1 THE LINUX/x86 BOOT PROTOCOL 2 --------------------------- 3 4On the x86 platform, the Linux kernel uses a rather complicated boot 5convention. This has evolved partially due to historical aspects, as 6well as the desire in the early days to have the kernel itself be a 7bootable image, the complicated PC memory model and due to changed 8expectations in the PC industry caused by the effective demise of 9real-mode DOS as a mainstream operating system. 10 11Currently, the following versions of the Linux/x86 boot protocol exist. 12 13Old kernels: zImage/Image support only. Some very early kernels 14 may not even support a command line. 15 16Protocol 2.00: (Kernel 1.3.73) Added bzImage and initrd support, as 17 well as a formalized way to communicate between the 18 boot loader and the kernel. setup.S made relocatable, 19 although the traditional setup area still assumed 20 writable. 21 22Protocol 2.01: (Kernel 1.3.76) Added a heap overrun warning. 23 24Protocol 2.02: (Kernel 2.4.0-test3-pre3) New command line protocol. 25 Lower the conventional memory ceiling. No overwrite 26 of the traditional setup area, thus making booting 27 safe for systems which use the EBDA from SMM or 32-bit 28 BIOS entry points. zImage deprecated but still 29 supported. 30 31Protocol 2.03: (Kernel 2.4.18-pre1) Explicitly makes the highest possible 32 initrd address available to the bootloader. 33 34Protocol 2.04: (Kernel 2.6.14) Extend the syssize field to four bytes. 35 36Protocol 2.05: (Kernel 2.6.20) Make protected mode kernel relocatable. 37 Introduce relocatable_kernel and kernel_alignment fields. 38 39Protocol 2.06: (Kernel 2.6.22) Added a field that contains the size of 40 the boot command line. 41 42Protocol 2.07: (Kernel 2.6.24) Added paravirtualised boot protocol. 43 Introduced hardware_subarch and hardware_subarch_data 44 and KEEP_SEGMENTS flag in load_flags. 45 46Protocol 2.08: (Kernel 2.6.26) Added crc32 checksum and ELF format 47 payload. Introduced payload_offset and payload_length 48 fields to aid in locating the payload. 49 50Protocol 2.09: (Kernel 2.6.26) Added a field of 64-bit physical 51 pointer to single linked list of struct setup_data. 52 53Protocol 2.10: (Kernel 2.6.31) Added a protocol for relaxed alignment 54 beyond the kernel_alignment added, new init_size and 55 pref_address fields. Added extended boot loader IDs. 56 57**** MEMORY LAYOUT 58 59The traditional memory map for the kernel loader, used for Image or 60zImage kernels, typically looks like: 61 62 | | 630A0000 +------------------------+ 64 | Reserved for BIOS | Do not use. Reserved for BIOS EBDA. 6509A000 +------------------------+ 66 | Command line | 67 | Stack/heap | For use by the kernel real-mode code. 68098000 +------------------------+ 69 | Kernel setup | The kernel real-mode code. 70090200 +------------------------+ 71 | Kernel boot sector | The kernel legacy boot sector. 72090000 +------------------------+ 73 | Protected-mode kernel | The bulk of the kernel image. 74010000 +------------------------+ 75 | Boot loader | <- Boot sector entry point 0000:7C00 76001000 +------------------------+ 77 | Reserved for MBR/BIOS | 78000800 +------------------------+ 79 | Typically used by MBR | 80000600 +------------------------+ 81 | BIOS use only | 82000000 +------------------------+ 83 84 85When using bzImage, the protected-mode kernel was relocated to 860x100000 ("high memory"), and the kernel real-mode block (boot sector, 87setup, and stack/heap) was made relocatable to any address between 880x10000 and end of low memory. Unfortunately, in protocols 2.00 and 892.01 the 0x90000+ memory range is still used internally by the kernel; 90the 2.02 protocol resolves that problem. 91 92It is desirable to keep the "memory ceiling" -- the highest point in 93low memory touched by the boot loader -- as low as possible, since 94some newer BIOSes have begun to allocate some rather large amounts of 95memory, called the Extended BIOS Data Area, near the top of low 96memory. The boot loader should use the "INT 12h" BIOS call to verify 97how much low memory is available. 98 99Unfortunately, if INT 12h reports that the amount of memory is too 100low, there is usually nothing the boot loader can do but to report an 101error to the user. The boot loader should therefore be designed to 102take up as little space in low memory as it reasonably can. For 103zImage or old bzImage kernels, which need data written into the 1040x90000 segment, the boot loader should make sure not to use memory 105above the 0x9A000 point; too many BIOSes will break above that point. 