harvey/sys/man/4/cfs

.TH CFS 4
.SH NAME
cfs \- cache file system
.SH SYNOPSIS
.B cfs
.B -s
.RB [ -rdS ]
.RB [ -f
.IR partition ]
.PP
.B cfs 
.B -a
.I netaddr
.RB [ -rdS ]
.RB [ -f
.IR partition ]
.RI [ mtpt ]
.SH DESCRIPTION
.I Cfs
is a user-level file server that caches information about remote
files onto a local disk.
It is normally started by the kernel at boot time, though users may start
it manually.
.I Cfs
is interposed between the kernel and a network connection to a
remote file server to improve the
efficiency of access across slow network connections such as modem
lines.
On each open of a file 
.I cfs
checks the consistency of cached information and discards any old
information for that file.
.PP
.I Cfs
mounts onto
.I mtpt
(default
.BR / )
after connecting to the file server.
.PP
The options are:
.TP
.B s
the connection to the remote file server is on file
descriptors 0 and 1.
.TP
.BI "a " netaddr
dial the destination
.I netaddr
to connect to a remote file server.
.TP
.B r
reformat the cache disk partition.
.TP
.B d
turn on debugging.
.TP
.B S
turn on statistics gathering.  A file called
.B cfsctl
at the root of the caching file system can be read to get
statistics concerning number of calls/bytes on client and server
sides and latencies.
.TP
.BI "f " partition
use file
.I partition
as the cache disk partition.
.PP
All 9P messages except
.BR read ,
.BR clone ,
and
.B walk
(see
.IR intro (5))
are passed through
.I cfs
unchanged to the remote server.
A
.B clone
followed immediately by a
.B walk
is converted into a
.BR clwalk .
If possible, a
.B read
is satisfied by cached data.
Otherwise, the file server is queried for any missing data.
.SH FILES
.TP
.B /dev/sdC0/cache
Default file used for storing cached data.
.SH SOURCE
.B /sys/src/cmd/cfs