]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
nvme: nvme-fc: move tagset removal to nvme_fc_delete_ctrl()
authorEwan D. Milne <emilne@redhat.com>
Mon, 10 Nov 2025 21:20:00 +0000 (16:20 -0500)
committerKeith Busch <kbusch@kernel.org>
Mon, 17 Nov 2025 17:00:42 +0000 (09:00 -0800)
commitea3442efabd0aa3930c5bab73c3901ef38ef6ac3
tree4add59c9d2dfa57b6a9d49587d08b6ed9d58b1cf
parent6d87cd5335784351280f82c47cc8a657271929c3
nvme: nvme-fc: move tagset removal to nvme_fc_delete_ctrl()

Now target is removed from nvme_fc_ctrl_free() which is the ctrl->ref
release handler. And even admin queue is unquiesced there, this way
is definitely wrong because the ctr->ref is grabbed when submitting
command.

And Marco observed that nvme_fc_ctrl_free() can be called from request
completion code path, and trigger kernel warning since request completes
from softirq context.

Fix the issue by moveing target removal into nvme_fc_delete_ctrl(),
which is also aligned with nvme-tcp and nvme-rdma.

Patch originally proposed by Ming Lei, then modified to move the tagset
removal down to after nvme_fc_delete_association() after further testing.

Cc: Marco Patalano <mpatalan@redhat.com>
Cc: Ewan Milne <emilne@redhat.com>
Cc: James Smart <james.smart@broadcom.com>
Cc: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Cc: stable@vger.kernel.org
Tested-by: Marco Patalano <mpatalan@redhat.com>
Reviewed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
drivers/nvme/host/fc.c