106 107For a modern bzImage kernel with boot protocol version >= 2.02, a 108memory layout like the following is suggested: 109 110 ~ ~ 111 | Protected-mode kernel | 112100000 +------------------------+ 113 | I/O memory hole | 1140A0000 +------------------------+ 115 | Reserved for BIOS | Leave as much as possible unused 116 ~ ~ 117 | Command line | (Can also be below the X+10000 mark) 118X+10000 +------------------------+ 119 | Stack/heap | For use by the kernel real-mode code. 120X+08000 +------------------------+ 121 | Kernel setup | The kernel real-mode code. 122 | Kernel boot sector | The kernel legacy boot sector. 123X +------------------------+ 124 | Boot loader | <- Boot sector entry point 0000:7C00 125001000 +------------------------+ 126 | Reserved for MBR/BIOS | 127000800 +------------------------+ 128 | Typically used by MBR | 129000600 +------------------------+ 130 | BIOS use only | 131000000 +------------------------+ 132 133... where the address X is as low as the design of the boot loader 134permits. 135 136 137**** THE REAL-MODE KERNEL HEADER 138 139In the following text, and anywhere in the kernel boot sequence, "a 140sector" refers to 512 bytes. It is independent of the actual sector 141size of the underlying medium. 142 143The first step in loading a Linux kernel should be to load the 144real-mode code (boot sector and setup code) and then examine the 145following header at offset 0x01f1. The real-mode code can total up to 14632K, although the boot loader may choose to load only the first two 147sectors (1K) and then examine the bootup sector size. 148 149The header looks like: 150 151Offset Proto Name Meaning 152/Size 153 15401F1/1 ALL(1 setup_sects The size of the setup in sectors 15501F2/2 ALL root_flags If set, the root is mounted readonly 15601F4/4 2.04+(2 syssize The size of the 32-bit code in 16-byte paras 15701F8/2 ALL ram_size DO NOT USE - for bootsect.S use only 15801FA/2 ALL vid_mode Video mode control 15901FC/2 ALL root_dev Default root device number 16001FE/2 ALL boot_flag 0xAA55 magic number 1610200/2 2.00+ jump Jump instruction 1620202/4 2.00+ header Magic signature "HdrS" 1630206/2 2.00+ version Boot protocol version supported 1640208/4 2.00+ realmode_swtch Boot loader hook (see below) 165020C/2 2.00+ start_sys_seg The load-low segment (0x1000) (obsolete) 166020E/2 2.00+ kernel_version Pointer to kernel version string 1670210/1 2.00+ type_of_loader Boot loader identifier 1680211/1 2.00+ loadflags Boot protocol option flags 1690212/2 2.00+ setup_move_size Move to high memory size (used with hooks) 1700214/4 2.00+ code32_start Boot loader hook (see below) 1710218/4 2.00+ ramdisk_image initrd load address (set by boot loader) 172021C/4 2.00+ ramdisk_size initrd size (set by boot loader) 1730220/4 2.00+ bootsect_kludge DO NOT USE - for bootsect.S use only 1740224/2 2.01+ heap_end_ptr Free memory after setup end 1750226/1 2.02+(3 ext_loader_ver Extended boot loader version 1760227/1 2.02+(3 ext_loader_type Extended boot loader ID 1770228/4 2.02+ cmd_line_ptr 32-bit pointer to the kernel command line 178022C/4 2.03+ ramdisk_max Highest legal initrd address 1790230/4 2.05+ kernel_alignment Physical addr alignment required for kernel 1800234/1 2.05+ relocatable_kernel Whether kernel is relocatable or not 1810235/1 2.10+ min_alignment Minimum alignment, as a power of two 1820236/2 N/A pad3 Unused 1830238/4 2.06+ cmdline_size Maximum size of the kernel command line 184023C/4 2.07+ hardware_subarch Hardware subarchitecture 1850240/8 2.07+ hardware_subarch_data Subarchitecture-specific data 1860248/4 2.08+ payload_offset Offset of kernel payload 187024C/4 2.08+ payload_length Length of kernel payload 1880250/8 2.09+ setup_data 64-bit physical pointer to linked list 189 of struct setup_data 1900258/8 2.10+ pref_address Preferred loading address 1910260/4 2.10+ init_size Linear memory required during initialization 192 193(1) For backwards compatibility, if the setup_sects field contains 0, the 194 real value is 4. 195 196(2) For boot protocol prior to 2.04, the upper two bytes of the syssize 197 field are unusable, which means the size of a bzImage kernel 198 cannot be determined. 199 200(3) Ignored, but safe to set, for boot protocols 2.02-2.09. 201 202If the "HdrS" (0x53726448) magic number is not found at offset 0x202, 203the boot protocol version is "old". Loading an old kernel, the 204following parameters should be assumed: 205 206 Image type = zImage 207 initrd not supported 208 Real-mode kernel must be located at 0x90000. 