1 THE LINUX/I386 BOOT PROTOCOL
2 ----------------------------
4 H. Peter Anvin <hpa@zytor.com>
7 On the i386 platform, the Linux kernel uses a rather complicated boot
8 convention. This has evolved partially due to historical aspects, as
9 well as the desire in the early days to have the kernel itself be a
10 bootable image, the complicated PC memory model and due to changed
11 expectations in the PC industry caused by the effective demise of
12 real-mode DOS as a mainstream operating system.
14 Currently, the following versions of the Linux/i386 boot protocol exist.
16 Old kernels: zImage/Image support only. Some very early kernels
17 may not even support a command line.
19 Protocol 2.00: (Kernel 1.3.73) Added bzImage and initrd support, as
20 well as a formalized way to communicate between the
21 boot loader and the kernel. setup.S made relocatable,
22 although the traditional setup area still assumed
25 Protocol 2.01: (Kernel 1.3.76) Added a heap overrun warning.
27 Protocol 2.02: (Kernel 2.4.0-test3-pre3) New command line protocol.
28 Lower the conventional memory ceiling. No overwrite
29 of the traditional setup area, thus making booting
30 safe for systems which use the EBDA from SMM or 32-bit
31 BIOS entry points. zImage deprecated but still
34 Protocol 2.03: (Kernel 2.4.18-pre1) Explicitly makes the highest possible
35 initrd address available to the bootloader.
37 Protocol 2.04: (Kernel 2.6.14) Extend the syssize field to four bytes.
39 Protocol 2.05: (Kernel 2.6.20) Make protected mode kernel relocatable.
40 Introduce relocatable_kernel and kernel_alignment fields.
42 Protocol 2.06: (Kernel 2.6.22) Added a field that contains the size of
48 The traditional memory map for the kernel loader, used for Image or
49 zImage kernels, typically looks like:
52 0A0000 +------------------------+
53 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA.
54 09A000 +------------------------+
56 | Stack/heap | For use by the kernel real-mode code.
57 098000 +------------------------+
58 | Kernel setup | The kernel real-mode code.
59 090200 +------------------------+
60 | Kernel boot sector | The kernel legacy boot sector.
61 090000 +------------------------+
62 | Protected-mode kernel | The bulk of the kernel image.
63 010000 +------------------------+
64 | Boot loader | <- Boot sector entry point 0000:7C00
65 001000 +------------------------+
66 | Reserved for MBR/BIOS |
67 000800 +------------------------+
68 | Typically used by MBR |
69 000600 +------------------------+
71 000000 +------------------------+
74 When using bzImage, the protected-mode kernel was relocated to
75 0x100000 ("high memory"), and the kernel real-mode block (boot sector,
76 setup, and stack/heap) was made relocatable to any address between
77 0x10000 and end of low memory. Unfortunately, in protocols 2.00 and
78 2.01 the 0x90000+ memory range is still used internally by the kernel;
79 the 2.02 protocol resolves that problem.
81 It is desirable to keep the "memory ceiling" -- the highest point in
82 low memory touched by the boot loader -- as low as possible, since
83 some newer BIOSes have begun to allocate some rather large amounts of
84 memory, called the Extended BIOS Data Area, near the top of low
85 memory. The boot loader should use the "INT 12h" BIOS call to verify
86 how much low memory is available.
88 Unfortunately, if INT 12h reports that the amount of memory is too
89 low, there is usually nothing the boot loader can do but to report an
90 error to the user. The boot loader should therefore be designed to
91 take up as little space in low memory as it reasonably can. For
92 zImage or old bzImage kernels, which need data written into the
93 0x90000 segment, the boot loader should make sure not to use memory
94 above the 0x9A000 point; too many BIOSes will break above that point.
96 For a modern bzImage kernel with boot protocol version >= 2.02, a
97 memory layout like the following is suggested:
100 | Protected-mode kernel |
101 100000 +------------------------+
103 0A0000 +------------------------+
104 | Reserved for BIOS | Leave as much as possible unused
106 | Command line | (Can also be below the X+10000 mark)
107 X+10000 +------------------------+
108 | Stack/heap | For use by the kernel real-mode code.
