1 THE LINUX/x86 BOOT PROTOCOL
2 ---------------------------
4 On the x86 platform, the Linux kernel uses a rather complicated boot
5 convention. This has evolved partially due to historical aspects, as
6 well as the desire in the early days to have the kernel itself be a
7 bootable image, the complicated PC memory model and due to changed
8 expectations in the PC industry caused by the effective demise of
9 real-mode DOS as a mainstream operating system.
11 Currently, the following versions of the Linux/x86 boot protocol exist.
13 Old kernels: zImage/Image support only. Some very early kernels
14 may not even support a command line.
16 Protocol 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
22 Protocol 2.01: (Kernel 1.3.76) Added a heap overrun warning.
24 Protocol 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
31 Protocol 2.03: (Kernel 2.4.18-pre1) Explicitly makes the highest possible
32 initrd address available to the bootloader.
34 Protocol 2.04: (Kernel 2.6.14) Extend the syssize field to four bytes.
36 Protocol 2.05: (Kernel 2.6.20) Make protected mode kernel relocatable.
37 Introduce relocatable_kernel and kernel_alignment fields.
39 Protocol 2.06: (Kernel 2.6.22) Added a field that contains the size of
40 the boot command line.
42 Protocol 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.
46 Protocol 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.
50 Protocol 2.09: (Kernel 2.6.26) Added a field of 64-bit physical
51 pointer to single linked list of struct setup_data.
53 Protocol 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.
57 Protocol 2.11: (Kernel 3.6) Added a field for offset of EFI handover
60 Protocol 2.12: (Kernel 3.8) Added the xloadflags field and extension fields
61 to struct boot_params for loading bzImage and ramdisk
64 Protocol 2.13: (Kernel 3.14) Support 32- and 64-bit flags being set in
65 xloadflags to support booting a 64-bit kernel from 32-bit
70 The traditional memory map for the kernel loader, used for Image or
71 zImage kernels, typically looks like:
74 0A0000 +------------------------+
75 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA.
76 09A000 +------------------------+
78 | Stack/heap | For use by the kernel real-mode code.
79 098000 +------------------------+
80 | Kernel setup | The kernel real-mode code.
81 090200 +------------------------+
82 | Kernel boot sector | The kernel legacy boot sector.
83 090000 +------------------------+
84 | Protected-mode kernel | The bulk of the kernel image.
85 010000 +------------------------+
86 | Boot loader | <- Boot sector entry point 0000:7C00
87 001000 +------------------------+
88 | Reserved for MBR/BIOS |
89 000800 +------------------------+
90 | Typically used by MBR |
91 000600 +------------------------+
93 000000 +------------------------+
96 When using bzImage, the protected-mode kernel was relocated to
97 0x100000 ("high memory"), and the kernel real-mode block (boot sector,
98 setup, and stack/heap) was made relocatable to any address between
99 0x10000 and end of low memory. Unfortunately, in protocols 2.00 and
100 2.01 the 0x90000+ memory range is still used internally by the kernel;
101 the 2.02 protocol resolves that problem.
103 It is desirable to keep the "memory ceiling" -- the highest point in
104 low memory touched by the boot loader -- as low as possible, since
105 some newer BIOSes have begun to allocate some rather large amounts of
106 memory, called the Extended BIOS Data Area, near the top of low
107 memory. The boot loader should use the "INT 12h" BIOS call to verify
108 how much low memory is available.
110 Unfortunately, if INT 12h reports that the amount of memory is too
111 low, there is usually nothing the boot loader can do but to report an
112 error to the user. The boot loader should therefore be designed to
113 take up as little space in low memory as it reasonably can. For
114 zImage or old bzImage kernels, which need data written into the
115 0x90000 segment, the boot loader should make sure not to use memory
116 above the 0x9A000 point; too many BIOSes will break above that point.
118 For a modern bzImage kernel with boot protocol version >= 2.02, a
119 memory layout like the following is suggested:
122 | Protected-mode kernel |
123 100000 +------------------------+
125 0A0000 +------------------------+
126 | Reserved for BIOS | Leave as much as possible unused
128 | Command line | (Can also be below the X+10000 mark)
129 X+10000 +------------------------+
130 | Stack/heap | For use by the kernel real-mode code.
131 X+08000 +------------------------+
132 | Kernel setup | The kernel real-mode code.
133 | Kernel boot sector | The kernel legacy boot sector.
