blob: c51314b1a463a35eb222d550409d0a29137149e2 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001 THE LINUX/I386 BOOT PROTOCOL
2 ----------------------------
3
4 H. Peter Anvin <hpa@zytor.com>
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -07005 Last update 2005-09-02
Linus Torvalds1da177e2005-04-16 15:20:36 -07006
7On the i386 platform, the Linux kernel uses a rather complicated boot
8convention. This has evolved partially due to historical aspects, as
9well as the desire in the early days to have the kernel itself be a
10bootable image, the complicated PC memory model and due to changed
11expectations in the PC industry caused by the effective demise of
12real-mode DOS as a mainstream operating system.
13
14Currently, four versions of the Linux/i386 boot protocol exist.
15
16Old kernels: zImage/Image support only. Some very early kernels
17 may not even support a command line.
18
19Protocol 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
23 writable.
24
25Protocol 2.01: (Kernel 1.3.76) Added a heap overrun warning.
26
27Protocol 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
32 supported.
33
34Protocol 2.03: (Kernel 2.4.18-pre1) Explicitly makes the highest possible
35 initrd address available to the bootloader.
36
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -070037Protocol 2.04: (Kernel 2.6.14) Extend the syssize field to four bytes.
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
40**** MEMORY LAYOUT
41
42The traditional memory map for the kernel loader, used for Image or
43zImage kernels, typically looks like:
44
45 | |
460A0000 +------------------------+
47 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA.
4809A000 +------------------------+
49 | Stack/heap/cmdline | For use by the kernel real-mode code.
50098000 +------------------------+
51 | Kernel setup | The kernel real-mode code.
52090200 +------------------------+
53 | Kernel boot sector | The kernel legacy boot sector.
54090000 +------------------------+
55 | Protected-mode kernel | The bulk of the kernel image.
56010000 +------------------------+
57 | Boot loader | <- Boot sector entry point 0000:7C00
58001000 +------------------------+
59 | Reserved for MBR/BIOS |
60000800 +------------------------+
61 | Typically used by MBR |
62000600 +------------------------+
63 | BIOS use only |
64000000 +------------------------+
65
66
67When using bzImage, the protected-mode kernel was relocated to
680x100000 ("high memory"), and the kernel real-mode block (boot sector,
69setup, and stack/heap) was made relocatable to any address between
700x10000 and end of low memory. Unfortunately, in protocols 2.00 and
712.01 the command line is still required to live in the 0x9XXXX memory
72range, and that memory range is still overwritten by the early kernel.
73The 2.02 protocol resolves that problem.
74
75It is desirable to keep the "memory ceiling" -- the highest point in
76low memory touched by the boot loader -- as low as possible, since
77some newer BIOSes have begun to allocate some rather large amounts of
78memory, called the Extended BIOS Data Area, near the top of low
79memory. The boot loader should use the "INT 12h" BIOS call to verify
80how much low memory is available.
81
82Unfortunately, if INT 12h reports that the amount of memory is too
83low, there is usually nothing the boot loader can do but to report an
84error to the user. The boot loader should therefore be designed to
85take up as little space in low memory as it reasonably can. For
86zImage or old bzImage kernels, which need data written into the
870x90000 segment, the boot loader should make sure not to use memory
88above the 0x9A000 point; too many BIOSes will break above that point.
89
90
91**** THE REAL-MODE KERNEL HEADER
92
93In the following text, and anywhere in the kernel boot sequence, "a
94sector" refers to 512 bytes. It is independent of the actual sector
95size of the underlying medium.
96
97The first step in loading a Linux kernel should be to load the
98real-mode code (boot sector and setup code) and then examine the
99following header at offset 0x01f1. The real-mode code can total up to
10032K, although the boot loader may choose to load only the first two
101sectors (1K) and then examine the bootup sector size.
