dotlockfile man page
dotlockfile — Utility to manage lockfiles
dotlockfile -l [-r retries] [-p] <-m | lockfile>
dotlockfile -l [-r retries] [-p] <-m | lockfile> cmd args...
dotlockfile -u | -t
dotlockfile is a command line utility to reliably create, test and remove lockfiles. It creates lockfiles reliably on local and NFS filesystems, because the crucial steps of testing for a preexisting lockfile and creating it are performed atomically by a single call to link(2). Manpage lockfile_create(3) describes the used algorithm.
dotlockfile is installed with attribute SETGID mail and thus can also be used to lock and unlock mailboxes even if the mailspool directory is only writable by group mail.
The name dotlockfile comes from the way mailboxes are locked for updates on a lot of UNIX systems. A lockfile is created with the same filename as the mailbox but with the string ".lock" appended.
The names dotlock and lockfile were already taken - hence the name dotlockfile :).
Create a lockfile if no preexisting valid lockfile is found, else wait and retry according to option -r. This option is the default, so it can be left off.
A lockfile is treated as valid,
· if it holds the process-id of a running process,
· or if it does not hold any process-id and has been touched less than 5 minutes ago (timestamp is younger than 5 minutes).
- -r retries
The number of times dotlockfile retries to acquire the lock if it failed the first time before giving up. The initial sleep after failing to acquire the lock is 5 seconds. After each retry the sleep intervall is increased incrementally by 5 seconds up to a maximum sleep of 60 seconds between tries. The default number of retries is 5. To try only once, use "-r 0". To try indefinitely, use "-r -1".
Remove a lockfile.
Touch an existing lockfile (update the timestamp). Useful for lockfiles on NFS filesystems. For lockfiles on local filesystems the -p option is preferable.
Write the process-id of the calling process (or dotlockfile itself if a command is executed) into the lockfile. Also when testing for an existing lockfile, check the contents for the process-id of a running process to verify if the lockfile is still valid. Obviously useful only for lockfiles on local filesystems.
Lock or unlock the current users mailbox. The path to the mailbox is the default system mailspool directory (usually /var/mail) with the username as gotten from getpwuid() appended. If the environment variable $MAIL is set, that is used instead. Then the string ".lock" is appended to get the name of the actual lockfile.
The lockfile to be created or removed. Must not be specified if the -m option is given.
Create lockfile, run the command , wait for it to exit, and remove lockfile.
Zero on success, and non-zero on failure. When locking (the default, or with the -l option) dotlockfile returns the same values as the library function lockfile_create(3). Unlocking a non-existant lockfile is not an error.
If a command is executed, the return value does not correspond with that of the command that was run. If the locking and unlocking was succesful, the exit status is always zero.
The lockfile is created exactly as named on the command line. The extension ".lock" is not automatically appended.
This utility is a lot like the lockfile(1) utility included with procmail, and the mutt_dotlock(1) utility included with mutt. However the command-line arguments differ, and so does the return status. It is believed, that dotlockfile is the most flexible implementation, since it automatically detects when it needs to use priviliges to lock a mailbox, and does it safely.
The above mentioned lockfile_create(3) manpage is present in the liblockfile-dev package.
Miquel van Smoorenburg, email@example.com