134 X +------------------------+
135 | Boot loader | <- Boot sector entry point 0000:7C00
136 001000 +------------------------+
137 | Reserved for MBR/BIOS |
138 000800 +------------------------+
139 | Typically used by MBR |
140 000600 +------------------------+
142 000000 +------------------------+
144 ... where the address X is as low as the design of the boot loader
148 **** THE REAL-MODE KERNEL HEADER
150 In the following text, and anywhere in the kernel boot sequence, "a
151 sector" refers to 512 bytes. It is independent of the actual sector
152 size of the underlying medium.
154 The first step in loading a Linux kernel should be to load the
155 real-mode code (boot sector and setup code) and then examine the
156 following header at offset 0x01f1. The real-mode code can total up to
157 32K, although the boot loader may choose to load only the first two
158 sectors (1K) and then examine the bootup sector size.
160 The header looks like:
162 Offset Proto Name Meaning
165 01F1/1 ALL(1 setup_sects The size of the setup in sectors
166 01F2/2 ALL root_flags If set, the root is mounted readonly
167 01F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras
168 01F8/2 ALL ram_size DO NOT USE - for bootsect.S use only
169 01FA/2 ALL vid_mode Video mode control
170 01FC/2 ALL root_dev Default root device number
171 01FE/2 ALL boot_flag 0xAA55 magic number
172 0200/2 2.00+ jump Jump instruction
173 0202/4 2.00+ header Magic signature "HdrS"
174 0206/2 2.00+ version Boot protocol version supported
175 0208/4 2.00+ realmode_swtch Boot loader hook (see below)
176 020C/2 2.00+ start_sys_seg The load-low segment (0x1000) (obsolete)
177 020E/2 2.00+ kernel_version Pointer to kernel version string
178 0210/1 2.00+ type_of_loader Boot loader identifier
179 0211/1 2.00+ loadflags Boot protocol option flags
180 0212/2 2.00+ setup_move_size Move to high memory size (used with hooks)
181 0214/4 2.00+ code32_start Boot loader hook (see below)
182 0218/4 2.00+ ramdisk_image initrd load address (set by boot loader)
183 021C/4 2.00+ ramdisk_size initrd size (set by boot loader)
184 0220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only
185 0224/2 2.01+ heap_end_ptr Free memory after setup end
186 0226/1 2.02+(3 ext_loader_ver Extended boot loader version
187 0227/1 2.02+(3 ext_loader_type Extended boot loader ID
188 0228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line
189 022C/4 2.03+ initrd_addr_max Highest legal initrd address
190 0230/4 2.05+ kernel_alignment Physical addr alignment required for kernel
191 0234/1 2.05+ relocatable_kernel Whether kernel is relocatable or not
192 0235/1 2.10+ min_alignment Minimum alignment, as a power of two
193 0236/2 2.12+ xloadflags Boot protocol option flags
194 0238/4 2.06+ cmdline_size Maximum size of the kernel command line
195 023C/4 2.07+ hardware_subarch Hardware subarchitecture
196 0240/8 2.07+ hardware_subarch_data Subarchitecture-specific data
197 0248/4 2.08+ payload_offset Offset of kernel payload
198 024C/4 2.08+ payload_length Length of kernel payload
199 0250/8 2.09+ setup_data 64-bit physical pointer to linked list
201 0258/8 2.10+ pref_address Preferred loading address
202 0260/4 2.10+ init_size Linear memory required during initialization
203 0264/4 2.11+ handover_offset Offset of handover entry point
205 (1) For backwards compatibility, if the setup_sects field contains 0, the
208 (2) For boot protocol prior to 2.04, the upper two bytes of the syssize
209 field are unusable, which means the size of a bzImage kernel
210 cannot be determined.
212 (3) Ignored, but safe to set, for boot protocols 2.02-2.09.
214 If the "HdrS" (0x53726448) magic number is not found at offset 0x202,
215 the boot protocol version is "old". Loading an old kernel, the
216 following parameters should be assumed:
220 Real-mode kernel must be located at 0x90000.
222 Otherwise, the "version" field contains the protocol version,
223 e.g. protocol version 2.01 will contain 0x0201 in this field. When
224 setting fields in the header, you must make sure only to set fields
225 supported by the protocol version in use.
228 **** DETAILS OF HEADER FIELDS
230 For each field, some are information from the kernel to the bootloader
231 ("read"), some are expected to be filled out by the bootloader
232 ("write"), and some are expected to be read and modified by the
233 bootloader ("modify").
235 All general purpose boot loaders should write the fields marked
236 (obligatory). Boot loaders who want to load the kernel at a
237 nonstandard address should fill in the fields marked (reloc); other
238 boot loaders can ignore those fields.
