boot.txt 17 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456
  1. THE LINUX/I386 BOOT PROTOCOL
  2. ----------------------------
  3. H. Peter Anvin <hpa@zytor.com>
  4. Last update 2007-01-26
  5. On the i386 platform, the Linux kernel uses a rather complicated boot
  6. convention. This has evolved partially due to historical aspects, as
  7. well as the desire in the early days to have the kernel itself be a
  8. bootable image, the complicated PC memory model and due to changed
  9. expectations in the PC industry caused by the effective demise of
  10. real-mode DOS as a mainstream operating system.
  11. Currently, four versions of the Linux/i386 boot protocol exist.
  12. Old kernels: zImage/Image support only. Some very early kernels
  13. may not even support a command line.
  14. Protocol 2.00: (Kernel 1.3.73) Added bzImage and initrd support, as
  15. well as a formalized way to communicate between the
  16. boot loader and the kernel. setup.S made relocatable,
  17. although the traditional setup area still assumed
  18. writable.
  19. Protocol 2.01: (Kernel 1.3.76) Added a heap overrun warning.
  20. Protocol 2.02: (Kernel 2.4.0-test3-pre3) New command line protocol.
  21. Lower the conventional memory ceiling. No overwrite
  22. of the traditional setup area, thus making booting
  23. safe for systems which use the EBDA from SMM or 32-bit
  24. BIOS entry points. zImage deprecated but still
  25. supported.
  26. Protocol 2.03: (Kernel 2.4.18-pre1) Explicitly makes the highest possible
  27. initrd address available to the bootloader.
  28. Protocol 2.04: (Kernel 2.6.14) Extend the syssize field to four bytes.
  29. Protocol 2.05: (Kernel 2.6.20) Make protected mode kernel relocatable.
  30. Introduce relocatable_kernel and kernel_alignment fields.
  31. **** MEMORY LAYOUT
  32. The traditional memory map for the kernel loader, used for Image or
  33. zImage kernels, typically looks like:
  34. | |
  35. 0A0000 +------------------------+
  36. | Reserved for BIOS | Do not use. Reserved for BIOS EBDA.
  37. 09A000 +------------------------+
  38. | Stack/heap/cmdline | For use by the kernel real-mode code.
  39. 098000 +------------------------+
  40. | Kernel setup | The kernel real-mode code.
  41. 090200 +------------------------+
  42. | Kernel boot sector | The kernel legacy boot sector.
  43. 090000 +------------------------+
  44. | Protected-mode kernel | The bulk of the kernel image.
  45. 010000 +------------------------+
  46. | Boot loader | <- Boot sector entry point 0000:7C00
  47. 001000 +------------------------+
  48. | Reserved for MBR/BIOS |
  49. 000800 +------------------------+
  50. | Typically used by MBR |
  51. 000600 +------------------------+
  52. | BIOS use only |
  53. 000000 +------------------------+
  54. When using bzImage, the protected-mode kernel was relocated to
  55. 0x100000 ("high memory"), and the kernel real-mode block (boot sector,
  56. setup, and stack/heap) was made relocatable to any address between
  57. 0x10000 and end of low memory. Unfortunately, in protocols 2.00 and
  58. 2.01 the command line is still required to live in the 0x9XXXX memory
  59. range, and that memory range is still overwritten by the early kernel.
  60. The 2.02 protocol resolves that problem.
  61. It is desirable to keep the "memory ceiling" -- the highest point in
  62. low memory touched by the boot loader -- as low as possible, since
  63. some newer BIOSes have begun to allocate some rather large amounts of
  64. memory, called the Extended BIOS Data Area, near the top of low
  65. memory. The boot loader should use the "INT 12h" BIOS call to verify
  66. how much low memory is available.
  67. Unfortunately, if INT 12h reports that the amount of memory is too
  68. low, there is usually nothing the boot loader can do but to report an
  69. error to the user. The boot loader should therefore be designed to
  70. take up as little space in low memory as it reasonably can. For
  71. zImage or old bzImage kernels, which need data written into the
  72. 0x90000 segment, the boot loader should make sure not to use memory
  73. above the 0x9A000 point; too many BIOSes will break above that point.
  74. **** THE REAL-MODE KERNEL HEADER
  75. In the following text, and anywhere in the kernel boot sequence, "a
  76. sector" refers to 512 bytes. It is independent of the actual sector
  77. size of the underlying medium.
  78. The first step in loading a Linux kernel should be to load the
  79. real-mode code (boot sector and setup code) and then examine the
  80. following header at offset 0x01f1. The real-mode code can total up to
  81. 32K, although the boot loader may choose to load only the first two
  82. sectors (1K) and then examine the bootup sector size.
