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.
42 The traditional memory map for the kernel loader, used for Image or
43 zImage kernels, typically looks like:
46 0A0000 +------------------------+
47 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA.
48 09A000 +------------------------+
49 | Stack/heap/cmdline | For use by the kernel real-mode code.
50 098000 +------------------------+
51 | Kernel setup | The kernel real-mode code.
52 090200 +------------------------+
53 | Kernel boot sector | The kernel legacy boot sector.
54 090000 +------------------------+
55 | Protected-mode kernel | The bulk of the kernel image.
56 010000 +------------------------+
57 | Boot loader | <- Boot sector entry point 0000:7C00
58 001000 +------------------------+
59 | Reserved for MBR/BIOS |
60 000800 +------------------------+
61 | Typically used by MBR |
62 000600 +------------------------+
64 000000 +------------------------+
67 When using bzImage, the protected-mode kernel was relocated to
68 0x100000 ("high memory"), and the kernel real-mode block (boot sector,
69 setup, and stack/heap) was made relocatable to any address between
70 0x10000 and end of low memory. Unfortunately, in protocols 2.00 and
71 2.01 the command line is still required to live in the 0x9XXXX memory
72 range, and that memory range is still overwritten by the early kernel.
73 The 2.02 protocol resolves that problem.
75 It is desirable to keep the "memory ceiling" -- the highest point in
76 low memory touched by the boot loader -- as low as possible, since
77 some newer BIOSes have begun to allocate some rather large amounts of
78 memory, called the Extended BIOS Data Area, near the top of low
79 memory. The boot loader should use the "INT 12h" BIOS call to verify
80 how much low memory is available.
82 Unfortunately, if INT 12h reports that the amount of memory is too
83 low, there is usually nothing the boot loader can do but to report an
84 error to the user. The boot loader should therefore be designed to
85 take up as little space in low memory as it reasonably can. For
86 zImage or old bzImage kernels, which need data written into the
87 0x90000 segment, the boot loader should make sure not to use memory
88 above the 0x9A000 point; too many BIOSes will break above that point.
91 **** THE REAL-MODE KERNEL HEADER
93 In the following text, and anywhere in the kernel boot sequence, "a
94 sector" refers to 512 bytes. It is independent of the actual sector
95 size of the underlying medium.
97 The first step in loading a Linux kernel should be to load the
98 real-mode code (boot sector and setup code) and then examine the
99 following header at offset 0x01f1. The real-mode code can total up to
100 32K, although the boot loader may choose to load only the first two
101 sectors (1K) and then examine the bootup sector size.
103 The header looks like:
105 Offset Proto Name Meaning
108 01F1/1 ALL(1 setup_sects The size of the setup in sectors
109 01F2/2 ALL root_flags If set, the root is mounted readonly
110 01F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras
111 01F8/2 ALL ram_size DO NOT USE - for bootsect.S use only
112 01FA/2 ALL vid_mode Video mode control
113 01FC/2 ALL root_dev Default root device number
114 01FE/2 ALL boot_flag 0xAA55 magic number
115 0200/2 2.00+ jump Jump instruction
116 0202/4 2.00+ header Magic signature "HdrS"
117 0206/2 2.00+ version Boot protocol version supported
118 0208/4 2.00+ realmode_swtch Boot loader hook (see below)
119 020C/2 2.00+ start_sys The load-low segment (0x1000) (obsolete)
120 020E/2 2.00+ kernel_version Pointer to kernel version string
121 0210/1 2.00+ type_of_loader Boot loader identifier
122 0211/1 2.00+ loadflags Boot protocol option flags
123 0212/2 2.00+ setup_move_size Move to high memory size (used with hooks)
124 0214/4 2.00+ code32_start Boot loader hook (see below)
125 0218/4 2.00+ ramdisk_image initrd load address (set by boot loader)
126 021C/4 2.00+ ramdisk_size initrd size (set by boot loader)
127 0220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only
128 0224/2 2.01+ heap_end_ptr Free memory after setup end
129 0226/2 N/A pad1 Unused
130 0228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line
131 022C/4 2.03+ initrd_addr_max Highest legal initrd address
133 (1) For backwards compatibility, if the setup_sects field contains 0, the
136 (2) For boot protocol prior to 2.04, the upper two bytes of the syssize
137 field are unusable, which means the size of a bzImage kernel
138 cannot be determined.
140 If the "HdrS" (0x53726448) magic number is not found at offset 0x202,
141 the boot protocol version is "old". Loading an old kernel, the
142 following parameters should be assumed:
146 Real-mode kernel must be located at 0x90000.
148 Otherwise, the "version" field contains the protocol version,
149 e.g. protocol version 2.01 will contain 0x0201 in this field. When
150 setting fields in the header, you must make sure only to set fields
151 supported by the protocol version in use.
