lockmail is a helper utility for working with mailbox files. Mailbox files must be locked to prevent other applications from modifying the mailbox at the same time. Different system use different locking conventions. lockmail uses two of the most common locking mechanisms in use, which should work reliably on most systems.
lockfile is the pathname to an existing mailbox file. By default, lockmail tries to lock the mailbox every five seconds (if the mailbox is already locked), and will give up after three minutes. After the mailbox is succesfully locked, lockmail runs program as a child process, with any optional arguments. When program terminates, lockmail removes the mailbox lock, and terminates itself.
If a regular lock fails, try a read-only lock. Use this option to lock mailbox files in a read-only directory.
If the lock attempt fails, try again for up to timeout seconds. The actual timeout is rounded up to the next five second interval (a lock attempt is tried every five seconds).
This section briefly describes the locking mechanism used by lockmail. lockmail uses three different locking conventions in order to maximize compatibility with other mail software: C-Client folder locks, dot-locks, and file locks.
Mail software based on the C-Client library creates lock files named /tmp/.dddddd.iiiiii. Here, dddddd and iiiiii are the device number and the inode number of the mailbox file (the st_dev and st_ino fields in the inode), in hexadecimal. If the process ID saved in the C-Client folder lock file is not valid, lockmail concludes that it's a stale lock file, and will remove it.
| NOTE: | A race condition exists where a C-Client process is killed after it creates a lock file, but before saving its process ID in the lock file. The race window is very small, but it exists. The C-Client library does not appear to ever clear out the lock file. lockmail attempts to resolve this race condition by deleting zero-length lock files that are at least five minutes old. |
lockmail also creates, and honors dot-lock files. Dot-lock files are first created as temporary files, then linked to lockfile.lock. The link operation fails if the dot-lock file already exists. lockmail uses an enhanced method of dot-locking, where its process ID, and the name of the server where lockmail is running is also saved in its dot-lock file. If the operation fails due to an existing dot-lock file that was created by another lockmail process on the same server, and the process ID no longer exists, this stale dot-lock file is removed immediately. In all other situations a dot-lock file older than five minutes is considered stale, and removed.
| NOTE: | A failure to create a dot-lock file is silently ignored if the reason for the failure is because lockmail does not have the write permission in the dot-lock file's directory. The incoming mail spool directory (usually /var/spool/mail) typically does not have global write permissions, so the attempt to create the dot-lock file in the spool directory will fail, and lockmail will be content with using file-locking only. |
The final locking mechanism lockmail uses is the operating system's file locking facility. If lockmail fails to obtain all three locks, lockmail will sleep for five seconds and try again. The only exception is a failure to create a dot-lock because of no write access to the dot-lock file's directory, which is ignored. If lockmail still fails to obtain all required locks in the amount of time specified by the -t option (or its default value), lockmail will terminate with the EX_TEMPFAIL exit code.
lockmail runs program after obtaining the last file lock, waits until program terminates, and releases all locks. program must terminate before any of the locks obtained by lockmail expire, and are considered stale. lockmail will then terminate with the same exit code as program.