109 X+08000 +------------------------+
110 | Kernel setup | The kernel real-mode code.
111 | Kernel boot sector | The kernel legacy boot sector.
112 X +------------------------+
113 | Boot loader | <- Boot sector entry point 0000:7C00
114 001000 +------------------------+
115 | Reserved for MBR/BIOS |
116 000800 +------------------------+
117 | Typically used by MBR |
118 000600 +------------------------+
120 000000 +------------------------+
122 ... where the address X is as low as the design of the boot loader
126 **** THE REAL-MODE KERNEL HEADER
128 In the following text, and anywhere in the kernel boot sequence, "a
129 sector" refers to 512 bytes. It is independent of the actual sector
130 size of the underlying medium.
132 The first step in loading a Linux kernel should be to load the
133 real-mode code (boot sector and setup code) and then examine the
134 following header at offset 0x01f1. The real-mode code can total up to
135 32K, although the boot loader may choose to load only the first two
136 sectors (1K) and then examine the bootup sector size.
138 The header looks like:
140 Offset Proto Name Meaning
143 01F1/1 ALL(1 setup_sects The size of the setup in sectors
144 01F2/2 ALL root_flags If set, the root is mounted readonly
145 01F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras
146 01F8/2 ALL ram_size DO NOT USE - for bootsect.S use only
147 01FA/2 ALL vid_mode Video mode control
148 01FC/2 ALL root_dev Default root device number
149 01FE/2 ALL boot_flag 0xAA55 magic number
150 0200/2 2.00+ jump Jump instruction
151 0202/4 2.00+ header Magic signature "HdrS"
152 0206/2 2.00+ version Boot protocol version supported
153 0208/4 2.00+ realmode_swtch Boot loader hook (see below)
154 020C/2 2.00+ start_sys The load-low segment (0x1000) (obsolete)
155 020E/2 2.00+ kernel_version Pointer to kernel version string
156 0210/1 2.00+ type_of_loader Boot loader identifier
157 0211/1 2.00+ loadflags Boot protocol option flags
158 0212/2 2.00+ setup_move_size Move to high memory size (used with hooks)
159 0214/4 2.00+ code32_start Boot loader hook (see below)
160 0218/4 2.00+ ramdisk_image initrd load address (set by boot loader)
161 021C/4 2.00+ ramdisk_size initrd size (set by boot loader)
162 0220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only
163 0224/2 2.01+ heap_end_ptr Free memory after setup end
164 0226/2 N/A pad1 Unused
165 0228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line
166 022C/4 2.03+ initrd_addr_max Highest legal initrd address
167 0230/4 2.05+ kernel_alignment Physical addr alignment required for kernel
168 0234/1 2.05+ relocatable_kernel Whether kernel is relocatable or not
169 0235/3 N/A pad2 Unused
170 0238/4 2.06+ cmdline_size Maximum size of the kernel command line
171 023C/4 2.07+ hardware_subarch Hardware subarchitecture
172 0240/8 2.07+ hardware_subarch_data Subarchitecture-specific data
174 (1) For backwards compatibility, if the setup_sects field contains 0, the
177 (2) For boot protocol prior to 2.04, the upper two bytes of the syssize
178 field are unusable, which means the size of a bzImage kernel
179 cannot be determined.
181 If the "HdrS" (0x53726448) magic number is not found at offset 0x202,
182 the boot protocol version is "old". Loading an old kernel, the
183 following parameters should be assumed:
187 Real-mode kernel must be located at 0x90000.
189 Otherwise, the "version" field contains the protocol version,
190 e.g. protocol version 2.01 will contain 0x0201 in this field. When
191 setting fields in the header, you must make sure only to set fields
192 supported by the protocol version in use.
195 **** DETAILS OF HEADER FIELDS
197 For each field, some are information from the kernel to the bootloader
198 ("read"), some are expected to be filled out by the bootloader
199 ("write"), and some are expected to be read and modified by the
200 bootloader ("modify").
202 All general purpose boot loaders should write the fields marked
203 (obligatory). Boot loaders who want to load the kernel at a
204 nonstandard address should fill in the fields marked (reloc); other
205 boot loaders can ignore those fields.
