Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!apple!bionet!ames!ucsd!helios.ee.lbl.gov!pasteur!buddy.Berkeley.EDU!deboor
From: deboor@buddy.Berkeley.EDU (Adam R de Boor)
Newsgroups: comp.unix.wizards
Subject: Re: Locks for exclusive access to resources
Summary: unlink ain't atomic
Message-ID: <14834@pasteur.Berkeley.EDU>
Date: 21 Jun 89 05:32:31 GMT
References: <1020@twwells.uucp> <1989Jun14.052648.20183@twwells.com>
Sender: news@pasteur.Berkeley.EDU
Reply-To: deboor@buddy.Berkeley.EDU.UUCP (Adam R de Boor)
Distribution: usa
Organization: University of California, Berkeley
Lines: 9


unlink isn't an atomic operation under NFS. You'd think it would be (you'd
think file creation would be atomic too...), but it isn't. I've had cases where
unlinks of the same file on two systems both succeeded, though *one* should
have returned ENOENT. I eventually went with an election algorithm (slightly
different problem -- the locking was to determine which server was to be
the master...)

a