Closed GoogleCodeExporter closed 8 years ago
Original comment by alex.ble...@gmail.com
on 6 Mar 2010 at 1:31
When compiled under the 'Debug' rather than 'Release', and without the
ZFS_DEBUG defined, I don't see a
panic when exporting pools. I do get a problem when unloading the ZFS kernel
extension though, which would
be problematic. So there's clearly some kind of stability issue here.
Original comment by alex.ble...@gmail.com
on 7 Mar 2010 at 9:48
The problem is that we're using the mt_cache to free up the mp, but I think by
then we've already nuked it.
The call (indirectly) ends up coming from here:
http://github.com/alblue/mac-zfs/blob/maczfs_72/usr/src/uts/common/os/kmem.c#L11
33
1133 kmem_cache_free(cp->cache_magtype->mt_cache, mp);
At that point, cp->cache_magtype->mt_cache is null. we thus try and access its
properties, and then fail with
the panic.
It's possibly coming from here:
http://github.com/alblue/mac-zfs/blob/maczfs_72/usr/src/uts/common/os/kmem.c#L29
11
2911 mtp->mt_cache = NULL;
In any case, by the time we get to the end of the routine, we don't have an
mt_cache any more, so we panic.
Original comment by alex.ble...@gmail.com
on 7 Mar 2010 at 11:31
OK, verified this change with both Release and Debug builds. I no longer see
this specific issue.
Original comment by alex.ble...@gmail.com
on 8 Mar 2010 at 12:08
Fixed at
http://github.com/alblue/mac-zfs/commit/090e1ba81cfbbc9aace0cbfb31dfc14258abfa19
Original comment by alex.ble...@gmail.com
on 8 Mar 2010 at 1:22
Original issue reported on code.google.com by
alex.ble...@gmail.com
on 6 Mar 2010 at 1:30