should be entered into the 932 vid_mode field, as it is used by the kernel before the command 933 line is parsed. 934 935 mem=<size> 936 <size> is an integer in C notation optionally followed by 937 (case insensitive) K, M, G, T, P or E (meaning << 10, << 20, 938 << 30, << 40, << 50 or << 60). This specifies the end of 939 memory to the kernel. This affects the possible placement of 940 an initrd, since an initrd should be placed near end of 941 memory. Note that this is an option to *both* the kernel and 942 the bootloader! 943 944 initrd=<file> 945 An initrd should be loaded. The meaning of <file> is 946 obviously bootloader-dependent, and some boot loaders 947 (e.g. LILO) do not have such a command. 948 949In addition, some boot loaders add the following options to the 950user-specified command line: 951 952 BOOT_IMAGE=<file> 953 The boot image which was loaded. Again, the meaning of <file> 954 is obviously bootloader-dependent. 955 956 auto 957 The kernel was booted without explicit user intervention. 958 959If these options are added by the boot loader, it is highly 960recommended that they are located *first*, before the user-specified 961or configuration-specified command line. Otherwise, "init=/bin/sh" 962gets confused by the "auto" option. 963 964 965**** RUNNING THE KERNEL 966 967The kernel is started by jumping to the kernel entry point, which is 968located at *segment* offset 0x20 from the start of the real mode 969kernel. This means that if you loaded your real-mode kernel code at 9700x90000, the kernel entry point is 9020:0000. 971 972At entry, ds = es = ss should point to the start of the real-mode 973kernel code (0x9000 if the code is loaded at 0x90000), sp should be 974set up properly, normally pointing to the top of the heap, and 975interrupts should be disabled. Furthermore, to guard against bugs in 976the kernel, it is recommended that the boot loader sets fs = gs = ds = 977es = ss. 978 979In our example from above, we would do: 980 981 /* Note: in the case of the "old" kernel protocol, base_ptr must 982 be == 0x90000 at this point; see the previous sample code */ 983 984 seg = base_ptr >> 4; 985 986 cli(); /* Enter with interrupts disabled! */ 987 988 /* Set up the real-mode kernel stack */ 989 _SS = seg; 990 _SP = heap_end; 991 992 _DS = _ES = _FS = _GS = seg; 993 jmp_far(seg+0x20, 0); /* Run the kernel */ 994 995If your boot sector accesses a floppy drive, it is recommended to 996switch off the floppy motor before running the kernel, since the 997kernel boot leaves interrupts off and thus the motor will not be 998switched off, especially if the loaded kernel has the floppy driver as 999a demand-loaded module! 1000 1001 1002**** ADVANCED BOOT LOADER HOOKS 1003 1004If the boot loader runs in a particularly hostile environment (such as 1005LOADLIN, which runs under DOS) it may be impossible to follow the 1006standard memory location requirements. Such a boot loader may use the 1007following hooks that, if set, are invoked by the kernel at the 1008appropriate time. The use of these hooks should probably be 1009considered an absolutely last resort! 1010 1011IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and 1012%edi across invocation. 1013 1014 realmode_swtch: 1015 A 16-bit real mode far subroutine invoked immediately before 1016 entering protected mode. The default routine disables NMI, so 1017 your routine should probably do so, too. 1018 1019 code32_start: 1020 A 32-bit flat-mode routine *jumped* to immediately after the 1021 transition to protected mode, but before the kernel is 1022 uncompressed. No segments, except CS, are guaranteed to be 1023 set up (current kernels do, but older ones do not); you should 1024 set them up to BOOT_DS (0x18) yourself. 1025 1026 After completing your hook, you should jump to the address 1027 that was in this field before your boot loader overwrote it 1028 (relocated, if appropriate.) 1029 1030 1031**** 32-bit BOOT PROTOCOL 1032 1033For machine with some new BIOS other than legacy BIOS, such as EFI, 1034LinuxBIOS, etc, and kexec, the 16-bit real mode setup code in kernel 1035based on legacy BIOS can not be used, so a 32-bit boot protocol needs 1036to be defined. 