[PATCH v6 0/5] nvme-tcp: Support receiving KeyUpdate requests

[PATCH v6 0/5] nvme-tcp: Support receiving KeyUpdate requests

[alistair23]

From: Alistair Francis <alistair.francis@wdc.com>

The TLS 1.3 specification allows the TLS client or server to send a
KeyUpdate. This is generally used when the sequence is about to
overflow or after a certain amount of bytes have been encrypted.

The TLS spec doesn't mandate the conditions though, so a KeyUpdate
can be sent by the TLS client or server at any time. This includes
when running NVMe-OF over a TLS 1.3 connection.

As such Linux should be able to handle a KeyUpdate event, as the
other NVMe side could initiate a KeyUpdate.

Upcoming WD NVMe-TCP hardware controllers implement TLS support
and send KeyUpdate requests.

This series builds on top of the existing TLS EKEYEXPIRED work,
which already detects a KeyUpdate request. We can now pass that
information up to the NVMe layer (target and host) and then pass
it up to userspace.

Userspace (ktls-utils) will need to save the connection state
in the keyring during the initial handshake. The kernel then
provides the key serial back to userspace when handling a
KeyUpdate. Userspace can use this to restore the connection
information and then update the keys, this final process
is similar to the initial handshake.

This series depends on the recvmsg() kernel patch:
https://lore.kernel.org/linux-nvme/2cbe1350-0bf5-4487-be33-1d317cb73acf@suse.de/T/#mf56283228ae6c93e37dfbf1c0f6263910217cd80

ktls-utils (tlshd) userspace patches are available at:
https://lore.kernel.org/kernel-tls-handshake/CAKmqyKNpFhPtM8HAkgRMKQA8_N7AgoeqaSTe2=0spPnb+Oz2ng@mail.gmail.com/T/#mb277f5c998282666d0f41cc02f4abf516fcc4e9c

Link: https://datatracker.ietf.org/doc/html/rfc8446#section-4.6.3

Based-on: 2cbe1350-0bf5-4487-be33-1d317cb73acf@suse.de

v6:
 - Don't free handshake request on completion (handshake_sk_destruct_req())
 - Add handshake_req_keyupdate() which reuses existing handshake request
   for a KeyUpdate
 - Other small improvements and tidyups
v5:
 - Cleanup code flow for nvme-tcp
 - When using recvmsg in the host code first check for MSG_CTRUNC
   in the msg_flags returned from recvmsg() and use that to determine
   if it's a control message
 - Drop clientkeyupdaterequest and serverkeyupdaterequest
v4:
 - Don't stop the keep-alive timer
 - Remove any support for sending a KeyUpdate
 - Add tls_client_keyupdate_psk()' and 'tls_server_keyupdate_psk()'
 - Code cleanups
 - Change order of patches
v3:
 - Rebase on the recvmsg() workflow patch
 - Add debugfs support for the host
 - Don't cancel an ongoing request
 - Ensure a request is destructed on completion
v2:
 - Change "key-serial" to "session-id"
 - Fix reported build failures
 - Drop tls_clear_err() function
 - Stop keep alive timer during KeyUpdate
 - Drop handshake message decoding in the NVMe layer

Alistair Francis (5):
  net/handshake: Store the key serial number on completion
  net/handshake: Define handshake_req_keyupdate
  net/handshake: Support KeyUpdate message types
  nvme-tcp: Support KeyUpdate
  nvmet-tcp: Support KeyUpdate

 Documentation/netlink/specs/handshake.yaml |  20 +-
 Documentation/networking/tls-handshake.rst |   1 +
 drivers/nvme/host/tcp.c                    | 111 ++++++++---
 drivers/nvme/target/tcp.c                  | 213 ++++++++++++++-------
 include/net/handshake.h                    |  11 +-
 include/uapi/linux/handshake.h             |  12 ++
 net/handshake/genl.c                       |   5 +-
 net/handshake/handshake.h                  |   2 +
 net/handshake/request.c                    |  95 +++++++++
 net/handshake/tlshd.c                      |  97 +++++++++-
 net/sunrpc/svcsock.c                       |   4 +-
 net/sunrpc/xprtsock.c                      |   4 +-
 12 files changed, 480 insertions(+), 95 deletions(-)

-- 
2.51.1

[PATCH v6 1/5] net/handshake: Store the key serial number on completion

[alistair23]

From: Alistair Francis <alistair.francis@wdc.com>

Allow userspace to include a key serial number when completing a
handshake with the HANDSHAKE_CMD_DONE command.

We then store this serial number and will provide it back to userspace
in the future. This allows userspace to save data to the keyring and
then restore that data later.

