[OpenAFS-devel] deadlock on linux-2.4

[Neulinger, Nathan]

works clean for me on both UP and SMP, but both on SMP kernel. Switched
the SMP box back to disk cache for testing this.=20

-- Nathan

------------------------------------------------------------
Nathan Neulinger                       EMail:  nneul@umr.edu
University of Missouri - Rolla         Phone: (573) 341-4841
Computing Services                       Fax: (573) 341-4216


> -----Original Message-----
> From: Pavel Semerad [mailto:semerad@ss1000.ms.mff.cuni.cz]=20
> Sent: Wednesday, June 05, 2002 10:50 AM
> To: openafs-devel@openafs.org
> Subject: Re: [OpenAFS-devel] deadlock on linux-2.4
>=20
>=20
> > >   I have found more info and created a temorary hack,=20
> which fixes it,
> > > but it will not be probably the correct solution.
> >=20
> > Thanks for tracking this down.  As you point out, we=20
> shouldn't obtain
> > vcache locks while holding any dcache locks, due to lock ordering.
> > The right solution is probably to delay calling=20
> afs_ProcessFS() until
> > after we've released the dcache entries; can you try the patch below
> > and see if it works for you?
>=20
>   It's OK. Tested on UP machine. I cannot test it on SMP now.
> >=20
> > > I don't know why other architectures are OK, may be it is=20
> not possible
> > > to call afs_StoreAllSegment and afs_UFSRead together on them.
> >=20
> > I suspect other architectures also have the same bug.  Note=20
> that this
> > only affects the development branch (the dcache locking=20
> code has only
> > been added recently), which is probably why this bug hasn't=20
> been seen
> > by others yet.
>=20
>   I have older version on other architectures, this is why it=20
> worked :-)
> >=20
> > -- kolya
> >=20
> Pavel
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-devel
>=20