209 210Otherwise, the "version" field contains the protocol version, 211e.g. protocol version 2.01 will contain 0x0201 in this field. When 212setting fields in the header, you must make sure only to set fields 213supported by the protocol version in use. 214 215 216**** DETAILS OF HEADER FIELDS 217 218For each field, some are information from the kernel to the bootloader 219("read"), some are expected to be filled out by the bootloader 220("write"), and some are expected to be read and modified by the 221bootloader ("modify"). 222 223All general purpose boot loaders should write the fields marked 224(obligatory). Boot loaders who want to load the kernel at a 225nonstandard address should fill in the fields marked (reloc); other 226boot loaders can ignore those fields. 227 228The byte order of all fields is littleendian (this is x86, after all.) 229 230Field name: setup_sects 231Type: read 232Offset/size: 0x1f1/1 233Protocol: ALL 234 235 The size of the setup code in 512-byte sectors. If this field is 236 0, the real value is 4. The real-mode code consists of the boot 237 sector (always one 512-byte sector) plus the setup code. 238 239Field name: root_flags 240Type: modify (optional) 241Offset/size: 0x1f2/2 242Protocol: ALL 243 244 If this field is nonzero, the root defaults to readonly. The use of 245 this field is deprecated; use the "ro" or "rw" options on the 246 command line instead. 247 248Field name: syssize 249Type: read 250Offset/size: 0x1f4/4 (protocol 2.04+) 0x1f4/2 (protocol ALL) 251Protocol: 2.04+ 252 253 The size of the protected-mode code in units of 16-byte paragraphs. 254 For protocol versions older than 2.04 this field is only two bytes 255 wide, and therefore cannot be trusted for the size of a kernel if 256 the LOAD_HIGH flag is set. 257 258Field name: ram_size 259Type: kernel internal 260Offset/size: 0x1f8/2 261Protocol: ALL 262 263 This field is obsolete. 264 265Field name: vid_mode 266Type: modify (obligatory) 267Offset/size: 0x1fa/2 268 269 Please see the section on SPECIAL COMMAND LINE OPTIONS. 270 271Field name: root_dev 272Type: modify (optional) 273Offset/size: 0x1fc/2 274Protocol: ALL 275 276 The default root device device number. The use of this field is 277 deprecated, use the "root=" option on the command line instead. 278 279Field name: boot_flag 280Type: read 281Offset/size: 0x1fe/2 282Protocol: ALL 283 284 Contains 0xAA55. This is the closest thing old Linux kernels have 285 to a magic number. 286 287Field name: jump 288Type: read 289Offset/size: 0x200/2 290Protocol: 2.00+ 291 292 Contains an x86 jump instruction, 0xEB followed by a signed offset 293 relative to byte 0x202. This can be used to determine the size of 294 the header. 295 296Field name: header 297Type: read 298Offset/size: 0x202/4 299Protocol: 2.00+ 300 301 Contains the magic number "HdrS" (0x53726448). 302 303Field name: version 304Type: read 305Offset/size: 0x206/2 306Protocol: 2.00+ 307 308 Contains the boot protocol version, in (major << 8)+minor format, 309 e.g. 0x0204 for version 2.04, and 0x0a11 for a hypothetical version 310 10.17. 311 312Field name: realmode_swtch 313Type: modify (optional) 314Offset/size: 0x208/4 315Protocol: 2.00+ 316 317 Boot loader hook (see ADVANCED BOOT LOADER HOOKS below.) 318 319Field name: start_sys_seg 320Type: read 321Offset/size: 0x20c/2 322Protocol: 2.00+ 323 324 The load low segment (0x1000). Obsolete. 325 326Field name: kernel_version 327Type: read 328Offset/size: 0x20e/2 329Protocol: 2.00+ 330 331 If set to a nonzero value, contains a pointer to a NUL-terminated 332 human-readable kernel version number string, less 0x200. This can 333 be used to display the kernel version to the user. This value 334 should be less than (0x200*setup_sects). 335 336 For example, if this value is set to 0x1c00, the kernel version 337 number string can be found at offset 0x1e00 in the kernel file. 338 This is a valid value if and only if the "setup_sects" field 339 contains the value 15 or higher, as: 340 341 0x1c00 < 15*0x200 (= 0x1e00) but 342 0x1c00 >= 14*0x200 (= 0x1c00) 343 344 0x1c00 >> 9 = 14, so the minimum value for setup_secs is 15. 345 346Field name: type_of_loader 347Type: write (obligatory) 348Offset/size: 0x210/1 349Protocol: 2.00+ 350 351 If your boot loader has an assigned id (see table below), enter 352 0xTV here, where T is an identifier for the boot loader and V is 353 a version number. Otherwise, enter 0xFF here. 354 355 For boot loader IDs above T = 0xD, write T = 0xE to this field and 356 write the extended ID minus 0x10 to the ext_loader_type field. 357 Similarly, the ext_loader_ver field can be used to provide more than 358 four bits for the bootloader version. 