102
103The header looks like:
104
105Offset Proto Name Meaning
106/Size
107
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -070010801F1/1 ALL(1 setup_sects The size of the setup in sectors
Linus Torvalds1da177e2005-04-16 15:20:36 -070010901F2/2 ALL root_flags If set, the root is mounted readonly
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -070011001F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras
Linus Torvalds1da177e2005-04-16 15:20:36 -070011101F8/2 ALL ram_size DO NOT USE - for bootsect.S use only
11201FA/2 ALL vid_mode Video mode control
11301FC/2 ALL root_dev Default root device number
11401FE/2 ALL boot_flag 0xAA55 magic number
1150200/2 2.00+ jump Jump instruction
1160202/4 2.00+ header Magic signature "HdrS"
1170206/2 2.00+ version Boot protocol version supported
1180208/4 2.00+ realmode_swtch Boot loader hook (see below)
119020C/2 2.00+ start_sys The load-low segment (0x1000) (obsolete)
120020E/2 2.00+ kernel_version Pointer to kernel version string
1210210/1 2.00+ type_of_loader Boot loader identifier
1220211/1 2.00+ loadflags Boot protocol option flags
1230212/2 2.00+ setup_move_size Move to high memory size (used with hooks)
1240214/4 2.00+ code32_start Boot loader hook (see below)
1250218/4 2.00+ ramdisk_image initrd load address (set by boot loader)
126021C/4 2.00+ ramdisk_size initrd size (set by boot loader)
1270220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only
1280224/2 2.01+ heap_end_ptr Free memory after setup end
1290226/2 N/A pad1 Unused
1300228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line
131022C/4 2.03+ initrd_addr_max Highest legal initrd address
132
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -0700133(1) For backwards compatibility, if the setup_sects field contains 0, the
134 real value is 4.
135
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.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139
140If the "HdrS" (0x53726448) magic number is not found at offset 0x202,
141the boot protocol version is "old". Loading an old kernel, the
142following parameters should be assumed:
143
144 Image type = zImage
145 initrd not supported
146 Real-mode kernel must be located at 0x90000.
147
148Otherwise, the "version" field contains the protocol version,
149e.g. protocol version 2.01 will contain 0x0201 in this field. When
150setting fields in the header, you must make sure only to set fields
151supported by the protocol version in use.
152
153The "kernel_version" field, if set to a nonzero value, contains a
154pointer to a null-terminated human-readable kernel version number
155string, less 0x200. This can be used to display the kernel version to
156the user. This value should be less than (0x200*setup_sects). For
157example, if this value is set to 0x1c00, the kernel version number
158string can be found at offset 0x1e00 in the kernel file. This is a
159valid value if and only if the "setup_sects" field contains the value
16014 or higher.
161
162Most boot loaders will simply load the kernel at its target address
163directly. Such boot loaders do not need to worry about filling in
164most of the fields in the header. The following fields should be
165filled out, however:
166
167 vid_mode:
168 Please see the section on SPECIAL COMMAND LINE OPTIONS.
169
170 type_of_loader:
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.
174
175 Assigned boot loader ids:
176 0 LILO
177 1 Loadlin
178 2 bootsect-loader
179 3 SYSLINUX
180 4 EtherBoot
181 5 ELILO
182 7 GRuB
183 8 U-BOOT
Jeremy Fitzhardinge354332e2006-09-12 20:35:57 -0700184 9 Xen
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185
186 Please contact <hpa@zytor.com> if you need a bootloader ID
187 value assigned.
188
189 loadflags, heap_end_ptr:
190 If the protocol version is 2.01 or higher, enter the
191 offset limit of the setup heap into heap_end_ptr and set the
192 0x80 bit (CAN_USE_HEAP) of loadflags. heap_end_ptr appears to
193 be relative to the start of setup (offset 0x0200).
194
195 setup_move_size:
196 When using protocol 2.00 or 2.01, if the real mode
197 kernel is not loaded at 0x90000, it gets moved there later in
198 the loading sequence. Fill in this field if you want
199 additional data (such as the kernel command line) moved in
200 addition to the real-mode kernel itself.
201
202 ramdisk_image, ramdisk_size:
203 If your boot loader has loaded an initial ramdisk (initrd),
204 set ramdisk_image to the 32-bit pointer to the ramdisk data
205 and the ramdisk_size to the size of the ramdisk data.
206
207 The initrd should typically be located as high in memory as
208 possible, as it may otherwise get overwritten by the early
209 kernel initialization sequence. However, it must never be
210 located above the address specified in the initrd_addr_max
211 field. The initrd should be at least 4K page aligned.
212
213 cmd_line_ptr:
214 If the protocol version is 2.02 or higher, this is a 32-bit
215 pointer to the kernel command line. The kernel command line
216 can be located anywhere between the end of setup and 0xA0000.
