[PATCH v6 0/1] multipathd: don't leave maps read-only after write protect is cleared

[PATCH v6 0/1] multipathd: don't leave maps read-only after write protect is cleared

[Brian Bunker]

This is v6. It follows Ben's review of v5: needs_ro_update() no longer
takes an ro value and no longer tries to avoid reading the paths' ro
state from sysfs. It recomputes force_readonly from all of the map's
paths on every call, which removes the special cases v5 needed. The
changes are listed at the end.

As agreed, tracking the ro state per path and dropping force_readonly
altogether, as Martin and Uday Shankar suggested, will be a separate
series on top of this fix.

A multipath map can stay read-only after write protect is cleared on
the storage, even though all of its paths are read-write. It happens
when the map's paths are removed and re-added while write protect is
set:

1. Write protect is set. The map is reloaded read-only.
   uev_update_path() sets force_readonly for that reload and clears it
   afterwards.

2. All of the map's paths are removed, and new paths are added while
   write protect is still set. ev_add_path() sees a read-only path and
   sets force_readonly = 1. Nothing ever clears it.

3. Write protect is cleared on the new paths, one after another. Each
   path's DISK_RO=0 uevent is ignored while any path of the map is
   still read-only.

4. The last path's uevent reloads the map. The reload sees the stale
   force_readonly = 1 and loads the map read-only, logging "reloaded
   (retval 0)".

force_readonly only works in one direction: ev_add_path() can set it
but never clears it. If the new paths had been added already writable,
ev_add_path() wouldn't have set the flag and the map wouldn't have
stuck. The paths have to be added while write protected and then
cleared in place.

After step 4 nothing reloads the map again. A rescan doesn't change the
paths' ro state, so it generates no DISK_RO uevent. And since
multipathd now sees the map as read-only, needs_ro_update() also
ignores the next DISK_RO=1 uevent for it.

If two paths' uevents happen to be handled after all paths are
writable, the map does recover: uev_update_path() clears
force_readonly after its reload, so the second reload is read-write.
That makes the problem look intermittent.

Example
-------

We hit this on RHEL 9 device-mapper-multipath 0.8.7, whose
ev_add_path() and uev_update_path() handle force_readonly the same way
as current master. The host's iSCSI sessions were torn down and
re-established while the array was write protecting the volume. From
the messages log, trimmed:

  05:48:27  ...114d8: reload [0 51200 multipath 0 1 alua 1 1
            service-time 0 6 1 135:272 1 66:624 1 69:720 1 129:560 1
            132:656 1 135:752 1]
  05:48:27  sdsd: map ...114d8 reloaded (retval 0)
  05:51:04  kernel: sd 24:0:0:48: [sdaer] Write Protect is on
  05:51:05  ...114d8: remaining active paths: 0
  05:51:07  kernel: sd 29:0:0:48: [sdaoi] Write Protect is on
            (and four more new paths)
  06:09:01  kernel: sd 24:0:0:48: [sdaer] Write Protect is off
  06:09:07  kernel: sd 29:0:0:48: [sdaoi] Write Protect is off
  06:09:10  sdaoi: update path write_protect to '0' (uevent)
  06:09:10  ...114d8: reload [0 51200 multipath 0 1 alua 1 1
            service-time 0 6 1 67:880 1 70:992 1 130:848 1 133:960 1
            65:1072 1 67:1056 1]
  06:09:10  sdaoi: map ...114d8 reloaded (retval 0)

The state afterwards, as recorded by the test:

  dm-106: ro
  sdaer: rw  sdagu: rw  sdaix: rw  sdala: rw  sdand: rw  sdaoi: rw

The patch
---------

The patch sets force_readonly right before each load from the state of
the map's paths. dm_addmap_create() clears it as soon as it has read
it, and dm_addmap_reload() clears it once the reload has succeeded.
needs_ro_update() sets it for the reload done for a change uevent,
from the ro state of all of the map's paths in sysfs, and
ev_add_path() and update_map() compute it from all of the map's paths.
With the patch, the 06:09:10 reload would have found no read-only path
and loaded the map read-write.

Changes since v5:

- needs_ro_update() no longer takes an ro value. As Ben suggested, it
  sets force_readonly from map_has_ro_path() on every call. If the map
  is already read-only and has to stay so, it clears the flag and
  returns false, since dm has to reopen the paths for any read-write
  reload anyway. Otherwise it returns true if the flag differs from
  the map's current read-only state. This fixes the v5 case where the
  map was read-write, a path was still read-only, and nothing was
  reloaded, and it drops the "keep the flag after a failed read-only
  reload" special case: the flag is recomputed on every change uevent,
  so a failed reload is retried on the next one.

- uev_update_path() no longer reads the path's ro state itself, from
  sysfs or from the uevent's DISK_RO. It just calls needs_ro_update().
  The uevent still only tells us that something changed, and sysfs,
  which the kernel updates before it sends the uevent, is the source
  of the ro state. The "update path write_protect" message now prints
  the new state from force_readonly.

- A path whose ro attribute can't be read in sysfs counts as not
  read-only, instead of making uev_update_path() do nothing.

- The rest of the patch (dm_addmap_create(), dm_addmap_reload(),
  update_map(), ev_add_path()) is unchanged from v5.

Changes since v4:

- dm_addmap_reload() no longer clears force_readonly up front and sets
  it back on failure. It clears the flag only once the reload and
  resume have succeeded, as Ben suggested. A failed reload leaves it
  set, so the next reload tries read-only again.

- uev_update_path() takes the path's ro state from sysfs instead of
  from the uevent's DISK_RO. The uevent only says that something
  changed. The kernel updates the ro attribute before it sends the
  uevent, so sysfs is current when the uevent is handled, while the
  uevent itself may have been queued behind later changes.

  Note that this changes behavior: any change uevent for a path in a
  map, not only one with DISK_RO, can now trigger a read-only reload
  if sysfs disagrees with the map. It also avoids the pointless
  reload that 6e40bf92 describes when a rescan's DISK_RO=0 and
  DISK_RO=1 uevents are handled after the rescan has finished, since
  sysfs then shows read-only for both. The cost is a sysfs read of
  each path of the map per change uevent, next to the pathinfo() call
  uev_update_path() already makes.

- dm_addmap_create() is unchanged and keeps the read-only shortcut.

Changes since v3:

- If a reload fails, dm_addmap_reload() sets force_readonly back to
  the value it had on entry. The map still has its old table, so a map
  that was supposed to become read-only is tried again on the next
  reload.

- Move the walk over the map's path groups into map_has_ro_path().

Changes since v2:

- In needs_ro_update(), fetch the dm info with dm_get_info() if the
  map has none, and return false if that fails, instead of assuming a
  reload is needed.

- Clear force_readonly in dm_addmap_create() and dm_addmap_reload() as
  soon as they have read it, instead of in dm_addmap() on success.

- I didn't add per-path tracking of the ro state. force_readonly is
  only computed from sysfs right before a load, which is rare, so the
  cost of reading sysfs there is small.

Changes since v1:

- Set force_readonly in needs_ro_update() when it returns true,
  instead of in uev_update_path().

- Clear force_readonly in dm_addmap() on success, as for need_reload,
  instead of after the loads in ev_add_path() and uev_update_path().

- In ev_add_path(), only compute force_readonly when need_do_map is
  set.

- Walk mpp->pg rather than mpp->paths in needs_ro_update(), since
  mpp->paths may be stale there. ev_add_path() and update_map() still
  use mpp->paths: both call adopt_paths() first, and mpp->pg doesn't
  contain the new paths until setup_map().

- Move sysfs_get_ro() above has_ro_path() instead of adding a forward
  declaration. The function is unchanged.

Brian Bunker (1):
  multipathd: don't leave maps read-only after write protect is cleared

 libmultipath/devmapper.c |  18 ++++-
 multipathd/main.c        | 144 ++++++++++++++++++++++++---------------
 2 files changed, 105 insertions(+), 57 deletions(-)

-- 
2.56.0

[PATCH v6 1/1] multipathd: don't leave maps read-only after write protect is cleared

[Brian Bunker]

ev_add_path() sets mpp->force_readonly when a newly added path is
read-only, but never clears it. If a map is created or reloaded while
its storage is write protected, the flag stays set. When write protect
is cleared, the reload that uev_update_path() does for the DISK_RO=0
uevent uses the stale flag and loads the map read-only, even though all
of its paths are read-write. uev_update_path() clears the flag after
that reload, so the map only recovers if another DISK_RO=0 uevent
reloads it again. If none comes, the map stays read-only, and since
mpp->dmi.read_only is then 1, needs_ro_update() also ignores later
DISK_RO=1 uevents for it.

Fix this by setting force_readonly before each load and clearing it
once the load has succeeded:

- uev_update_path() no longer takes the path's ro state from the
  uevent's DISK_RO. The uevent only says that something changed. The
  kernel updates the ro attribute before it sends the uevent, so sysfs
  is current when the uevent is handled, while the uevent itself may
  have been queued behind later changes. It just calls
  needs_ro_update(), which reads the ro state of all of the map's paths
  from sysfs.

- needs_ro_update() no longer takes an ro value. It sets
  force_readonly from the ro state of the map's paths, and returns true
  if that differs from the map's current state. If the map is already
  read-only and has to stay so, it clears the flag and returns false,
  since dm has to reopen the paths for any read-write reload anyway. If
  the map has no dm info, it fetches it, and returns false if that
  fails, instead of assuming a reload is needed. Since the flag is
  recomputed on every call, a failed read-only reload is retried on the
  next uevent without any special handling.

- ev_add_path() sets it from all of the map's paths rather than only
  the path being added, and only when it is about to load the map
  (need_do_map). mpp->paths is up to date there after adopt_paths(),
  while mpp->pg doesn't contain the new paths yet.

- update_map(), which does reloads that were delayed while the map was
  waiting for udev (UDEV_WAIT_RELOAD), recomputes it from the map's
  paths before setting up the map.

- dm_addmap_create() clears it as soon as it has read it. A failed
  create leaves no map behind, and the next create computes the flag
  again.

- dm_addmap_reload() clears it once the reload and resume have
  succeeded. If the reload fails, the map keeps its old table, and the
  flag stays set so that the next reload tries again.

Signed-off-by: Brian Bunker <brian@purestorage.com>
---
 libmultipath/devmapper.c |  18 ++++-
 multipathd/main.c        | 144 ++++++++++++++++++++++++---------------
 2 files changed, 105 insertions(+), 57 deletions(-)

diff --git a/libmultipath/devmapper.c b/libmultipath/devmapper.c
index e3870f59..34e0f640 100644
--- a/libmultipath/devmapper.c
+++ b/libmultipath/devmapper.c
@@ -539,10 +539,15 @@ static uint16_t build_udev_flags(const struct multipath *mpp, int reload)
 
 int dm_addmap_create (struct multipath *mpp, char *params)
 {
-	int ro;
+	int ro = mpp->force_readonly ? 1 : 0;
 	uint16_t udev_flags = build_udev_flags(mpp, 0);
 
-	for (ro = mpp->force_readonly ? 1 : 0; ro <= 1; ro++) {
+	/*
+	 * force_readonly only applies to this load. Clear it, so that it
+	 * can't affect later loads, even if this one fails.
+	 */
+	mpp->force_readonly = 0;
+	for (; ro <= 1; ro++) {
 		int err;
 
 		if (dm_addmap(DM_DEVICE_CREATE, TGT_MPATH, mpp, params, ro,
@@ -599,8 +604,15 @@ int dm_addmap_reload(struct multipath *mpp, char *params, int flush)
 		r = dm_simplecmd(DM_DEVICE_RESUME, mpp->alias,
 				 DMFL_NEED_SYNC | (flush ? 0 : DMFL_NO_FLUSH),
 				 udev_flags);
-	if (r)
+	if (r) {
+		/*
+		 * force_readonly only applies to this reload. If the reload
+		 * fails, the map keeps its old table, and the flag stays set
+		 * so that the next reload tries again.
+		 */
+		mpp->force_readonly = 0;
 		return r;
+	}
 
 	/* If the resume failed, dm will leave the device suspended, and
 	 * drop the new table, so doing a second resume will try using
diff --git a/multipathd/main.c b/multipathd/main.c
index 9bce4d6b..f3b0cb5f 100644
--- a/multipathd/main.c
+++ b/multipathd/main.c
@@ -709,6 +709,55 @@ void cleanup_reset_vec(struct vector_s **v)
 	vector_reset(*v);
 }
 
+static int
+sysfs_get_ro (struct path *pp)
+{
+	int ro;
+	char buff[3]; /* Either "0\n\0" or "1\n\0" */
+
+	if (!pp->udev)
+		return -1;
+
+	if (!sysfs_attr_get_value_ok(pp->udev, "ro", buff, sizeof(buff))) {
+		condlog(3, "%s: Cannot read ro attribute in sysfs", pp->dev);
+		return -1;
+	}
+
+	if (sscanf(buff, "%d\n", &ro) != 1 || ro < 0 || ro > 1) {
+		condlog(3, "%s: Cannot parse ro attribute", pp->dev);
+		return -1;
+	}
+
+	return ro;
+}
+
+/* A map must be loaded read-only if any of its paths is read-only. */
+static bool
+has_ro_path(vector paths)
+{
+	struct path *pp;
+	unsigned int i;
+
+	vector_foreach_slot (paths, pp, i) {
+		if (sysfs_get_ro(pp) == 1)
+			return true;
+	}
+	return false;
+}
+
+static bool
+map_has_ro_path(struct multipath *mpp)
+{
+	struct pathgroup *pgp;
+	unsigned int i;
+
+	vector_foreach_slot (mpp->pg, pgp, i) {
+		if (has_ro_path(pgp->paths))
+			return true;
+	}
+	return false;
+}
+
 static int
 update_map (struct multipath *mpp, struct vectors *vecs, int new_map)
 {
@@ -732,6 +781,14 @@ retry:
 
 	mpp->action = ACT_RELOAD;
 