359 360 For example, for T = 0x15, V = 0x234, write: 361 362 type_of_loader <- 0xE4 363 ext_loader_type <- 0x05 364 ext_loader_ver <- 0x23 365 366 Assigned boot loader ids: 367 0 LILO (0x00 reserved for pre-2.00 bootloader) 368 1 Loadlin 369 2 bootsect-loader (0x20, all other values reserved) 370 3 Syslinux 371 4 Etherboot/gPXE 372 5 ELILO 373 7 GRUB 374 8 U-Boot 375 9 Xen 376 A Gujin 377 B Qemu 378 C Arcturus Networks uCbootloader 379 E Extended (see ext_loader_type) 380 F Special (0xFF = undefined) 381 382 Please contact <hpa@zytor.com> if you need a bootloader ID 383 value assigned. 384 385Field name: loadflags 386Type: modify (obligatory) 387Offset/size: 0x211/1 388Protocol: 2.00+ 389 390 This field is a bitmask. 391 392 Bit 0 (read): LOADED_HIGH 393 - If 0, the protected-mode code is loaded at 0x10000. 394 - If 1, the protected-mode code is loaded at 0x100000. 395 396 Bit 5 (write): QUIET_FLAG 397 - If 0, print early messages. 398 - If 1, suppress early messages. 399 This requests to the kernel (decompressor and early 400 kernel) to not write early messages that require 401 accessing the display hardware directly. 402 403 Bit 6 (write): KEEP_SEGMENTS 404 Protocol: 2.07+ 405 - If 0, reload the segment registers in the 32bit entry point. 406 - If 1, do not reload the segment registers in the 32bit entry point. 407 Assume that %cs %ds %ss %es are all set to flat segments with 408 a base of 0 (or the equivalent for their environment). 409 410 Bit 7 (write): CAN_USE_HEAP 411 Set this bit to 1 to indicate that the value entered in the 412 heap_end_ptr is valid. If this field is clear, some setup code 413 functionality will be disabled. 414 415Field name: setup_move_size 416Type: modify (obligatory) 417Offset/size: 0x212/2 418Protocol: 2.00-2.01 419 420 When using protocol 2.00 or 2.01, if the real mode kernel is not 421 loaded at 0x90000, it gets moved there later in the loading 422 sequence. Fill in this field if you want additional data (such as 423 the kernel command line) moved in addition to the real-mode kernel 424 itself. 425 426 The unit is bytes starting with the beginning of the boot sector. 427 428 This field is can be ignored when the protocol is 2.02 or higher, or 429 if the real-mode code is loaded at 0x90000. 430 431Field name: code32_start 432Type: modify (optional, reloc) 433Offset/size: 0x214/4 434Protocol: 2.00+ 435 436 The address to jump to in protected mode. This defaults to the load 437 address of the kernel, and can be used by the boot loader to 438 determine the proper load address. 439 440 This field can be modified for two purposes: 441 442 1. as a boot loader hook (see ADVANCED BOOT LOADER HOOKS below.) 443 444 2. if a bootloader which does not install a hook loads a 445 relocatable kernel at a nonstandard address it will have to modify 446 this field to point to the load address. 447 448Field name: ramdisk_image 449Type: write (obligatory) 450Offset/size: 0x218/4 451Protocol: 2.00+ 452 453 The 32-bit linear address of the initial ramdisk or ramfs. Leave at 454 zero if there is no initial ramdisk/ramfs. 455 456Field name: ramdisk_size 457Type: write (obligatory) 458Offset/size: 0x21c/4 459Protocol: 2.00+ 460 461 Size of the initial ramdisk or ramfs. Leave at zero if there is no 462 initial ramdisk/ramfs. 463 464Field name: bootsect_kludge 465Type: kernel internal 466Offset/size: 0x220/4 467Protocol: 2.00+ 468 469 This field is obsolete. 470 471Field name: heap_end_ptr 472Type: write (obligatory) 473Offset/size: 0x224/2 474Protocol: 2.01+ 475 476 Set this field to the offset (from the beginning of the real-mode 477 code) of the end of the setup stack/heap, minus 0x0200. 478 479Field name: ext_loader_ver 480Type: write (optional) 481Offset/size: 0x226/1 482Protocol: 2.02+ 483 484 This field is used as an extension of the version number in the 485 type_of_loader field. The total version number is considered to be 486 (type_of_loader & 0x0f) + (ext_loader_ver << 4). 487 488 The use of this field is boot loader specific. If not written, it 489 is zero. 490 491 Kernels prior to 2.6.31 did not recognize this field, but it is safe 492 to write for protocol version 2.02 or higher. 493 494Field name: ext_loader_type 495Type: write (obligatory if (type_of_loader & 0xf0) == 0xe0) 496Offset/size: 0x227/1 497Protocol: 2.02+ 498 499 This field is used as an extension of the type number in 500 type_of_loader field. If the type in type_of_loader is 0xE, then 501 the actual type is (ext_loader_type + 0x10). 502 503 This field is ignored if the type in type_of_loader is not 0xE. 504 505 Kernels prior to 2.6.31 did not recognize this field, but it is safe 506 to write for protocol version 2.02 or higher. 507 508Field name: cmd_line_ptr 509Type: write (obligatory) 510Offset/size: 0x228/4 511Protocol: 2.02+ 512 513 Set this field to the linear address of the kernel command line. 514 The kernel command line can be located anywhere between the end of 515 the setup heap and 0xA0000; it does not have to be located in the 516 same 64K segment as the real-mode code itself. 