This will be used to support the TLS KeyUpdate operation, as now
userspace can resume information about a established session.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Hannes Reincke <hare@suse.de>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
---
v6:
 - Add the "ta_" and "th_" prefixs
v5:
 - Change name to "handshake session ID"
v4:
 - No change
v3:
 - No change
v2:
 - Change "key-serial" to "session-id"

 Documentation/netlink/specs/handshake.yaml |  4 ++++
 Documentation/networking/tls-handshake.rst |  1 +
 drivers/nvme/host/tcp.c                    |  3 ++-
 drivers/nvme/target/tcp.c                  |  3 ++-
 include/net/handshake.h                    |  5 ++++-
 include/uapi/linux/handshake.h             |  1 +
 net/handshake/genl.c                       |  5 +++--
 net/handshake/tlshd.c                      | 15 +++++++++++++--
 net/sunrpc/svcsock.c                       |  4 +++-
 net/sunrpc/xprtsock.c                      |  4 +++-
 10 files changed, 36 insertions(+), 9 deletions(-)

diff --git a/Documentation/netlink/specs/handshake.yaml b/Documentation/netlink/specs/handshake.yaml
index 95c3fade7a8d..a273bc74d26f 100644
--- a/Documentation/netlink/specs/handshake.yaml
+++ b/Documentation/netlink/specs/handshake.yaml
@@ -87,6 +87,9 @@ attribute-sets:
         name: remote-auth
         type: u32
         multi-attr: true
+      -
+        name: session-id
+        type: u32
 
 operations:
   list:
@@ -123,6 +126,7 @@ operations:
             - status
             - sockfd
             - remote-auth
+            - session-id
 
 mcast-groups:
   list:
diff --git a/Documentation/networking/tls-handshake.rst b/Documentation/networking/tls-handshake.rst
index 6f5ea1646a47..c58b3c8b16c1 100644
--- a/Documentation/networking/tls-handshake.rst
+++ b/Documentation/networking/tls-handshake.rst
@@ -60,6 +60,7 @@ fills in a structure that contains the parameters of the request:
         key_serial_t    ta_my_privkey;
         unsigned int    ta_num_peerids;
         key_serial_t    ta_my_peerids[5];
+        key_serial_t    ta_handshake_session_id;
   };
 
 The @ta_sock field references an open and connected socket. The consumer
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 9058ea64b89c..024d02248831 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1694,7 +1694,8 @@ static void nvme_tcp_set_queue_io_cpu(struct nvme_tcp_queue *queue)
 		qid, queue->io_cpu);
 }
 
-static void nvme_tcp_tls_done(void *data, int status, key_serial_t pskid)
+static void nvme_tcp_tls_done(void *data, int status, key_serial_t pskid,
+			      key_serial_t handshake_session_id)
 {
 	struct nvme_tcp_queue *queue = data;
 	struct nvme_tcp_ctrl *ctrl = queue->ctrl;
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 470bf37e5a63..7f8516892359 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1780,7 +1780,8 @@ static int nvmet_tcp_tls_key_lookup(struct nvmet_tcp_queue *queue,
 }
 
 static void nvmet_tcp_tls_handshake_done(void *data, int status,
-					 key_serial_t peerid)
+					 key_serial_t peerid,
+					 key_serial_t handshake_session_id)
 {
 	struct nvmet_tcp_queue *queue = data;
 
diff --git a/include/net/handshake.h b/include/net/handshake.h
index 8ebd4f9ed26e..d9b2411d5523 100644
--- a/include/net/handshake.h
+++ b/include/net/handshake.h
@@ -15,10 +15,12 @@ enum {
 	TLS_NO_PEERID = 0,
 	TLS_NO_CERT = 0,
 	TLS_NO_PRIVKEY = 0,
+	TLS_NO_SESSION_ID = 0,
 };
 
 typedef void	(*tls_done_func_t)(void *data, int status,
-				   key_serial_t peerid);
+				   key_serial_t peerid,
+				   key_serial_t handshake_session_id);
 
 struct tls_handshake_args {
 	struct socket		*ta_sock;
@@ -31,6 +33,7 @@ struct tls_handshake_args {
 	key_serial_t		ta_my_privkey;
 	unsigned int		ta_num_peerids;
 	key_serial_t		ta_my_peerids[5];
+	key_serial_t		ta_handshake_session_id;
 };
 
 int tls_client_hello_anon(const struct tls_handshake_args *args, gfp_t flags);
diff --git a/include/uapi/linux/handshake.h b/include/uapi/linux/handshake.h
index 662e7de46c54..b68ffbaa5f31 100644
--- a/include/uapi/linux/handshake.h
+++ b/include/uapi/linux/handshake.h
@@ -55,6 +55,7 @@ enum {
 	HANDSHAKE_A_DONE_STATUS = 1,
 	HANDSHAKE_A_DONE_SOCKFD,
 	HANDSHAKE_A_DONE_REMOTE_AUTH,
+	HANDSHAKE_A_DONE_SESSION_ID,
 
 	__HANDSHAKE_A_DONE_MAX,
 	HANDSHAKE_A_DONE_MAX = (__HANDSHAKE_A_DONE_MAX - 1)
diff --git a/net/handshake/genl.c b/net/handshake/genl.c
index f55d14d7b726..6cdce7e5dbc0 100644
--- a/net/handshake/genl.c
+++ b/net/handshake/genl.c
@@ -16,10 +16,11 @@ static const struct nla_policy handshake_accept_nl_policy[HANDSHAKE_A_ACCEPT_HAN
 };
 
