Date: Fri, 12 Feb 2021 21:15:17 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 253471] Change 348906: vdev_geom_attach_by_guids() breaks other use cases Message-ID: <bug-253471-227@https.bugs.freebsd.org/bugzilla/>
next in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D253471 Bug ID: 253471 Summary: Change 348906: vdev_geom_attach_by_guids() breaks other use cases Product: Base System Version: 12.2-STABLE Hardware: Any OS: Any Status: New Severity: Affects Only Me Priority: --- Component: kern Assignee: bugs@FreeBSD.org Reporter: daveb@spectralogic.com Change MFC r344316, SVN Importer: FreeBSD/base rev 348906 breaks the following scenario: - Assume 3 blank, all ZFS label cleared, disks in the system. - Shutdown, remove one of the blanks, put in on the shelf, reboot. - Create a mirror pool on the 2 remaining disk, shutdown. - Replace one of the mirror disks with the with the previously removed blank, put the zfs mirror member on the shelf, reboot. All these machinations are required to ensure the blank disk enumerates the same as the mirror disk member (has the same vdev_path). zpool status will now report the pool as a healthy "ONLINE" with a missing disk: # zpool status -x pool: pool200 state: ONLINE status: One or more devices could not be used because the label is missing = or invalid. Sufficient replicas exist for the pool to continue functioning in a degraded state. action: Replace the device using 'zpool replace'. see: http://illumos.org/msg/ZFS-8000-4J scan: none requested config: NAME STATE READ WRITE CKSUM pool200 ONLINE 0 0 0 mirror-0 ONLINE 0 0 0 1165392866195823403 UNAVAIL 0 0 0 was /dev/da8 da9 ONLINE 0 0 0 This is because change 348906 allows vdev_geom_attach_by_guids() to attach a vdev that has a "NO_MATCH" guid but satisfies the requested vdev_path: } else if (match =3D=3D best_match) { /* match =3D NO_MATCH and best_match is initialized to NO_MATCH; * if the paths match we will attach it. */ if (strcmp(pp->name, vdpath) =3D=3D 0) { best_pp =3D pp; } } >From ZFS_LOG() via vdev_geom_open_by_guids() and others: vdev_geom_attach:240[1]: Attaching to da8. vdev_geom_attach:309[1]: Created consumer for da8. vdev_geom_open_by_guids:797[1]: Attach by guid [17284178551510092527:1165392866195823403] succeeded, provider da8. Later back in vdev_validate(), this condition never gets vdev_propagate()ed when detected because for this one error, vdev_set_state() is called with "is_open" true: if ((label =3D vdev_label_read_config(vd, txg)) =3D=3D NULL) { vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN, VDEV_AUX_BAD_LABEL); return (0); } It is curious to me that for this one error path vdev_validate() calls vdev_set_state(..., B_TRUE, ...) whereas all other are B_FALSE. For my purposes, reverting 348906 fixes the issue, but changing=20 vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN, VDEV_AUX_BAD_LABEL); to vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN, VDEV_AUX_BAD_LABEL); also solves the problem. --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-253471-227>