LOCKD(8) BSD System Manager's Manual LOCKD(8)NAME
lockd - NFS lock daemon
SYNOPSIS
lockd [-d debug-level] [-g minutes]
DESCRIPTION
Lockd is a daemon which provides record-level locking services in an NFS
environment. It should be run at boot time by /etc/rc (see rc(8)), just
before mountd.
The options are as follows:
-d Specify the debug level:
a log arguments
c log cached host connection behavior
r log rpc calls
b log background activity
A turn on all debug flags
-g Specify the number of minutes of grace period on startup. During
this time period no new locks may be obtained, only reclaimed
locks will be granted. The default is 2 minutes, and anything
greater than 10 is treated as an error.
Versions 1, 3 and 4 of the Network Lock Manager (NLM) protocol are sup-
ported.
Two processes are created to implement the NLM lock protocol: One manages
the NFS client side and performs local file locking operations based on
requests forwarded from lock managers running on NFS client machines.
The other manages the NFS server side, and forwards local lock requests
on files in NFS mounted filesystems to their respective NFS servers.
Both processes are single threaded, but extensive use of non-blocking RPC
calls are used to maximize the overlapping of operations. They maintain
a parent child relationship, with the parent recording its pid in
/var/run/lockd.pid. The parent will forward SIGINT and SIGHUP to the
child, which will cause both of them to drop all local locks held for NFS
clients, and to generally cleanup and then exit. Whenever lockd is not
running, local locking operations on NFS mounted filesystems will return
EOPNOTSUPP.
When lockd is started, NFS clients and servers which were involved in
prior locking operations will be notified (via statd(8)), and during the
grace period locks, will be reestablished (if possible).
Error conditions are logged to syslog, using log level LOG_ERR and facil-
ity LOG_DAEMON.
The statd daemon must be invoked before the lockd daemon is run.
FILES
/usr/include/rpcsvc/nlm_prot.x
RPC protocol specification for the network lock manager (NLM) pro-
tocol.
/var/run/lockd.pid
The process id (PID) of the parent lockd process.
/var/run/lock
The fifo used by the client side lockd process to read local lock-
ing requests from the kernel.
SEE ALSOfcntl(2), flock(2), statd(8)BUGS
Locks are not reestablished from the NFS client side when the NFS server
has rebooted.
If a remote host will not be coming back on-line, there is no way to in-
form lockd that it should quit re-trying to the host.
STANDARDS
The implementation is based on the X/Open CAE Specification, Document
number C525, "Protocols for X/Open Interworking: XNFS, Version 3".
BSDI BSD/OS March 17, 1998 2