240 The byte order of all fields is littleendian (this is x86, after all.)
242 Field name: setup_sects
247 The size of the setup code in 512-byte sectors. If this field is
248 0, the real value is 4. The real-mode code consists of the boot
249 sector (always one 512-byte sector) plus the setup code.
251 Field name: root_flags
252 Type: modify (optional)
256 If this field is nonzero, the root defaults to readonly. The use of
257 this field is deprecated; use the "ro" or "rw" options on the
258 command line instead.
262 Offset/size: 0x1f4/4 (protocol 2.04+) 0x1f4/2 (protocol ALL)
265 The size of the protected-mode code in units of 16-byte paragraphs.
266 For protocol versions older than 2.04 this field is only two bytes
267 wide, and therefore cannot be trusted for the size of a kernel if
268 the LOAD_HIGH flag is set.
271 Type: kernel internal
275 This field is obsolete.
278 Type: modify (obligatory)
281 Please see the section on SPECIAL COMMAND LINE OPTIONS.
284 Type: modify (optional)
288 The default root device device number. The use of this field is
289 deprecated, use the "root=" option on the command line instead.
291 Field name: boot_flag
296 Contains 0xAA55. This is the closest thing old Linux kernels have
304 Contains an x86 jump instruction, 0xEB followed by a signed offset
305 relative to byte 0x202. This can be used to determine the size of
313 Contains the magic number "HdrS" (0x53726448).
320 Contains the boot protocol version, in (major << 8)+minor format,
321 e.g. 0x0204 for version 2.04, and 0x0a11 for a hypothetical version
324 Field name: realmode_swtch
325 Type: modify (optional)
329 Boot loader hook (see ADVANCED BOOT LOADER HOOKS below.)
331 Field name: start_sys_seg
336 The load low segment (0x1000). Obsolete.
338 Field name: kernel_version
343 If set to a nonzero value, contains a pointer to a NUL-terminated
344 human-readable kernel version number string, less 0x200. This can
345 be used to display the kernel version to the user. This value
346 should be less than (0x200*setup_sects).
348 For example, if this value is set to 0x1c00, the kernel version
349 number string can be found at offset 0x1e00 in the kernel file.
350 This is a valid value if and only if the "setup_sects" field
351 contains the value 15 or higher, as:
353 0x1c00 < 15*0x200 (= 0x1e00) but
354 0x1c00 >= 14*0x200 (= 0x1c00)
356 0x1c00 >> 9 = 14, so the minimum value for setup_secs is 15.
358 Field name: type_of_loader
359 Type: write (obligatory)
363 If your boot loader has an assigned id (see table below), enter
364 0xTV here, where T is an identifier for the boot loader and V is
365 a version number. Otherwise, enter 0xFF here.
367 For boot loader IDs above T = 0xD, write T = 0xE to this field and
368 write the extended ID minus 0x10 to the ext_loader_type field.
369 Similarly, the ext_loader_ver field can be used to provide more than
370 four bits for the bootloader version.
372 For example, for T = 0x15, V = 0x234, write:
374 type_of_loader <- 0xE4
375 ext_loader_type <- 0x05
376 ext_loader_ver <- 0x23
378 Assigned boot loader ids (hexadecimal):
380 0 LILO (0x00 reserved for pre-2.00 bootloader)
382 2 bootsect-loader (0x20, all other values reserved)
384 4 Etherboot/gPXE/iPXE
391 C Arcturus Networks uCbootloader
393 E Extended (see ext_loader_type)
394 F Special (0xFF = undefined)
396 11 Minimal Linux Bootloader <http://sebastian-plotz.blogspot.de>
397 12 OVMF UEFI virtualization stack
399 Please contact <hpa@zytor.com> if you need a bootloader ID
402 Field name: loadflags
403 Type: modify (obligatory)
407 This field is a bitmask.
409 Bit 0 (read): LOADED_HIGH
410 - If 0, the protected-mode code is loaded at 0x10000.
411 - If 1, the protected-mode code is loaded at 0x100000.
413 Bit 1 (kernel internal): KASLR_FLAG
414 - Used internally by the compressed kernel to communicate
415 KASLR status to kernel proper.
417 If 0, KASLR disabled.
419 Bit 5 (write): QUIET_FLAG
420 - If 0, print early messages.
421 - If 1, suppress early messages.
422 This requests to the kernel (decompressor and early
423 kernel) to not write early messages that require
424 accessing the display hardware directly.
426 Bit 6 (write): KEEP_SEGMENTS
428 - If 0, reload the segment registers in the 32bit entry point.
429 - If 1, do not reload the segment registers in the 32bit entry point.
