EXTLINUX is a Syslinux variant which boots from a Linux filesystem.
[4.00+] FAT12/16/32, ext2/3/4, Btrfs,
[4.06+] FAT12/16/32, NTFS, ext2/3/4, Btrfs,
[5.01+] FAT12/16/32, NTFS, ext2/3/4, Btrfs, XFS,
[6.03+] FAT12/16/32, NTFS, ext2/3/4, Btrfs, XFS, UFS/FFS,
It works the same way as SYSLINUX, with a few slight modifications.
1. The installer runs on a *mounted* filesystem. Run the extlinux installer on the directory in which you want EXTLINUX installed:
extlinux --install /boot
Specify --install (-i) to install for the first time, or --update (-U) to upgrade a previous installation.
NOTE: This does not have to be the root directory of a filesystem. If "/boot" is a mount point of a supported filesystem, then you can do:
mkdir -p /boot/extlinux extlinux --install /boot/extlinux
... to create a subdirectory and install EXTLINUX in it.
2. The configuration file is called "extlinux.conf", and is expected to be found in the same directory as EXTLINUX is installed in.
Since v.4.02, syslinux.cfg and [/boot/]syslinux/ are generic terms and also valid for EXTLINUX. extlinux.conf and [/boot/]extlinux/ take precedence for EXTLINUX if they are present.
3. Pathnames can be absolute or relative; if absolute (with a leading slash), they are relative to the root of the filesystem on which EXTLINUX is installed (/boot in the example above); if relative, they are relative to the Current Working Directory - initially the EXTLINUX directory (where extlinux.conf - or the alternative syslinux.cfg - is located).
EXTLINUX supports subdirectories, but the total path length is limited to 255 characters.
4. EXTLINUX now supports symbolic links. However, extremely long symbolic links might hit the pathname limit. Also, please note that absolute symbolic links are interpreted from the root *of the filesystem*, which might be different from how the running system would interpret it (e.g. in the case of a separate /boot partition.) Therefore, use relative symbolic links if at all possible.
5. EXTLINUX now has "boot-once" support. The boot-once information is stored in an on-disk datastructure, part of ldlinux.sys, called the "Auxillary Data Vector". The Auxilliary Data Vector is also available to c32 modules that want to store small amounts of information.
To set the boot-once information:
extlinux --once 'command' /boot/extlinux
where "command" is any command you could enter at the Syslinux command line. It will be executed on the next boot and then erased.
To clear the boot-once information:
extlinux --clear-once /boot/extlinux
If EXTLINUX is used on a RAID-1, this is recommended, since under certain circumstances a RAID-1 rebuild can "resurrect" the boot-once information otherwise.
To clear the entire Auxillary Data Vector:
extlinux --reset-adv /boot/extlinux
This will erase all data stored in the ADV, including boot-once.
The --once, --clear-once, and --reset-adv commands can be combined with --install or --update, if desired. The ADV is preserved across updates, unless --reset-adv is specified.
Note that EXTLINUX installs in the filesystem partition like a well-behaved bootloader :). Thus, it needs a master boot record in the partition table; the mbr.bin shipped with SYSLINUX should work well. To install it:
cat mbr.bin > /dev/XXX
... where /dev/XXX is the appropriate master device, e.g. /dev/hda, and make sure the correct partition is set as "active".
If you have multiple disks in a software RAID configuration, the preferred way to boot is:
- Create a separate RAID-1 partition for /boot.
- Note that the Linux RAID-1 driver can span as many disks as you wish.
- Install the MBR on *each disk*, and mark the RAID-1 partition as "active".
- Run "
extlinux --raid --install /boot" to install EXTLINUX.
- This will install it on all the drives in the RAID-1 set, which means you can boot any combination of drives in any order.
It is not required to re-run the extlinux installer after installing new kernels. If you are using ext3 journalling, however, it might be desirable to do so, since running the extlinux installer will flush the log. Otherwise a dirty shutdown could cause some of the new kernel image to still be in the log. This is a general problem for boot loaders on journalling filesystems; it is not specific to EXTLINUX. The "sync" command does not flush the log on the ext3 filesystem.
The Syslinux Project boot loaders support chain loading other operating systems via a separate module, chain.c32. To use it, specify a LABEL in the configuration file with KERNEL chain.c32 and APPEND [hd|fd]<number> [<partition>].
# Windows CE/ME/NT, a very dense operating system. # Second partition (2) on the first hard disk (hd0); # Linux would *typically* call this /dev/hda2 or /dev/sda2. LABEL cement KERNEL chain.c32 APPEND hd0 2
See also README.menu.