Date: Sat, 21 May 2016 02:29:35 +0000 (UTC) From: Alan Somers <asomers@FreeBSD.org> To: src-committers@freebsd.org, svn-src-projects@freebsd.org Subject: svn commit: r300357 - projects/zfsd/head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs Message-ID: <201605210229.u4L2TZkZ049156@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: asomers Date: Sat May 21 02:29:35 2016 New Revision: 300357 URL: https://svnweb.freebsd.org/changeset/base/300357 Log: Fix the resource.fs.zfs.statechange message It had a number of problems: * It was only being emitted on a transition to the HEALTHY state. That made it impossible for zfsd to take actions based on drives getting sicker. * It compared the new state to vdev_prevstate, which is the state that the vdev had the last time it was opened. That doesn't make sense, because a vdev can change state multiple times without being reopened. * vdev_set_state contains logic that will change the device's new state based on various conditions. However, the statechange event was being posted _before_ that logic took effect. Now it's being posted after. Sponsored by: Spectra Logic Corp Modified: projects/zfsd/head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c Modified: projects/zfsd/head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c ============================================================================== --- projects/zfsd/head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c Sat May 21 02:14:11 2016 (r300356) +++ projects/zfsd/head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c Sat May 21 02:29:35 2016 (r300357) @@ -3366,19 +3366,6 @@ vdev_set_state(vdev_t *vd, boolean_t iso vd->vdev_ops->vdev_op_leaf) vd->vdev_ops->vdev_op_close(vd); - /* - * If we have brought this vdev back into service, we need - * to notify fmd so that it can gracefully repair any outstanding - * cases due to a missing device. We do this in all cases, even those - * that probably don't correlate to a repaired fault. This is sure to - * catch all cases, and we let the zfs-retire agent sort it out. If - * this is a transient state it's OK, as the retire agent will - * double-check the state of the vdev before repairing it. - */ - if (state == VDEV_STATE_HEALTHY && vd->vdev_ops->vdev_op_leaf && - vd->vdev_prevstate != state) - zfs_post_state_change(spa, vd); - if (vd->vdev_removed && state == VDEV_STATE_CANT_OPEN && (aux == VDEV_AUX_OPEN_FAILED || vd->vdev_checkremove)) { @@ -3459,6 +3446,16 @@ vdev_set_state(vdev_t *vd, boolean_t iso vd->vdev_removed = B_FALSE; } + /* + * Notify the fmd of the state change. Be verbose and post + * notifications even for stuff that's not important; the fmd agent can + * sort it out. Don't emit state change events for non-leaf vdevs since + * they can't change state on their own. The FMD can check their state + * if it wants to when it sees that a leaf vdev had a state change. + */ + if (vd->vdev_ops->vdev_op_leaf) + zfs_post_state_change(spa, vd); + if (!isopen && vd->vdev_parent) vdev_propagate_state(vd->vdev_parent); }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201605210229.u4L2TZkZ049156>