517 518 Fill in this field even if your boot loader does not support a 519 command line, in which case you can point this to an empty string 520 (or better yet, to the string "auto".) If this field is left at 521 zero, the kernel will assume that your boot loader does not support 522 the 2.02+ protocol. 523 524Field name: ramdisk_max 525Type: read 526Offset/size: 0x22c/4 527Protocol: 2.03+ 528 529 The maximum address that may be occupied by the initial 530 ramdisk/ramfs contents. For boot protocols 2.02 or earlier, this 531 field is not present, and the maximum address is 0x37FFFFFF. (This 532 address is defined as the address of the highest safe byte, so if 533 your ramdisk is exactly 131072 bytes long and this field is 534 0x37FFFFFF, you can start your ramdisk at 0x37FE0000.) 535 536Field name: kernel_alignment 537Type: read/modify (reloc) 538Offset/size: 0x230/4 539Protocol: 2.05+ (read), 2.10+ (modify) 540 541 Alignment unit required by the kernel (if relocatable_kernel is 542 true.) A relocatable kernel that is loaded at an alignment 543 incompatible with the value in this field will be realigned during 544 kernel initialization. 545 546 Starting with protocol version 2.10, this reflects the kernel 547 alignment preferred for optimal performance; it is possible for the 548 loader to modify this field to permit a lesser alignment. See the 549 min_alignment and pref_address field below. 550 551Field name: relocatable_kernel 552Type: read (reloc) 553Offset/size: 0x234/1 554Protocol: 2.05+ 555 556 If this field is nonzero, the protected-mode part of the kernel can 557 be loaded at any address that satisfies the kernel_alignment field. 558 After loading, the boot loader must set the code32_start field to 559 point to the loaded code, or to a boot loader hook. 560 561Field name: min_alignment 562Type: read (reloc) 563Offset/size: 0x235/1 564Protocol: 2.10+ 565 566 This field, if nonzero, indicates as a power of two the minimum 567 alignment required, as opposed to preferred, by the kernel to boot. 568 If a boot loader makes use of this field, it should update the 569 kernel_alignment field with the alignment unit desired; typically: 570 571 kernel_alignment = 1 << min_alignment 572 573 There may be a considerable performance cost with an excessively 574 misaligned kernel. Therefore, a loader should typically try each 575 power-of-two alignment from kernel_alignment down to this alignment. 576 577Field name: cmdline_size 578Type: read 579Offset/size: 0x238/4 580Protocol: 2.06+ 581 582 The maximum size of the command line without the terminating 583 zero. This means that the command line can contain at most 584 cmdline_size characters. With protocol version 2.05 and earlier, the 585 maximum size was 255. 586 587Field name: hardware_subarch 588Type: write (optional, defaults to x86/PC) 589Offset/size: 0x23c/4 590Protocol: 2.07+ 591 592 In a paravirtualized environment the hardware low level architectural 593 pieces such as interrupt handling, page table handling, and 594 accessing process control registers needs to be done differently. 595 596 This field allows the bootloader to inform the kernel we are in one 597 one of those environments. 598 599 0x00000000 The default x86/PC environment 600 0x00000001 lguest 601 0x00000002 Xen 602 0x00000003 Moorestown MID 603 0x00000004 CE4100 TV Platform 604 605Field name: hardware_subarch_data 606Type: write (subarch-dependent) 607Offset/size: 0x240/8 608Protocol: 2.07+ 609 610 A pointer to data that is specific to hardware subarch 611 This field is currently unused for the default x86/PC environment, 612 do not modify. 613 614Field name: payload_offset 615Type: read 616Offset/size: 0x248/4 617Protocol: 2.08+ 618 619 If non-zero then this field contains the offset from the beginning 620 of the protected-mode code to the payload. 621 622 The payload may be compressed. The format of both the compressed and 623 uncompressed data should be determined using the standard magic 624 numbers. The currently supported compression formats are gzip 625 (magic numbers 1F 8B or 1F 9E), bzip2 (magic number 42 5A), LZMA 626 (magic number 5D 00), and XZ (magic number FD 37). The uncompressed 627 payload is currently always ELF (magic number 7F 45 4C 46). 628 629Field name: payload_length 630Type: read 631Offset/size: 0x24c/4 632Protocol: 2.08+ 633 634 The length of the payload. 