1445 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3===========================4The Linux/x86 Boot Protocol5===========================6 7On the x86 platform, the Linux kernel uses a rather complicated boot8convention. This has evolved partially due to historical aspects, as9well as the desire in the early days to have the kernel itself be a10bootable image, the complicated PC memory model and due to changed11expectations in the PC industry caused by the effective demise of12real-mode DOS as a mainstream operating system.13 14Currently, the following versions of the Linux/x86 boot protocol exist.15 16============= ============================================================17Old kernels zImage/Image support only. Some very early kernels18 may not even support a command line.19 20Protocol 2.00 (Kernel 1.3.73) Added bzImage and initrd support, as21 well as a formalized way to communicate between the22 boot loader and the kernel. setup.S made relocatable,23 although the traditional setup area still assumed24 writable.25 26Protocol 2.01 (Kernel 1.3.76) Added a heap overrun warning.27 28Protocol 2.02 (Kernel 2.4.0-test3-pre3) New command line protocol.29 Lower the conventional memory ceiling. No overwrite30 of the traditional setup area, thus making booting31 safe for systems which use the EBDA from SMM or 32-bit32 BIOS entry points. zImage deprecated but still33 supported.34 35Protocol 2.03 (Kernel 2.4.18-pre1) Explicitly makes the highest possible36 initrd address available to the bootloader.37 38Protocol 2.04 (Kernel 2.6.14) Extend the syssize field to four bytes.39 40Protocol 2.05 (Kernel 2.6.20) Make protected mode kernel relocatable.41 Introduce relocatable_kernel and kernel_alignment fields.42 43Protocol 2.06 (Kernel 2.6.22) Added a field that contains the size of44 the boot command line.45 46Protocol 2.07 (Kernel 2.6.24) Added paravirtualised boot protocol.47 Introduced hardware_subarch and hardware_subarch_data48 and KEEP_SEGMENTS flag in load_flags.49 50Protocol 2.08 (Kernel 2.6.26) Added crc32 checksum and ELF format51 payload. Introduced payload_offset and payload_length52 fields to aid in locating the payload.53 54Protocol 2.09 (Kernel 2.6.26) Added a field of 64-bit physical55 pointer to single linked list of struct setup_data.56 57Protocol 2.10 (Kernel 2.6.31) Added a protocol for relaxed alignment58 beyond the kernel_alignment added, new init_size and59 pref_address fields. Added extended boot loader IDs.60 61Protocol 2.11 (Kernel 3.6) Added a field for offset of EFI handover62 protocol entry point.63 64Protocol 2.12 (Kernel 3.8) Added the xloadflags field and extension fields65 to struct boot_params for loading bzImage and ramdisk66 above 4G in 64bit.67 68Protocol 2.13 (Kernel 3.14) Support 32- and 64-bit flags being set in69 xloadflags to support booting a 64-bit kernel from 32-bit70 EFI71 72Protocol 2.14 BURNT BY INCORRECT COMMIT73 ae7e1238e68f2a472a125673ab506d49158c188974 ("x86/boot: Add ACPI RSDP address to setup_header")75 DO NOT USE!!! ASSUME SAME AS 2.13.76 77Protocol 2.15 (Kernel 5.5) Added the kernel_info and kernel_info.setup_type_max.78============= ============================================================79 80 .. note::81 The protocol version number should be changed only if the setup header82 is changed. There is no need to update the version number if boot_params83 or kernel_info are changed. Additionally, it is recommended to use84 xloadflags (in this case the protocol version number should not be85 updated either) or kernel_info to communicate supported Linux kernel86 features to the boot loader. Due to very limited space available in87 the original setup header every update to it should be considered88 with great care. Starting from the protocol 2.15 the primary way to89 communicate things to the boot loader is the kernel_info.90 91 92Memory Layout93=============94 95The traditional memory map for the kernel loader, used for Image or96zImage kernels, typically looks like::97 98 | |99 0A0000 +------------------------+100 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA.101 09A000 +------------------------+102 | Command line |103 | Stack/heap | For use by the kernel real-mode code.104 098000 +------------------------+105 | Kernel setup | The kernel real-mode code.106 090200 +------------------------+107 | Kernel boot sector | The kernel legacy boot sector.108 090000 +------------------------+109 | Protected-mode kernel | The bulk of the kernel image.110 010000 +------------------------+111 | Boot loader | <- Boot sector entry point 0000:7C00112 001000 +------------------------+113 | Reserved for MBR/BIOS |114 000800 +------------------------+115 | Typically used by MBR |116 000600 +------------------------+117 | BIOS use only |118 000000 +------------------------+119 120When using bzImage, the protected-mode kernel was relocated to1210x100000 ("high memory"), and the kernel real-mode block (boot sector,122setup, and stack/heap) was made relocatable to any address between1230x10000 and end of low memory. Unfortunately, in protocols 2.00 and1242.01 the 0x90000+ memory range is still used internally by the kernel;125the 2.02 protocol resolves that problem.126 127It is desirable to keep the "memory ceiling" -- the highest point in128low memory touched by the boot loader -- as low as possible, since129some newer BIOSes have begun to allocate some rather large amounts of130memory, called the Extended BIOS Data Area, near the top of low131memory. The boot loader should use the "INT 12h" BIOS call to verify132how much low memory is available.133 134Unfortunately, if INT 12h reports that the amount of memory is too135low, there is usually nothing the boot loader can do but to report an136error to the user. The boot loader should therefore be designed to137take up as little space in low memory as it reasonably can. For138zImage or old bzImage kernels, which need data written into the1390x90000 segment, the boot loader should make sure not to use memory140above the 0x9A000 point; too many BIOSes will break above that point.141 142For a modern bzImage kernel with boot protocol version >= 2.02, a143memory layout like the following is suggested::144 145 ~ ~146 | Protected-mode kernel |147 100000 +------------------------+148 | I/O memory hole |149 0A0000 +------------------------+150 | Reserved for BIOS | Leave as much as possible unused151 ~ ~152 | Command line | (Can also be below the X+10000 mark)153 X+10000 +------------------------+154 | Stack/heap | For use by the kernel real-mode code.155 X+08000 +------------------------+156 | Kernel setup | The kernel real-mode code.157 | Kernel boot sector | The kernel legacy boot sector.158 X +------------------------+159 | Boot loader | <- Boot sector entry point 0000:7C00160 001000 +------------------------+161 | Reserved for MBR/BIOS |162 000800 +------------------------+163 | Typically used by MBR |164 000600 +------------------------+165 | BIOS use only |166 000000 +------------------------+167 168 ... where the address X is as low as the design of the boot loader permits.169 170 171The Real-Mode Kernel Header172===========================173 174In the following text, and anywhere in the kernel boot sequence, "a175sector" refers to 512 bytes. It is independent of the actual sector176size of the underlying medium.177 178The first step in loading a Linux kernel should be to load the179real-mode code (boot sector and setup code) and then examine the180following header at offset 0x01f1. The real-mode code can total up to18132K, although the boot loader may choose to load only the first two182sectors (1K) and then examine the bootup sector size.183 184The header looks like:185 186=========== ======== ===================== ============================================187Offset/Size Proto Name Meaning188=========== ======== ===================== ============================================18901F1/1 ALL(1) setup_sects The size of the setup in sectors19001F2/2 ALL root_flags If set, the root is mounted readonly19101F4/4 2.04+(2) syssize The size of the 32-bit code in 16-byte paras19201F8/2 ALL ram_size DO NOT USE - for bootsect.S use only19301FA/2 ALL vid_mode Video mode control19401FC/2 ALL root_dev Default root device number19501FE/2 ALL boot_flag 0xAA55 magic number1960200/2 2.00+ jump Jump instruction1970202/4 2.00+ header Magic signature "HdrS"1980206/2 2.00+ version Boot protocol version supported1990208/4 2.00+ realmode_swtch Boot loader hook (see below)200020C/2 2.00+ start_sys_seg The load-low segment (0x1000) (obsolete)201020E/2 2.00+ kernel_version Pointer to kernel version string2020210/1 2.00+ type_of_loader Boot loader identifier2030211/1 2.00+ loadflags Boot protocol option flags2040212/2 2.00+ setup_move_size Move to high memory size (used with hooks)2050214/4 2.00+ code32_start Boot loader hook (see below)2060218/4 2.00+ ramdisk_image initrd load address (set by boot loader)207021C/4 2.00+ ramdisk_size initrd size (set by boot loader)2080220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only2090224/2 2.01+ heap_end_ptr Free memory after setup end2100226/1 2.02+(3) ext_loader_ver Extended boot loader version2110227/1 2.02+(3) ext_loader_type Extended boot loader ID2120228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line213022C/4 2.03+ initrd_addr_max Highest legal initrd address2140230/4 2.05+ kernel_alignment Physical addr alignment required for kernel2150234/1 2.05+ relocatable_kernel Whether kernel is relocatable or not2160235/1 2.10+ min_alignment Minimum alignment, as a power of two2170236/2 2.12+ xloadflags Boot protocol option flags2180238/4 2.06+ cmdline_size Maximum size of the kernel command line219023C/4 2.07+ hardware_subarch Hardware subarchitecture2200240/8 2.07+ hardware_subarch_data Subarchitecture-specific data2210248/4 2.08+ payload_offset Offset of kernel payload222024C/4 2.08+ payload_length Length of kernel payload2230250/8 2.09+ setup_data 64-bit physical pointer to linked list224 of struct setup_data2250258/8 2.10+ pref_address Preferred loading address2260260/4 2.10+ init_size Linear memory required during initialization2270264/4 2.11+ handover_offset Offset of handover entry point2280268/4 2.15+ kernel_info_offset Offset of the kernel_info229=========== ======== ===================== ============================================230 231.. note::232 (1) For backwards compatibility, if the setup_sects field contains 0, the233 real value is 4.234 235 (2) For boot protocol prior to 2.04, the upper two bytes of the syssize236 field are unusable, which means the size of a bzImage kernel237 cannot be determined.238 239 (3) Ignored, but safe to set, for boot protocols 2.02-2.09.240 241If the "HdrS" (0x53726448) magic number is not found at offset 0x202,242the boot protocol version is "old". Loading an old kernel, the243following parameters should be assumed::244 245 Image type = zImage246 initrd not supported247 Real-mode kernel must be located at 0x90000.248 249Otherwise, the "version" field contains the protocol version,250e.g. protocol version 2.01 will contain 0x0201 in this field. When251setting fields in the header, you must make sure only to set fields252supported by the protocol version in use.253 254 255Details of Header Fields256========================257 258For each field, some are information from the kernel to the bootloader259("read"), some are expected to be filled out by the bootloader260("write"), and some are expected to be read and modified by the261bootloader ("modify").262 263All general purpose boot loaders should write the fields marked264(obligatory). Boot loaders who want to load the kernel at a265nonstandard address should fill in the fields marked (reloc); other266boot loaders can ignore those fields.267 268The byte order of all fields is littleendian (this is x86, after all.)269 270============ ===========271Field name: setup_sects272Type: read273Offset/size: 0x1f1/1274Protocol: ALL275============ ===========276 277 The size of the setup code in 512-byte sectors. If this field is278 0, the real value is 4. The real-mode code consists of the boot279 sector (always one 512-byte sector) plus the setup code.280 281============ =================282Field name: root_flags283Type: modify (optional)284Offset/size: 0x1f2/2285Protocol: ALL286============ =================287 288 If this field is nonzero, the root defaults to readonly. The use of289 this field is deprecated; use the "ro" or "rw" options on the290 command line instead.291 292============ ===============================================293Field name: syssize294Type: read295Offset/size: 0x1f4/4 (protocol 2.04+) 0x1f4/2 (protocol ALL)296Protocol: 2.04+297============ ===============================================298 299 The size of the protected-mode code in units of 16-byte paragraphs.300 For protocol versions older than 2.04 this field is only two bytes301 wide, and therefore cannot be trusted for the size of a kernel if302 the LOAD_HIGH flag is set.303 304============ ===============305Field name: ram_size306Type: kernel internal307Offset/size: 0x1f8/2308Protocol: ALL309============ ===============310 311 This field is obsolete.312 313============ ===================314Field name: vid_mode315Type: modify (obligatory)316Offset/size: 0x1fa/2317============ ===================318 319 Please see the section on SPECIAL COMMAND LINE OPTIONS.320 