430 Assume that %cs %ds %ss %es are all set to flat segments with
431 a base of 0 (or the equivalent for their environment).
433 Bit 7 (write): CAN_USE_HEAP
434 Set this bit to 1 to indicate that the value entered in the
435 heap_end_ptr is valid. If this field is clear, some setup code
436 functionality will be disabled.
438 Field name: setup_move_size
439 Type: modify (obligatory)
443 When using protocol 2.00 or 2.01, if the real mode kernel is not
444 loaded at 0x90000, it gets moved there later in the loading
445 sequence. Fill in this field if you want additional data (such as
446 the kernel command line) moved in addition to the real-mode kernel
449 The unit is bytes starting with the beginning of the boot sector.
451 This field is can be ignored when the protocol is 2.02 or higher, or
452 if the real-mode code is loaded at 0x90000.
454 Field name: code32_start
455 Type: modify (optional, reloc)
459 The address to jump to in protected mode. This defaults to the load
460 address of the kernel, and can be used by the boot loader to
461 determine the proper load address.
463 This field can be modified for two purposes:
465 1. as a boot loader hook (see ADVANCED BOOT LOADER HOOKS below.)
467 2. if a bootloader which does not install a hook loads a
468 relocatable kernel at a nonstandard address it will have to modify
469 this field to point to the load address.
471 Field name: ramdisk_image
472 Type: write (obligatory)
476 The 32-bit linear address of the initial ramdisk or ramfs. Leave at
477 zero if there is no initial ramdisk/ramfs.
479 Field name: ramdisk_size
480 Type: write (obligatory)
484 Size of the initial ramdisk or ramfs. Leave at zero if there is no
485 initial ramdisk/ramfs.
487 Field name: bootsect_kludge
488 Type: kernel internal
492 This field is obsolete.
494 Field name: heap_end_ptr
495 Type: write (obligatory)
499 Set this field to the offset (from the beginning of the real-mode
500 code) of the end of the setup stack/heap, minus 0x0200.
502 Field name: ext_loader_ver
503 Type: write (optional)
507 This field is used as an extension of the version number in the
508 type_of_loader field. The total version number is considered to be
509 (type_of_loader & 0x0f) + (ext_loader_ver << 4).
511 The use of this field is boot loader specific. If not written, it
514 Kernels prior to 2.6.31 did not recognize this field, but it is safe
515 to write for protocol version 2.02 or higher.
517 Field name: ext_loader_type
518 Type: write (obligatory if (type_of_loader & 0xf0) == 0xe0)
522 This field is used as an extension of the type number in
523 type_of_loader field. If the type in type_of_loader is 0xE, then
524 the actual type is (ext_loader_type + 0x10).
526 This field is ignored if the type in type_of_loader is not 0xE.
528 Kernels prior to 2.6.31 did not recognize this field, but it is safe
529 to write for protocol version 2.02 or higher.
531 Field name: cmd_line_ptr
532 Type: write (obligatory)
536 Set this field to the linear address of the kernel command line.
537 The kernel command line can be located anywhere between the end of
538 the setup heap and 0xA0000; it does not have to be located in the
539 same 64K segment as the real-mode code itself.
541 Fill in this field even if your boot loader does not support a
542 command line, in which case you can point this to an empty string
543 (or better yet, to the string "auto".) If this field is left at
544 zero, the kernel will assume that your boot loader does not support
547 Field name: initrd_addr_max
552 The maximum address that may be occupied by the initial
553 ramdisk/ramfs contents. For boot protocols 2.02 or earlier, this
554 field is not present, and the maximum address is 0x37FFFFFF. (This
555 address is defined as the address of the highest safe byte, so if
556 your ramdisk is exactly 131072 bytes long and this field is
557 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)
559 Field name: kernel_alignment
560 Type: read/modify (reloc)
562 Protocol: 2.05+ (read), 2.10+ (modify)
564 Alignment unit required by the kernel (if relocatable_kernel is
565 true.) A relocatable kernel that is loaded at an alignment
566 incompatible with the value in this field will be realigned during
567 kernel initialization.
569 Starting with protocol version 2.10, this reflects the kernel
570 alignment preferred for optimal performance; it is possible for the
571 loader to modify this field to permit a lesser alignment. See the
572 min_alignment and pref_address field below.
574 Field name: relocatable_kernel
579 If this field is nonzero, the protected-mode part of the kernel can
580 be loaded at any address that satisfies the kernel_alignment field.
581 After loading, the boot loader must set the code32_start field to
582 point to the loaded code, or to a boot loader hook.
584 Field name: min_alignment
589 This field, if nonzero, indicates as a power of two the minimum
590 alignment required, as opposed to preferred, by the kernel to boot.
