THE LINUX/I386 BOOT PROTOCOL

system888net

                          THE LINUX/I386 BOOT PROTOCOL
                     ----------------------------
                    H. Peter Anvin
                        Last update 2007-05-16
On the i386 platform, the Linux kernel uses a rather complicated boot
convention.  This has evolved partially due to historical aspects, as
well as the desire in the early days to have the kernel itself be a
bootable image, the complicated PC memory model and due to changed
expectations in the PC industry caused by the effective demise of
real-mode DOS as a mainstream operating system.
Currently, the following versions of the Linux/i386 boot protocol exist.
Old kernels:        zImage/Image support only.  Some very early kernels
                may not even support a command line.
Protocol 2.00:        (Kernel 1.3.73) Added bzImage and initrd support, as
                well as a formalized way to communicate between the
                boot loader and the kernel.  setup.S made relocatable,
                although the traditional setup area still assumed
                writable.
Protocol 2.01:        (Kernel 1.3.76) Added a heap overrun warning.
Protocol 2.02:        (Kernel 2.4.0-test3-pre3) New command line protocol.
                Lower the conventional memory ceiling.        No overwrite
                of the traditional setup area, thus making booting
                safe for systems which use the EBDA from SMM or 32-bit
                BIOS entry points.  zImage deprecated but still
                supported.
Protocol 2.03:        (Kernel 2.4.18-pre1) Explicitly makes the highest possible
                initrd address available to the bootloader.
Protocol 2.04:        (Kernel 2.6.14) Extend the syssize field to four bytes.
Protocol 2.05:        (Kernel 2.6.20) Make protected mode kernel relocatable.
                Introduce relocatable_kernel and kernel_alignment fields.
Protocol 2.06:        (Kernel 2.6.22) Added a field that contains the size of
                the boot command line
**** MEMORY LAYOUT
The traditional memory map for the kernel loader, used for Image or
zImage kernels, typically looks like:
        |                        |
0A0000  +------------------------+
        |  Reserved for BIOS     |        Do not use.  Reserved for BIOS EBDA.
09A000  +------------------------+
        |  Command line          |
        |  Stack/heap            |        For use by the kernel real-mode code.
098000  +------------------------+      
        |  Kernel setup          |        The kernel real-mode code.
090200  +------------------------+
        |  Kernel boot sector    |        The kernel legacy boot sector.
090000  +------------------------+
        |  Protected-mode kernel |        The bulk of the kernel image.
010000  +------------------------+
        |  Boot loader           |        = 2.02, a
memory layout like the following is suggested:
        ~                        ~
        |  Protected-mode kernel |
100000  +------------------------+
        |  I/O memory hole       |
0A0000  +------------------------+
        |  Reserved for BIOS     |        Leave as much as possible unused
        ~                        ~
        |  Command line          |        (Can also be below the X+10000 mark)
X+10000 +------------------------+
        |  Stack/heap            |        For use by the kernel real-mode code.
X+08000 +------------------------+      
        |  Kernel setup          |        The kernel real-mode code.
        |  Kernel boot sector    |        The kernel legacy boot sector.
X       +------------------------+
        |  Boot loader           |         if you need a bootloader ID
  value assigned.
Field name:        loadflags
Type:                modify (obligatory)
Offset/size:        0x211/1
Protocol:        2.00+
  This field is a bitmask.
  Bit 0 (read):        LOADED_HIGH
        - If 0, the protected-mode code is loaded at 0x10000.
        - If 1, the protected-mode code is loaded at 0x100000.
  Bit 7 (write): CAN_USE_HEAP
        Set this bit to 1 to indicate that the value entered in the
        heap_end_ptr is valid.  If this field is clear, some setup code
        functionality will be disabled.
Field name:        setup_move_size
Type:                modify (obligatory)
Offset/size:        0x212/2
Protocol:        2.00-2.01
  When using protocol 2.00 or 2.01, if the real mode kernel is not
  loaded at 0x90000, it gets moved there later in the loading
  sequence.  Fill in this field if you want additional data (such as
  the kernel command line) moved in addition to the real-mode kernel
  itself.
  The unit is bytes starting with the beginning of the boot sector.