321============ =================322Field name: root_dev323Type: modify (optional)324Offset/size: 0x1fc/2325Protocol: ALL326============ =================327 328 The default root device device number. The use of this field is329 deprecated, use the "root=" option on the command line instead.330 331============ =========332Field name: boot_flag333Type: read334Offset/size: 0x1fe/2335Protocol: ALL336============ =========337 338 Contains 0xAA55. This is the closest thing old Linux kernels have339 to a magic number.340 341============ =======342Field name: jump343Type: read344Offset/size: 0x200/2345Protocol: 2.00+346============ =======347 348 Contains an x86 jump instruction, 0xEB followed by a signed offset349 relative to byte 0x202. This can be used to determine the size of350 the header.351 352============ =======353Field name: header354Type: read355Offset/size: 0x202/4356Protocol: 2.00+357============ =======358 359 Contains the magic number "HdrS" (0x53726448).360 361============ =======362Field name: version363Type: read364Offset/size: 0x206/2365Protocol: 2.00+366============ =======367 368 Contains the boot protocol version, in (major << 8)+minor format,369 e.g. 0x0204 for version 2.04, and 0x0a11 for a hypothetical version370 10.17.371 372============ =================373Field name: realmode_swtch374Type: modify (optional)375Offset/size: 0x208/4376Protocol: 2.00+377============ =================378 379 Boot loader hook (see ADVANCED BOOT LOADER HOOKS below.)380 381============ =============382Field name: start_sys_seg383Type: read384Offset/size: 0x20c/2385Protocol: 2.00+386============ =============387 388 The load low segment (0x1000). Obsolete.389 390============ ==============391Field name: kernel_version392Type: read393Offset/size: 0x20e/2394Protocol: 2.00+395============ ==============396 397 If set to a nonzero value, contains a pointer to a NUL-terminated398 human-readable kernel version number string, less 0x200. This can399 be used to display the kernel version to the user. This value400 should be less than (0x200*setup_sects).401 402 For example, if this value is set to 0x1c00, the kernel version403 number string can be found at offset 0x1e00 in the kernel file.404 This is a valid value if and only if the "setup_sects" field405 contains the value 15 or higher, as::406 407 0x1c00 < 15*0x200 (= 0x1e00) but408 0x1c00 >= 14*0x200 (= 0x1c00)409 410 0x1c00 >> 9 = 14, So the minimum value for setup_secs is 15.411 412============ ==================413Field name: type_of_loader414Type: write (obligatory)415Offset/size: 0x210/1416Protocol: 2.00+417============ ==================418 419 If your boot loader has an assigned id (see table below), enter420 0xTV here, where T is an identifier for the boot loader and V is421 a version number. Otherwise, enter 0xFF here.422 423 For boot loader IDs above T = 0xD, write T = 0xE to this field and424 write the extended ID minus 0x10 to the ext_loader_type field.425 Similarly, the ext_loader_ver field can be used to provide more than426 four bits for the bootloader version.427 428 For example, for T = 0x15, V = 0x234, write::429 430 type_of_loader <- 0xE4431 ext_loader_type <- 0x05432 ext_loader_ver <- 0x23433 434 Assigned boot loader ids (hexadecimal):435 436 == =======================================437 0 LILO438 (0x00 reserved for pre-2.00 bootloader)439 1 Loadlin440 2 bootsect-loader441 (0x20, all other values reserved)442 3 Syslinux443 4 Etherboot/gPXE/iPXE444 5 ELILO445 7 GRUB446 8 U-Boot447 9 Xen448 A Gujin449 B Qemu450 C Arcturus Networks uCbootloader451 D kexec-tools452 E Extended (see ext_loader_type)453 F Special (0xFF = undefined)454 10 Reserved455 11 Minimal Linux Bootloader456 <http://sebastian-plotz.blogspot.de>457 12 OVMF UEFI virtualization stack458 13 barebox459 == =======================================460 461 Please contact <hpa@zytor.com> if you need a bootloader ID value assigned.462 463============ ===================464Field name: loadflags465Type: modify (obligatory)466Offset/size: 0x211/1467Protocol: 2.00+468============ ===================469 470 This field is a bitmask.471 472 Bit 0 (read): LOADED_HIGH473 474 - If 0, the protected-mode code is loaded at 0x10000.475 - If 1, the protected-mode code is loaded at 0x100000.476 477 Bit 1 (kernel internal): KASLR_FLAG478 479 - Used internally by the compressed kernel to communicate480 KASLR status to kernel proper.481 482 - If 1, KASLR enabled.483 - If 0, KASLR disabled.484 485 Bit 5 (write): QUIET_FLAG486 487 - If 0, print early messages.488 - If 1, suppress early messages.489 490 This requests to the kernel (decompressor and early491 kernel) to not write early messages that require492 accessing the display hardware directly.493 494 Bit 6 (obsolete): KEEP_SEGMENTS495 496 Protocol: 2.07+497 498 - This flag is obsolete.499 500 Bit 7 (write): CAN_USE_HEAP501 502 Set this bit to 1 to indicate that the value entered in the503 heap_end_ptr is valid. If this field is clear, some setup code504 functionality will be disabled.505 506 507============ ===================508Field name: setup_move_size509Type: modify (obligatory)510Offset/size: 0x212/2511Protocol: 2.00-2.01512============ ===================513 514 When using protocol 2.00 or 2.01, if the real mode kernel is not515 loaded at 0x90000, it gets moved there later in the loading516 sequence. Fill in this field if you want additional data (such as517 the kernel command line) moved in addition to the real-mode kernel518 itself.519 520 The unit is bytes starting with the beginning of the boot sector.521 522 This field is can be ignored when the protocol is 2.02 or higher, or523 if the real-mode code is loaded at 0x90000.524 525============ ========================526Field name: code32_start527Type: modify (optional, reloc)528Offset/size: 0x214/4529Protocol: 2.00+530============ ========================531 532 The address to jump to in protected mode. This defaults to the load533 address of the kernel, and can be used by the boot loader to534 determine the proper load address.535 536 This field can be modified for two purposes:537 538 1. as a boot loader hook (see Advanced Boot Loader Hooks below.)539 540 2. if a bootloader which does not install a hook loads a541 relocatable kernel at a nonstandard address it will have to modify542 this field to point to the load address.543 544============ ==================545Field name: ramdisk_image546Type: write (obligatory)547Offset/size: 0x218/4548Protocol: 2.00+549============ ==================550 551 The 32-bit linear address of the initial ramdisk or ramfs. Leave at552 zero if there is no initial ramdisk/ramfs.553 554============ ==================555Field