153 The "kernel_version" field, if set to a nonzero value, contains a
154 pointer to a null-terminated human-readable kernel version number
155 string, less 0x200. This can be used to display the kernel version to
156 the user. This value should be less than (0x200*setup_sects). For
157 example, if this value is set to 0x1c00, the kernel version number
158 string can be found at offset 0x1e00 in the kernel file. This is a
159 valid value if and only if the "setup_sects" field contains the value
162 Most boot loaders will simply load the kernel at its target address
163 directly. Such boot loaders do not need to worry about filling in
164 most of the fields in the header. The following fields should be
168 Please see the section on SPECIAL COMMAND LINE OPTIONS.
171 If your boot loader has an assigned id (see table below), enter
172 0xTV here, where T is an identifier for the boot loader and V is
173 a version number. Otherwise, enter 0xFF here.
175 Assigned boot loader ids:
185 Please contact <hpa@zytor.com> if you need a bootloader ID
188 loadflags, heap_end_ptr:
189 If the protocol version is 2.01 or higher, enter the
190 offset limit of the setup heap into heap_end_ptr and set the
191 0x80 bit (CAN_USE_HEAP) of loadflags. heap_end_ptr appears to
192 be relative to the start of setup (offset 0x0200).
195 When using protocol 2.00 or 2.01, if the real mode
196 kernel is not loaded at 0x90000, it gets moved there later in
197 the loading sequence. Fill in this field if you want
198 additional data (such as the kernel command line) moved in
199 addition to the real-mode kernel itself.
201 ramdisk_image, ramdisk_size:
202 If your boot loader has loaded an initial ramdisk (initrd),
203 set ramdisk_image to the 32-bit pointer to the ramdisk data
204 and the ramdisk_size to the size of the ramdisk data.
206 The initrd should typically be located as high in memory as
207 possible, as it may otherwise get overwritten by the early
208 kernel initialization sequence. However, it must never be
209 located above the address specified in the initrd_addr_max
210 field. The initrd should be at least 4K page aligned.
213 If the protocol version is 2.02 or higher, this is a 32-bit
214 pointer to the kernel command line. The kernel command line
215 can be located anywhere between the end of setup and 0xA0000.
216 Fill in this field even if your boot loader does not support a
217 command line, in which case you can point this to an empty
218 string (or better yet, to the string "auto".) If this field
219 is left at zero, the kernel will assume that your boot loader
220 does not support the 2.02+ protocol.
223 The maximum address that may be occupied by the initrd
224 contents. For boot protocols 2.02 or earlier, this field is
225 not present, and the maximum address is 0x37FFFFFF. (This
226 address is defined as the address of the highest safe byte, so
227 if your ramdisk is exactly 131072 bytes long and this field is
228 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)
231 **** THE KERNEL COMMAND LINE
233 The kernel command line has become an important way for the boot
234 loader to communicate with the kernel. Some of its options are also
235 relevant to the boot loader itself, see "special command line options"
238 The kernel command line is a null-terminated string currently up to
239 255 characters long, plus the final null. A string that is too long
240 will be automatically truncated by the kernel, a boot loader may allow
241 a longer command line to be passed to permit future kernels to extend
244 If the boot protocol version is 2.02 or later, the address of the
245 kernel command line is given by the header field cmd_line_ptr (see
246 above.) This address can be anywhere between the end of the setup
249 If the protocol version is *not* 2.02 or higher, the kernel
250 command line is entered using the following protocol:
252 At offset 0x0020 (word), "cmd_line_magic", enter the magic
255 At offset 0x0022 (word), "cmd_line_offset", enter the offset
256 of the kernel command line (relative to the start of the
259 The kernel command line *must* be within the memory region
260 covered by setup_move_size, so you may need to adjust this
264 **** SAMPLE BOOT CONFIGURATION
266 As a sample configuration, assume the following layout of the real
267 mode segment (this is a typical, and recommended layout):
269 0x0000-0x7FFF Real mode kernel
270 0x8000-0x8FFF Stack and heap
271 0x9000-0x90FF Kernel command line
273 Such a boot loader should enter the following fields in the header:
275 unsigned long base_ptr; /* base address for real-mode segment */
277 if ( setup_sects == 0 ) {
281 if ( protocol >= 0x0200 ) {
282 type_of_loader = <type code>;
283 if ( loading_initrd ) {
284 ramdisk_image = <initrd_address>;
285 ramdisk_size = <initrd_size>;
287 if ( protocol >= 0x0201 ) {
288 heap_end_ptr = 0x9000 - 0x200;
289 loadflags |= 0x80; /* CAN_USE_HEAP */
291 if ( protocol >= 0x0202 ) {
292 cmd_line_ptr = base_ptr + 0x9000;
294 cmd_line_magic = 0xA33F;
295 cmd_line_offset = 0x9000;
296 setup_move_size = 0x9100;
299 /* Very old kernel */
301 cmd_line_magic = 0xA33F;
302 cmd_line_offset = 0x9000;
304 /* A very old kernel MUST have its real-mode code
307 if ( base_ptr != 0x90000 ) {
308 /* Copy the real-mode kernel */
309 memcpy(0x90000, base_ptr, (setup_sects+1)*512);
310 /* Copy the command line */
311 memcpy(0x99000, base_ptr+0x9000, 256);
313 base_ptr = 0x90000; /* Relocated */
316 /* It is recommended to clear memory up to the 32K mark */
317 memset(0x90000 + (setup_sects+1)*512, 0,
318 (64-(setup_sects+1))*512);
322 **** LOADING THE REST OF THE KERNEL
324 The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512
325 in the kernel file (again, if setup_sects == 0 the real value is 4.)