 /* HANDSHAKE_CMD_DONE - do */
-static const struct nla_policy handshake_done_nl_policy[HANDSHAKE_A_DONE_REMOTE_AUTH + 1] = {
+static const struct nla_policy handshake_done_nl_policy[HANDSHAKE_A_DONE_SESSION_ID + 1] = {
 	[HANDSHAKE_A_DONE_STATUS] = { .type = NLA_U32, },
 	[HANDSHAKE_A_DONE_SOCKFD] = { .type = NLA_S32, },
 	[HANDSHAKE_A_DONE_REMOTE_AUTH] = { .type = NLA_U32, },
+	[HANDSHAKE_A_DONE_SESSION_ID] = { .type = NLA_U32, },
 };
 
 /* Ops table for handshake */
@@ -35,7 +36,7 @@ static const struct genl_split_ops handshake_nl_ops[] = {
 		.cmd		= HANDSHAKE_CMD_DONE,
 		.doit		= handshake_nl_done_doit,
 		.policy		= handshake_done_nl_policy,
-		.maxattr	= HANDSHAKE_A_DONE_REMOTE_AUTH,
+		.maxattr	= HANDSHAKE_A_DONE_SESSION_ID,
 		.flags		= GENL_CMD_CAP_DO,
 	},
 };
diff --git a/net/handshake/tlshd.c b/net/handshake/tlshd.c
index 8f9532a15f43..e72f45bdc226 100644
--- a/net/handshake/tlshd.c
+++ b/net/handshake/tlshd.c
@@ -26,7 +26,8 @@
 
 struct tls_handshake_req {
 	void			(*th_consumer_done)(void *data, int status,
-						    key_serial_t peerid);
+						    key_serial_t peerid,
+						    key_serial_t handshake_session_id);
 	void			*th_consumer_data;
 
 	int			th_type;
@@ -39,6 +40,8 @@ struct tls_handshake_req {
 
 	unsigned int		th_num_peerids;
 	key_serial_t		th_peerid[5];
+
+	key_serial_t		th_handshake_session_id;
 };
 
 static struct tls_handshake_req *
@@ -55,6 +58,7 @@ tls_handshake_req_init(struct handshake_req *req,
 	treq->th_num_peerids = 0;
 	treq->th_certificate = TLS_NO_CERT;
 	treq->th_privkey = TLS_NO_PRIVKEY;
+	treq->th_handshake_session_id = TLS_NO_SESSION_ID;
 	return treq;
 }
 
@@ -83,6 +87,13 @@ static void tls_handshake_remote_peerids(struct tls_handshake_req *treq,
 		if (i >= treq->th_num_peerids)
 			break;
 	}
+
+	nla_for_each_attr(nla, head, len, rem) {
+		if (nla_type(nla) == HANDSHAKE_A_DONE_SESSION_ID) {
+			treq->th_handshake_session_id = nla_get_u32(nla);
+			break;
+		}
+	}
 }
 
 /**
@@ -105,7 +116,7 @@ static void tls_handshake_done(struct handshake_req *req,
 		set_bit(HANDSHAKE_F_REQ_SESSION, &req->hr_flags);
 
 	treq->th_consumer_done(treq->th_consumer_data, -status,
-			       treq->th_peerid[0]);
+			       treq->th_peerid[0], treq->th_handshake_session_id);
 }
 
 #if IS_ENABLED(CONFIG_KEYS)
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 7b90abc5cf0e..2401b4c757f6 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -444,13 +444,15 @@ static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
  * @data: address of xprt to wake
  * @status: status of handshake
  * @peerid: serial number of key containing the remote peer's identity
+ * @handshake_session_id: serial number of the userspace session ID
  *
  * If a security policy is specified as an export option, we don't
  * have a specific export here to check. So we set a "TLS session
  * is present" flag on the xprt and let an upper layer enforce local
  * security policy.
  */
-static void svc_tcp_handshake_done(void *data, int status, key_serial_t peerid)
+static void svc_tcp_handshake_done(void *data, int status, key_serial_t peerid,
+				   key_serial_t handshake_session_id)
 {
 	struct svc_xprt *xprt = data;
 	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 3aa987e7f072..5c6e7543f293 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2589,9 +2589,11 @@ static int xs_tcp_tls_finish_connecting(struct rpc_xprt *lower_xprt,
  * @data: address of xprt to wake
  * @status: status of handshake
  * @peerid: serial number of key containing the remote's identity
+ * @handshake_session_id: serial number of the userspace session ID
  *
  */
-static void xs_tls_handshake_done(void *data, int status, key_serial_t peerid)
+static void xs_tls_handshake_done(void *data, int status, key_serial_t peerid,
+				  key_serial_t handshake_session_id)
 {
 	struct rpc_xprt *lower_xprt = data;
 	struct sock_xprt *lower_transport =
-- 
2.51.1

[PATCH v6 2/5] net/handshake: Define handshake_req_keyupdate

[alistair23]

From: Alistair Francis <alistair.francis@wdc.com>

Add a new handshake_req_keyupdate() function which is similar to the
existing handshake_req_submit().

The new handshake_req_keyupdate() does not add the request to the hash
table (unlike handshake_req_submit()) but instead uses the existing
request from the initial handshake.

