[PATCH] locks: don't unnecessarily fail posix lock operations
authorMiklos Szeredi <miklos@szeredi.hu>
Fri, 23 Jun 2006 09:05:09 +0000 (02:05 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Fri, 23 Jun 2006 14:43:02 +0000 (07:43 -0700)
commit0d9a490abe1f69fda220f7866f6f23af41daa128
tree499b7f9c44bb5a079f8ba303fefba402e603c314
parent090d2b185d8680fc26a2eaf4245d4171dcf4baf1
[PATCH] locks: don't unnecessarily fail posix lock operations

posix_lock_file() was too cautious, failing operations on OOM, even if they
didn't actually require an allocation.

This has the disadvantage, that a failing unlock on process exit could lead to
a memory leak.  There are two possibilites for this:

- filesystem implements .lock() and calls back to posix_lock_file().  On
cleanup of files_struct locks_remove_posix() is called which should remove all
locks belonging to files_struct.  However if filesystem calls
posix_lock_file() which fails, then those locks will never be freed.

- if a file is closed while a lock is blocked, then after acquiring
fcntl_setlk() will undo the lock.  But this unlock itself might fail on OOM,
again possibly leaking the lock.

The solution is to move the checking of the allocations until after it is sure
that they will be needed.  This will solve the above problem since unlock will
always succeed unless it splits an existing region.

Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
fs/locks.c