635 636Field name: setup_data 637Type: write (special) 638Offset/size: 0x250/8 639Protocol: 2.09+ 640 641 The 64-bit physical pointer to NULL terminated single linked list of 642 struct setup_data. This is used to define a more extensible boot 643 parameters passing mechanism. The definition of struct setup_data is 644 as follow: 645 646 struct setup_data { 647 u64 next; 648 u32 type; 649 u32 len; 650 u8 data[0]; 651 }; 652 653 Where, the next is a 64-bit physical pointer to the next node of 654 linked list, the next field of the last node is 0; the type is used 655 to identify the contents of data; the len is the length of data 656 field; the data holds the real payload. 657 658 This list may be modified at a number of points during the bootup 659 process. Therefore, when modifying this list one should always make 660 sure to consider the case where the linked list already contains 661 entries. 662 663Field name: pref_address 664Type: read (reloc) 665Offset/size: 0x258/8 666Protocol: 2.10+ 667 668 This field, if nonzero, represents a preferred load address for the 669 kernel. A relocating bootloader should attempt to load at this 670 address if possible. 671 672 A non-relocatable kernel will unconditionally move itself and to run 673 at this address. 674 675Field name: init_size 676Type: read 677Offset/size: 0x260/4 678 679 This field indicates the amount of linear contiguous memory starting 680 at the kernel runtime start address that the kernel needs before it 681 is capable of examining its memory map. This is not the same thing 682 as the total amount of memory the kernel needs to boot, but it can 683 be used by a relocating boot loader to help select a safe load 684 address for the kernel. 685 686 The kernel runtime start address is determined by the following algorithm: 687 688 if (relocatable_kernel) 689 runtime_start = align_up(load_address, kernel_alignment) 690 else 691 runtime_start = pref_address 692 693 694**** THE IMAGE CHECKSUM 695 696From boot protocol version 2.08 onwards the CRC-32 is calculated over 697the entire file using the characteristic polynomial 0x04C11DB7 and an 698initial remainder of 0xffffffff. The checksum is appended to the 699file; therefore the CRC of the file up to the limit specified in the 700syssize field of the header is always 0. 701 702 703**** THE KERNEL COMMAND LINE 704 705The kernel command line has become an important way for the boot 706loader to communicate with the kernel. Some of its options are also 707relevant to the boot loader itself, see "special command line options" 708below. 709 710The kernel command line is a null-terminated string. The maximum 711length can be retrieved from the field cmdline_size. Before protocol 712version 2.06, the maximum was 255 characters. A string that is too 713long will be automatically truncated by the kernel. 714 715If the boot protocol version is 2.02 or later, the address of the 716kernel command line is given by the header field cmd_line_ptr (see 717above.) This address can be anywhere between the end of the setup 718heap and 0xA0000. 719 720If the protocol version is *not* 2.02 or higher, the kernel 721command line is entered using the following protocol: 722 723 At offset 0x0020 (word), "cmd_line_magic", enter the magic 724 number 0xA33F. 725 726 At offset 0x0022 (word), "cmd_line_offset", enter the offset 727 of the kernel command line (relative to the start of the 728 real-mode kernel). 729 730 The kernel command line *must* be within the memory region 731 covered by setup_move_size, so you may need to adjust this 732 field. 733 734 735**** MEMORY LAYOUT OF THE REAL-MODE CODE 736 737The real-mode code requires a stack/heap to be set up, as well as 738memory allocated for the kernel command line. This needs to be done 739in the real-mode accessible memory in bottom megabyte. 740 741It should be noted that modern machines often have a sizable Extended 742BIOS Data Area (EBDA). As a result, it is advisable to use as little 743of the low megabyte as possible. 744 745Unfortunately, under the following circumstances the 0x90000 memory 746segment has to be used: 747 748 - When loading a zImage kernel ((loadflags & 0x01) == 0). 749 - When loading a 2.01 or earlier boot protocol kernel. 750 751 -> For the 2.00 and 2.01 boot protocols, the real-mode code 752 can be loaded at another address, but it is internally 753 relocated to 0x90000. For the "old" protocol, the 754 real-mode code must be loaded at 0x90000. 755 756When loading at 0x90000, avoid using memory above 0x9a000. 757 758For boot protocol 2.02 or higher, the command line does not have to be 759located in the same 64K segment as the real-mode setup code; it is 760thus permitted to give the stack/heap the full 64K segment and locate 761the command line above it. 762 763The kernel command line should not be located below the real-mode 764code, nor should it be located in high memory. 