During the initial handshake handshake_req_submit() will add the request
to the hash table. The request will not be removed from the hash table
unless the socket is closed (reference count hits zero).

After the initial handshake handshake_req_keyupdate() can be used to re-use
the existing request in the hash table to trigger a KeyUpdate with
userspace.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
---
v6:
 - New patch

 net/handshake/handshake.h |  2 +
 net/handshake/request.c   | 95 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 97 insertions(+)

diff --git a/net/handshake/handshake.h b/net/handshake/handshake.h
index a48163765a7a..04feacd1e21d 100644
--- a/net/handshake/handshake.h
+++ b/net/handshake/handshake.h
@@ -84,6 +84,8 @@ void handshake_req_hash_destroy(void);
 void *handshake_req_private(struct handshake_req *req);
 struct handshake_req *handshake_req_hash_lookup(struct sock *sk);
 struct handshake_req *handshake_req_next(struct handshake_net *hn, int class);
+int handshake_req_keyupdate(struct socket *sock, struct handshake_req *req,
+			 gfp_t flags);
 int handshake_req_submit(struct socket *sock, struct handshake_req *req,
 			 gfp_t flags);
 void handshake_complete(struct handshake_req *req, unsigned int status,
diff --git a/net/handshake/request.c b/net/handshake/request.c
index 274d2c89b6b2..916caab88fe0 100644
--- a/net/handshake/request.c
+++ b/net/handshake/request.c
@@ -196,6 +196,101 @@ struct handshake_req *handshake_req_next(struct handshake_net *hn, int class)
 }
 EXPORT_SYMBOL_IF_KUNIT(handshake_req_next);
 
+/**
+ * handshake_req_keyupdate - Submit a KeyUpdate request
+ * @sock: open socket on which to perform the handshake
+ * @req: handshake arguments, this must already be allocated and exist
+ * in the hash table, which happens as part of handshake_req_submit()
+ * @flags: memory allocation flags
+ *
+ * Return values:
+ *   %0: Request queued
+ *   %-EINVAL: Invalid argument
+ *   %-EBUSY: A handshake is already under way for this socket
+ *   %-ESRCH: No handshake agent is available
+ *   %-EAGAIN: Too many pending handshake requests
+ *   %-ENOMEM: Failed to allocate memory
+ *   %-EMSGSIZE: Failed to construct notification message
+ *   %-EOPNOTSUPP: Handshake module not initialized
+ *
+ * A zero return value from handshake_req_submit() means that
+ * exactly one subsequent completion callback is guaranteed.
+ *
+ * A negative return value from handshake_req_submit() means that
+ * no completion callback will be done and that @req has been
+ * destroyed.
+ */
+int handshake_req_keyupdate(struct socket *sock, struct handshake_req *req,
+			    gfp_t flags)
+{
+	struct handshake_net *hn;
+	struct net *net;
+	struct handshake_req *req_lookup;
+	int ret;
+
+	if (!sock || !req || !sock->file) {
+		kfree(req);
+		return -EINVAL;
+	}
+
+	req->hr_sk = sock->sk;
+	if (!req->hr_sk) {
+		kfree(req);
+		return -EINVAL;
+	}
+	req->hr_odestruct = req->hr_sk->sk_destruct;
+	req->hr_sk->sk_destruct = handshake_sk_destruct;
+
+	ret = -EOPNOTSUPP;
+	net = sock_net(req->hr_sk);
+	hn = handshake_pernet(net);
+	if (!hn)
+		goto out_err;
+
+	ret = -EAGAIN;
+	if (READ_ONCE(hn->hn_pending) >= hn->hn_pending_max)
+		goto out_err;
+
+	spin_lock(&hn->hn_lock);
+	ret = -EOPNOTSUPP;
+	if (test_bit(HANDSHAKE_F_NET_DRAINING, &hn->hn_flags))
+		goto out_unlock;
+	ret = -EBUSY;
+
+	req_lookup = handshake_req_hash_lookup(sock->sk);
+	if (!req_lookup)
+		goto out_unlock;
+
+	if (req_lookup != req)
+		goto out_unlock;
+	if (!__add_pending_locked(hn, req))
+		goto out_unlock;
+	spin_unlock(&hn->hn_lock);
+
+	test_and_clear_bit(HANDSHAKE_F_REQ_COMPLETED, &req->hr_flags);
+
+	ret = handshake_genl_notify(net, req->hr_proto, flags);
+	if (ret) {
+		trace_handshake_notify_err(net, req, req->hr_sk, ret);
+		if (remove_pending(hn, req))
+			goto out_err;
+	}
+
+	/* Prevent socket release while a handshake request is pending */
+	sock_hold(req->hr_sk);
+
+	trace_handshake_submit(net, req, req->hr_sk);
+	return 0;
+
+out_unlock:
+	spin_unlock(&hn->hn_lock);
+out_err:
+	trace_handshake_submit_err(net, req, req->hr_sk, ret);
+	handshake_req_destroy(req);
+	return ret;
+}
+EXPORT_SYMBOL(handshake_req_keyupdate);
+
 /**
  * handshake_req_submit - Submit a handshake request
  * @sock: open socket on which to perform the handshake
-- 
2.51.1

[PATCH v6 3/5] net/handshake: Support KeyUpdate message types

[alistair23]

From: Alistair Francis <alistair.francis@wdc.com>

When reporting the msg-type to userspace let's also support reporting
KeyUpdate events. This supports reporting a client/server event and if
the other side requested a KeyUpdateRequest.