  83. The header looks like:
  84. Offset Proto Name Meaning
  85. /Size
  86. 01F1/1 ALL(1 setup_sects The size of the setup in sectors
  87. 01F2/2 ALL root_flags If set, the root is mounted readonly
  88. 01F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras
  89. 01F8/2 ALL ram_size DO NOT USE - for bootsect.S use only
  90. 01FA/2 ALL vid_mode Video mode control
  91. 01FC/2 ALL root_dev Default root device number
  92. 01FE/2 ALL boot_flag 0xAA55 magic number
  93. 0200/2 2.00+ jump Jump instruction
  94. 0202/4 2.00+ header Magic signature "HdrS"
  95. 0206/2 2.00+ version Boot protocol version supported
  96. 0208/4 2.00+ realmode_swtch Boot loader hook (see below)
  97. 020C/2 2.00+ start_sys The load-low segment (0x1000) (obsolete)
  98. 020E/2 2.00+ kernel_version Pointer to kernel version string
  99. 0210/1 2.00+ type_of_loader Boot loader identifier
  100. 0211/1 2.00+ loadflags Boot protocol option flags
  101. 0212/2 2.00+ setup_move_size Move to high memory size (used with hooks)
  102. 0214/4 2.00+ code32_start Boot loader hook (see below)
  103. 0218/4 2.00+ ramdisk_image initrd load address (set by boot loader)
  104. 021C/4 2.00+ ramdisk_size initrd size (set by boot loader)
  105. 0220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only
  106. 0224/2 2.01+ heap_end_ptr Free memory after setup end
  107. 0226/2 N/A pad1 Unused
  108. 0228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line
  109. 022C/4 2.03+ initrd_addr_max Highest legal initrd address
  110. 0230/4 2.05+ kernel_alignment Physical addr alignment required for kernel
  111. 0234/1 2.05+ relocatable_kernel Whether kernel is relocatable or not
  112. (1) For backwards compatibility, if the setup_sects field contains 0, the
  113. real value is 4.
  114. (2) For boot protocol prior to 2.04, the upper two bytes of the syssize
  115. field are unusable, which means the size of a bzImage kernel
  116. cannot be determined.
  117. If the "HdrS" (0x53726448) magic number is not found at offset 0x202,
  118. the boot protocol version is "old". Loading an old kernel, the
  119. following parameters should be assumed:
  120. Image type = zImage
  121. initrd not supported
  122. Real-mode kernel must be located at 0x90000.
  123. Otherwise, the "version" field contains the protocol version,
  124. e.g. protocol version 2.01 will contain 0x0201 in this field. When
  125. setting fields in the header, you must make sure only to set fields
  126. supported by the protocol version in use.
  127. The "kernel_version" field, if set to a nonzero value, contains a
  128. pointer to a null-terminated human-readable kernel version number
  129. string, less 0x200. This can be used to display the kernel version to
  130. the user. This value should be less than (0x200*setup_sects). For
  131. example, if this value is set to 0x1c00, the kernel version number
  132. string can be found at offset 0x1e00 in the kernel file. This is a
  133. valid value if and only if the "setup_sects" field contains the value
  134. 14 or higher.
  135. Most boot loaders will simply load the kernel at its target address
  136. directly. Such boot loaders do not need to worry about filling in
  137. most of the fields in the header. The following fields should be
  138. filled out, however:
  139. vid_mode:
  140. Please see the section on SPECIAL COMMAND LINE OPTIONS.
  141. type_of_loader:
  142. If your boot loader has an assigned id (see table below), enter
  143. 0xTV here, where T is an identifier for the boot loader and V is
  144. a version number. Otherwise, enter 0xFF here.
  145. Assigned boot loader ids:
  146. 0 LILO
  147. 1 Loadlin
  148. 2 bootsect-loader
  149. 3 SYSLINUX
  150. 4 EtherBoot
  151. 5 ELILO
  152. 7 GRuB
  153. 8 U-BOOT
  154. 9 Xen
  155. A Gujin
  156. Please contact <hpa@zytor.com> if you need a bootloader ID
  157. value assigned.
  158. loadflags, heap_end_ptr:
  159. If the protocol version is 2.01 or higher, enter the
  160. offset limit of the setup heap into heap_end_ptr and set the
  161. 0x80 bit (CAN_USE_HEAP) of loadflags. heap_end_ptr appears to
  162. be relative to the start of setup (offset 0x0200).
  163. setup_move_size:
  164. When using protocol 2.00 or 2.01, if the real mode
  165. kernel is not loaded at 0x90000, it gets moved there later in
  166. the loading sequence. Fill in this field if you want
  167. additional data (such as the kernel command line) moved in
  168. addition to the real-mode kernel itself.
  169. ramdisk_image, ramdisk_size:
  170. If your boot loader has loaded an initial ramdisk (initrd),
  171. set ramdisk_image to the 32-bit pointer to the ramdisk data
  172. and the ramdisk_size to the size of the ramdisk data.