765 766 767**** SAMPLE BOOT CONFIGURATION 768 769As a sample configuration, assume the following layout of the real 770mode segment: 771 772 When loading below 0x90000, use the entire segment: 773 774 0x0000-0x7fff Real mode kernel 775 0x8000-0xdfff Stack and heap 776 0xe000-0xffff Kernel command line 777 778 When loading at 0x90000 OR the protocol version is 2.01 or earlier: 779 780 0x0000-0x7fff Real mode kernel 781 0x8000-0x97ff Stack and heap 782 0x9800-0x9fff Kernel command line 783 784Such a boot loader should enter the following fields in the header: 785 786 unsigned long base_ptr; /* base address for real-mode segment */ 787 788 if ( setup_sects == 0 ) { 789 setup_sects = 4; 790 } 791 792 if ( protocol >= 0x0200 ) { 793 type_of_loader = <type code>; 794 if ( loading_initrd ) { 795 ramdisk_image = <initrd_address>; 796 ramdisk_size = <initrd_size>; 797 } 798 799 if ( protocol >= 0x0202 && loadflags & 0x01 ) 800 heap_end = 0xe000; 801 else 802 heap_end = 0x9800; 803 804 if ( protocol >= 0x0201 ) { 805 heap_end_ptr = heap_end - 0x200; 806 loadflags |= 0x80; /* CAN_USE_HEAP */ 807 } 808 809 if ( protocol >= 0x0202 ) { 810 cmd_line_ptr = base_ptr + heap_end; 811 strcpy(cmd_line_ptr, cmdline); 812 } else { 813 cmd_line_magic = 0xA33F; 814 cmd_line_offset = heap_end; 815 setup_move_size = heap_end + strlen(cmdline)+1; 816 strcpy(base_ptr+cmd_line_offset, cmdline); 817 } 818 } else { 819 /* Very old kernel */ 820 821 heap_end = 0x9800; 822 823 cmd_line_magic = 0xA33F; 824 cmd_line_offset = heap_end; 825 826 /* A very old kernel MUST have its real-mode code 827 loaded at 0x90000 */ 828 829 if ( base_ptr != 0x90000 ) { 830 /* Copy the real-mode kernel */ 831 memcpy(0x90000, base_ptr, (setup_sects+1)*512); 832 base_ptr = 0x90000; /* Relocated */ 833 } 834 835 strcpy(0x90000+cmd_line_offset, cmdline); 836 837 /* It is recommended to clear memory up to the 32K mark */ 838 memset(0x90000 + (setup_sects+1)*512, 0, 839 (64-(setup_sects+1))*512); 840 } 841 842 843**** LOADING THE REST OF THE KERNEL 844 845The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512 846in the kernel file (again, if setup_sects == 0 the real value is 4.) 847It should be loaded at address 0x10000 for Image/zImage kernels and 8480x100000 for bzImage kernels. 849 850The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01 851bit (LOAD_HIGH) in the loadflags field is set: 852 853 is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01); 854 load_address = is_bzImage ? 0x100000 : 0x10000; 855 856Note that Image/zImage kernels can be up to 512K in size, and thus use 857the entire 0x10000-0x90000 range of memory. This means it is pretty 858much a requirement for these kernels to load the real-mode part at 8590x90000. bzImage kernels allow much more flexibility. 860 861 862**** SPECIAL COMMAND LINE OPTIONS 863 864If the command line provided by the boot loader is entered by the 865user, the user may expect the following command line options to work. 866They should normally not be deleted from the kernel command line even 867though not all of them are actually meaningful to the kernel. Boot 868loader authors who need additional command line options for the boot 869loader itself should get them registered in 870Documentation/kernel-parameters.txt to make sure they will not 871conflict with actual kernel options now or in the future. 872 873 vga=<mode> 874 <mode> here is either an integer (in C notation, either 875 decimal, octal, or hexadecimal) or one of the strings 876 "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask" 877 (meaning 0xFFFD). This value should be entered into the 878 vid_mode field, as it is used by the kernel before the command 879 line is parsed. 880 881 mem=<size> 882 <size> is an integer in C notation optionally followed by 883 (case insensitive) K, M, G, T, P or E (meaning << 10, << 20, 884 << 30, << 40, << 50 or << 60). This specifies the end of 885 memory to the kernel. This affects the possible placement of 886 an initrd, since an initrd should be placed near end of 887 memory. Note that this is an option to *both* the kernel and 888 the bootloader! 889 890 initrd=<file> 891 An initrd should be loaded. The meaning of <file> is 892 obviously bootloader-dependent, and some boot loaders 893 (e.g. LILO) do not have such a command. 894 895In addition, some boot loaders add the following options to the 896user-specified command line: 897 898 BOOT_IMAGE=<file> 899 The boot image which was loaded. Again, the meaning of <file> 900 is obviously bootloader-dependent. 901 902 auto 903 The kernel was booted without explicit user intervention. 