591 If a boot loader makes use of this field, it should update the
592 kernel_alignment field with the alignment unit desired; typically:
594 kernel_alignment = 1 << min_alignment
596 There may be a considerable performance cost with an excessively
597 misaligned kernel. Therefore, a loader should typically try each
598 power-of-two alignment from kernel_alignment down to this alignment.
600 Field name: xloadflags
605 This field is a bitmask.
607 Bit 0 (read): XLF_KERNEL_64
608 - If 1, this kernel has the legacy 64-bit entry point at 0x200.
610 Bit 1 (read): XLF_CAN_BE_LOADED_ABOVE_4G
611 - If 1, kernel/boot_params/cmdline/ramdisk can be above 4G.
613 Bit 2 (read): XLF_EFI_HANDOVER_32
614 - If 1, the kernel supports the 32-bit EFI handoff entry point
615 given at handover_offset.
617 Bit 3 (read): XLF_EFI_HANDOVER_64
618 - If 1, the kernel supports the 64-bit EFI handoff entry point
619 given at handover_offset + 0x200.
621 Bit 4 (read): XLF_EFI_KEXEC
622 - If 1, the kernel supports kexec EFI boot with EFI runtime support.
624 Field name: cmdline_size
629 The maximum size of the command line without the terminating
630 zero. This means that the command line can contain at most
631 cmdline_size characters. With protocol version 2.05 and earlier, the
632 maximum size was 255.
634 Field name: hardware_subarch
635 Type: write (optional, defaults to x86/PC)
639 In a paravirtualized environment the hardware low level architectural
640 pieces such as interrupt handling, page table handling, and
641 accessing process control registers needs to be done differently.
643 This field allows the bootloader to inform the kernel we are in one
644 one of those environments.
646 0x00000000 The default x86/PC environment
649 0x00000003 Moorestown MID
650 0x00000004 CE4100 TV Platform
652 Field name: hardware_subarch_data
653 Type: write (subarch-dependent)
657 A pointer to data that is specific to hardware subarch
658 This field is currently unused for the default x86/PC environment,
661 Field name: payload_offset
666 If non-zero then this field contains the offset from the beginning
667 of the protected-mode code to the payload.
669 The payload may be compressed. The format of both the compressed and
670 uncompressed data should be determined using the standard magic
671 numbers. The currently supported compression formats are gzip
672 (magic numbers 1F 8B or 1F 9E), bzip2 (magic number 42 5A), LZMA
673 (magic number 5D 00), XZ (magic number FD 37), and LZ4 (magic number
674 02 21). The uncompressed payload is currently always ELF (magic
677 Field name: payload_length
682 The length of the payload.
684 Field name: setup_data
685 Type: write (special)
689 The 64-bit physical pointer to NULL terminated single linked list of
690 struct setup_data. This is used to define a more extensible boot
691 parameters passing mechanism. The definition of struct setup_data is
701 Where, the next is a 64-bit physical pointer to the next node of
702 linked list, the next field of the last node is 0; the type is used
703 to identify the contents of data; the len is the length of data
704 field; the data holds the real payload.
706 This list may be modified at a number of points during the bootup
707 process. Therefore, when modifying this list one should always make
708 sure to consider the case where the linked list already contains
711 Field name: pref_address
716 This field, if nonzero, represents a preferred load address for the
717 kernel. A relocating bootloader should attempt to load at this
720 A non-relocatable kernel will unconditionally move itself and to run
723 Field name: init_size
727 This field indicates the amount of linear contiguous memory starting
728 at the kernel runtime start address that the kernel needs before it
729 is capable of examining its memory map. This is not the same thing
730 as the total amount of memory the kernel needs to boot, but it can
731 be used by a relocating boot loader to help select a safe load
732 address for the kernel.
734 The kernel runtime start address is determined by the following algorithm:
736 if (relocatable_kernel)
737 runtime_start = align_up(load_address, kernel_alignment)
739 runtime_start = pref_address
741 Field name: handover_offset
745 This field is the offset from the beginning of the kernel image to
746 the EFI handover protocol entry point. Boot loaders using the EFI
747 handover protocol to boot the kernel should jump to this offset.
749 See EFI HANDOVER PROTOCOL below for more details.
752 **** THE IMAGE CHECKSUM
754 From boot protocol version 2.08 onwards the CRC-32 is calculated over
755 the entire file using the characteristic polynomial 0x04C11DB7 and an
756 initial remainder of 0xffffffff. The checksum is appended to the
757 file; therefore the CRC of the file up to the limit specified in the
758 syssize field of the header is always 0.