  
  This field is can be ignored when the protocol is 2.02 or higher, or
  if the real-mode code is loaded at 0x90000.
Field name:        code32_start
Type:                modify (optional, reloc)
Offset/size:        0x214/4
Protocol:        2.00+
  The address to jump to in protected mode.  This defaults to the load
  address of the kernel, and can be used by the boot loader to
  determine the proper load address.
  This field can be modified for two purposes:
  1. as a boot loader hook (see separate chapter.)
  2. if a bootloader which does not install a hook loads a
     relocatable kernel at a nonstandard address it will have to modify
     this field to point to the load address.
Field name:        ramdisk_image
Type:                write (obligatory)
Offset/size:        0x218/4
Protocol:        2.00+
  The 32-bit linear address of the initial ramdisk or ramfs.  Leave at
  zero if there is no initial ramdisk/ramfs.
Field name:        ramdisk_size
Type:                write (obligatory)
Offset/size:        0x21c/4
Protocol:        2.00+
  Size of the initial ramdisk or ramfs.  Leave at zero if there is no
  initial ramdisk/ramfs.
Field name:        bootsect_kludge
Type:                kernel internal
Offset/size:        0x220/4
Protocol:        2.00+
  This field is obsolete.
Field name:        heap_end_ptr
Type:                write (obligatory)
Offset/size:        0x224/2
Protocol:        2.01+
  Set this field to the offset (from the beginning of the real-mode
  code) of the end of the setup stack/heap, minus 0x0200.
Field name:        cmd_line_ptr
Type:                write (obligatory)
Offset/size:        0x228/4
Protocol:        2.02+
  Set this field to the linear address of the kernel command line.
  The kernel command line can be located anywhere between the end of
  the setup heap and 0xA0000; it does not have to be located in the
  same 64K segment as the real-mode code itself.
  Fill in this field even if your boot loader does not support a
  command line, in which case you can point this to an empty string
  (or better yet, to the string "auto".)  If this field is left at
  zero, the kernel will assume that your boot loader does not support
  the 2.02+ protocol.
Field name:        initrd_addr_max
Type:                read
Offset/size:        0x22c/4
Protocol:        2.03+
  The maximum address that may be occupied by the initial
  ramdisk/ramfs contents.  For boot protocols 2.02 or earlier, this
  field is not present, and the maximum address is 0x37FFFFFF.  (This
  address is defined as the address of the highest safe byte, so if
  your ramdisk is exactly 131072 bytes long and this field is
  0x37FFFFFF, you can start your ramdisk at 0x37FE0000.)
Field name:        kernel_alignment
Type:                read (reloc)
Offset/size:        0x230/4
Protocol:        2.05+
  Alignment unit required by the kernel (if relocatable_kernel is true.)
Field name:        relocatable_kernel
Type:                read (reloc)
Offset/size:        0x234/1
Protocol:        2.05+
  If this field is nonzero, the protected-mode part of the kernel can
  be loaded at any address that satisfies the kernel_alignment field.
  After loading, the boot loader must set the code32_start field to
  point to the loaded code, or to a boot loader hook.
Field name:        cmdline_size
Type:                read
Offset/size:        0x238/4
Protocol:        2.06+
  The maximum size of the command line without the terminating
  zero. This means that the command line can contain at most
  cmdline_size characters. With protocol version 2.05 and earlier, the
  maximum size was 255.
**** THE KERNEL COMMAND LINE
The kernel command line has become an important way for the boot
loader to communicate with the kernel.  Some of its options are also
relevant to the boot loader itself, see "special command line options"
below.
The kernel command line is a null-terminated string. The maximum
length can be retrieved from the field cmdline_size.  Before protocol
version 2.06, the maximum was 255 characters.  A string that is too
long will be automatically truncated by the kernel.
If the boot protocol version is 2.02 or later, the address of the
kernel command line is given by the header field cmd_line_ptr (see
above.)  This address can be anywhere between the end of the setup
heap and 0xA0000.
If the protocol version is *not* 2.02 or higher, the kernel
command line is entered using the following protocol:
        At offset 0x0020 (word), "cmd_line_magic", enter the magic
        number 0xA33F.
        At offset 0x0022 (word), "cmd_line_offset", enter the offset
        of the kernel command line (relative to the start of the
        real-mode kernel).
