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, four 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.
38 Protocol 2.05: (Kernel 2.6.20) Make protected mode kernel relocatable.
39 Introduce relocatable_kernel and kernel_alignment fields.
44 The traditional memory map for the kernel loader, used for Image or
45 zImage kernels, typically looks like:
48 0A0000 +------------------------+
49 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA.
50 09A000 +------------------------+
51 | Stack/heap/cmdline | For use by the kernel real-mode code.
52 098000 +------------------------+
53 | Kernel setup | The kernel real-mode code.
54 090200 +------------------------+
55 | Kernel boot sector | The kernel legacy boot sector.
56 090000 +------------------------+
57 | Protected-mode kernel | The bulk of the kernel image.
58 010000 +------------------------+
59 | Boot loader | <- Boot sector entry point 0000:7C00
60 001000 +------------------------+
61 | Reserved for MBR/BIOS |
62 000800 +------------------------+
63 | Typically used by MBR |
64 000600 +------------------------+
66 000000 +------------------------+
69 When using bzImage, the protected-mode kernel was relocated to
70 0x100000 ("high memory"), and the kernel real-mode block (boot sector,
71 setup, and stack/heap) was made relocatable to any address between
72 0x10000 and end of low memory. Unfortunately, in protocols 2.00 and
73 2.01 the command line is still required to live in the 0x9XXXX memory
74 range, and that memory range is still overwritten by the early kernel.
75 The 2.02 protocol resolves that problem.
77 It is desirable to keep the "memory ceiling" -- the highest point in
78 low memory touched by the boot loader -- as low as possible, since
79 some newer BIOSes have begun to allocate some rather large amounts of
80 memory, called the Extended BIOS Data Area, near the top of low
81 memory. The boot loader should use the "INT 12h" BIOS call to verify
82 how much low memory is available.
84 Unfortunately, if INT 12h reports that the amount of memory is too
85 low, there is usually nothing the boot loader can do but to report an
86 error to the user. The boot loader should therefore be designed to
87 take up as little space in low memory as it reasonably can. For
88 zImage or old bzImage kernels, which need data written into the
89 0x90000 segment, the boot loader should make sure not to use memory
90 above the 0x9A000 point; too many BIOSes will break above that point.
93 **** THE REAL-MODE KERNEL HEADER
95 In the following text, and anywhere in the kernel boot sequence, "a
96 sector" refers to 512 bytes. It is independent of the actual sector
97 size of the underlying medium.
99 The first step in loading a Linux kernel should be to load the
100 real-mode code (boot sector and setup code) and then examine the
101 following header at offset 0x01f1. The real-mode code can total up to
102 32K, although the boot loader may choose to load only the first two
103 sectors (1K) and then examine the bootup sector size.
105 The header looks like:
107 Offset Proto Name Meaning
110 01F1/1 ALL(1 setup_sects The size of the setup in sectors
111 01F2/2 ALL root_flags If set, the root is mounted readonly
112 01F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras
113 01F8/2 ALL ram_size DO NOT USE - for bootsect.S use only
114 01FA/2 ALL vid_mode Video mode control
115 01FC/2 ALL root_dev Default root device number
116 01FE/2 ALL boot_flag 0xAA55 magic number
117 0200/2 2.00+ jump Jump instruction
118 0202/4 2.00+ header Magic signature "HdrS"
119 0206/2 2.00+ version Boot protocol version supported
120 0208/4 2.00+ realmode_swtch Boot loader hook (see below)
121 020C/2 2.00+ start_sys The load-low segment (0x1000) (obsolete)
122 020E/2 2.00+ kernel_version Pointer to kernel version string
123 0210/1 2.00+ type_of_loader Boot loader identifier
124 0211/1 2.00+ loadflags Boot protocol option flags
125 0212/2 2.00+ setup_move_size Move to high memory size (used with hooks)
126 0214/4 2.00+ code32_start Boot loader hook (see below)
127 0218/4 2.00+ ramdisk_image initrd load address (set by boot loader)
128 021C/4 2.00+ ramdisk_size initrd size (set by boot loader)
129 0220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only
130 0224/2 2.01+ heap_end_ptr Free memory after setup end
131 0226/2 N/A pad1 Unused
132 0228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line
133 022C/4 2.03+ initrd_addr_max Highest legal initrd address
134 0230/4 2.05+ kernel_alignment Physical addr alignment required for kernel
135 0234/1 2.05+ relocatable_kernel Whether kernel is relocatable or not
137 (1) For backwards compatibility, if the setup_sects field contains 0, the
140 (2) For boot protocol prior to 2.04, the upper two bytes of the syssize
141 field are unusable, which means the size of a bzImage kernel
142 cannot be determined.
144 If the "HdrS" (0x53726448) magic number is not found at offset 0x202,
145 the boot protocol version is "old". Loading an old kernel, the
146 following parameters should be assumed:
150 Real-mode kernel must be located at 0x90000.
152 Otherwise, the "version" field contains the protocol version,
153 e.g. protocol version 2.01 will contain 0x0201 in this field. When
154 setting fields in the header, you must make sure only to set fields
155 supported by the protocol version in use.