761 **** THE KERNEL COMMAND LINE
763 The kernel command line has become an important way for the boot
764 loader to communicate with the kernel. Some of its options are also
765 relevant to the boot loader itself, see "special command line options"
768 The kernel command line is a null-terminated string. The maximum
769 length can be retrieved from the field cmdline_size. Before protocol
770 version 2.06, the maximum was 255 characters. A string that is too
771 long will be automatically truncated by the kernel.
773 If the boot protocol version is 2.02 or later, the address of the
774 kernel command line is given by the header field cmd_line_ptr (see
775 above.) This address can be anywhere between the end of the setup
778 If the protocol version is *not* 2.02 or higher, the kernel
779 command line is entered using the following protocol:
781 At offset 0x0020 (word), "cmd_line_magic", enter the magic
784 At offset 0x0022 (word), "cmd_line_offset", enter the offset
785 of the kernel command line (relative to the start of the
788 The kernel command line *must* be within the memory region
789 covered by setup_move_size, so you may need to adjust this
793 **** MEMORY LAYOUT OF THE REAL-MODE CODE
795 The real-mode code requires a stack/heap to be set up, as well as
796 memory allocated for the kernel command line. This needs to be done
797 in the real-mode accessible memory in bottom megabyte.
799 It should be noted that modern machines often have a sizable Extended
800 BIOS Data Area (EBDA). As a result, it is advisable to use as little
801 of the low megabyte as possible.
803 Unfortunately, under the following circumstances the 0x90000 memory
804 segment has to be used:
806 - When loading a zImage kernel ((loadflags & 0x01) == 0).
807 - When loading a 2.01 or earlier boot protocol kernel.
809 -> For the 2.00 and 2.01 boot protocols, the real-mode code
810 can be loaded at another address, but it is internally
811 relocated to 0x90000. For the "old" protocol, the
812 real-mode code must be loaded at 0x90000.
814 When loading at 0x90000, avoid using memory above 0x9a000.
816 For boot protocol 2.02 or higher, the command line does not have to be
817 located in the same 64K segment as the real-mode setup code; it is
818 thus permitted to give the stack/heap the full 64K segment and locate
819 the command line above it.
821 The kernel command line should not be located below the real-mode
822 code, nor should it be located in high memory.
825 **** SAMPLE BOOT CONFIGURATION
827 As a sample configuration, assume the following layout of the real
830 When loading below 0x90000, use the entire segment:
832 0x0000-0x7fff Real mode kernel
833 0x8000-0xdfff Stack and heap
834 0xe000-0xffff Kernel command line
836 When loading at 0x90000 OR the protocol version is 2.01 or earlier:
838 0x0000-0x7fff Real mode kernel
839 0x8000-0x97ff Stack and heap
840 0x9800-0x9fff Kernel command line
842 Such a boot loader should enter the following fields in the header:
844 unsigned long base_ptr; /* base address for real-mode segment */
846 if ( setup_sects == 0 ) {
850 if ( protocol >= 0x0200 ) {
851 type_of_loader = <type code>;
852 if ( loading_initrd ) {
853 ramdisk_image = <initrd_address>;
854 ramdisk_size = <initrd_size>;
857 if ( protocol >= 0x0202 && loadflags & 0x01 )
862 if ( protocol >= 0x0201 ) {
863 heap_end_ptr = heap_end - 0x200;
864 loadflags |= 0x80; /* CAN_USE_HEAP */
867 if ( protocol >= 0x0202 ) {
868 cmd_line_ptr = base_ptr + heap_end;
869 strcpy(cmd_line_ptr, cmdline);
871 cmd_line_magic = 0xA33F;
872 cmd_line_offset = heap_end;
873 setup_move_size = heap_end + strlen(cmdline)+1;
874 strcpy(base_ptr+cmd_line_offset, cmdline);
877 /* Very old kernel */
881 cmd_line_magic = 0xA33F;
882 cmd_line_offset = heap_end;
884 /* A very old kernel MUST have its real-mode code
887 if ( base_ptr != 0x90000 ) {
888 /* Copy the real-mode kernel */
889 memcpy(0x90000, base_ptr, (setup_sects+1)*512);
890 base_ptr = 0x90000; /* Relocated */
893 strcpy(0x90000+cmd_line_offset, cmdline);
895 /* It is recommended to clear memory up to the 32K mark */
896 memset(0x90000 + (setup_sects+1)*512, 0,
897 (64-(setup_sects+1))*512);
901 **** LOADING THE REST OF THE KERNEL
903 The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512
904 in the kernel file (again, if setup_sects == 0 the real value is 4.)