Link: https://datatracker.ietf.org/doc/html/rfc8446#section-4.6.3
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
---
v6:
 - Init th_key_update_request field to 0
v5:
 - Drop clientkeyupdaterequest and serverkeyupdaterequest
v4:
 - Don't overload existing functions, instead create new ones
v3:
 - Fixup yamllint and kernel-doc failures

 Documentation/netlink/specs/handshake.yaml | 16 ++++-
 drivers/nvme/host/tcp.c                    | 15 +++-
 drivers/nvme/target/tcp.c                  | 10 ++-
 include/net/handshake.h                    |  6 ++
 include/uapi/linux/handshake.h             | 11 +++
 net/handshake/tlshd.c                      | 84 +++++++++++++++++++++-
 6 files changed, 134 insertions(+), 8 deletions(-)

diff --git a/Documentation/netlink/specs/handshake.yaml b/Documentation/netlink/specs/handshake.yaml
index a273bc74d26f..2f77216c8ddf 100644
--- a/Documentation/netlink/specs/handshake.yaml
+++ b/Documentation/netlink/specs/handshake.yaml
@@ -21,12 +21,18 @@ definitions:
     type: enum
     name: msg-type
     value-start: 0
-    entries: [unspec, clienthello, serverhello]
+    entries: [unspec, clienthello, serverhello, clientkeyupdate,
+              serverkeyupdate]
   -
     type: enum
     name: auth
     value-start: 0
     entries: [unspec, unauth, psk, x509]
+  -
+    type: enum
+    name: key-update-type
+    value-start: 0
+    entries: [unspec, send, received, received_request_update]
 
 attribute-sets:
   -
@@ -74,6 +80,13 @@ attribute-sets:
       -
         name: keyring
         type: u32
+      -
+        name: key-update-request
+        type: u32
+        enum: key-update-type
+      -
+        name: session-id
+        type: u32
   -
     name: done
     attributes:
@@ -116,6 +129,7 @@ operations:
             - certificate
             - peername
             - keyring
+            - session-id
     -
       name: done
       doc: Handler reports handshake completion
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 024d02248831..4797a4532b0d 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -20,6 +20,7 @@
 #include <linux/iov_iter.h>
 #include <net/busy_poll.h>
 #include <trace/events/sock.h>
+#include <uapi/linux/handshake.h>
 
 #include "nvme.h"
 #include "fabrics.h"
@@ -206,6 +207,10 @@ static struct workqueue_struct *nvme_tcp_wq;
 static const struct blk_mq_ops nvme_tcp_mq_ops;
 static const struct blk_mq_ops nvme_tcp_admin_mq_ops;
 static int nvme_tcp_try_send(struct nvme_tcp_queue *queue);
+static int nvme_tcp_start_tls(struct nvme_ctrl *nctrl,
+			      struct nvme_tcp_queue *queue,
+			      key_serial_t pskid,
+			      enum handshake_key_update_type keyupdate);
 
 static inline struct nvme_tcp_ctrl *to_tcp_ctrl(struct nvme_ctrl *ctrl)
 {
@@ -1729,7 +1734,8 @@ static void nvme_tcp_tls_done(void *data, int status, key_serial_t pskid,
 
 static int nvme_tcp_start_tls(struct nvme_ctrl *nctrl,
 			      struct nvme_tcp_queue *queue,
-			      key_serial_t pskid)
+			      key_serial_t pskid,
+			      enum handshake_key_update_type keyupdate)
 {
 	int qid = nvme_tcp_queue_id(queue);
 	int ret;
@@ -1751,7 +1757,10 @@ static int nvme_tcp_start_tls(struct nvme_ctrl *nctrl,
 	args.ta_timeout_ms = tls_handshake_timeout * 1000;
 	queue->tls_err = -EOPNOTSUPP;
 	init_completion(&queue->tls_complete);
-	ret = tls_client_hello_psk(&args, GFP_KERNEL);
+	if (keyupdate == HANDSHAKE_KEY_UPDATE_TYPE_UNSPEC)
+		ret = tls_client_hello_psk(&args, GFP_KERNEL);
+	else
+		ret = tls_client_keyupdate_psk(&args, GFP_KERNEL, keyupdate);
 	if (ret) {
 		dev_err(nctrl->device, "queue %d: failed to start TLS: %d\n",
 			qid, ret);
@@ -1901,7 +1910,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl, int qid,
 
 	/* If PSKs are configured try to start TLS */
 	if (nvme_tcp_tls_configured(nctrl) && pskid) {
-		ret = nvme_tcp_start_tls(nctrl, queue, pskid);
+		ret = nvme_tcp_start_tls(nctrl, queue, pskid, HANDSHAKE_KEY_UPDATE_TYPE_UNSPEC);
 		if (ret)
 			goto err_init_connect;
 	}
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 7f8516892359..818efdeccef1 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1833,7 +1833,8 @@ static void nvmet_tcp_tls_handshake_timeout(struct work_struct *w)
 	kref_put(&queue->kref, nvmet_tcp_release_queue);
 }
 
