[manual index][section index]


kfs - disk file system


mount {disk/kfs [-r [ -b bsize ] ] [-c] [-A] [-P] [-R] [-W] [ -n name ] file} dir


Kfs implements a hierarchical Inferno file system within an existing file, which is typically a disk or flash memory partition. It gives access to it through the Styx protocol on its standard input, and the contents can be mounted directly on a given dir as shown above. The file system format is the same as that used by the kfs command of Plan 9, except that the modification user ID is implemented.

The -r option causes the file system to be reset to an initially empty state (`reamed'). Permission checking is turned off, to allow any desired permissions and file ownership to be set. (In other words, the -W and -P options are also set by default.) The file system block size is set to the bsize given by the -b option (default: 1024 bytes), which must be a multiple of 512 and not greater than 16k bytes. The block size is stored in the file and need not be given again. The storage representation is always little-endian.

Otherwise, the file system is checked if required, unless the -c option is given.

The contents of the file system can be provided by using commands such as mkdir(1), cp(1) and rm(1) in dir, or built from a description using mkfs(8).

The mapping between user names and internal IDs within the file system is established by the file adm/users (within the file system itself) as described by users(6), which kfs reads when it starts. If no such file exists, as for instance when the file system is initially empty, kfs uses a minimal set corresponding to the following users(6) file:


Any users(6) file used with kfs should include entries for at least adm, none, and noworld as above (although group membership can vary).

Kfs can optionally serve a control file, for use by kfscmd(8). If the -n option is given, kfs creates a channel /chan/kfs.name.cmd and accepts commands on it from the user that started kfs.

Other options are:

do not update access times; useful when running a file system in flash over ftl(3), to avoid excessive wear
suppress permission checking
file system is read only
allow wstat (see sys-stat(2) or stat(5)) to make arbitrary changes to user and group fields


Create an empty file system in the file kfs.file. Because the file system will be no larger than the existing file's size, and the file is assumed not to be a device file, use zeros(1) to prepare a file with 2048 blocks of 1024 bytes each:

zeros 1024 2048 >kfs.file
mount -c {disk/kfs -r kfs.file} /n/local

The -c option to mount allows files to be created in /n/local.




dd(1), zeros(1), flash(3), ftl(3), logfs(3), sd(3), users(6), kfscmd(8), mkfs(8)


Because the file system format is the same as Plan 9's kfs, this one also does not support file names longer than 27 bytes. It likewise cannot cope with files bigger than 2⁳ⁱ-1 bytes.

KFS(4 ) Rev:  Thu Feb 15 14:43:41 GMT 2007