      
        The kernel command line *must* be within the memory region
        covered by setup_move_size, so you may need to adjust this
        field.
**** MEMORY LAYOUT OF THE REAL-MODE CODE
The real-mode code requires a stack/heap to be set up, as well as
memory allocated for the kernel command line.  This needs to be done
in the real-mode accessible memory in bottom megabyte.
It should be noted that modern machines often have a sizable Extended
BIOS Data Area (EBDA).  As a result, it is advisable to use as little
of the low megabyte as possible.
Unfortunately, under the following circumstances the 0x90000 memory
segment has to be used:
        - When loading a zImage kernel ((loadflags & 0x01) == 0).
        - When loading a 2.01 or earlier boot protocol kernel.
          -> For the 2.00 and 2.01 boot protocols, the real-mode code
             can be loaded at another address, but it is internally
             relocated to 0x90000.  For the "old" protocol, the
             real-mode code must be loaded at 0x90000.
When loading at 0x90000, avoid using memory above 0x9a000.
For boot protocol 2.02 or higher, the command line does not have to be
located in the same 64K segment as the real-mode setup code; it is
thus permitted to give the stack/heap the full 64K segment and locate
the command line above it.
The kernel command line should not be located below the real-mode
code, nor should it be located in high memory.
**** SAMPLE BOOT CONFIGURATION
As a sample configuration, assume the following layout of the real
mode segment:
    When loading below 0x90000, use the entire segment:
        0x0000-0x7fff        Real mode kernel
        0x8000-0xdfff        Stack and heap
        0xe000-0xffff        Kernel command line
    When loading at 0x90000 OR the protocol version is 2.01 or earlier:
        0x0000-0x7fff        Real mode kernel
        0x8000-0x97ff        Stack and heap
        0x9800-0x9fff        Kernel command line
Such a boot loader should enter the following fields in the header:
        unsigned long base_ptr;        /* base address for real-mode segment */
        if ( setup_sects == 0 ) {
                setup_sects = 4;
        }
        if ( protocol >= 0x0200 ) {
                type_of_loader = ;
                if ( loading_initrd ) {
                        ramdisk_image = ;
                        ramdisk_size = ;
                }
                if ( protocol >= 0x0202 && loadflags & 0x01 )
                        heap_end = 0xe000;
                else
                        heap_end = 0x9800;
                if ( protocol >= 0x0201 ) {
                        heap_end_ptr = heap_end - 0x200;
                        loadflags |= 0x80; /* CAN_USE_HEAP */
                }
                if ( protocol >= 0x0202 ) {
                        cmd_line_ptr = base_ptr + heap_end;
                        strcpy(cmd_line_ptr, cmdline);
                } else {
                        cmd_line_magic        = 0xA33F;
                        cmd_line_offset = heap_end;
                        setup_move_size = heap_end + strlen(cmdline)+1;
                        strcpy(base_ptr+cmd_line_offset, cmdline);
                }
        } else {
                /* Very old kernel */
                heap_end = 0x9800;
                cmd_line_magic        = 0xA33F;
                cmd_line_offset = heap_end;
                /* A very old kernel MUST have its real-mode code
                   loaded at 0x90000 */
                if ( base_ptr != 0x90000 ) {
                        /* Copy the real-mode kernel */
                        memcpy(0x90000, base_ptr, (setup_sects+1)*512);
                        base_ptr = 0x90000;                 /* Relocated */
                }
                strcpy(0x90000+cmd_line_offset, cmdline);
                /* It is recommended to clear memory up to the 32K mark */
                memset(0x90000 + (setup_sects+1)*512, 0,
                       (64-(setup_sects+1))*512);
        }
**** LOADING THE REST OF THE KERNEL
The 32-bit (non-real-mode) kernel starts at offset (setup_sects+1)*512
in the kernel file (again, if setup_sects == 0 the real value is 4.)
It should be loaded at address 0x10000 for Image/zImage kernels and
0x100000 for bzImage kernels.