217 Fill in this field even if your boot loader does not support a
218 command line, in which case you can point this to an empty
219 string (or better yet, to the string "auto".) If this field
220 is left at zero, the kernel will assume that your boot loader
221 does not support the 2.02+ protocol.
222
223 ramdisk_max:
224 The maximum address that may be occupied by the initrd
225 contents. For boot protocols 2.02 or earlier, this field is
226 not present, and the maximum address is 0x37FFFFFF. (This
227 address is defined as the address of the highest safe byte, so
228 if your ramdisk is exactly 131072 bytes long and this field is
229 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)
230
231
232**** THE KERNEL COMMAND LINE
233
234The kernel command line has become an important way for the boot
235loader to communicate with the kernel. Some of its options are also
236relevant to the boot loader itself, see "special command line options"
237below.
238
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -0700239The kernel command line is a null-terminated string currently up to
240255 characters long, plus the final null. A string that is too long
241will be automatically truncated by the kernel, a boot loader may allow
242a longer command line to be passed to permit future kernels to extend
243this limit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
245If the boot protocol version is 2.02 or later, the address of the
246kernel command line is given by the header field cmd_line_ptr (see
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -0700247above.) This address can be anywhere between the end of the setup
248heap and 0xA0000.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
250If the protocol version is *not* 2.02 or higher, the kernel
251command line is entered using the following protocol:
252
253 At offset 0x0020 (word), "cmd_line_magic", enter the magic
254 number 0xA33F.
255
256 At offset 0x0022 (word), "cmd_line_offset", enter the offset
257 of the kernel command line (relative to the start of the
258 real-mode kernel).
259
260 The kernel command line *must* be within the memory region
261 covered by setup_move_size, so you may need to adjust this
262 field.
263
264
265**** SAMPLE BOOT CONFIGURATION
266
267As a sample configuration, assume the following layout of the real
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -0700268mode segment (this is a typical, and recommended layout):
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
270 0x0000-0x7FFF Real mode kernel
271 0x8000-0x8FFF Stack and heap
272 0x9000-0x90FF Kernel command line
273
274Such a boot loader should enter the following fields in the header:
275
276 unsigned long base_ptr; /* base address for real-mode segment */
277
278 if ( setup_sects == 0 ) {
279 setup_sects = 4;
280 }
281
282 if ( protocol >= 0x0200 ) {
283 type_of_loader = <type code>;
284 if ( loading_initrd ) {
285 ramdisk_image = <initrd_address>;
286 ramdisk_size = <initrd_size>;
287 }
288 if ( protocol >= 0x0201 ) {
289 heap_end_ptr = 0x9000 - 0x200;
290 loadflags |= 0x80; /* CAN_USE_HEAP */
291 }
292 if ( protocol >= 0x0202 ) {
293 cmd_line_ptr = base_ptr + 0x9000;
294 } else {
295 cmd_line_magic = 0xA33F;
296 cmd_line_offset = 0x9000;
297 setup_move_size = 0x9100;
298 }
299 } else {
300 /* Very old kernel */
301
302 cmd_line_magic = 0xA33F;
303 cmd_line_offset = 0x9000;
304
305 /* A very old kernel MUST have its real-mode code
306 loaded at 0x90000 */
307
308 if ( base_ptr != 0x90000 ) {
309 /* Copy the real-mode kernel */
310 memcpy(0x90000, base_ptr, (setup_sects+1)*512);
311 /* Copy the command line */
312 memcpy(0x99000, base_ptr+0x9000, 256);
313
314 base_ptr = 0x90000; /* Relocated */
315 }
316
317 /* It is recommended to clear memory up to the 32K mark */
318 memset(0x90000 + (setup_sects+1)*512, 0,
319 (64-(setup_sects+1))*512);
320 }
321
322
323**** LOADING THE REST OF THE KERNEL
324
H. Peter Anvinf8eeaaf2005-09-06 15:17:24 -0700325The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512
326in the kernel file (again, if setup_sects == 0 the real value is 4.)
327It should be loaded at address 0x10000 for Image/zImage kernels and
Linus Torvalds1da177e2005-04-16 15:20:36 -07003280x100000 for bzImage kernels.
329
330The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01
331bit (LOAD_HIGH) in the loadflags field is set:
332
333 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);
334 load_address = is_bzImage ? 0x100000 : 0x10000;
335
336Note that Image/zImage kernels can be up to 512K in size, and thus use
337the entire 0x10000-0x90000 range of memory. This means it is pretty
338much a requirement for these kernels to load the real-mode part at
3390x90000. bzImage kernels allow much more flexibility.
