On Tue, 2001-12-11 at 10:11, Matthijs van der Klip wrote:
> >>At the time there was a bug which caused an infinite loop in the kernel
> >>when dmapi was turned on. Since dmapi did not have any real application
> >>at the time, we decided to turn it off. Later kernels fix this and it
> >>is safe to turn it on. If you turn it on in the 1.0.2 kernel rpms you
> >>will probably hit the bug.
> >
> >Ah. Thanks for the info. I've rebuilt a kernel with dmapi and have
> >enabled it on a test filesystem. I have not yet hit the bug but will
> >be watching for it...
>
>
> Hi Steve,
>
> I think I have hit the bug. I can receive exactly one event for a
> filesystem. After that it locks up with a 'resource busy' error.
> Does this sound like the 1.0.2 dmapi bug?
Hmm, not sure it does, you tend to get a 100% busy cpu if you hit
the bug, this sounds more like a DMAPI programming issue to me,
in which case Dean will have to answer this.
>
> Are the 2.4.14 kernels in 'testing' RH based? I need XFS, DMAPI as well
> as tux and ipvs in the same kernel.
Ugh, the only redhat kernels we do are the ones released as 1.0.1 and
1.0.2 XFS, the general patches and cvs trees are all based on Linus
kernels. For 1.0.2 there are 2.4.7 and 2.4.9 based redhat kernels.
I have attached the patch which fixes the dmapi related hang.
Steve
>
>
> Best regards,
>
> Matthijs van der Klip
> NOS - Dutch Public Broadcasting Organisation
--
Steve Lord voice: +1-651-683-3511
Principal Engineer, Filesystem Software email: lord@xxxxxxx
patch.dmapi
Description: Text document
|