Skip to contentSkip to conversation

ublk: two fixes

PATCH Applied Ming Lei 2 participants 4 messages Apr 9, 2025

Subsystems (2)
lore
Share & tools
lore

Ming Lei

author
#0 lore raw

v2 cover ublk: two fixes

Hello Jens,

The 1st patch fixes one kernel panic issue when running IO vs. remove devices on ublk zc.

The 2nd one fixes one regression introduced in this cycle.

Thanks,

V2:
- fix 2/2 commit log & fixes tag (Uday)

Ming Lei (2):
ublk: fix handling recovery & reissue in ublk_abort_queue()
ublk: don't fail request for recovery & reissue in case of

    ubq->canceling
 drivers/block/ublk_drv.c | 39 +++++++++++++++++++++++++++++++--------
 1 file changed, 31 insertions(+), 8 deletions(-)
Signature · 1 line
2.47.0
author
#1 lore raw

PATCH v2 1/2 ublk: fix handling recovery & reissue in ublk_abort_queue()

Commit 8284066946e6 ("ublk: grab request reference when the request is handled by userspace") doesn't grab request reference in case of recovery reissue. Then the request can be requeued & re-dispatch & failed when canceling uring command.

If it is one zc request, the request can be freed before io_uring returns the zc buffer back, then cause kernel panic:

[  126.773061] BUG: kernel NULL pointer dereference, address: 00000000000000c8
[  126.773657] #PF: supervisor read access in kernel mode
[  126.774052] #PF: error_code(0x0000) - not-present page
[  126.774455] PGD 0 P4D 0
[  126.774698] Oops: Oops: 0000 [#1] SMP NOPTI
[  126.775034] CPU: 13 UID: 0 PID: 1612 Comm: kworker/u64:55 Not tainted 6.14.0_blk+ #182 PREEMPT(full)
[  126.775676] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-1.fc39 04/01/2014
[  126.776275] Workqueue: iou_exit io_ring_exit_work
[  126.776651] RIP: 0010:ublk_io_release+0x14/0x130 [ublk_drv]

Fixes it by always grabbing request reference for aborting the request.

Fixes 8284066946e6 ("ublk: grab request reference when the request is handled by userspace")
---
 drivers/block/ublk_drv.c | 30 ++++++++++++++++++++++++++----
 1 file changed, 26 insertions(+), 4 deletions(-)
drivers/block/ublk_drv.c +26 -4
index 2fd05c1bd30b..41bed67508f2 100644
@@ -1140,6 +1140,25 @@ static void ublk_complete_rq(struct kref *ref)
 	__ublk_complete_rq(req);
 }
 
+static void ublk_do_fail_rq(struct request *req)
+{
+	struct ublk_queue *ubq = req->mq_hctx->driver_data;
+
+	if (ublk_nosrv_should_reissue_outstanding(ubq->dev))
+		blk_mq_requeue_request(req, false);
+	else
+		__ublk_complete_rq(req);
+}
+
+static void ublk_fail_rq_fn(struct kref *ref)
+{
+	struct ublk_rq_data *data = container_of(ref, struct ublk_rq_data,
+			ref);
+	struct request *req = blk_mq_rq_from_pdu(data);
+
+	ublk_do_fail_rq(req);
+}
+
 /*
  * Since ublk_rq_task_work_cb always fails requests immediately during
  * exiting, __ublk_fail_req() is only called from abort context during
@@ -1153,10 +1172,13 @@ static void __ublk_fail_req(struct ublk_queue *ubq, struct ublk_io *io,
 {
 	WARN_ON_ONCE(io->flags & UBLK_IO_FLAG_ACTIVE);
 
-	if (ublk_nosrv_should_reissue_outstanding(ubq->dev))
-		blk_mq_requeue_request(req, false);
-	else
-		ublk_put_req_ref(ubq, req);
+	if (ublk_need_req_ref(ubq)) {
+		struct ublk_rq_data *data = blk_mq_rq_to_pdu(req);
+
+		kref_put(&data->ref, ublk_fail_rq_fn);
+	} else {
+		ublk_do_fail_rq(req);
+	}
 }
 
 static void ubq_complete_io_cmd(struct ublk_io *io, int res,
Signature · 1 line
2.47.0
author
#2 lore raw

PATCH v2 2/2 ublk: don't fail request for recovery & reissue in case of ubq->canceling

In reply to Ming Lei

ubq->canceling is set with request queue quiesced when io_uring context is exiting. USER_RECOVERY or !RECOVERY_FAIL_IO requires request to be re-queued and re-dispatch after device is recovered.

However commit d796cea7b9f3 ("ublk: implement ->queue_rqs()") still may fail any request in case of ubq->canceling, this way breaks USER_RECOVERY or !RECOVERY_FAIL_IO.

Fix it by calling __ublk_abort_rq() in case of ubq->canceling.

---
 drivers/block/ublk_drv.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)
drivers/block/ublk_drv.c +5 -4
index 41bed67508f2..d6ca2f1097ad 100644
@@ -1371,7 +1371,8 @@ static enum blk_eh_timer_return ublk_timeout(struct request *rq)
 	return BLK_EH_RESET_TIMER;
 }
 
-static blk_status_t ublk_prep_req(struct ublk_queue *ubq, struct request *rq)
+static blk_status_t ublk_prep_req(struct ublk_queue *ubq, struct request *rq,
+				  bool check_cancel)
 {
 	blk_status_t res;
 
@@ -1390,7 +1391,7 @@ static blk_status_t ublk_prep_req(struct ublk_queue *ubq, struct request *rq)
 	if (ublk_nosrv_should_queue_io(ubq) && unlikely(ubq->force_abort))
 		return BLK_STS_IOERR;
 
-	if (unlikely(ubq->canceling))
+	if (check_cancel && unlikely(ubq->canceling))
 		return BLK_STS_IOERR;
 
 	/* fill iod to slot in io cmd buffer */
@@ -1409,7 +1410,7 @@ static blk_status_t ublk_queue_rq(struct blk_mq_hw_ctx *hctx,
 	struct request *rq = bd->rq;
 	blk_status_t res;
 
-	res = ublk_prep_req(ubq, rq);
+	res = ublk_prep_req(ubq, rq, false);
 	if (res != BLK_STS_OK)
 		return res;
 
@@ -1441,7 +1442,7 @@ static void ublk_queue_rqs(struct rq_list *rqlist)
 			ublk_queue_cmd_list(ubq, &submit_list);
 		ubq = this_q;
 
-		if (ublk_prep_req(ubq, req) == BLK_STS_OK)
+		if (ublk_prep_req(ubq, req, true) == BLK_STS_OK)
 			rq_list_add_tail(&submit_list, req);
 		else
 			rq_list_add_tail(&requeue_list, req);
Signature · 1 line
2.47.0
maintainer Applied
#3 lore raw
In reply to Ming Lei
Quoted 8 lines · Ming Lei

The 1st patch fixes one kernel panic issue when running IO vs. remove devices on ublk zc.

The 2nd one fixes one regression introduced in this cycle.

Thanks,

[...]

Applied, thanks!

[1/2] ublk: fix handling recovery & reissue in ublk_abort_queue()

      commit: 6ee6bd5d4fce502a5b5a2ea805e9ff16e6aa890f
[2/2] ublk: don't fail request for recovery & reissue in case of ubq->canceling
      commit: 18461f2a02be04f8bbbe3b37fecfc702e3fa5bc2

Best regards,

Signature · 1 line
Jens Axboe