Skip to contentSkip to conversation

Patch "drbd: fix request leak introduced by locking/atomic, kref: Kill kref_sub()" has been added to the 4.11-stable tree

PATCH [email protected] 1 participant 1 message May 23, 2017

Subsystems (3)
Stable branch Block layer DRBD driver
lore
Share & tools
lore

[email protected]

author maintainer Applied
#0 lore raw

This is a note to let you know that I've just added the patch titled

drbd: fix request leak introduced by locking/atomic, kref: Kill kref_sub()

to the 4.11-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:
drbd-fix-request-leak-introduced-by-locking-atomic-kref-kill-kref_sub.patch
and it can be found in the queue-4.11 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

Quoted 1 line

From a00ebd1cf12c378a1d4f7a1d6daf1d76c1eaad82 Mon Sep 17 00:00:00 2001

From: Lars Ellenberg <[email protected]>
Date: Thu, 11 May 2017 10:21:46 +0200
Subject: drbd: fix request leak introduced by locking/atomic, kref: Kill kref_sub()

From: Lars Ellenberg <[email protected]>

commit a00ebd1cf12c378a1d4f7a1d6daf1d76c1eaad82 upstream.

When killing kref_sub(), the unconditional additional kref_get() was not properly paired with the necessary kref_put(), causing a leak of struct drbd_requests (~ 224 Bytes) per submitted bio, and breaking DRBD in general, as the destructor of those "drbd_requests" does more than just the mempoll_free().

Fixes bdfafc4ffdd2 ("locking/atomic, kref: Kill kref_sub()")
Signed-off-by Lars Ellenberg <[email protected]>
Signed-off-by Jens Axboe <[email protected]>
Signed-off-by Greg Kroah-Hartman <[email protected]>
---
 drivers/block/drbd/drbd_req.c |   27 +++++++++++++++------------
 1 file changed, 15 insertions(+), 12 deletions(-)
drivers/block/drbd/drbd_req.c +15 -12
@@ -314,24 +314,32 @@ void drbd_req_complete(struct drbd_reque
 }
 
 /* still holds resource->req_lock */
-static int drbd_req_put_completion_ref(struct drbd_request *req, struct bio_and_error *m, int put)
+static void drbd_req_put_completion_ref(struct drbd_request *req, struct bio_and_error *m, int put)
 {
 	struct drbd_device *device = req->device;
 	D_ASSERT(device, m || (req->rq_state & RQ_POSTPONED));
 
+	if (!put)
+		return;
+
 	if (!atomic_sub_and_test(put, &req->completion_ref))
-		return 0;
+		return;
 
 	drbd_req_complete(req, m);
 
+	/* local completion may still come in later,
+	 * we need to keep the req object around. */
+	if (req->rq_state & RQ_LOCAL_ABORTED)
+		return;
+
 	if (req->rq_state & RQ_POSTPONED) {
 		/* don't destroy the req object just yet,
 		 * but queue it for retry */
 		drbd_restart_request(req);
-		return 0;
+		return;
 	}
 
-	return 1;
+	kref_put(&req->kref, drbd_req_destroy);
 }
 
 static void set_if_null_req_next(struct drbd_peer_device *peer_device, struct drbd_request *req)

@@ -518,12 +526,8 @@ static void mod_rq_state(struct drbd_req
 	if (req->i.waiting)
 		wake_up(&device->misc_wait);
 
-	if (c_put) {
-		if (drbd_req_put_completion_ref(req, m, c_put))
-			kref_put(&req->kref, drbd_req_destroy);
-	} else {
-		kref_put(&req->kref, drbd_req_destroy);
-	}
+	drbd_req_put_completion_ref(req, m, c_put);
+	kref_put(&req->kref, drbd_req_destroy);
 }
 
 static void drbd_report_io_error(struct drbd_device *device, struct drbd_request *req)

@@ -1363,8 +1367,7 @@ nodata:
 	}
 
 out:
-	if (drbd_req_put_completion_ref(req, &m, 1))
-		kref_put(&req->kref, drbd_req_destroy);
+	drbd_req_put_completion_ref(req, &m, 1);
 	spin_unlock_irq(&resource->req_lock);
 
 	/* Even though above is a kref_put(), this is safe.

Patches currently in stable-queue which might be from [email protected] are

queue-4.11/drbd-fix-request-leak-introduced-by-locking-atomic-kref-kill-kref_sub.patch