New ZFSv28 patchset for 8-STABLE: Kernel Panic

[jhell]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 12/28/2010 18:20, Martin Matuska wrote:
> Please don't consider these patches as production-ready.
> What we want to do is find and resolve as many bugs as possible.

I completely agree with Martin here. If your running it then your
willing to loose what you have if you have not taken precaution to save
your data somewhere else. Even though with that said ZFS does a pretty
fine job of ensuring that nothing happens to it, it is still best
practice to have a copy somewhere other than "IN THAT BOX" ;)

Another note too, I think I read that you mentioned using the L2ARC and
slog device on the same disk.... You simply shouldn't do this it could
be contributing to the real cause and there is absolutely no gain in
either sanity or performance and you will end up bottle-necking your system.

> 
> To help us fix these bugs, a way to reproduce the bug from a clean start
> (e.g. in virtualbox) would be great and speed up finding the cause for
> the problem.
> 
> Your problem looks like some sort of deadlock. In your case, when you
> experiene the hang, try running "procstat -k -k PID" in another shell
> (console). That will give us valuable information.
> 

Martin,

	I agree with the above that it may be some sort of live or dead lock
problem in this case. It would be awesome to know if some of the
following sysctl(8)'s values are and how this reacts when set to the
opposite of their current values.

vfs.zfs.l2arc_noprefetch:
vfs.zfs.dedup.prefetch:
vfs.zfs.prefetch_disable

The reason why I say this is on one of my personal systems that I toy
with the box cannot make it very long with prefetch enabled on either
v15 or v28 after some 'unknown' commit to world on stable/8. Now this
may actually be just a contributing factor that makes it happen sooner
than it normally would but probably also directly relates to the exact
problem. I would love to see this go away as I had been using the L2ARC
with prefetch enabled for a long time and now all of a sudden just
plainly does not work correctly.

I also have about 19 core.txt.NN files from when this started happening
with various stack traces. If you would like these just let me know and
Ill privately mail them to you.


Regards,

- -- 

 jhell,v - JJH48-ARIN
-----BEGIN PGP SIGNATURE-----

iQEbBAEBAgAGBQJNGsDLAAoJEJBXh4mJ2FR+TqkH8wVFQKiU/C6L+F4Y3/ClScQD
b4s0IkC1B+bHl9eD6Hhxif/1iKj1w9clYvuLt8ageDF98KTB9GCRjuh48VswdtPQ
FQtDRTj1pGzWPxmOn2Nrf7qrFnymqZk+qoTBX8A1nDvrSe41Mqp82ue9E7nZ1ipg
Dz9k5F8J+WxUAZYLHxtxLvYEa19/hvG1K5LOpRKIU0iycsqaywezFflTGDcR5lT8
A50ic9sZ21jr87CK45TLv1Wmu+kDgpy2j1x77bYTDGoAzQMlPcOENO8st8EobcWB
eIwwXIjtRwOKF4rSxoqwwYxOM4ek+tK4p6ZnO1uLipNXMB+zJTjs//GV6Xp3TA==
=Io4+
-----END PGP SIGNATURE-----