207 The byte order of all fields is littleendian (this is x86, after all.)
209 Field name: setup_sects
214 The size of the setup code in 512-byte sectors. If this field is
215 0, the real value is 4. The real-mode code consists of the boot
216 sector (always one 512-byte sector) plus the setup code.
218 Field name: root_flags
219 Type: modify (optional)
223 If this field is nonzero, the root defaults to readonly. The use of
224 this field is deprecated; use the "ro" or "rw" options on the
225 command line instead.
229 Offset/size: 0x1f4/4 (protocol 2.04+) 0x1f4/2 (protocol ALL)
232 The size of the protected-mode code in units of 16-byte paragraphs.
233 For protocol versions older than 2.04 this field is only two bytes
234 wide, and therefore cannot be trusted for the size of a kernel if
235 the LOAD_HIGH flag is set.
238 Type: kernel internal
242 This field is obsolete.
245 Type: modify (obligatory)
248 Please see the section on SPECIAL COMMAND LINE OPTIONS.
251 Type: modify (optional)
255 The default root device device number. The use of this field is
256 deprecated, use the "root=" option on the command line instead.
258 Field name: boot_flag
263 Contains 0xAA55. This is the closest thing old Linux kernels have
271 Contains an x86 jump instruction, 0xEB followed by a signed offset
272 relative to byte 0x202. This can be used to determine the size of
280 Contains the magic number "HdrS" (0x53726448).
287 Contains the boot protocol version, in (major << 8)+minor format,
288 e.g. 0x0204 for version 2.04, and 0x0a11 for a hypothetical version
291 Field name: readmode_swtch
292 Type: modify (optional)
296 Boot loader hook (see ADVANCED BOOT LOADER HOOKS below.)
298 Field name: start_sys
303 The load low segment (0x1000). Obsolete.
305 Field name: kernel_version
310 If set to a nonzero value, contains a pointer to a NUL-terminated
311 human-readable kernel version number string, less 0x200. This can
312 be used to display the kernel version to the user. This value
313 should be less than (0x200*setup_sects).
315 For example, if this value is set to 0x1c00, the kernel version
316 number string can be found at offset 0x1e00 in the kernel file.
317 This is a valid value if and only if the "setup_sects" field
318 contains the value 15 or higher, as:
320 0x1c00 < 15*0x200 (= 0x1e00) but
321 0x1c00 >= 14*0x200 (= 0x1c00)
323 0x1c00 >> 9 = 14, so the minimum value for setup_secs is 15.
325 Field name: type_of_loader
326 Type: write (obligatory)
330 If your boot loader has an assigned id (see table below), enter
331 0xTV here, where T is an identifier for the boot loader and V is
332 a version number. Otherwise, enter 0xFF here.
334 Assigned boot loader ids:
335 0 LILO (0x00 reserved for pre-2.00 bootloader)
337 2 bootsect-loader (0x20, all other values reserved)
347 Please contact <hpa@zytor.com> if you need a bootloader ID
350 Field name: loadflags
351 Type: modify (obligatory)
355 This field is a bitmask.
357 Bit 0 (read): LOADED_HIGH
358 - If 0, the protected-mode code is loaded at 0x10000.
359 - If 1, the protected-mode code is loaded at 0x100000.
361 Bit 6 (write): KEEP_SEGMENTS
363 - if 0, reload the segment registers in the 32bit entry point.
364 - if 1, do not reload the segment registers in the 32bit entry point.
365 Assume that %cs %ds %ss %es are all set to flat segments with
366 a base of 0 (or the equivalent for their environment).
368 Bit 7 (write): CAN_USE_HEAP
369 Set this bit to 1 to indicate that the value entered in the
370 heap_end_ptr is valid. If this field is clear, some setup code
371 functionality will be disabled.
