On Thu, 2008-01-03 at 11:10 +0100, Michael Kerrisk wrote: > Hi Trond, > > Trond Myklebust wrote: > > On Wed, 2008-01-02 at 10:55 +0100, Michael Kerrisk wrote: > >> Hey Trond, > >> > >> Would you take a quick look at > >> > >> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=453946 > >> > >> Do you have some insights on this? What should the man page say about > >> current support of O_EXCL on NFS? > >> > >> Cheers, > >> > >> Michael > > > > Hi Michael, > > > > Timo is correct. O_EXCL has been supported for NFSv3 (and NFSv4) ever > > since the 2.6 series kernels came out. > > > > However, please note that > > * it is not supported for any 2.4.x kernels, since that would > > require some VFS re-engineering that is not compatible with the > > 'stable kernel' rules. > > * As Timo states, the NFSv2 protocol cannot support O_EXCL > > semantics. > > Thanks. I've rewritten the text as shown below. > > I also have another question: is O_APPEND supported in NFSv3 or NFSv4? > > > I think the second point warrants keeping the link() workaround mention, > > though, since people who are interested in writing completely portable > > software may still find it useful. For instance, a lot of mail delivery > > setups are still using the type of locking scheme described there. > > Yes, that makes sense. > > For man-pages-2.75, I rewrote the paragraph as shown here. Could you > confirm that this is okay? > > O_EXCL is not supported on NFS file systems before > NFSv3 or on Linux before kernel 2.6; it is sup-
I'd probably just say 'not supported on NFSv2'. There is not NFSv1, so this would be quite unambiguous. > ported on Linux 2.6 and later, with NFSv3 or > later. In environments where NFS O_EXCL support > is not provided, programs that rely on it for per- > forming locking tasks will contain a race condi- > tion. Portable programs that want to perform > atomic file locking using a lockfile, and need to > avoid reliance on NFS support for O_EXCL, can cre- > ate a unique file on the same file system (e.g., > incorporating hostname and PID), and use link(2) > to make a link to the lockfile. If link(2) > returns 0, the lock is successful. Otherwise, use > stat(2) on the unique file to check if its link > count has increased to 2, in which case the lock > is also successful. Looks good otherwise. Cheers Trond -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