name: ramdisk_size556Type: write (obligatory)557Offset/size: 0x21c/4558Protocol: 2.00+559============ ==================560 561 Size of the initial ramdisk or ramfs. Leave at zero if there is no562 initial ramdisk/ramfs.563 564============ ===============565Field name: bootsect_kludge566Type: kernel internal567Offset/size: 0x220/4568Protocol: 2.00+569============ ===============570 571 This field is obsolete.572 573============ ==================574Field name: heap_end_ptr575Type: write (obligatory)576Offset/size: 0x224/2577Protocol: 2.01+578============ ==================579 580 Set this field to the offset (from the beginning of the real-mode581 code) of the end of the setup stack/heap, minus 0x0200.582 583============ ================584Field name: ext_loader_ver585Type: write (optional)586Offset/size: 0x226/1587Protocol: 2.02+588============ ================589 590 This field is used as an extension of the version number in the591 type_of_loader field. The total version number is considered to be592 (type_of_loader & 0x0f) + (ext_loader_ver << 4).593 594 The use of this field is boot loader specific. If not written, it595 is zero.596 597 Kernels prior to 2.6.31 did not recognize this field, but it is safe598 to write for protocol version 2.02 or higher.599 600============ =====================================================601Field name: ext_loader_type602Type: write (obligatory if (type_of_loader & 0xf0) == 0xe0)603Offset/size: 0x227/1604Protocol: 2.02+605============ =====================================================606 607 This field is used as an extension of the type number in608 type_of_loader field. If the type in type_of_loader is 0xE, then609 the actual type is (ext_loader_type + 0x10).610 611 This field is ignored if the type in type_of_loader is not 0xE.612 613 Kernels prior to 2.6.31 did not recognize this field, but it is safe614 to write for protocol version 2.02 or higher.615 616============ ==================617Field name: cmd_line_ptr618Type: write (obligatory)619Offset/size: 0x228/4620Protocol: 2.02+621============ ==================622 623 Set this field to the linear address of the kernel command line.624 The kernel command line can be located anywhere between the end of625 the setup heap and 0xA0000; it does not have to be located in the626 same 64K segment as the real-mode code itself.627 628 Fill in this field even if your boot loader does not support a629 command line, in which case you can point this to an empty string630 (or better yet, to the string "auto".) If this field is left at631 zero, the kernel will assume that your boot loader does not support632 the 2.02+ protocol.633 634============ ===============635Field name: initrd_addr_max636Type: read637Offset/size: 0x22c/4638Protocol: 2.03+639============ ===============640 641 The maximum address that may be occupied by the initial642 ramdisk/ramfs contents. For boot protocols 2.02 or earlier, this643 field is not present, and the maximum address is 0x37FFFFFF. (This644 address is defined as the address of the highest safe byte, so if645 your ramdisk is exactly 131072 bytes long and this field is646 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)647 648============ ============================649Field name: kernel_alignment650Type: read/modify (reloc)651Offset/size: 0x230/4652Protocol: 2.05+ (read), 2.10+ (modify)653============ ============================654 655 Alignment unit required by the kernel (if relocatable_kernel is656 true.) A relocatable kernel that is loaded at an alignment657 incompatible with the value in this field will be realigned during658 kernel initialization.659 660 Starting with protocol version 2.10, this reflects the kernel661 alignment preferred for optimal performance; it is possible for the662 loader to modify this field to permit a lesser alignment. See the663 min_alignment and pref_address field below.664 665============ ==================666Field name: relocatable_kernel667Type: read (reloc)668Offset/size: 0x234/1669Protocol: 2.05+670============ ==================671 672 If this field is nonzero, the protected-mode part of the kernel can673 be loaded at any address that satisfies the kernel_alignment field.674 After loading, the boot loader must set the code32_start field to675 point to the loaded code, or to a boot loader hook.676 677============ =============678Field name: min_alignment679Type: read (reloc)680Offset/size: 0x235/1681Protocol: 2.10+682============ =============683 684 This field, if nonzero, indicates as a power of two the minimum685 alignment required, as opposed to preferred, by the kernel to boot.686 If a boot loader makes use of this field, it should update the687 kernel_alignment field with the alignment unit desired; typically::688 689 kernel_alignment = 1 << min_alignment690 691 There may be a considerable performance cost with an excessively692 misaligned kernel. Therefore, a loader should typically try each693 power-of-two alignment from kernel_alignment down to this alignment.694 695============ ==========696Field name: xloadflags697Type: read698Offset/size: 0x236/2699Protocol: 2.12+700============ ==========701 702 This field is a bitmask.703 704 Bit 0 (read): XLF_KERNEL_64705 706 - If 1, this kernel has the legacy 64-bit entry point at 0x200.707 708 Bit 1 (read): XLF_CAN_BE_LOADED_ABOVE_4G709 710 - If 1, kernel/boot_params/cmdline/ramdisk can be above 4G.711 712 Bit 2 (read): XLF_EFI_HANDOVER_32713 714 - If 1, the kernel supports the 32-bit EFI handoff entry point715 given at handover_offset.716 717 Bit 3 (read): XLF_EFI_HANDOVER_64718 719 - If 1, the kernel supports the 64-bit EFI handoff entry point720 given at handover_offset + 0x200.721 722 Bit 4 (read): XLF_EFI_KEXEC723 724 - If 1, the kernel supports kexec EFI boot with EFI runtime support.725 726 727============ ============728Field name: cmdline_size729Type: read730Offset/size: 0x238/4731Protocol: 2.06+732============ ============733 734 The maximum size of the command line without the terminating735 zero. This means that the command line can contain at most736 cmdline_size characters. With protocol version 2.05 and earlier, the737 maximum size was 255.738 739============ ====================================740Field name: hardware_subarch741Type: write (optional, defaults to x86/PC)742Offset/size: 0x23c/4743Protocol: 2.07+744============ ====================================745 746 In a paravirtualized environment the hardware low level architectural747 pieces such as interrupt handling, page table handling, and748 accessing process control registers needs to be done differently.749 750 This field allows the bootloader to inform the kernel we are in one751 