-static int nvmet_tcp_tls_handshake(struct nvmet_tcp_queue *queue)
+static int nvmet_tcp_tls_handshake(struct nvmet_tcp_queue *queue,
+				   enum handshake_key_update_type keyupdate)
 {
 	int ret = -EOPNOTSUPP;
 	struct tls_handshake_args args;
@@ -1852,7 +1853,10 @@ static int nvmet_tcp_tls_handshake(struct nvmet_tcp_queue *queue)
 	args.ta_keyring = key_serial(queue->port->nport->keyring);
 	args.ta_timeout_ms = tls_handshake_timeout * 1000;
 
-	ret = tls_server_hello_psk(&args, GFP_KERNEL);
+	if (keyupdate == HANDSHAKE_KEY_UPDATE_TYPE_UNSPEC)
+		ret = tls_server_hello_psk(&args, GFP_KERNEL);
+	else
+		ret = tls_server_keyupdate_psk(&args, GFP_KERNEL, keyupdate);
 	if (ret) {
 		kref_put(&queue->kref, nvmet_tcp_release_queue);
 		pr_err("failed to start TLS, err=%d\n", ret);
@@ -1934,7 +1938,7 @@ static void nvmet_tcp_alloc_queue(struct nvmet_tcp_port *port,
 		sk->sk_data_ready = port->data_ready;
 		write_unlock_bh(&sk->sk_callback_lock);
 		if (!nvmet_tcp_try_peek_pdu(queue)) {
-			if (!nvmet_tcp_tls_handshake(queue))
+			if (!nvmet_tcp_tls_handshake(queue, HANDSHAKE_KEY_UPDATE_TYPE_UNSPEC))
 				return;
 			/* TLS handshake failed, terminate the connection */
 			goto out_destroy_sq;
diff --git a/include/net/handshake.h b/include/net/handshake.h
index d9b2411d5523..54fb101202d2 100644
--- a/include/net/handshake.h
+++ b/include/net/handshake.h
@@ -10,6 +10,8 @@
 #ifndef _NET_HANDSHAKE_H
 #define _NET_HANDSHAKE_H
 
+#include <uapi/linux/handshake.h>
+
 enum {
 	TLS_NO_KEYRING = 0,
 	TLS_NO_PEERID = 0,
@@ -39,8 +41,12 @@ struct tls_handshake_args {
 int tls_client_hello_anon(const struct tls_handshake_args *args, gfp_t flags);
 int tls_client_hello_x509(const struct tls_handshake_args *args, gfp_t flags);
 int tls_client_hello_psk(const struct tls_handshake_args *args, gfp_t flags);
+int tls_client_keyupdate_psk(const struct tls_handshake_args *args, gfp_t flags,
+			     enum handshake_key_update_type keyupdate);
 int tls_server_hello_x509(const struct tls_handshake_args *args, gfp_t flags);
 int tls_server_hello_psk(const struct tls_handshake_args *args, gfp_t flags);
+int tls_server_keyupdate_psk(const struct tls_handshake_args *args, gfp_t flags,
+			     enum handshake_key_update_type keyupdate);
 
 bool tls_handshake_cancel(struct sock *sk);
 void tls_handshake_close(struct socket *sock);
diff --git a/include/uapi/linux/handshake.h b/include/uapi/linux/handshake.h
index b68ffbaa5f31..483815a064f0 100644
--- a/include/uapi/linux/handshake.h
+++ b/include/uapi/linux/handshake.h
@@ -19,6 +19,8 @@ enum handshake_msg_type {
 	HANDSHAKE_MSG_TYPE_UNSPEC,
 	HANDSHAKE_MSG_TYPE_CLIENTHELLO,
 	HANDSHAKE_MSG_TYPE_SERVERHELLO,
+	HANDSHAKE_MSG_TYPE_CLIENTKEYUPDATE,
+	HANDSHAKE_MSG_TYPE_SERVERKEYUPDATE,
 };
 
 enum handshake_auth {
@@ -28,6 +30,13 @@ enum handshake_auth {
 	HANDSHAKE_AUTH_X509,
 };
 
+enum handshake_key_update_type {
+	HANDSHAKE_KEY_UPDATE_TYPE_UNSPEC,
+	HANDSHAKE_KEY_UPDATE_TYPE_SEND,
+	HANDSHAKE_KEY_UPDATE_TYPE_RECEIVED,
+	HANDSHAKE_KEY_UPDATE_TYPE_RECEIVED_REQUEST_UPDATE,
+};
+
 enum {
 	HANDSHAKE_A_X509_CERT = 1,
 	HANDSHAKE_A_X509_PRIVKEY,
@@ -46,6 +55,8 @@ enum {
 	HANDSHAKE_A_ACCEPT_CERTIFICATE,
 	HANDSHAKE_A_ACCEPT_PEERNAME,
 	HANDSHAKE_A_ACCEPT_KEYRING,
+	HANDSHAKE_A_ACCEPT_KEY_UPDATE_REQUEST,
+	HANDSHAKE_A_ACCEPT_SESSION_ID,
 