+	/*
+	 * This may be a delayed reload, and the paths' write protect
+	 * state may have changed since force_readonly was last set.
+	 * Recompute it from the current paths. mpp->paths is up to date
+	 * after adopt_paths(), while mpp->pg isn't until setup_map().
+	 */
+	mpp->force_readonly = has_ro_path(mpp->paths);
+
 	if (setup_map(mpp, &params, vecs)) {
 		condlog(0, "%s: failed to setup new map in update", mpp->alias);
 		retries = -1;
@@ -1270,28 +1327,6 @@ out:
 	return ret;
 }
 
-static int
-sysfs_get_ro (struct path *pp)
-{
-	int ro;
-	char buff[3]; /* Either "0\n\0" or "1\n\0" */
-
-	if (!pp->udev)
-		return -1;
-
-	if (!sysfs_attr_get_value_ok(pp->udev, "ro", buff, sizeof(buff))) {
-		condlog(3, "%s: Cannot read ro attribute in sysfs", pp->dev);
-		return -1;
-	}
-
-	if (sscanf(buff, "%d\n", &ro) != 1 || ro < 0 || ro > 1) {
-		condlog(3, "%s: Cannot parse ro attribute", pp->dev);
-		return -1;
-	}
-
-	return ro;
-}
-
 /*
  * returns:
  * 0: added
@@ -1305,7 +1340,6 @@ ev_add_path (struct path * pp, struct vectors * vecs, int need_do_map)
 	int retries = 3;
 	int start_waiter = 0;
 	int ret;
-	int ro;
 	unsigned char prflag = PR_UNSET;
 
 	/*
@@ -1370,14 +1404,16 @@ rescan:
 			goto fail; /* leave path added to pathvec */
 	}
 
-	/* ro check - if new path is ro, force map to be ro as well */
-	ro = sysfs_get_ro(pp);
-	if (ro == 1)
-		mpp->force_readonly = 1;
-
 	if (!need_do_map)
 		return 0;
 
+	/*
+	 * Load the map read-only if any of its paths is read-only, not
+	 * just the one being added. mpp->paths is up to date after
+	 * adopt_paths(), while mpp->pg doesn't have the new paths yet.
+	 */
+	mpp->force_readonly = has_ro_path(mpp->paths);
+
 	if (!dm_map_present(mpp->alias)) {
 		mpp->action = ACT_CREATE;
 		start_waiter = 1;
@@ -1594,27 +1630,27 @@ finish_path_init(struct path *pp, struct vectors * vecs)
 }
 
 static bool
-needs_ro_update(struct multipath *mpp, int ro)
+needs_ro_update(struct multipath *mpp)
 {
-	struct pathgroup * pgp;
-	struct path * pp;
-	unsigned int i, j;
-
-	if (!mpp || ro < 0)
+	if (!mpp)
 		return false;
-	if (!has_dm_info(mpp))
-		return true;
-	if (mpp->dmi.read_only == ro)
+	if (!has_dm_info(mpp) &&
+	    dm_get_info(mpp->alias, &mpp->dmi) != DMP_OK) {
+		condlog(2, "%s: failed to get dm info", mpp->alias);
 		return false;
-	if (ro == 1)
-		return true;
-	vector_foreach_slot (mpp->pg, pgp, i) {
-		vector_foreach_slot (pgp->paths, pp, j) {
-			if (sysfs_get_ro(pp) == 1)
-				return false;
-		}
 	}
-	return true;
+	mpp->force_readonly = map_has_ro_path(mpp);
+
+	/*
+	 * The map is read-only and has to stay so. dm has to reopen its
+	 * paths for any read-write reload anyway, so the flag isn't needed.
+	 */
+	if (mpp->force_readonly && mpp->dmi.read_only) {
+		mpp->force_readonly = 0;
+		return false;
+	}
+	/* The current map state doesn't match the paths' state. Update it. */
+	return (mpp->force_readonly != mpp->dmi.read_only);
 }
 
 int resize_map(struct multipath *mpp, unsigned long long size,
@@ -1652,7 +1688,7 @@ out:
 static int
 uev_update_path (struct uevent *uev, struct vectors * vecs)
 {
-	int ro, retval = 0, rc;
+	int retval = 0, rc;
 	struct path * pp;
 	struct config *conf;
 	int needs_reinit = 0;
@@ -1725,22 +1761,22 @@ uev_update_path (struct uevent *uev, struct vectors * vecs)
 			pthread_cleanup_pop(1);
 		}
 