905 It should be loaded at address 0x10000 for Image/zImage kernels and
906 0x100000 for bzImage kernels.
908 The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01
909 bit (LOAD_HIGH) in the loadflags field is set:
911 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);
912 load_address = is_bzImage ? 0x100000 : 0x10000;
914 Note that Image/zImage kernels can be up to 512K in size, and thus use
915 the entire 0x10000-0x90000 range of memory. This means it is pretty
916 much a requirement for these kernels to load the real-mode part at
917 0x90000. bzImage kernels allow much more flexibility.
920 **** SPECIAL COMMAND LINE OPTIONS
922 If the command line provided by the boot loader is entered by the
923 user, the user may expect the following command line options to work.
924 They should normally not be deleted from the kernel command line even
925 though not all of them are actually meaningful to the kernel. Boot
926 loader authors who need additional command line options for the boot
927 loader itself should get them registered in
928 Documentation/admin-guide/kernel-parameters.rst to make sure they will not
929 conflict with actual kernel options now or in the future.
932 <mode> here is either an integer (in C notation, either
933 decimal, octal, or hexadecimal) or one of the strings
934 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"
935 (meaning 0xFFFD). This value should be entered into the
936 vid_mode field, as it is used by the kernel before the command
940 <size> is an integer in C notation optionally followed by
941 (case insensitive) K, M, G, T, P or E (meaning << 10, << 20,
942 << 30, << 40, << 50 or << 60). This specifies the end of
943 memory to the kernel. This affects the possible placement of
944 an initrd, since an initrd should be placed near end of
945 memory. Note that this is an option to *both* the kernel and
949 An initrd should be loaded. The meaning of <file> is
950 obviously bootloader-dependent, and some boot loaders
951 (e.g. LILO) do not have such a command.
953 In addition, some boot loaders add the following options to the
954 user-specified command line:
957 The boot image which was loaded. Again, the meaning of <file>
958 is obviously bootloader-dependent.
961 The kernel was booted without explicit user intervention.
963 If these options are added by the boot loader, it is highly
964 recommended that they are located *first*, before the user-specified
965 or configuration-specified command line. Otherwise, "init=/bin/sh"
966 gets confused by the "auto" option.
969 **** RUNNING THE KERNEL
971 The kernel is started by jumping to the kernel entry point, which is
972 located at *segment* offset 0x20 from the start of the real mode
973 kernel. This means that if you loaded your real-mode kernel code at
974 0x90000, the kernel entry point is 9020:0000.
976 At entry, ds = es = ss should point to the start of the real-mode
977 kernel code (0x9000 if the code is loaded at 0x90000), sp should be
978 set up properly, normally pointing to the top of the heap, and
979 interrupts should be disabled. Furthermore, to guard against bugs in
980 the kernel, it is recommended that the boot loader sets fs = gs = ds =
983 In our example from above, we would do:
985 /* Note: in the case of the "old" kernel protocol, base_ptr must
986 be == 0x90000 at this point; see the previous sample code */
990 cli(); /* Enter with interrupts disabled! */
992 /* Set up the real-mode kernel stack */
996 _DS = _ES = _FS = _GS = seg;
997 jmp_far(seg+0x20, 0); /* Run the kernel */
999 If your boot sector accesses a floppy drive, it is recommended to
1000 switch off the floppy motor before running the kernel, since the
1001 kernel boot leaves interrupts off and thus the motor will not be
1002 switched off, especially if the loaded kernel has the floppy driver as
1003 a demand-loaded module!
1006 **** ADVANCED BOOT LOADER HOOKS
1008 If the boot loader runs in a particularly hostile environment (such as
1009 LOADLIN, which runs under DOS) it may be impossible to follow the
1010 standard memory location requirements. Such a boot loader may use the
1011 following hooks that, if set, are invoked by the kernel at the
1012 appropriate time. The use of these hooks should probably be
1013 considered an absolutely last resort!
1015 IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and
1016 %edi across invocation.
1019 A 16-bit real mode far subroutine invoked immediately before
1020 entering protected mode. The default routine disables NMI, so
1021 your routine should probably do so, too.
1024 A 32-bit flat-mode routine *jumped* to immediately after the
1025 transition to protected mode, but before the kernel is
1026 uncompressed. No segments, except CS, are guaranteed to be
1027 set up (current kernels do, but older ones do not); you should
1028 set them up to BOOT_DS (0x18) yourself.
1030 After completing your hook, you should jump to the address
1031 that was in this field before your boot loader overwrote it
1032 (relocated, if appropriate.)