157 The "kernel_version" field, if set to a nonzero value, contains a
158 pointer to a null-terminated human-readable kernel version number
159 string, less 0x200. This can be used to display the kernel version to
160 the user. This value should be less than (0x200*setup_sects). For
161 example, if this value is set to 0x1c00, the kernel version number
162 string can be found at offset 0x1e00 in the kernel file. This is a
163 valid value if and only if the "setup_sects" field contains the value
166 Most boot loaders will simply load the kernel at its target address
167 directly. Such boot loaders do not need to worry about filling in
168 most of the fields in the header. The following fields should be
172 Please see the section on SPECIAL COMMAND LINE OPTIONS.
175 If your boot loader has an assigned id (see table below), enter
176 0xTV here, where T is an identifier for the boot loader and V is
177 a version number. Otherwise, enter 0xFF here.
179 Assigned boot loader ids:
191 Please contact <hpa@zytor.com> if you need a bootloader ID
194 loadflags, heap_end_ptr:
195 If the protocol version is 2.01 or higher, enter the
196 offset limit of the setup heap into heap_end_ptr and set the
197 0x80 bit (CAN_USE_HEAP) of loadflags. heap_end_ptr appears to
198 be relative to the start of setup (offset 0x0200).
201 When using protocol 2.00 or 2.01, if the real mode
202 kernel is not loaded at 0x90000, it gets moved there later in
203 the loading sequence. Fill in this field if you want
204 additional data (such as the kernel command line) moved in
205 addition to the real-mode kernel itself.
207 ramdisk_image, ramdisk_size:
208 If your boot loader has loaded an initial ramdisk (initrd),
209 set ramdisk_image to the 32-bit pointer to the ramdisk data
210 and the ramdisk_size to the size of the ramdisk data.
212 The initrd should typically be located as high in memory as
213 possible, as it may otherwise get overwritten by the early
214 kernel initialization sequence. However, it must never be
215 located above the address specified in the initrd_addr_max
216 field. The initrd should be at least 4K page aligned.
219 If the protocol version is 2.02 or higher, this is a 32-bit
220 pointer to the kernel command line. The kernel command line
221 can be located anywhere between the end of setup and 0xA0000.
222 Fill in this field even if your boot loader does not support a
223 command line, in which case you can point this to an empty
224 string (or better yet, to the string "auto".) If this field
225 is left at zero, the kernel will assume that your boot loader
226 does not support the 2.02+ protocol.
229 The maximum address that may be occupied by the initrd
230 contents. For boot protocols 2.02 or earlier, this field is
231 not present, and the maximum address is 0x37FFFFFF. (This
232 address is defined as the address of the highest safe byte, so
233 if your ramdisk is exactly 131072 bytes long and this field is
234 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)
237 **** THE KERNEL COMMAND LINE
239 The kernel command line has become an important way for the boot
240 loader to communicate with the kernel. Some of its options are also
241 relevant to the boot loader itself, see "special command line options"
244 The kernel command line is a null-terminated string currently up to
245 255 characters long, plus the final null. A string that is too long
246 will be automatically truncated by the kernel, a boot loader may allow
247 a longer command line to be passed to permit future kernels to extend
250 If the boot protocol version is 2.02 or later, the address of the
251 kernel command line is given by the header field cmd_line_ptr (see
252 above.) This address can be anywhere between the end of the setup
255 If the protocol version is *not* 2.02 or higher, the kernel
256 command line is entered using the following protocol:
258 At offset 0x0020 (word), "cmd_line_magic", enter the magic
261 At offset 0x0022 (word), "cmd_line_offset", enter the offset
262 of the kernel command line (relative to the start of the
265 The kernel command line *must* be within the memory region
266 covered by setup_move_size, so you may need to adjust this
270 **** SAMPLE BOOT CONFIGURATION
272 As a sample configuration, assume the following layout of the real
273 mode segment (this is a typical, and recommended layout):
275 0x0000-0x7FFF Real mode kernel
276 0x8000-0x8FFF Stack and heap
277 0x9000-0x90FF Kernel command line
279 Such a boot loader should enter the following fields in the header:
281 unsigned long base_ptr; /* base address for real-mode segment */
283 if ( setup_sects == 0 ) {
287 if ( protocol >= 0x0200 ) {
288 type_of_loader = <type code>;
289 if ( loading_initrd ) {
290 ramdisk_image = <initrd_address>;
291 ramdisk_size = <initrd_size>;
293 if ( protocol >= 0x0201 ) {
294 heap_end_ptr = 0x9000 - 0x200;
295 loadflags |= 0x80; /* CAN_USE_HEAP */
297 if ( protocol >= 0x0202 ) {
298 cmd_line_ptr = base_ptr + 0x9000;
300 cmd_line_magic = 0xA33F;
301 cmd_line_offset = 0x9000;
302 setup_move_size = 0x9100;
305 /* Very old kernel */
307 cmd_line_magic = 0xA33F;
308 cmd_line_offset = 0x9000;
310 /* A very old kernel MUST have its real-mode code
313 if ( base_ptr != 0x90000 ) {
314 /* Copy the real-mode kernel */
315 memcpy(0x90000, base_ptr, (setup_sects+1)*512);
316 /* Copy the command line */
317 memcpy(0x99000, base_ptr+0x9000, 256);
319 base_ptr = 0x90000; /* Relocated */
322 /* It is recommended to clear memory up to the 32K mark */
323 memset(0x90000 + (setup_sects+1)*512, 0,
324 (64-(setup_sects+1))*512);
328 **** LOADING THE REST OF THE KERNEL
330 The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512
331 in the kernel file (again, if setup_sects == 0 the real value is 4.)