-		ro = uevent_get_disk_ro(uev);
-		if (needs_ro_update(mpp, ro)) {
-			condlog(2, "%s: update path write_protect to '%d' (uevent)", uev->kernel, ro);
+		/*
+		 * The uevent only tells us that something changed. Check the
+		 * paths' ro state in sysfs, which the kernel updates before
+		 * sending the uevent, instead of trusting the uevent's DISK_RO.
+		 */
+		if (needs_ro_update(mpp)) {
+			condlog(2, "%s: update path write_protect to '%d' (uevent)", uev->kernel, mpp->force_readonly);
 
 			if (mpp->wait_for_udev != UDEV_WAIT_DONE)
 				mpp->wait_for_udev = UDEV_WAIT_RELOAD;
 			else {
-				if (ro == 1)
-					pp->mpp->force_readonly = 1;
 				retval = reload_and_sync_map(mpp, vecs);
 				if (retval == 2)
 					condlog(2, "%s: map removed during reload", pp->dev);
-				else {
-					pp->mpp->force_readonly = 0;
+				else
 					condlog(2, "%s: map %s reloaded (retval %d)", uev->kernel, mpp->alias, retval);
-				}
 			}
 		}
 		if (auto_resize != AUTO_RESIZE_NEVER && mpp &&
-- 
2.56.0

Re: [PATCH v6 1/1] multipathd: don't leave maps read-only after write protect is cleared

[Benjamin Marzinski]

On Mon, Nov 02, 2026 at 04:31:50PM -0800, Brian Bunker wrote:
> ev_add_path() sets mpp->force_readonly when a newly added path is
> read-only, but never clears it. If a map is created or reloaded while
> its storage is write protected, the flag stays set. When write protect
> is cleared, the reload that uev_update_path() does for the DISK_RO=0
> uevent uses the stale flag and loads the map read-only, even though all
> of its paths are read-write. uev_update_path() clears the flag after
> that reload, so the map only recovers if another DISK_RO=0 uevent
> reloads it again. If none comes, the map stays read-only, and since
> mpp->dmi.read_only is then 1, needs_ro_update() also ignores later
> DISK_RO=1 uevents for it.
> 
> Fix this by setting force_readonly before each load and clearing it
> once the load has succeeded:
> 
> - uev_update_path() no longer takes the path's ro state from the
>   uevent's DISK_RO. The uevent only says that something changed. The
>   kernel updates the ro attribute before it sends the uevent, so sysfs
>   is current when the uevent is handled, while the uevent itself may
>   have been queued behind later changes. It just calls
>   needs_ro_update(), which reads the ro state of all of the map's paths
>   from sysfs.
> 
> - needs_ro_update() no longer takes an ro value. It sets
>   force_readonly from the ro state of the map's paths, and returns true
>   if that differs from the map's current state. If the map is already
>   read-only and has to stay so, it clears the flag and returns false,
>   since dm has to reopen the paths for any read-write reload anyway. If
>   the map has no dm info, it fetches it, and returns false if that
>   fails, instead of assuming a reload is needed. Since the flag is
>   recomputed on every call, a failed read-only reload is retried on the
>   next uevent without any special handling.
> 
> - ev_add_path() sets it from all of the map's paths rather than only
>   the path being added, and only when it is about to load the map
>   (need_do_map). mpp->paths is up to date there after adopt_paths(),
>   while mpp->pg doesn't contain the new paths yet.
> 
> - update_map(), which does reloads that were delayed while the map was
>   waiting for udev (UDEV_WAIT_RELOAD), recomputes it from the map's
>   paths before setting up the map.
> 
> - dm_addmap_create() clears it as soon as it has read it. A failed
>   create leaves no map behind, and the next create computes the flag
>   again.
> 
> - dm_addmap_reload() clears it once the reload and resume have
>   succeeded. If the reload fails, the map keeps its old table, and the
>   flag stays set so that the next reload tries again.
> 
> Signed-off-by: Brian Bunker <brian@purestorage.com>
Thanks for dealing with all the revisions

Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>