904 905If these options are added by the boot loader, it is highly 906recommended that they are located *first*, before the user-specified 907or configuration-specified command line. Otherwise, "init=/bin/sh" 908gets confused by the "auto" option. 909 910 911**** RUNNING THE KERNEL 912 913The kernel is started by jumping to the kernel entry point, which is 914located at *segment* offset 0x20 from the start of the real mode 915kernel. This means that if you loaded your real-mode kernel code at 9160x90000, the kernel entry point is 9020:0000. 917 918At entry, ds = es = ss should point to the start of the real-mode 919kernel code (0x9000 if the code is loaded at 0x90000), sp should be 920set up properly, normally pointing to the top of the heap, and 921interrupts should be disabled. Furthermore, to guard against bugs in 922the kernel, it is recommended that the boot loader sets fs = gs = ds = 923es = ss. 924 925In our example from above, we would do: 926 927 /* Note: in the case of the "old" kernel protocol, base_ptr must 928 be == 0x90000 at this point; see the previous sample code */ 929 930 seg = base_ptr >> 4; 931 932 cli(); /* Enter with interrupts disabled! */ 933 934 /* Set up the real-mode kernel stack */ 935 _SS = seg; 936 _SP = heap_end; 937 938 _DS = _ES = _FS = _GS = seg; 939 jmp_far(seg+0x20, 0); /* Run the kernel */ 940 941If your boot sector accesses a floppy drive, it is recommended to 942switch off the floppy motor before running the kernel, since the 943kernel boot leaves interrupts off and thus the motor will not be 944switched off, especially if the loaded kernel has the floppy driver as 945a demand-loaded module! 946 947 948**** ADVANCED BOOT LOADER HOOKS 949 950If the boot loader runs in a particularly hostile environment (such as 951LOADLIN, which runs under DOS) it may be impossible to follow the 952standard memory location requirements. Such a boot loader may use the 953following hooks that, if set, are invoked by the kernel at the 954appropriate time. The use of these hooks should probably be 955considered an absolutely last resort! 956 957IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and 958%edi across invocation. 959 960 realmode_swtch: 961 A 16-bit real mode far subroutine invoked immediately before 962 entering protected mode. The default routine disables NMI, so 963 your routine should probably do so, too. 964 965 code32_start: 966 A 32-bit flat-mode routine *jumped* to immediately after the 967 transition to protected mode, but before the kernel is 968 uncompressed. No segments, except CS, are guaranteed to be 969 set up (current kernels do, but older ones do not); you should 970 set them up to BOOT_DS (0x18) yourself. 971 972 After completing your hook, you should jump to the address 973 that was in this field before your boot loader overwrote it 974 (relocated, if appropriate.) 975 976 977**** 32-bit BOOT PROTOCOL 978 979For machine with some new BIOS other than legacy BIOS, such as EFI, 980LinuxBIOS, etc, and kexec, the 16-bit real mode setup code in kernel 981based on legacy BIOS can not be used, so a 32-bit boot protocol needs 982to be defined. 983 984In 32-bit boot protocol, the first step in loading a Linux kernel 985should be to setup the boot parameters (struct boot_params, 986traditionally known as "zero page"). The memory for struct boot_params 987should be allocated and initialized to all zero. Then the setup header 988from offset 0x01f1 of kernel image on should be loaded into struct 989boot_params and examined. The end of setup header can be calculated as 990follow: 991 992 0x0202 + byte value at offset 0x0201 993 994In addition to read/modify/write the setup header of the struct 995boot_params as that of 16-bit boot protocol, the boot loader should 996also fill the additional fields of the struct boot_params as that 997described in zero-page.txt. 998 999After setupping the struct boot_params, the boot loader can load the 100032/64-bit kernel in the same way as that of 16-bit boot protocol. 1001 1002In 32-bit boot protocol, the kernel is started by jumping to the 100332-bit kernel entry point, which is the start address of loaded 100432/64-bit kernel. 1005 1006At entry, the CPU must be in 32-bit protected mode with paging 1007disabled; a GDT must be loaded with the descriptors for selectors 1008__BOOT_CS(0x10) and __BOOT_DS(0x18); both descriptors must be 4G flat 1009segment; __BOOS_CS must have execute/read permission, and __BOOT_DS 1010must have read/write permission; CS must be __BOOT_CS and DS, ES, SS 1011must be __BOOT_DS; interrupt must be disabled; %esi must hold the base 1012address of the struct boot_params; %ebp, %edi and %ebx must be zero. 1013