[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#50872: Release-critical Bugreport for November 26, 1999



> On Mon, 29 Nov 1999, Anders Hammarquist wrote:
> 
> > > > Package: nfs-common (main)
> > > > Maintainer: Anders Hammarquist <iko@debian.org>
> > > >   50872  nfs-common: NFS file locking broken with LinUX 2.2.x

> I just had a look at this bug report and the problem looks to me like
> the following:
> The nfs client in kernel 2.2.* uses NFS locking by default
> The nfs client in kernel 2.0.* did _not_ use NFS locking by default.
> So if you use 2.2.* on the client with a server which doesn't support
> locking (like the user space nfsd), this will result in an error
> message on the client, if you try to lock a file via NFS using
> fcntl().
> 
> You can simply work around this error by adding the option "nolock" to
> mount options of the NFS filesystem which is mounted from a non
> locking server.
> 
> So I don't think, that this is a bug in nfs-common but simply a
> changed behavior in the Linux kernel, which may need some adjustment
> in the fstab.

His first problem almost certainly stems from using a 2.2 kernel with
an NFS server that does not support locking. His other problem (with
mail files on an HP box) I'm guessing is the result of a missing
statd somewhere, but I would like to have it confirmed. HP's NFS should
support locking...

> Maybe Debian should give users who upgrade from slink (kernel 2.0.*)
> to potato (kernel 2.2.*) some hint, if the server still doesn't
> support locking, but I'm not sure how this can be automated, because
> it depends on every single server, which is accessed in fstab,
> whether the nolock option has to be added or not.

This would be good to put in some release notes somewhere, and in
the upgrade instructions. Automating it is probably not a good idea
though, since you can't reliably tell if the server supports locking until
after you have done the upgrade (and even then it may change if the
user changes from userspace NFS to knfs).

/Anders

-- 
 -- Of course I'm crazy, but that doesn't mean I'm wrong.
Anders Hammarquist                                  | iko@cd.chalmers.se
Physics student                                     | Hem: +46 31 47 69 27
Chalmers University of Technology, G|teborg, Sweden | Mob: +46 707 27 86 87



Reply to: