Hi! It seems, I've found a bug in the ZFS v28 on the latest stable: if we have a snapshot with some files having an extended attributes, then attempt to read an extended attributes's value leads to a well reproducible kernel panic.
The part of backtrace follows: #6 0xffffffff804bbe44 in calltrap () at /usr/src/sys/amd64/amd64/exception.S:228 #7 0xffffffff80950ea7 in zil_commit (zilog=0x0, foid=5795917) at /usr/src/sys/modules/zfs/../../cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c:1497 #8 0xffffffff80979e6b in zfs_freebsd_read (ap=Variable "ap" is not available.) at /usr/src/sys/modules/zfs/../../cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c:622 #9 0xffffffff80979750 in zfs_getextattr (ap=0xffffff80dd5d8820) at vnode_if.h:384 #10 0xffffffff8038921b in extattr_get_vp (vp=0xffffff0056a01588, attrnamespace=1, attrname=0xffffff80dd5d89a0 "DOSATTRIB", data=Variable "data" is not available.) at vnode_if.h:1332 It seems that ZIL isn't available for snapshots, but zfs_freebsd_read doesn't check this when calling zil_commit. The attached patch fixes this issue. Can anybody confirm this? -- Alexander Zagrebin
--- sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c.orig 2011-08-01 23:04:07.358173627 +0400 +++ sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c 2011-08-02 00:10:02.674585604 +0400 @@ -618,7 +618,8 @@ zfs_read(vnode_t *vp, uio_t *uio, int io /* * If we're in FRSYNC mode, sync out this znode before reading it. */ - if (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zfsvfs->z_log && + (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) zil_commit(zfsvfs->z_log, zp->z_id); /*
_______________________________________________ freebsd-current@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-current To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"