one of those environments.752 753 ========== ==============================754 0x00000000 The default x86/PC environment755 0x00000001 lguest756 0x00000002 Xen757 0x00000003 Moorestown MID758 0x00000004 CE4100 TV Platform759 ========== ==============================760 761============ =========================762Field name: hardware_subarch_data763Type: write (subarch-dependent)764Offset/size: 0x240/8765Protocol: 2.07+766============ =========================767 768 A pointer to data that is specific to hardware subarch769 This field is currently unused for the default x86/PC environment,770 do not modify.771 772============ ==============773Field name: payload_offset774Type: read775Offset/size: 0x248/4776Protocol: 2.08+777============ ==============778 779 If non-zero then this field contains the offset from the beginning780 of the protected-mode code to the payload.781 782 The payload may be compressed. The format of both the compressed and783 uncompressed data should be determined using the standard magic784 numbers. The currently supported compression formats are gzip785 (magic numbers 1F 8B or 1F 9E), bzip2 (magic number 42 5A), LZMA786 (magic number 5D 00), XZ (magic number FD 37), LZ4 (magic number787 02 21) and ZSTD (magic number 28 B5). The uncompressed payload is788 currently always ELF (magic number 7F 45 4C 46).789 790============ ==============791Field name: payload_length792Type: read793Offset/size: 0x24c/4794Protocol: 2.08+795============ ==============796 797 The length of the payload.798 799============ ===============800Field name: setup_data801Type: write (special)802Offset/size: 0x250/8803Protocol: 2.09+804============ ===============805 806 The 64-bit physical pointer to NULL terminated single linked list of807 struct setup_data. This is used to define a more extensible boot808 parameters passing mechanism. The definition of struct setup_data is809 as follow::810 811 struct setup_data {812 u64 next;813 u32 type;814 u32 len;815 u8 data[0];816 };817 818 Where, the next is a 64-bit physical pointer to the next node of819 linked list, the next field of the last node is 0; the type is used820 to identify the contents of data; the len is the length of data821 field; the data holds the real payload.822 823 This list may be modified at a number of points during the bootup824 process. Therefore, when modifying this list one should always make825 sure to consider the case where the linked list already contains826 entries.827 828 The setup_data is a bit awkward to use for extremely large data objects,829 both because the setup_data header has to be adjacent to the data object830 and because it has a 32-bit length field. However, it is important that831 intermediate stages of the boot process have a way to identify which832 chunks of memory are occupied by kernel data.833 834 Thus setup_indirect struct and SETUP_INDIRECT type were introduced in835 protocol 2.15::836 837 struct setup_indirect {838 __u32 type;839 __u32 reserved; /* Reserved, must be set to zero. */840 __u64 len;841 __u64 addr;842 };843 844 The type member is a SETUP_INDIRECT | SETUP_* type. However, it cannot be845 SETUP_INDIRECT itself since making the setup_indirect a tree structure846 could require a lot of stack space in something that needs to parse it847 and stack space can be limited in boot contexts.848 849 Let's give an example how to point to SETUP_E820_EXT data using setup_indirect.850 In this case setup_data and setup_indirect will look like this::851 852 struct setup_data {853 __u64 next = 0 or <addr_of_next_setup_data_struct>;854 __u32 type = SETUP_INDIRECT;855 __u32 len = sizeof(setup_indirect);856 __u8 data[sizeof(setup_indirect)] = struct setup_indirect {857 __u32 type = SETUP_INDIRECT | SETUP_E820_EXT;858 __u32 reserved = 0;859 __u64 len = <len_of_SETUP_E820_EXT_data>;860 __u64 addr = <addr_of_SETUP_E820_EXT_data>;861 }862 }863 864.. note::865 SETUP_INDIRECT | SETUP_NONE objects cannot be properly distinguished866 from SETUP_INDIRECT itself. So, this kind of objects cannot be provided867 by the bootloaders.868 869============ ============870Field name: pref_address871Type: read (reloc)872Offset/size: 0x258/8873Protocol: 2.10+874============ ============875 876 This field, if nonzero, represents a preferred load address for the877 kernel. A relocating bootloader should attempt to load at this878 address if possible.879 880 A non-relocatable kernel will unconditionally move itself and to run881 at this address. A relocatable kernel will move itself to this address if it882 loaded below this address.883 884============ =======885Field name: init_size886Type: read887Offset/size: 0x260/4888============ =======889 890 This field indicates the amount of linear contiguous memory starting891 at the kernel runtime start address that the kernel needs before it892 is capable of examining its memory map. This is not the same thing893 as the total amount of memory the kernel needs to boot, but it can894 be used by a relocating boot loader to help select a safe load895 address for the kernel.896 897 The kernel runtime start address is determined by the following algorithm::898 899 if (relocatable_kernel)900 runtime_start = align_up(load_address, kernel_alignment)901 else902 runtime_start = pref_address903 904============ ===============905Field name: handover_offset906Type: read907Offset/size: 0x264/4908============ ===============909 910 This field is the offset from the beginning of the kernel image to911 the EFI handover protocol entry point. Boot loaders using the EFI912 handover protocol to boot the kernel should jump to this offset.913 914 See EFI HANDOVER PROTOCOL below for more details.915 916============ ==================917Field name: kernel_info_offset918Type: read919Offset/size: 0x268/4920Protocol: 2.15+921============ ==================922 923 This field is the offset from the beginning of the kernel image to the924 kernel_info. The kernel_info structure is embedded in the Linux image925 in the uncompressed protected mode region.926 927 928The kernel_info929===============930 931The relationships between the headers are analogous to the various data932sections:933 934 setup_header = .data935 boot_params/setup_data = .bss936 937What is missing from the above list? That's right:938 939 kernel_info = .rodata940 941We have been (ab)using .data for things that could go into .rodata or .bss for942a long time, for lack of alternatives and -- especially early on -- inertia.943Also, the BIOS stub is responsible for creating boot_params, so it isn't944available to a BIOS-based loader (setup_data is, though).945 946setup_header is permanently limited to 144 bytes due to the reach of