1037 1038In 32-bit boot protocol, the first step in loading a Linux kernel 1039should be to setup the boot parameters (struct boot_params, 1040traditionally known as "zero page"). The memory for struct boot_params 1041should be allocated and initialized to all zero. Then the setup header 1042from offset 0x01f1 of kernel image on should be loaded into struct 1043boot_params and examined. The end of setup header can be calculated as 1044follow: 1045 1046 0x0202 + byte value at offset 0x0201 1047 1048In addition to read/modify/write the setup header of the struct 1049boot_params as that of 16-bit boot protocol, the boot loader should 1050also fill the additional fields of the struct boot_params as that 1051described in zero-page.txt. 1052 1053After setting up the struct boot_params, the boot loader can load the 105432/64-bit kernel in the same way as that of 16-bit boot protocol. 1055 1056In 32-bit boot protocol, the kernel is started by jumping to the 105732-bit kernel entry point, which is the start address of loaded 105832/64-bit kernel. 1059 1060At entry, the CPU must be in 32-bit protected mode with paging 1061disabled; a GDT must be loaded with the descriptors for selectors 1062__BOOT_CS(0x10) and __BOOT_DS(0x18); both descriptors must be 4G flat 1063segment; __BOOT_CS must have execute/read permission, and __BOOT_DS 1064must have read/write permission; CS must be __BOOT_CS and DS, ES, SS 1065must be __BOOT_DS; interrupt must be disabled; %esi must hold the base 1066address of the struct boot_params; %ebp, %edi and %ebx must be zero. 1067 1068**** 64-bit BOOT PROTOCOL 1069 1070For machine with 64bit cpus and 64bit kernel, we could use 64bit bootloader 1071and we need a 64-bit boot protocol. 1072 1073In 64-bit boot protocol, the first step in loading a Linux kernel 1074should be to setup the boot parameters (struct boot_params, 1075traditionally known as "zero page"). The memory for struct boot_params 1076could be allocated anywhere (even above 4G) and initialized to all zero. 1077Then, the setup header at offset 0x01f1 of kernel image on should be 1078loaded into struct boot_params and examined. The end of setup header 1079can be calculated as follows: 1080 1081 0x0202 + byte value at offset 0x0201 1082 1083In addition to read/modify/write the setup header of the struct 1084boot_params as that of 16-bit boot protocol, the boot loader should 1085also fill the additional fields of the struct boot_params as described 1086in zero-page.txt. 1087 1088After setting up the struct boot_params, the boot loader can load 108964-bit kernel in the same way as that of 16-bit boot protocol, but 1090kernel could be loaded above 4G. 1091 1092In 64-bit boot protocol, the kernel is started by jumping to the 109364-bit kernel entry point, which is the start address of loaded 109464-bit kernel plus 0x200. 1095 1096At entry, the CPU must be in 64-bit mode with paging enabled. 1097The range with setup_header.init_size from start address of loaded 1098kernel and zero page and command line buffer get ident mapping; 1099a GDT must be loaded with the descriptors for selectors 1100__BOOT_CS(0x10) and __BOOT_DS(0x18); both descriptors must be 4G flat 1101segment; __BOOT_CS must have execute/read permission, and __BOOT_DS 1102must have read/write permission; CS must be __BOOT_CS and DS, ES, SS 1103must be __BOOT_DS; interrupt must be disabled; %rsi must hold the base 1104address of the struct boot_params. 1105 1106**** EFI HANDOVER PROTOCOL 1107 1108This protocol allows boot loaders to defer initialisation to the EFI 1109boot stub. The boot loader is required to load the kernel/initrd(s) 1110from the boot media and jump to the EFI handover protocol entry point 1111which is hdr->handover_offset bytes from the beginning of 1112startup_{32,64}. 1113 1114The function prototype for the handover entry point looks like this, 1115 1116 efi_main(void *handle, efi_system_table_t *table, struct boot_params *bp) 1117 1118'handle' is the EFI image handle passed to the boot loader by the EFI 1119firmware, 'table' is the EFI system table - these are the first two 1120arguments of the "handoff state" as described in section 2.3 of the 1121UEFI specification. 'bp' is the boot loader-allocated boot params. 1122 1123The boot loader *must* fill out the following fields in bp, 1124 1125 o hdr.code32_start 1126 o hdr.cmd_line_ptr 1127 o hdr.cmdline_size 1128 o hdr.ramdisk_image (if applicable) 1129 o hdr.ramdisk_size (if applicable) 1130 1131All other fields should be zero. 1132