This is a note to let you know that I've just added the patch titled
rbd: retrieve and check lock owner twice before blocklisting
to the 6.1-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary
The filename of the patch is:
rbd-retrieve-and-check-lock-owner-twice-before-blocklisting.patch
and it can be found in the queue-6.1 subdirectory.
If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@xxxxxxxxxxxxxxx> know about it.
>From 588159009d5b7a09c3e5904cffddbe4a4e170301 Mon Sep 17 00:00:00 2001
From: Ilya Dryomov <idryomov@xxxxxxxxx>
Date: Sat, 22 Jul 2023 20:28:08 +0200
Subject: rbd: retrieve and check lock owner twice before blocklisting
From: Ilya Dryomov <idryomov@xxxxxxxxx>
commit 588159009d5b7a09c3e5904cffddbe4a4e170301 upstream.
An attempt to acquire exclusive lock can race with the current lock
owner closing the image:
1. lock is held by client123, rbd_lock() returns -EBUSY
2. get_lock_owner_info() returns client123 instance details
3. client123 closes the image, lock is released
4. find_watcher() returns 0 as there is no matching watcher anymore
5. client123 instance gets erroneously blocklisted
Particularly impacted is mirror snapshot scheduler in snapshot-based
mirroring since it happens to open and close images a lot (images are
opened only for as long as it takes to take the next mirror snapshot,
the same client instance is used for all images).
To reduce the potential for erroneous blocklisting, retrieve the lock
owner again after find_watcher() returns 0. If it's still there, make
sure it matches the previously detected lock owner.
Cc: stable@xxxxxxxxxxxxxxx # f38cb9d9c204: rbd: make get_lock_owner_info() return a single locker or NULL
Cc: stable@xxxxxxxxxxxxxxx # 8ff2c64c9765: rbd: harden get_lock_owner_info() a bit
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: Ilya Dryomov <idryomov@xxxxxxxxx>
Reviewed-by: Dongsheng Yang <dongsheng.yang@xxxxxxxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
---
drivers/block/rbd.c | 25 +++++++++++++++++++++++--
1 file changed, 23 insertions(+), 2 deletions(-)
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -3850,6 +3850,15 @@ static void wake_lock_waiters(struct rbd
list_splice_tail_init(&rbd_dev->acquiring_list, &rbd_dev->running_list);
}
+static bool locker_equal(const struct ceph_locker *lhs,
+ const struct ceph_locker *rhs)
+{
+ return lhs->id.name.type == rhs->id.name.type &&
+ lhs->id.name.num == rhs->id.name.num &&
+ !strcmp(lhs->id.cookie, rhs->id.cookie) &&
+ ceph_addr_equal_no_type(&lhs->info.addr, &rhs->info.addr);
+}
+
static void free_locker(struct ceph_locker *locker)
{
if (locker)
@@ -3970,11 +3979,11 @@ out:
static int rbd_try_lock(struct rbd_device *rbd_dev)
{
struct ceph_client *client = rbd_dev->rbd_client->client;
- struct ceph_locker *locker;
+ struct ceph_locker *locker, *refreshed_locker;
int ret;
for (;;) {
- locker = NULL;
+ locker = refreshed_locker = NULL;
ret = rbd_lock(rbd_dev);
if (ret != -EBUSY)
@@ -3994,6 +4003,16 @@ static int rbd_try_lock(struct rbd_devic
if (ret)
goto out; /* request lock or error */
+ refreshed_locker = get_lock_owner_info(rbd_dev);
+ if (IS_ERR(refreshed_locker)) {
+ ret = PTR_ERR(refreshed_locker);
+ refreshed_locker = NULL;
+ goto out;
+ }
+ if (!refreshed_locker ||
+ !locker_equal(locker, refreshed_locker))
+ goto again;
+
rbd_warn(rbd_dev, "breaking header lock owned by %s%llu",
ENTITY_NAME(locker->id.name));
@@ -4015,10 +4034,12 @@ static int rbd_try_lock(struct rbd_devic
}
again:
+ free_locker(refreshed_locker);
free_locker(locker);
}
out:
+ free_locker(refreshed_locker);
free_locker(locker);
return ret;
}
Patches currently in stable-queue which might be from idryomov@xxxxxxxxx are
queue-6.1/rbd-retrieve-and-check-lock-owner-twice-before-blocklisting.patch
queue-6.1/rbd-make-get_lock_owner_info-return-a-single-locker-or-null.patch
queue-6.1/rbd-harden-get_lock_owner_info-a-bit.patch
queue-6.1/ceph-never-send-metrics-if-disable_send_metrics-is-set.patch