1035 **** 32-bit BOOT PROTOCOL
1037 For machine with some new BIOS other than legacy BIOS, such as EFI,
1038 LinuxBIOS, etc, and kexec, the 16-bit real mode setup code in kernel
1039 based on legacy BIOS can not be used, so a 32-bit boot protocol needs
1042 In 32-bit boot protocol, the first step in loading a Linux kernel
1043 should be to setup the boot parameters (struct boot_params,
1044 traditionally known as "zero page"). The memory for struct boot_params
1045 should be allocated and initialized to all zero. Then the setup header
1046 from offset 0x01f1 of kernel image on should be loaded into struct
1047 boot_params and examined. The end of setup header can be calculated as
1050 0x0202 + byte value at offset 0x0201
1052 In addition to read/modify/write the setup header of the struct
1053 boot_params as that of 16-bit boot protocol, the boot loader should
1054 also fill the additional fields of the struct boot_params as that
1055 described in zero-page.txt.
1057 After setting up the struct boot_params, the boot loader can load the
1058 32/64-bit kernel in the same way as that of 16-bit boot protocol.
1060 In 32-bit boot protocol, the kernel is started by jumping to the
1061 32-bit kernel entry point, which is the start address of loaded
1064 At entry, the CPU must be in 32-bit protected mode with paging
1065 disabled; a GDT must be loaded with the descriptors for selectors
1066 __BOOT_CS(0x10) and __BOOT_DS(0x18); both descriptors must be 4G flat
1067 segment; __BOOT_CS must have execute/read permission, and __BOOT_DS
1068 must have read/write permission; CS must be __BOOT_CS and DS, ES, SS
1069 must be __BOOT_DS; interrupt must be disabled; %esi must hold the base
1070 address of the struct boot_params; %ebp, %edi and %ebx must be zero.
1072 **** 64-bit BOOT PROTOCOL
1074 For machine with 64bit cpus and 64bit kernel, we could use 64bit bootloader
1075 and we need a 64-bit boot protocol.
1077 In 64-bit boot protocol, the first step in loading a Linux kernel
1078 should be to setup the boot parameters (struct boot_params,
1079 traditionally known as "zero page"). The memory for struct boot_params
1080 could be allocated anywhere (even above 4G) and initialized to all zero.
1081 Then, the setup header at offset 0x01f1 of kernel image on should be
1082 loaded into struct boot_params and examined. The end of setup header
1083 can be calculated as follows:
1085 0x0202 + byte value at offset 0x0201
1087 In addition to read/modify/write the setup header of the struct
1088 boot_params as that of 16-bit boot protocol, the boot loader should
1089 also fill the additional fields of the struct boot_params as described
1092 After setting up the struct boot_params, the boot loader can load
1093 64-bit kernel in the same way as that of 16-bit boot protocol, but
1094 kernel could be loaded above 4G.
1096 In 64-bit boot protocol, the kernel is started by jumping to the
1097 64-bit kernel entry point, which is the start address of loaded
1098 64-bit kernel plus 0x200.
1100 At entry, the CPU must be in 64-bit mode with paging enabled.
1101 The range with setup_header.init_size from start address of loaded
1102 kernel and zero page and command line buffer get ident mapping;
1103 a GDT must be loaded with the descriptors for selectors
1104 __BOOT_CS(0x10) and __BOOT_DS(0x18); both descriptors must be 4G flat
1105 segment; __BOOT_CS must have execute/read permission, and __BOOT_DS
1106 must have read/write permission; CS must be __BOOT_CS and DS, ES, SS
1107 must be __BOOT_DS; interrupt must be disabled; %rsi must hold the base
1108 address of the struct boot_params.
1110 **** EFI HANDOVER PROTOCOL
1112 This protocol allows boot loaders to defer initialisation to the EFI
1113 boot stub. The boot loader is required to load the kernel/initrd(s)
1114 from the boot media and jump to the EFI handover protocol entry point
1115 which is hdr->handover_offset bytes from the beginning of
1118 The function prototype for the handover entry point looks like this,
1120 efi_main(void *handle, efi_system_table_t *table, struct boot_params *bp)
1122 'handle' is the EFI image handle passed to the boot loader by the EFI
1123 firmware, 'table' is the EFI system table - these are the first two
1124 arguments of the "handoff state" as described in section 2.3 of the
1125 UEFI specification. 'bp' is the boot loader-allocated boot params.
1127 The boot loader *must* fill out the following fields in bp,
1131 o hdr.ramdisk_image (if applicable)
1132 o hdr.ramdisk_size (if applicable)
1134 All other fields should be zero.