373 Field name: setup_move_size
374 Type: modify (obligatory)
378 When using protocol 2.00 or 2.01, if the real mode kernel is not
379 loaded at 0x90000, it gets moved there later in the loading
380 sequence. Fill in this field if you want additional data (such as
381 the kernel command line) moved in addition to the real-mode kernel
384 The unit is bytes starting with the beginning of the boot sector.
386 This field is can be ignored when the protocol is 2.02 or higher, or
387 if the real-mode code is loaded at 0x90000.
389 Field name: code32_start
390 Type: modify (optional, reloc)
394 The address to jump to in protected mode. This defaults to the load
395 address of the kernel, and can be used by the boot loader to
396 determine the proper load address.
398 This field can be modified for two purposes:
400 1. as a boot loader hook (see ADVANCED BOOT LOADER HOOKS below.)
402 2. if a bootloader which does not install a hook loads a
403 relocatable kernel at a nonstandard address it will have to modify
404 this field to point to the load address.
406 Field name: ramdisk_image
407 Type: write (obligatory)
411 The 32-bit linear address of the initial ramdisk or ramfs. Leave at
412 zero if there is no initial ramdisk/ramfs.
414 Field name: ramdisk_size
415 Type: write (obligatory)
419 Size of the initial ramdisk or ramfs. Leave at zero if there is no
420 initial ramdisk/ramfs.
422 Field name: bootsect_kludge
423 Type: kernel internal
427 This field is obsolete.
429 Field name: heap_end_ptr
430 Type: write (obligatory)
434 Set this field to the offset (from the beginning of the real-mode
435 code) of the end of the setup stack/heap, minus 0x0200.
437 Field name: cmd_line_ptr
438 Type: write (obligatory)
442 Set this field to the linear address of the kernel command line.
443 The kernel command line can be located anywhere between the end of
444 the setup heap and 0xA0000; it does not have to be located in the
445 same 64K segment as the real-mode code itself.
447 Fill in this field even if your boot loader does not support a
448 command line, in which case you can point this to an empty string
449 (or better yet, to the string "auto".) If this field is left at
450 zero, the kernel will assume that your boot loader does not support
453 Field name: initrd_addr_max
458 The maximum address that may be occupied by the initial
459 ramdisk/ramfs contents. For boot protocols 2.02 or earlier, this
460 field is not present, and the maximum address is 0x37FFFFFF. (This
461 address is defined as the address of the highest safe byte, so if
462 your ramdisk is exactly 131072 bytes long and this field is
463 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)
465 Field name: kernel_alignment
470 Alignment unit required by the kernel (if relocatable_kernel is true.)
472 Field name: relocatable_kernel
477 If this field is nonzero, the protected-mode part of the kernel can
478 be loaded at any address that satisfies the kernel_alignment field.
479 After loading, the boot loader must set the code32_start field to
480 point to the loaded code, or to a boot loader hook.
482 Field name: cmdline_size
487 The maximum size of the command line without the terminating
488 zero. This means that the command line can contain at most
489 cmdline_size characters. With protocol version 2.05 and earlier, the
490 maximum size was 255.
492 Field name: hardware_subarch
497 In a paravirtualized environment the hardware low level architectural
498 pieces such as interrupt handling, page table handling, and
499 accessing process control registers needs to be done differently.
501 This field allows the bootloader to inform the kernel we are in one
502 one of those environments.
504 0x00000000 The default x86/PC environment
508 Field name: hardware_subarch_data
513 A pointer to data that is specific to hardware subarch
516 **** THE KERNEL COMMAND LINE
518 The kernel command line has become an important way for the boot
519 loader to communicate with the kernel. Some of its options are also
520 relevant to the boot loader itself, see "special command line options"
523 The kernel command line is a null-terminated string. The maximum
524 length can be retrieved from the field cmdline_size. Before protocol
525 version 2.06, the maximum was 255 characters. A string that is too
526 long will be automatically truncated by the kernel.
528 If the boot protocol version is 2.02 or later, the address of the
529 kernel command line is given by the header field cmd_line_ptr (see
530 above.) This address can be anywhere between the end of the setup
533 If the protocol version is *not* 2.02 or higher, the kernel
534 command line is entered using the following protocol:
536 At offset 0x0020 (word), "cmd_line_magic", enter the magic
539 At offset 0x0022 (word), "cmd_line_offset", enter the offset
540 of the kernel command line (relative to the start of the
543 The kernel command line *must* be within the memory region
544 covered by setup_move_size, so you may need to adjust this
548 **** MEMORY LAYOUT OF THE REAL-MODE CODE
550 The real-mode code requires a stack/heap to be set up, as well as
551 memory allocated for the kernel command line. This needs to be done
552 in the real-mode accessible memory in bottom megabyte.