 	__HANDSHAKE_A_ACCEPT_MAX,
 	HANDSHAKE_A_ACCEPT_MAX = (__HANDSHAKE_A_ACCEPT_MAX - 1)
diff --git a/net/handshake/tlshd.c b/net/handshake/tlshd.c
index e72f45bdc226..d102211e9d77 100644
--- a/net/handshake/tlshd.c
+++ b/net/handshake/tlshd.c
@@ -41,6 +41,7 @@ struct tls_handshake_req {
 	unsigned int		th_num_peerids;
 	key_serial_t		th_peerid[5];
 
+	unsigned int		th_key_update_request;
 	key_serial_t		th_handshake_session_id;
 };
 
@@ -58,7 +59,9 @@ tls_handshake_req_init(struct handshake_req *req,
 	treq->th_num_peerids = 0;
 	treq->th_certificate = TLS_NO_CERT;
 	treq->th_privkey = TLS_NO_PRIVKEY;
-	treq->th_handshake_session_id = TLS_NO_SESSION_ID;
+	treq->th_key_update_request = 0;
+	treq->th_handshake_session_id = args->ta_handshake_session_id;
+
 	return treq;
 }
 
@@ -265,6 +268,16 @@ static int tls_handshake_accept(struct handshake_req *req,
 		break;
 	}
 
+	ret = nla_put_u32(msg, HANDSHAKE_A_ACCEPT_SESSION_ID,
+			  treq->th_handshake_session_id);
+	if (ret < 0)
+		goto out_cancel;
+
+	ret = nla_put_u32(msg, HANDSHAKE_A_ACCEPT_KEY_UPDATE_REQUEST,
+			  treq->th_key_update_request);
+	if (ret < 0)
+		goto out_cancel;
+
 	genlmsg_end(msg, hdr);
 	return genlmsg_reply(msg, info);
 
@@ -373,6 +386,44 @@ int tls_client_hello_psk(const struct tls_handshake_args *args, gfp_t flags)
 }
 EXPORT_SYMBOL(tls_client_hello_psk);
 
+/**
+ * tls_client_keyupdate_psk - request a PSK-based TLS handshake on a socket
+ * @args: socket and handshake parameters for this request
+ * @flags: memory allocation control flags
+ * @keyupdate: specifies the type of KeyUpdate operation
+ *
+ * Return values:
+ *   %0: Handshake request enqueue; ->done will be called when complete
+ *   %-EINVAL: Wrong number of local peer IDs
+ *   %-ESRCH: No user agent is available
+ *   %-ENOMEM: Memory allocation failed
+ */
+int tls_client_keyupdate_psk(const struct tls_handshake_args *args, gfp_t flags,
+			     enum handshake_key_update_type keyupdate)
+{
+	struct tls_handshake_req *treq;
+	struct handshake_req *req;
+	unsigned int i;
+
+	if (!args->ta_num_peerids ||
+	    args->ta_num_peerids > ARRAY_SIZE(treq->th_peerid))
+		return -EINVAL;
+
+	req = handshake_req_hash_lookup(args->ta_sock->sk);
+	if (!req)
+		return -ENOMEM;
+	treq = tls_handshake_req_init(req, args);
+	treq->th_type = HANDSHAKE_MSG_TYPE_CLIENTKEYUPDATE;
+	treq->th_key_update_request = keyupdate;
+	treq->th_auth_mode = HANDSHAKE_AUTH_PSK;
+	treq->th_num_peerids = args->ta_num_peerids;
+	for (i = 0; i < args->ta_num_peerids; i++)
+		treq->th_peerid[i] = args->ta_my_peerids[i];
+
+	return handshake_req_keyupdate(args->ta_sock, req, flags);
+}
+EXPORT_SYMBOL(tls_client_keyupdate_psk);
+
 /**
  * tls_server_hello_x509 - request a server TLS handshake on a socket
  * @args: socket and handshake parameters for this request
@@ -429,6 +480,37 @@ int tls_server_hello_psk(const struct tls_handshake_args *args, gfp_t flags)
 }
 EXPORT_SYMBOL(tls_server_hello_psk);
 
+/**
+ * tls_server_keyupdate_psk - request a server TLS KeyUpdate on a socket
+ * @args: socket and handshake parameters for this request
+ * @flags: memory allocation control flags
+ * @keyupdate: specifies the type of KeyUpdate operation
+ *
+ * Return values:
+ *   %0: Handshake request enqueue; ->done will be called when complete
+ *   %-ESRCH: No user agent is available
+ *   %-ENOMEM: Memory allocation failed
+ */
+int tls_server_keyupdate_psk(const struct tls_handshake_args *args, gfp_t flags,
+			     enum handshake_key_update_type keyupdate)
+{
+	struct tls_handshake_req *treq;
+	struct handshake_req *req;
+
+	req = handshake_req_hash_lookup(args->ta_sock->sk);
+	if (!req)
+		return -ENOMEM;
+	treq = tls_handshake_req_init(req, args);
+	treq->th_type = HANDSHAKE_MSG_TYPE_SERVERKEYUPDATE;
+	treq->th_key_update_request = keyupdate;
+	treq->th_auth_mode = HANDSHAKE_AUTH_PSK;
+	treq->th_num_peerids = 1;
+	treq->th_peerid[0] = args->ta_my_peerids[0];
+
+	return handshake_req_keyupdate(args->ta_sock, req, flags);
+}
+EXPORT_SYMBOL(tls_server_keyupdate_psk);
+
 /**
  * tls_handshake_cancel - cancel a pending handshake
  * @sk: socket on which there is an ongoing handshake
-- 
2.51.1