332 It should be loaded at address 0x10000 for Image/zImage kernels and
333 0x100000 for bzImage kernels.
335 The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01
336 bit (LOAD_HIGH) in the loadflags field is set:
338 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);
339 load_address = is_bzImage ? 0x100000 : 0x10000;
341 Note that Image/zImage kernels can be up to 512K in size, and thus use
342 the entire 0x10000-0x90000 range of memory. This means it is pretty
343 much a requirement for these kernels to load the real-mode part at
344 0x90000. bzImage kernels allow much more flexibility.
347 **** SPECIAL COMMAND LINE OPTIONS
349 If the command line provided by the boot loader is entered by the
350 user, the user may expect the following command line options to work.
351 They should normally not be deleted from the kernel command line even
352 though not all of them are actually meaningful to the kernel. Boot
353 loader authors who need additional command line options for the boot
354 loader itself should get them registered in
355 Documentation/kernel-parameters.txt to make sure they will not
356 conflict with actual kernel options now or in the future.
359 <mode> here is either an integer (in C notation, either
360 decimal, octal, or hexadecimal) or one of the strings
361 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"
362 (meaning 0xFFFD). This value should be entered into the
363 vid_mode field, as it is used by the kernel before the command
367 <size> is an integer in C notation optionally followed by K, M
368 or G (meaning << 10, << 20 or << 30). This specifies the end
369 of memory to the kernel. This affects the possible placement
370 of an initrd, since an initrd should be placed near end of
371 memory. Note that this is an option to *both* the kernel and
375 An initrd should be loaded. The meaning of <file> is
376 obviously bootloader-dependent, and some boot loaders
377 (e.g. LILO) do not have such a command.
379 In addition, some boot loaders add the following options to the
380 user-specified command line:
383 The boot image which was loaded. Again, the meaning of <file>
384 is obviously bootloader-dependent.
387 The kernel was booted without explicit user intervention.
389 If these options are added by the boot loader, it is highly
390 recommended that they are located *first*, before the user-specified
391 or configuration-specified command line. Otherwise, "init=/bin/sh"
392 gets confused by the "auto" option.
395 **** RUNNING THE KERNEL
397 The kernel is started by jumping to the kernel entry point, which is
398 located at *segment* offset 0x20 from the start of the real mode
399 kernel. This means that if you loaded your real-mode kernel code at
400 0x90000, the kernel entry point is 9020:0000.
402 At entry, ds = es = ss should point to the start of the real-mode
403 kernel code (0x9000 if the code is loaded at 0x90000), sp should be
404 set up properly, normally pointing to the top of the heap, and
405 interrupts should be disabled. Furthermore, to guard against bugs in
406 the kernel, it is recommended that the boot loader sets fs = gs = ds =
409 In our example from above, we would do:
411 /* Note: in the case of the "old" kernel protocol, base_ptr must
412 be == 0x90000 at this point; see the previous sample code */
416 cli(); /* Enter with interrupts disabled! */
418 /* Set up the real-mode kernel stack */
420 _SP = 0x9000; /* Load SP immediately after loading SS! */
422 _DS = _ES = _FS = _GS = seg;
423 jmp_far(seg+0x20, 0); /* Run the kernel */
425 If your boot sector accesses a floppy drive, it is recommended to
426 switch off the floppy motor before running the kernel, since the
427 kernel boot leaves interrupts off and thus the motor will not be
428 switched off, especially if the loaded kernel has the floppy driver as
429 a demand-loaded module!
432 **** ADVANCED BOOT TIME HOOKS
434 If the boot loader runs in a particularly hostile environment (such as
435 LOADLIN, which runs under DOS) it may be impossible to follow the
436 standard memory location requirements. Such a boot loader may use the
437 following hooks that, if set, are invoked by the kernel at the
438 appropriate time. The use of these hooks should probably be
439 considered an absolutely last resort!
441 IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and
442 %edi across invocation.
445 A 16-bit real mode far subroutine invoked immediately before
446 entering protected mode. The default routine disables NMI, so
447 your routine should probably do so, too.
450 A 32-bit flat-mode routine *jumped* to immediately after the
451 transition to protected mode, but before the kernel is
452 uncompressed. No segments, except CS, are set up; you should
453 set them up to KERNEL_DS (0x18) yourself.
455 After completing your hook, you should jump to the address
456 that was in this field before your boot loader overwrote it.