554 It should be noted that modern machines often have a sizable Extended
555 BIOS Data Area (EBDA). As a result, it is advisable to use as little
556 of the low megabyte as possible.
558 Unfortunately, under the following circumstances the 0x90000 memory
559 segment has to be used:
561 - When loading a zImage kernel ((loadflags & 0x01) == 0).
562 - When loading a 2.01 or earlier boot protocol kernel.
564 -> For the 2.00 and 2.01 boot protocols, the real-mode code
565 can be loaded at another address, but it is internally
566 relocated to 0x90000. For the "old" protocol, the
567 real-mode code must be loaded at 0x90000.
569 When loading at 0x90000, avoid using memory above 0x9a000.
571 For boot protocol 2.02 or higher, the command line does not have to be
572 located in the same 64K segment as the real-mode setup code; it is
573 thus permitted to give the stack/heap the full 64K segment and locate
574 the command line above it.
576 The kernel command line should not be located below the real-mode
577 code, nor should it be located in high memory.
580 **** SAMPLE BOOT CONFIGURATION
582 As a sample configuration, assume the following layout of the real
585 When loading below 0x90000, use the entire segment:
587 0x0000-0x7fff Real mode kernel
588 0x8000-0xdfff Stack and heap
589 0xe000-0xffff Kernel command line
591 When loading at 0x90000 OR the protocol version is 2.01 or earlier:
593 0x0000-0x7fff Real mode kernel
594 0x8000-0x97ff Stack and heap
595 0x9800-0x9fff Kernel command line
597 Such a boot loader should enter the following fields in the header:
599 unsigned long base_ptr; /* base address for real-mode segment */
601 if ( setup_sects == 0 ) {
605 if ( protocol >= 0x0200 ) {
606 type_of_loader = <type code>;
607 if ( loading_initrd ) {
608 ramdisk_image = <initrd_address>;
609 ramdisk_size = <initrd_size>;
612 if ( protocol >= 0x0202 && loadflags & 0x01 )
617 if ( protocol >= 0x0201 ) {
618 heap_end_ptr = heap_end - 0x200;
619 loadflags |= 0x80; /* CAN_USE_HEAP */
622 if ( protocol >= 0x0202 ) {
623 cmd_line_ptr = base_ptr + heap_end;
624 strcpy(cmd_line_ptr, cmdline);
626 cmd_line_magic = 0xA33F;
627 cmd_line_offset = heap_end;
628 setup_move_size = heap_end + strlen(cmdline)+1;
629 strcpy(base_ptr+cmd_line_offset, cmdline);
632 /* Very old kernel */
636 cmd_line_magic = 0xA33F;
637 cmd_line_offset = heap_end;
639 /* A very old kernel MUST have its real-mode code
642 if ( base_ptr != 0x90000 ) {
643 /* Copy the real-mode kernel */
644 memcpy(0x90000, base_ptr, (setup_sects+1)*512);
645 base_ptr = 0x90000; /* Relocated */
648 strcpy(0x90000+cmd_line_offset, cmdline);
650 /* It is recommended to clear memory up to the 32K mark */
651 memset(0x90000 + (setup_sects+1)*512, 0,
652 (64-(setup_sects+1))*512);
656 **** LOADING THE REST OF THE KERNEL
658 The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512
659 in the kernel file (again, if setup_sects == 0 the real value is 4.)
660 It should be loaded at address 0x10000 for Image/zImage kernels and
661 0x100000 for bzImage kernels.
663 The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01
664 bit (LOAD_HIGH) in the loadflags field is set:
666 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);
667 load_address = is_bzImage ? 0x100000 : 0x10000;
669 Note that Image/zImage kernels can be up to 512K in size, and thus use
670 the entire 0x10000-0x90000 range of memory. This means it is pretty
671 much a requirement for these kernels to load the real-mode part at
672 0x90000. bzImage kernels allow much more flexibility.