  173. The initrd should typically be located as high in memory as
  174. possible, as it may otherwise get overwritten by the early
  175. kernel initialization sequence. However, it must never be
  176. located above the address specified in the initrd_addr_max
  177. field. The initrd should be at least 4K page aligned.
  178. cmd_line_ptr:
  179. If the protocol version is 2.02 or higher, this is a 32-bit
  180. pointer to the kernel command line. The kernel command line
  181. can be located anywhere between the end of setup and 0xA0000.
  182. Fill in this field even if your boot loader does not support a
  183. command line, in which case you can point this to an empty
  184. string (or better yet, to the string "auto".) If this field
  185. is left at zero, the kernel will assume that your boot loader
  186. does not support the 2.02+ protocol.
  187. ramdisk_max:
  188. The maximum address that may be occupied by the initrd
  189. contents. For boot protocols 2.02 or earlier, this field is
  190. not present, and the maximum address is 0x37FFFFFF. (This
  191. address is defined as the address of the highest safe byte, so
  192. if your ramdisk is exactly 131072 bytes long and this field is
  193. 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)
  194. **** THE KERNEL COMMAND LINE
  195. The kernel command line has become an important way for the boot
  196. loader to communicate with the kernel. Some of its options are also
  197. relevant to the boot loader itself, see "special command line options"
  198. below.
  199. The kernel command line is a null-terminated string currently up to
  200. 255 characters long, plus the final null. A string that is too long
  201. will be automatically truncated by the kernel, a boot loader may allow
  202. a longer command line to be passed to permit future kernels to extend
  203. this limit.
  204. If the boot protocol version is 2.02 or later, the address of the
  205. kernel command line is given by the header field cmd_line_ptr (see
  206. above.) This address can be anywhere between the end of the setup
  207. heap and 0xA0000.
  208. If the protocol version is *not* 2.02 or higher, the kernel
  209. command line is entered using the following protocol:
  210. At offset 0x0020 (word), "cmd_line_magic", enter the magic
  211. number 0xA33F.
  212. At offset 0x0022 (word), "cmd_line_offset", enter the offset
  213. of the kernel command line (relative to the start of the
  214. real-mode kernel).
  215. The kernel command line *must* be within the memory region
  216. covered by setup_move_size, so you may need to adjust this
  217. field.
  218. **** SAMPLE BOOT CONFIGURATION
  219. As a sample configuration, assume the following layout of the real
  220. mode segment (this is a typical, and recommended layout):
  221. 0x0000-0x7FFF Real mode kernel
  222. 0x8000-0x8FFF Stack and heap
  223. 0x9000-0x90FF Kernel command line
  224. Such a boot loader should enter the following fields in the header:
  225. unsigned long base_ptr; /* base address for real-mode segment */
  226. if ( setup_sects == 0 ) {
  227. setup_sects = 4;
  228. }
  229. if ( protocol >= 0x0200 ) {
  230. type_of_loader = <type code>;
  231. if ( loading_initrd ) {
  232. ramdisk_image = <initrd_address>;
  233. ramdisk_size = <initrd_size>;
  234. }
  235. if ( protocol >= 0x0201 ) {
  236. heap_end_ptr = 0x9000 - 0x200;
  237. loadflags |= 0x80; /* CAN_USE_HEAP */
  238. }
  239. if ( protocol >= 0x0202 ) {
  240. cmd_line_ptr = base_ptr + 0x9000;
  241. } else {
  242. cmd_line_magic = 0xA33F;
  243. cmd_line_offset = 0x9000;
  244. setup_move_size = 0x9100;
  245. }
  246. } else {
  247. /* Very old kernel */
  248. cmd_line_magic = 0xA33F;
  249. cmd_line_offset = 0x9000;
  250. /* A very old kernel MUST have its real-mode code
  251. loaded at 0x90000 */
  252. if ( base_ptr != 0x90000 ) {
  253. /* Copy the real-mode kernel */
  254. memcpy(0x90000, base_ptr, (setup_sects+1)*512);
  255. /* Copy the command line */
  256. memcpy(0x99000, base_ptr+0x9000, 256);
  257. base_ptr = 0x90000; /* Relocated */
  258. }
  259. /* It is recommended to clear memory up to the 32K mark */
  260. memset(0x90000 + (setup_sects+1)*512, 0,
  261. (64-(setup_sects+1))*512);
  262. }
  263. **** LOADING THE REST OF THE KERNEL
  264. The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512
  265. in the kernel file (again, if setup_sects == 0 the real value is 4.)
  266. It should be loaded at address 0x10000 for Image/zImage kernels and
  267. 0x100000 for bzImage kernels.