326 It should be loaded at address 0x10000 for Image/zImage kernels and
327 0x100000 for bzImage kernels.
329 The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01
330 bit (LOAD_HIGH) in the loadflags field is set:
332 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);
333 load_address = is_bzImage ? 0x100000 : 0x10000;
335 Note that Image/zImage kernels can be up to 512K in size, and thus use
336 the entire 0x10000-0x90000 range of memory. This means it is pretty
337 much a requirement for these kernels to load the real-mode part at
338 0x90000. bzImage kernels allow much more flexibility.
341 **** SPECIAL COMMAND LINE OPTIONS
343 If the command line provided by the boot loader is entered by the
344 user, the user may expect the following command line options to work.
345 They should normally not be deleted from the kernel command line even
346 though not all of them are actually meaningful to the kernel. Boot
347 loader authors who need additional command line options for the boot
348 loader itself should get them registered in
349 Documentation/kernel-parameters.txt to make sure they will not
350 conflict with actual kernel options now or in the future.
353 <mode> here is either an integer (in C notation, either
354 decimal, octal, or hexadecimal) or one of the strings
355 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"
356 (meaning 0xFFFD). This value should be entered into the
357 vid_mode field, as it is used by the kernel before the command
361 <size> is an integer in C notation optionally followed by K, M
362 or G (meaning << 10, << 20 or << 30). This specifies the end
363 of memory to the kernel. This affects the possible placement
364 of an initrd, since an initrd should be placed near end of
365 memory. Note that this is an option to *both* the kernel and
369 An initrd should be loaded. The meaning of <file> is
370 obviously bootloader-dependent, and some boot loaders
371 (e.g. LILO) do not have such a command.
373 In addition, some boot loaders add the following options to the
374 user-specified command line:
377 The boot image which was loaded. Again, the meaning of <file>
378 is obviously bootloader-dependent.
381 The kernel was booted without explicit user intervention.
383 If these options are added by the boot loader, it is highly
384 recommended that they are located *first*, before the user-specified
385 or configuration-specified command line. Otherwise, "init=/bin/sh"
386 gets confused by the "auto" option.
389 **** RUNNING THE KERNEL
391 The kernel is started by jumping to the kernel entry point, which is
392 located at *segment* offset 0x20 from the start of the real mode
393 kernel. This means that if you loaded your real-mode kernel code at
394 0x90000, the kernel entry point is 9020:0000.
396 At entry, ds = es = ss should point to the start of the real-mode
397 kernel code (0x9000 if the code is loaded at 0x90000), sp should be
398 set up properly, normally pointing to the top of the heap, and
399 interrupts should be disabled. Furthermore, to guard against bugs in
400 the kernel, it is recommended that the boot loader sets fs = gs = ds =
403 In our example from above, we would do:
405 /* Note: in the case of the "old" kernel protocol, base_ptr must
406 be == 0x90000 at this point; see the previous sample code */
410 cli(); /* Enter with interrupts disabled! */
412 /* Set up the real-mode kernel stack */
414 _SP = 0x9000; /* Load SP immediately after loading SS! */
416 _DS = _ES = _FS = _GS = seg;
417 jmp_far(seg+0x20, 0); /* Run the kernel */
419 If your boot sector accesses a floppy drive, it is recommended to
420 switch off the floppy motor before running the kernel, since the
421 kernel boot leaves interrupts off and thus the motor will not be
422 switched off, especially if the loaded kernel has the floppy driver as
423 a demand-loaded module!
426 **** ADVANCED BOOT TIME HOOKS
428 If the boot loader runs in a particularly hostile environment (such as
429 LOADLIN, which runs under DOS) it may be impossible to follow the
430 standard memory location requirements. Such a boot loader may use the
431 following hooks that, if set, are invoked by the kernel at the
432 appropriate time. The use of these hooks should probably be
433 considered an absolutely last resort!
435 IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and
436 %edi across invocation.
439 A 16-bit real mode far subroutine invoked immediately before
440 entering protected mode. The default routine disables NMI, so
441 your routine should probably do so, too.
444 A 32-bit flat-mode routine *jumped* to immediately after the
445 transition to protected mode, but before the kernel is
446 uncompressed. No segments, except CS, are set up; you should
447 set them up to KERNEL_DS (0x18) yourself.
449 After completing your hook, you should jump to the address
450 that was in this field before your boot loader overwrote it.