the9472-byte jump field, which doubles as a length field for the structure, combined948with the size of the "hole" in struct boot_params that a protected-mode loader949or the BIOS stub has to copy it into. It is currently 119 bytes long, which950leaves us with 25 very precious bytes. This isn't something that can be fixed951without revising the boot protocol entirely, breaking backwards compatibility.952 953boot_params proper is limited to 4096 bytes, but can be arbitrarily extended954by adding setup_data entries. It cannot be used to communicate properties of955the kernel image, because it is .bss and has no image-provided content.956 957kernel_info solves this by providing an extensible place for information about958the kernel image. It is readonly, because the kernel cannot rely on a959bootloader copying its contents anywhere, but that is OK; if it becomes960necessary it can still contain data items that an enabled bootloader would be961expected to copy into a setup_data chunk.962 963All kernel_info data should be part of this structure. Fixed size data have to964be put before kernel_info_var_len_data label. Variable size data have to be put965after kernel_info_var_len_data label. Each chunk of variable size data has to966be prefixed with header/magic and its size, e.g.::967 968 kernel_info:969 .ascii "LToP" /* Header, Linux top (structure). */970 .long kernel_info_var_len_data - kernel_info971 .long kernel_info_end - kernel_info972 .long 0x01234567 /* Some fixed size data for the bootloaders. */973 kernel_info_var_len_data:974 example_struct: /* Some variable size data for the bootloaders. */975 .ascii "0123" /* Header/Magic. */976 .long example_struct_end - example_struct977 .ascii "Struct"978 .long 0x89012345979 example_struct_end:980 example_strings: /* Some variable size data for the bootloaders. */981 .ascii "ABCD" /* Header/Magic. */982 .long example_strings_end - example_strings983 .asciz "String_0"984 .asciz "String_1"985 example_strings_end:986 kernel_info_end:987 988This way the kernel_info is self-contained blob.989 990.. note::991 Each variable size data header/magic can be any 4-character string,992 without \0 at the end of the string, which does not collide with993 existing variable length data headers/magics.994 995 996Details of the kernel_info Fields997=================================998 999============ ========1000Field name: header1001Offset/size: 0x0000/41002============ ========1003 1004 Contains the magic number "LToP" (0x506f544c).1005 1006============ ========1007Field name: size1008Offset/size: 0x0004/41009============ ========1010 1011 This field contains the size of the kernel_info including kernel_info.header.1012 It does not count kernel_info.kernel_info_var_len_data size. This field should be1013 used by the bootloaders to detect supported fixed size fields in the kernel_info1014 and beginning of kernel_info.kernel_info_var_len_data.1015 1016============ ========1017Field name: size_total1018Offset/size: 0x0008/41019============ ========1020 1021 This field contains the size of the kernel_info including kernel_info.header1022 and kernel_info.kernel_info_var_len_data.1023 1024============ ==============1025Field name: setup_type_max1026Offset/size: 0x000c/41027============ ==============1028 1029 This field contains maximal allowed type for setup_data and setup_indirect structs.1030 1031 1032The Image Checksum1033==================1034 1035From boot protocol version 2.08 onwards the CRC-32 is calculated over1036the entire file using the characteristic polynomial 0x04C11DB7 and an1037initial remainder of 0xffffffff. The checksum is appended to the1038file; therefore the CRC of the file up to the limit specified in the1039syssize field of the header is always 0.1040 1041 1042The Kernel Command Line1043=======================1044 1045The kernel command line has become an important way for the boot1046loader to communicate with the kernel. Some of its options are also1047relevant to the boot loader itself, see "special command line options"1048below.1049 1050The kernel command line is a null-terminated string. The maximum1051length can be retrieved from the field cmdline_size. Before protocol1052version 2.06, the maximum was 255 characters. A string that is too1053long will be automatically truncated by the kernel.1054 1055If the boot protocol version is 2.02 or later, the address of the1056kernel command line is given by the header field cmd_line_ptr (see1057above.) This address can be anywhere between the end of the setup1058heap and 0xA0000.1059 1060If the protocol version is *not* 2.02 or higher, the kernel1061command line is entered using the following protocol:1062 1063 - At offset 0x0020 (word), "cmd_line_magic", enter the magic1064 number 0xA33F.1065 1066 - At offset 0x0022 (word), "cmd_line_offset", enter the offset1067 of the kernel command line (relative to the start of the1068 real-mode kernel).1069 1070 - The kernel command line *must* be within the memory region1071 covered by setup_move_size, so you may need to adjust this1072 field.1073 1074 1075Memory Layout of The Real-Mode Code1076===================================1077 1078The real-mode code requires a stack/heap to be set up, as well as1079memory allocated for the kernel command line. This needs to be done1080in the real-mode accessible memory in bottom megabyte.1081 1082It should be noted that modern machines often have a sizable Extended1083BIOS Data Area (EBDA). As a result, it is advisable to use as little1084of the low megabyte as possible.1085 1086Unfortunately, under the following circumstances the 0x90000 memory1087segment has to be used:1088 1089 - When loading a zImage kernel ((loadflags & 0x01) == 0).1090 - When loading a 2.01 or earlier boot protocol kernel.1091 1092.. note::1093 For the 2.00 and 2.01 boot protocols, the real-mode code1094 can be loaded at another address, but it is internally1095 relocated to 0x90000. For the "old" protocol, the1096 real-mode code must be loaded at 0x90000.1097 1098When loading at 0x90000, avoid using memory above 0x9a000.1099 1100For boot protocol 2.02 or higher, the command line does not have to be1101located in the same 64K segment as the real-mode setup code; it is1102thus permitted to give the stack/heap the full 64K segment and locate1103the command line above it.1104 1105The kernel command line should not be located below the real-mode1106code, nor should it be located in high memory.1107 1108 1109Sample Boot Configuration1110=========================1111 1112As a sample configuration, assume the following layout of the real1113mode segment.1114 1115 When loading below 0x90000, use the entire segment:1116 1117 ============= ===================1118 0x0000-0x7fff Real mode kernel1119 0x8000-0xdfff Stack and heap1120 0xe000-0xffff Kernel command line1121 ============= ===================1122 1123 When loading at 0x90000 OR the protocol version is 2.01 or earlier:1124 1125 ============= ===================1126 0x0000-0x7fff Real mode kernel1127 0x8000-0x97ff Stack and heap1128 0x9800-0x9fff Kernel command line1129 ============= ===================1130 1131Such a boot loader should enter the following fields in the header::1132 1133 unsigned long base_ptr; /* base address for real-mode segment */1134 1135 if ( setup_sects == 0 ) {1136 setup_sects = 4;1137 }1138 1139 if ( protocol >= 0x0200 ) {1140 type_of_loader = <type code>;1141 if ( loading_initrd ) {1142 ramdisk_image = <initrd_address>;1143 ramdisk_size = <initrd_size>;1144 }1145 1146 if ( protocol >= 0x0202 && loadflags & 0x01 )1147 heap_end = 0xe000;1148 else1149 heap_end = 0x9800;1150 1151 if ( protocol >= 0x0201 ) {1152 heap_end_ptr = heap_end - 0x200;1153 loadflags |= 0x80; /* CAN_USE_HEAP */1154 }1155 1156 if ( protocol >= 0x0202 ) {1157 cmd_line_ptr = base_ptr + heap_end;1158 strcpy(cmd_line_ptr, cmdline);1159 } else {1160 cmd_line_magic = 0xA33F;1161 cmd_line_offset = heap_end;1162 setup_move_size = heap_end + strlen(cmdline)+1;1163 strcpy(base_ptr+cmd_line_offset, cmdline);1164 }1165 } else {1166 /* Very old kernel */1167 1168 heap_end = 0x9800;1169 1170 cmd_line_magic = 0xA33F;1171 cmd_line_offset = heap_end;1172 1173 /* A very old kernel MUST have its real-mode code1174 loaded at 0x90000 */1175 1176 if ( base_ptr != 0x90000 ) {1177 /* Copy the real-mode kernel */1178 memcpy(0x90000, base_ptr, (setup_sects+1)*512);1179 base_ptr = 0x90000; /* Relocated */1180 }1181 1182 strcpy(0x90000+cmd_line_offset, cmdline);1183 1184 /* It is recommended to clear memory up to the 32K mark */1185 memset(0x90000 + (setup_sects+1)*512, 0,1186 (64-(setup_sects+1))*512);1187 }1188 1189 1190Loading The Rest of The Kernel1191==============================1192 1193The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*5121194in the kernel file (again, if setup_sects == 0 the real value is 4.)1195It should be loaded at address 0x10000 for Image/zImage kernels and11960x100000 for bzImage kernels.1197 1198The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x011199bit (LOAD_HIGH) in the loadflags field is set::1200 1201 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);1202 load_address = is_bzImage ? 0x100000 : 0x10000;1203 1204Note that Image/zImage kernels can be up to 512K in size, and thus use1205the entire 0x10000-0x90000 range of memory. This means it is pretty1206much a requirement for these kernels to load the real-mode part at12070x90000. bzImage kernels allow much more flexibility.1208 1209Special Command Line Options1210============================1211 1212If the command line provided by the boot loader is entered by the1213user, the user may expect the following command line options to work.1214They should normally not be deleted from the kernel command line even1215though not all of them are actually meaningful to the kernel. Boot1216loader authors who need additional command line options for the boot1217loader itself should get them registered in1218Documentation/admin-guide/kernel-parameters.rst to make sure they will not1219conflict with actual kernel options now or in the future.1220 1221 vga=<mode>1222 <mode> here is either an integer (in C notation, either1223 decimal, octal, or hexadecimal) or one of the strings1224 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"1225 (meaning 0xFFFD). This value should be entered into the1226 vid_mode field, as it is used by the kernel before the command1227 line is parsed.1228 1229 mem=<size>1230 <size> is an integer in C notation optionally followed by1231 (case insensitive) K, M, G, T, P or E (meaning << 10, << 20,1232 << 30, << 40, << 50 or << 60). This specifies the end of1233 memory to the kernel. This affects the possible placement of1234 an initrd, since an initrd should be placed near end of1235 memory. Note that this is an option to *both* the kernel and1236 the bootloader!1237 1238 initrd=<file>1239 An initrd should be loaded. The meaning of <file> is1240 obviously bootloader-dependent, and some boot loaders1241 (e.g. LILO) do not have such a command.1242 1243In addition, some boot loaders add the following options to the1244user-specified command line:1245 1246 BOOT_IMAGE=<file>1247 The boot image which was loaded. Again, the meaning of <file>1248 is obviously bootloader-dependent.1249 1250 auto1251 The kernel was booted without explicit user intervention.1252 1253If these options are added by the boot loader, it is highly1254recommended that they are located *first*, before the user-specified1255or configuration-specified command line. Otherwise, "init=/bin/sh"1256gets confused by the "auto" option.1257 1258 1259Running the Kernel1260==================1261 1262The kernel is started by jumping to the kernel entry point, which is1263located at *segment* offset 0x20 from the start of the real mode1264kernel. This means that if you loaded your real-mode kernel code at12650x90000, the kernel entry point is 9020:0000.1266 1267At entry, ds = es = ss should point to the start of the real-mode1268kernel code (0x9000 if the code is loaded at 0x90000), sp should be1269set up properly, normally pointing to the top of the heap, and1270interrupts should be disabled. Furthermore, to guard against bugs in1271the kernel, it is recommended that the boot loader sets fs = gs = ds =1272es = ss.1273 1274In our example from above, we would do::1275 1276 /* Note: in the case of the "old" kernel protocol, base_ptr must1277 be == 0x90000 at this point; see the previous sample code */1278 1279 seg = base_ptr >> 4;1280 1281 cli(); /* Enter with interrupts disabled! */1282 1283 /* Set up the real-mode kernel stack */1284 _SS = seg;1285 _SP = heap_end;1286 1287 _DS = _ES = _FS = _GS = seg;1288 jmp_far(seg+0x20, 0); /* Run the kernel */1289 1290If your boot sector accesses a floppy drive, it is recommended to1291switch off the floppy motor before running the kernel, since the1292kernel boot leaves interrupts off and thus the motor will not be1293switched off, especially if the loaded kernel has the floppy driver as1294a demand-loaded module!1295 1296 1297Advanced Boot Loader Hooks1298==========================1299 1300If the boot loader runs in a particularly hostile environment (such as1301LOADLIN, which runs under DOS) it may be impossible to follow the1302standard memory location requirements. Such a boot loader may use the1303following hooks that, if set, are invoked by the kernel at