On 5/19/25 04:53, Shinichiro Kawasaki wrote:
On May 18, 2025 / 13:25, Hannes Reinecke wrote:
On 5/18/25 12:01, Sagi Grimberg wrote:
On 16/05/2025 15:31, Shinichiro Kawasaki wrote:
Hello all,
Using the kernel v6.15-rc6 and the latest blktests (git hash
613b8377e4d3), I
observe the test case nvme/063 fails with tcp transport. Kernel
reported WARN in
blk_mq_unquiesce_queue and KASAN sauf in blk_mq_queue_tag_busy_iter
[1]. The
failure is recreated in stable manner on my test nodes.
The test case script had a bug then this failure was not found until
the bug get
fixed. I tried the kernel v6.15-rc1, and observed the same failure
symptom. This
test case cannot be run with the kernel v6.14, since it does not
have secure
concatenation feature.
Actions for fix will be appreciated.
Hannes, did you encounter this?
No; I would think it's an artifact due to multipath not being enabled.
Shin'ichiro, can you reproduce it with CONFIG_NVME_MULTIPATH on?
I tried both CONFIG_NVME_MULTIPATH on and off, and the failure was recreated
regardless of the config value. Of note is that sometimes it is recreated at
the first test case run, and sometimes it is required to repeat the test case a
few times.
FYI, I attached the kernel config which I used to recreate the failure.
Hmm.
Can you check with this patch (on top of the previous one):
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 55569eb7770b..43d86e8c6df3 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -2197,6 +2197,7 @@ static int nvme_tcp_configure_io_queues(struct
nvme_ctrl *ctrl, bool new)
blk_mq_update_nr_hw_queues(ctrl->tagset,
ctrl->queue_count - 1);
nvme_unfreeze(ctrl);
+ nvme_quiesce_io_queues(ctrl);
}
/*
Thanks.
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@xxxxxxx +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich
