[PATCH] nvme-tcp: fix request PDU size when header digest is enabled

[PATCH] nvme-tcp: fix request PDU size when header digest is enabled

[Christian Aurich Zanettini Martins]

Since commit 61b0ed3be81b ("nvme-tcp: parallelize I/O queue allocation
and startup") the I/O tag set is allocated before the I/O queues.
nvme_tcp_init_request() sizes the PDU with nvme_tcp_hdgst_len(queue),
but queue->hdr_digest is only set later by nvme_tcp_alloc_queue(), so
every I/O PDU is allocated without room for the header digest.

With hdr_digest enabled, the digest is then written 4 bytes past the
PDU, over the header of the next PDU in the page_frag cache, and the
target drops the connection on the first I/O:

  nvmet_tcp: unexpected pdu type 159
  nvmet: ctrl 1 fatal error occurred!

Queues that a later reconnect adds for the first time have the same
problem.

Use the controller option, which is what nvme_tcp_alloc_queue() copies
into the queue.

Fixes: 61b0ed3be81b ("nvme-tcp: parallelize I/O queue allocation and startup")
Signed-off-by: Christian Aurich Zanettini Martins <christian.aurichzm@gmail.com>
---
Based on nvme-7.4 (aaedcff90148).

Tested in QEMU with nvmet-tcp over loopback and parallel I/O: with
hdr_digest the connection failed on the first I/O before the fix; after
it, no errors and the data matched, also with data_digest and with a
reconnect growing the queues from 2 to 4 (attr_qid_max).

 drivers/nvme/host/tcp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index f876375dffd2..7733aa092544 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -566,7 +566,7 @@ static int nvme_tcp_init_request(struct blk_mq_tag_set *set,
 	struct nvme_tcp_cmd_pdu *pdu;
 	int queue_idx = (set == &ctrl->tag_set) ? hctx_idx + 1 : 0;
 	struct nvme_tcp_queue *queue = &ctrl->queues[queue_idx];
-	u8 hdgst = nvme_tcp_hdgst_len(queue);
+	u8 hdgst = ctrl->ctrl.opts->hdr_digest ? NVME_TCP_DIGEST_LENGTH : 0;
 
 	mutex_lock(&queue->pf_cache_lock);
 	req->pdu = page_frag_alloc(&queue->pf_cache,
-- 
2.43.0

Re: [PATCH] nvme-tcp: fix request PDU size when header digest is enabled

[Keith Busch]

On Mon, Oct 26, 2026 at 03:46:21PM -0300, Christian Aurich Zanettini Martins wrote:
> @@ -566,7 +566,7 @@ static int nvme_tcp_init_request(struct blk_mq_tag_set *set,
>  	struct nvme_tcp_cmd_pdu *pdu;
>  	int queue_idx = (set == &ctrl->tag_set) ? hctx_idx + 1 : 0;
>  	struct nvme_tcp_queue *queue = &ctrl->queues[queue_idx];
> -	u8 hdgst = nvme_tcp_hdgst_len(queue);
> +	u8 hdgst = ctrl->ctrl.opts->hdr_digest ? NVME_TCP_DIGEST_LENGTH : 0;

I think everything this function needs should be initialized prior to
this function being called.

>  	mutex_lock(&queue->pf_cache_lock);

See where mutex_init is called for this queue's pf_cache_lock. Can you
move the queue's hdr_digest setting to that location instead?