  268. The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01
  269. bit (LOAD_HIGH) in the loadflags field is set:
  270. is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);
  271. load_address = is_bzImage ? 0x100000 : 0x10000;
  272. Note that Image/zImage kernels can be up to 512K in size, and thus use
  273. the entire 0x10000-0x90000 range of memory. This means it is pretty
  274. much a requirement for these kernels to load the real-mode part at
  275. 0x90000. bzImage kernels allow much more flexibility.
  276. **** SPECIAL COMMAND LINE OPTIONS
  277. If the command line provided by the boot loader is entered by the
  278. user, the user may expect the following command line options to work.
  279. They should normally not be deleted from the kernel command line even
  280. though not all of them are actually meaningful to the kernel. Boot
  281. loader authors who need additional command line options for the boot
  282. loader itself should get them registered in
  283. Documentation/kernel-parameters.txt to make sure they will not
  284. conflict with actual kernel options now or in the future.
  285. vga=<mode>
  286. <mode> here is either an integer (in C notation, either
  287. decimal, octal, or hexadecimal) or one of the strings
  288. "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"
  289. (meaning 0xFFFD). This value should be entered into the
  290. vid_mode field, as it is used by the kernel before the command
  291. line is parsed.
  292. mem=<size>
  293. <size> is an integer in C notation optionally followed by K, M
  294. or G (meaning << 10, << 20 or << 30). This specifies the end
  295. of memory to the kernel. This affects the possible placement
  296. of an initrd, since an initrd should be placed near end of
  297. memory. Note that this is an option to *both* the kernel and
  298. the bootloader!
  299. initrd=<file>
  300. An initrd should be loaded. The meaning of <file> is
  301. obviously bootloader-dependent, and some boot loaders
  302. (e.g. LILO) do not have such a command.
  303. In addition, some boot loaders add the following options to the
  304. user-specified command line:
  305. BOOT_IMAGE=<file>
  306. The boot image which was loaded. Again, the meaning of <file>
  307. is obviously bootloader-dependent.
  308. auto
  309. The kernel was booted without explicit user intervention.
  310. If these options are added by the boot loader, it is highly
  311. recommended that they are located *first*, before the user-specified
  312. or configuration-specified command line. Otherwise, "init=/bin/sh"
  313. gets confused by the "auto" option.
  314. **** RUNNING THE KERNEL
  315. The kernel is started by jumping to the kernel entry point, which is
  316. located at *segment* offset 0x20 from the start of the real mode
  317. kernel. This means that if you loaded your real-mode kernel code at
  318. 0x90000, the kernel entry point is 9020:0000.
  319. At entry, ds = es = ss should point to the start of the real-mode
  320. kernel code (0x9000 if the code is loaded at 0x90000), sp should be
  321. set up properly, normally pointing to the top of the heap, and
  322. interrupts should be disabled. Furthermore, to guard against bugs in
  323. the kernel, it is recommended that the boot loader sets fs = gs = ds =
  324. es = ss.
  325. In our example from above, we would do:
  326. /* Note: in the case of the "old" kernel protocol, base_ptr must
  327. be == 0x90000 at this point; see the previous sample code */
  328. seg = base_ptr >> 4;
  329. cli(); /* Enter with interrupts disabled! */
  330. /* Set up the real-mode kernel stack */
  331. _SS = seg;
  332. _SP = 0x9000; /* Load SP immediately after loading SS! */
  333. _DS = _ES = _FS = _GS = seg;
  334. jmp_far(seg+0x20, 0); /* Run the kernel */
  335. If your boot sector accesses a floppy drive, it is recommended to
  336. switch off the floppy motor before running the kernel, since the
  337. kernel boot leaves interrupts off and thus the motor will not be
  338. switched off, especially if the loaded kernel has the floppy driver as
  339. a demand-loaded module!
  340. **** ADVANCED BOOT TIME HOOKS
  341. If the boot loader runs in a particularly hostile environment (such as
  342. LOADLIN, which runs under DOS) it may be impossible to follow the
  343. standard memory location requirements. Such a boot loader may use the
  344. following hooks that, if set, are invoked by the kernel at the
  345. appropriate time. The use of these hooks should probably be
  346. considered an absolutely last resort!
  347. IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and
  348. %edi across invocation.
  349. realmode_swtch:
  350. A 16-bit real mode far subroutine invoked immediately before
  351. entering protected mode. The default routine disables NMI, so
  352. your routine should probably do so, too.
  353. code32_start:
  354. A 32-bit flat-mode routine *jumped* to immediately after the
  355. transition to protected mode, but before the kernel is
  356. uncompressed. No segments, except CS, are set up; you should
  357. set them up to KERNEL_DS (0x18) yourself.
  358. After completing your hook, you should jump to the address
  359. that was in this field before your boot loader overwrote it.