The kernel is a bzImage kernel if the protocol >= 2.00 and the 0x01
bit (LOAD_HIGH) in the loadflags field is set:
        is_bzImage = (protocol >= 0x0200) && (loadflags & 0x01);
        load_address = is_bzImage ? 0x100000 : 0x10000;
Note that Image/zImage kernels can be up to 512K in size, and thus use
the entire 0x10000-0x90000 range of memory.  This means it is pretty
much a requirement for these kernels to load the real-mode part at
0x90000.  bzImage kernels allow much more flexibility.
**** SPECIAL COMMAND LINE OPTIONS
If the command line provided by the boot loader is entered by the
user, the user may expect the following command line options to work.
They should normally not be deleted from the kernel command line even
though not all of them are actually meaningful to the kernel.  Boot
loader authors who need additional command line options for the boot
loader itself should get them registered in
Documentation/kernel-parameters.txt to make sure they will not
conflict with actual kernel options now or in the future.
  vga=
         here is either an integer (in C notation, either
        decimal, octal, or hexadecimal) or one of the strings
        "normal" (meaning 0xFFFF), "ext" (meaning 0xFFFE) or "ask"
        (meaning 0xFFFD).  This value should be entered into the
        vid_mode field, as it is used by the kernel before the command
        line is parsed.
  mem=
         is an integer in C notation optionally followed by
        (case insensitive) K, M, G, T, P or E (meaning
        An initrd should be loaded.  The meaning of  is
        obviously bootloader-dependent, and some boot loaders
        (e.g. LILO) do not have such a command.
In addition, some boot loaders add the following options to the
user-specified command line:
  BOOT_IMAGE=
        The boot image which was loaded.  Again, the meaning of
        is obviously bootloader-dependent.
  auto
        The kernel was booted without explicit user intervention.
If these options are added by the boot loader, it is highly
recommended that they are located *first*, before the user-specified
or configuration-specified command line.  Otherwise, "init=/bin/sh"
gets confused by the "auto" option.
**** RUNNING THE KERNEL
The kernel is started by jumping to the kernel entry point, which is
located at *segment* offset 0x20 from the start of the real mode
kernel.  This means that if you loaded your real-mode kernel code at
0x90000, the kernel entry point is 9020:0000.
At entry, ds = es = ss should point to the start of the real-mode
kernel code (0x9000 if the code is loaded at 0x90000), sp should be
set up properly, normally pointing to the top of the heap, and
interrupts should be disabled.  Furthermore, to guard against bugs in
the kernel, it is recommended that the boot loader sets fs = gs = ds =
es = ss.
In our example from above, we would do:
        /* Note: in the case of the "old" kernel protocol, base_ptr must
           be == 0x90000 at this point; see the previous sample code */
        seg = base_ptr >> 4;
        cli();        /* Enter with interrupts disabled! */
        /* Set up the real-mode kernel stack */
        _SS = seg;
        _SP = heap_end;
        _DS = _ES = _FS = _GS = seg;
        jmp_far(seg+0x20, 0);        /* Run the kernel */
If your boot sector accesses a floppy drive, it is recommended to
switch off the floppy motor before running the kernel, since the
kernel boot leaves interrupts off and thus the motor will not be
switched off, especially if the loaded kernel has the floppy driver as
a demand-loaded module!
**** ADVANCED BOOT TIME HOOKS
If the boot loader runs in a particularly hostile environment (such as
LOADLIN, which runs under DOS) it may be impossible to follow the
standard memory location requirements.  Such a boot loader may use the
following hooks that, if set, are invoked by the kernel at the
appropriate time.  The use of these hooks should probably be
considered an absolutely last resort!
IMPORTANT: All the hooks are required to preserve %esp, %ebp, %esi and
%edi across invocation.
  realmode_swtch:
        A 16-bit real mode far subroutine invoked immediately before
        entering protected mode.  The default routine disables NMI, so
        your routine should probably do so, too.
  code32_start:
        A 32-bit flat-mode routine *jumped* to immediately after the
        transition to protected mode, but before the kernel is
        uncompressed.  No segments, except CS, are guaranteed to be
        set up (current kernels do, but older ones do not); you should
        set them up to BOOT_DS (0x18) yourself.
        After completing your hook, you should jump to the address
        that was in this field before your boot loader overwrote it.


本文来自ChinaUnix博客，如果查看原文请点：http://blog.chinaunix.net/u2/86872/showart_1848997.html