340
341
342**** SPECIAL COMMAND LINE OPTIONS
343
344If the command line provided by the boot loader is entered by the
345user, the user may expect the following command line options to work.
346They should normally not be deleted from the kernel command line even
347though not all of them are actually meaningful to the kernel. Boot
348loader authors who need additional command line options for the boot
349loader itself should get them registered in
350Documentation/kernel-parameters.txt to make sure they will not
351conflict with actual kernel options now or in the future.
352
353 vga=<mode>
354 <mode> here is either an integer (in C notation, either
355 decimal, octal, or hexadecimal) or one of the strings
356 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"
357 (meaning 0xFFFD). This value should be entered into the
358 vid_mode field, as it is used by the kernel before the command
359 line is parsed.
360
361 mem=<size>
362 <size> is an integer in C notation optionally followed by K, M
363 or G (meaning << 10, << 20 or << 30). This specifies the end
364 of memory to the kernel. This affects the possible placement
365 of an initrd, since an initrd should be placed near end of
366 memory. Note that this is an option to *both* the kernel and
367 the bootloader!
368
369 initrd=<file>
370 An initrd should be loaded. The meaning of <file> is
371 obviously bootloader-dependent, and some boot loaders
372 (e.g. LILO) do not have such a command.
373
374In addition, some boot loaders add the following options to the
375user-specified command line:
376
377 BOOT_IMAGE=<file>
378 The boot image which was loaded. Again, the meaning of <file>
379 is obviously bootloader-dependent.
380
381 auto
382 The kernel was booted without explicit user intervention.
383
384If these options are added by the boot loader, it is highly
385recommended that they are located *first*, before the user-specified
386or configuration-specified command line. Otherwise, "init=/bin/sh"
387gets confused by the "auto" option.
388
389
390**** RUNNING THE KERNEL
391
392The kernel is started by jumping to the kernel entry point, which is
393located at *segment* offset 0x20 from the start of the real mode
394kernel. This means that if you loaded your real-mode kernel code at
3950x90000, the kernel entry point is 9020:0000.
396
397At entry, ds = es = ss should point to the start of the real-mode
398kernel code (0x9000 if the code is loaded at 0x90000), sp should be
399set up properly, normally pointing to the top of the heap, and
400interrupts should be disabled. Furthermore, to guard against bugs in
401the kernel, it is recommended that the boot loader sets fs = gs = ds =
402es = ss.
403
404In our example from above, we would do:
405
406 /* Note: in the case of the "old" kernel protocol, base_ptr must
407 be == 0x90000 at this point; see the previous sample code */
408
409 seg = base_ptr >> 4;
410
411 cli(); /* Enter with interrupts disabled! */
412
413 /* Set up the real-mode kernel stack */
414 _SS = seg;
415 _SP = 0x9000; /* Load SP immediately after loading SS! */
416
417 _DS = _ES = _FS = _GS = seg;
418 jmp_far(seg+0x20, 0); /* Run the kernel */
419
420If your boot sector accesses a floppy drive, it is recommended to
421switch off the floppy motor before running the kernel, since the
422kernel boot leaves interrupts off and thus the motor will not be
423switched off, especially if the loaded kernel has the floppy driver as
424a demand-loaded module!
425
426
427**** ADVANCED BOOT TIME HOOKS
428
429If the boot loader runs in a particularly hostile environment (such as
430LOADLIN, which runs under DOS) it may be impossible to follow the
431standard memory location requirements. Such a boot loader may use the
432following hooks that, if set, are invoked by the kernel at the
433appropriate time. The use of these hooks should probably be
434considered an absolutely last resort!
435
436IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and
437%edi across invocation.
438
439 realmode_swtch:
440 A 16-bit real mode far subroutine invoked immediately before
441 entering protected mode. The default routine disables NMI, so
442 your routine should probably do so, too.
443
444 code32_start:
445 A 32-bit flat-mode routine *jumped* to immediately after the
446 transition to protected mode, but before the kernel is
447 uncompressed. No segments, except CS, are set up; you should
448 set them up to KERNEL_DS (0x18) yourself.
449
450 After completing your hook, you should jump to the address
451 that was in this field before your boot loader overwrote it.