675 **** SPECIAL COMMAND LINE OPTIONS
677 If the command line provided by the boot loader is entered by the
678 user, the user may expect the following command line options to work.
679 They should normally not be deleted from the kernel command line even
680 though not all of them are actually meaningful to the kernel. Boot
681 loader authors who need additional command line options for the boot
682 loader itself should get them registered in
683 Documentation/kernel-parameters.txt to make sure they will not
684 conflict with actual kernel options now or in the future.
687 <mode> here is either an integer (in C notation, either
688 decimal, octal, or hexadecimal) or one of the strings
689 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"
690 (meaning 0xFFFD). This value should be entered into the
691 vid_mode field, as it is used by the kernel before the command
695 <size> is an integer in C notation optionally followed by
696 (case insensitive) K, M, G, T, P or E (meaning << 10, << 20,
697 << 30, << 40, << 50 or << 60). This specifies the end of
698 memory to the kernel. This affects the possible placement of
699 an initrd, since an initrd should be placed near end of
700 memory. Note that this is an option to *both* the kernel and
704 An initrd should be loaded. The meaning of <file> is
705 obviously bootloader-dependent, and some boot loaders
706 (e.g. LILO) do not have such a command.
708 In addition, some boot loaders add the following options to the
709 user-specified command line:
712 The boot image which was loaded. Again, the meaning of <file>
713 is obviously bootloader-dependent.
716 The kernel was booted without explicit user intervention.
718 If these options are added by the boot loader, it is highly
719 recommended that they are located *first*, before the user-specified
720 or configuration-specified command line. Otherwise, "init=/bin/sh"
721 gets confused by the "auto" option.
724 **** RUNNING THE KERNEL
726 The kernel is started by jumping to the kernel entry point, which is
727 located at *segment* offset 0x20 from the start of the real mode
728 kernel. This means that if you loaded your real-mode kernel code at
729 0x90000, the kernel entry point is 9020:0000.
731 At entry, ds = es = ss should point to the start of the real-mode
732 kernel code (0x9000 if the code is loaded at 0x90000), sp should be
733 set up properly, normally pointing to the top of the heap, and
734 interrupts should be disabled. Furthermore, to guard against bugs in
735 the kernel, it is recommended that the boot loader sets fs = gs = ds =
738 In our example from above, we would do:
740 /* Note: in the case of the "old" kernel protocol, base_ptr must
741 be == 0x90000 at this point; see the previous sample code */
745 cli(); /* Enter with interrupts disabled! */
747 /* Set up the real-mode kernel stack */
751 _DS = _ES = _FS = _GS = seg;
752 jmp_far(seg+0x20, 0); /* Run the kernel */
754 If your boot sector accesses a floppy drive, it is recommended to
755 switch off the floppy motor before running the kernel, since the
756 kernel boot leaves interrupts off and thus the motor will not be
757 switched off, especially if the loaded kernel has the floppy driver as
758 a demand-loaded module!
761 **** ADVANCED BOOT LOADER HOOKS
763 If the boot loader runs in a particularly hostile environment (such as
764 LOADLIN, which runs under DOS) it may be impossible to follow the
765 standard memory location requirements. Such a boot loader may use the
766 following hooks that, if set, are invoked by the kernel at the
767 appropriate time. The use of these hooks should probably be
768 considered an absolutely last resort!
770 IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and
771 %edi across invocation.
774 A 16-bit real mode far subroutine invoked immediately before
775 entering protected mode. The default routine disables NMI, so
776 your routine should probably do so, too.
779 A 32-bit flat-mode routine *jumped* to immediately after the
780 transition to protected mode, but before the kernel is
781 uncompressed. No segments, except CS, are guaranteed to be
782 set up (current kernels do, but older ones do not); you should
783 set them up to BOOT_DS (0x18) yourself.
785 After completing your hook, you should jump to the address
786 that was in this field before your boot loader overwrote it
787 (relocated, if appropriate.)