the1304appropriate time. The use of these hooks should probably be1305considered an absolutely last resort!1306 1307IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and1308%edi across invocation.1309 1310 realmode_swtch:1311 A 16-bit real mode far subroutine invoked immediately before1312 entering protected mode. The default routine disables NMI, so1313 your routine should probably do so, too.1314 1315 code32_start:1316 A 32-bit flat-mode routine *jumped* to immediately after the1317 transition to protected mode, but before the kernel is1318 uncompressed. No segments, except CS, are guaranteed to be1319 set up (current kernels do, but older ones do not); you should1320 set them up to BOOT_DS (0x18) yourself.1321 1322 After completing your hook, you should jump to the address1323 that was in this field before your boot loader overwrote it1324 (relocated, if appropriate.)1325 1326 132732-bit Boot Protocol1328====================1329 1330For machine with some new BIOS other than legacy BIOS, such as EFI,1331LinuxBIOS, etc, and kexec, the 16-bit real mode setup code in kernel1332based on legacy BIOS can not be used, so a 32-bit boot protocol needs1333to be defined.1334 1335In 32-bit boot protocol, the first step in loading a Linux kernel1336should be to setup the boot parameters (struct boot_params,1337traditionally known as "zero page"). The memory for struct boot_params1338should be allocated and initialized to all zero. Then the setup header1339from offset 0x01f1 of kernel image on should be loaded into struct1340boot_params and examined. The end of setup header can be calculated as1341follow::1342 1343 0x0202 + byte value at offset 0x02011344 1345In addition to read/modify/write the setup header of the struct1346boot_params as that of 16-bit boot protocol, the boot loader should1347also fill the additional fields of the struct boot_params as1348described in chapter Documentation/arch/x86/zero-page.rst.1349 1350After setting up the struct boot_params, the boot loader can load the135132/64-bit kernel in the same way as that of 16-bit boot protocol.1352 1353In 32-bit boot protocol, the kernel is started by jumping to the135432-bit kernel entry point, which is the start address of loaded135532/64-bit kernel.1356 1357At entry, the CPU must be in 32-bit protected mode with paging1358disabled; a GDT must be loaded with the descriptors for selectors1359__BOOT_CS(0x10) and __BOOT_DS(0x18); both descriptors must be 4G flat1360segment; __BOOT_CS must have execute/read permission, and __BOOT_DS1361must have read/write permission; CS must be __BOOT_CS and DS, ES, SS1362must be __BOOT_DS; interrupt must be disabled; %esi must hold the base1363address of the struct boot_params; %ebp, %edi and %ebx must be zero.1364 136564-bit Boot Protocol1366====================1367 1368For machine with 64bit cpus and 64bit kernel, we could use 64bit bootloader1369and we need a 64-bit boot protocol.1370 1371In 64-bit boot protocol, the first step in loading a Linux kernel1372should be to setup the boot parameters (struct boot_params,1373traditionally known as "zero page"). The memory for struct boot_params1374could be allocated anywhere (even above 4G) and initialized to all zero.1375Then, the setup header at offset 0x01f1 of kernel image on should be1376loaded into struct boot_params and examined. The end of setup header1377can be calculated as follows::1378 1379 0x0202 + byte value at offset 0x02011380 1381In addition to read/modify/write the setup header of the struct1382boot_params as that of 16-bit boot protocol, the boot loader should1383also fill the additional fields of the struct boot_params as described1384in chapter Documentation/arch/x86/zero-page.rst.1385 1386After setting up the struct boot_params, the boot loader can load138764-bit kernel in the same way as that of 16-bit boot protocol, but1388kernel could be loaded above 4G.1389 1390In 64-bit boot protocol, the kernel is started by jumping to the139164-bit kernel entry point, which is the start address of loaded139264-bit kernel plus 0x200.1393 1394At entry, the CPU must be in 64-bit mode with paging enabled.1395The range with setup_header.init_size from start address of loaded1396kernel and zero page and command line buffer get ident mapping;1397a GDT must be loaded with the descriptors for selectors1398__BOOT_CS(0x10) and __BOOT_DS(0x18); both descriptors must be 4G flat1399segment; __BOOT_CS must have execute/read permission, and __BOOT_DS1400must have read/write permission; CS must be __BOOT_CS and DS, ES, SS1401must be __BOOT_DS; interrupt must be disabled; %rsi must hold the base1402address of the struct boot_params.1403 1404EFI Handover Protocol (deprecated)1405==================================1406 1407This protocol allows boot loaders to defer initialisation to the EFI1408boot stub. The boot loader is required to load the kernel/initrd(s)1409from the boot media and jump to the EFI handover protocol entry point1410which is hdr->handover_offset bytes from the beginning of1411startup_{32,64}.1412 1413The boot loader MUST respect the kernel's PE/COFF metadata when it comes1414to section alignment, the memory footprint of the executable image beyond1415the size of the file itself, and any other aspect of the PE/COFF header1416that may affect correct operation of the image as a PE/COFF binary in the1417execution context provided by the EFI firmware.1418 1419The function prototype for the handover entry point looks like this::1420 1421 efi_stub_entry(void *handle, efi_system_table_t *table, struct boot_params *bp)1422 1423'handle' is the EFI image handle passed to the boot loader by the EFI1424firmware, 'table' is the EFI system table - these are the first two1425arguments of the "handoff state" as described in section 2.3 of the1426UEFI specification. 'bp' is the boot loader-allocated boot params.1427 1428The boot loader *must* fill out the following fields in bp::1429 1430 - hdr.cmd_line_ptr1431 - hdr.ramdisk_image (if applicable)1432 - hdr.ramdisk_size (if applicable)1433 1434All other fields should be zero.1435 1436NOTE: The EFI Handover Protocol is deprecated in favour of the ordinary PE/COFF1437 entry point, combined with the LINUX_EFI_INITRD_MEDIA_GUID based initrd1438 loading protocol (refer to [0] for an example of the bootloader side of1439 this), which removes the need for any knowledge on the part of the EFI1440 bootloader regarding the internal representation of boot_params or any1441 requirements/limitations regarding the placement of the command line1442 and ramdisk in memory, or the placement of the kernel image itself.1443 1444[0] https://github.com/u-boot/u-boot/commit/ec80b4735a593961fe701cc3a5d717d4739b0fd01445