[PATCH v6 4/5] nvme-tcp: Support KeyUpdate

[alistair23]

From: Alistair Francis <alistair.francis@wdc.com>

If the nvme_tcp_try_send() or nvme_tcp_try_recv() functions return
EKEYEXPIRED then the underlying TLS keys need to be updated. This occurs
on an KeyUpdate event as described in RFC8446
https://datatracker.ietf.org/doc/html/rfc8446#section-4.6.3.

If the NVMe Target (TLS server) initiates a KeyUpdate this patch will
allow the NVMe layer to process the KeyUpdate request and forward the
request to userspace. Userspace must then update the key to keep the
connection alive.

This patch allows us to handle the NVMe target sending a KeyUpdate
request without aborting the connection. At this time we don't support
initiating a KeyUpdate.

Based-on: 2cbe1350-0bf5-4487-be33-1d317cb73acf@suse.de
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
---
v6:
 - Don't use `struct nvme_tcp_hdr` to determine TLS_HANDSHAKE_KEYUPDATE,
   instead look at the cmsg fields.
 - Don't flush async_event_work
v5:
 - Cleanup code flow
 - Check for MSG_CTRUNC in the msg_flags return from recvmsg
   and use that to determine if it's a control message
v4:
 - Remove all support for initiating KeyUpdate
 - Don't call cancel_work() when updating keys
v3:
 - Don't cancel existing handshake requests
v2:
 - Don't change the state
 - Use a helper function for KeyUpdates
 - Continue sending in nvme_tcp_send_all() after a KeyUpdate
 - Remove command message using recvmsg

 drivers/nvme/host/tcp.c | 93 +++++++++++++++++++++++++++++++++--------
 1 file changed, 75 insertions(+), 18 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 4797a4532b0d..c0966929f97b 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -172,6 +172,7 @@ struct nvme_tcp_queue {
 	bool			tls_enabled;
 	u32			rcv_crc;
 	u32			snd_crc;
+	key_serial_t		handshake_session_id;
 	__le32			exp_ddgst;
 	__le32			recv_ddgst;
 	struct completion       tls_complete;
@@ -858,7 +859,10 @@ static void nvme_tcp_handle_c2h_term(struct nvme_tcp_queue *queue,
 static int nvme_tcp_recvmsg_pdu(struct nvme_tcp_queue *queue)
 {
 	char *pdu = queue->pdu;
+	char cbuf[CMSG_LEN(sizeof(char))] = {};
 	struct msghdr msg = {
+		.msg_control = cbuf,
+		.msg_controllen = sizeof(cbuf),
 		.msg_flags = MSG_DONTWAIT,
 	};
 	struct kvec iov = {
@@ -873,6 +877,12 @@ static int nvme_tcp_recvmsg_pdu(struct nvme_tcp_queue *queue)
 	if (ret <= 0)
 		return ret;
 
+	if (cbuf[CMSG_LEN(sizeof(char)) - 1] == TLS_RECORD_TYPE_HANDSHAKE &&
+	    pdu[queue->pdu_offset] == TLS_HANDSHAKE_KEYUPDATE) {
+		dev_err(queue->ctrl->ctrl.device, "KeyUpdate message\n");
+		return 1;
+	}
+
 	queue->pdu_remaining -= ret;
 	queue->pdu_offset += ret;
 	if (queue->pdu_remaining)
@@ -944,14 +954,14 @@ static int nvme_tcp_recvmsg_data(struct nvme_tcp_queue *queue)
 	struct request *rq =
 		nvme_cid_to_rq(nvme_tcp_tagset(queue), pdu->command_id);
 	struct nvme_tcp_request *req = blk_mq_rq_to_pdu(rq);
+	char cbuf[CMSG_LEN(sizeof(char))] = {};
+	struct msghdr msg;
+	int ret;
 
 	if (nvme_tcp_recv_state(queue) != NVME_TCP_RECV_DATA)
 		return 0;
 
 	while (queue->data_remaining) {
-		struct msghdr msg;
-		int ret;
-
 		if (!iov_iter_count(&req->iter)) {
 			req->curr_bio = req->curr_bio->bi_next;
 
@@ -975,11 +985,8 @@ static int nvme_tcp_recvmsg_data(struct nvme_tcp_queue *queue)
 		msg.msg_flags = MSG_DONTWAIT;
 
 		ret = sock_recvmsg(queue->sock, &msg, msg.msg_flags);
-		if (ret < 0) {
-			dev_err(queue->ctrl->ctrl.device,
-				"queue %d failed to receive request %#x data",
-				nvme_tcp_queue_id(queue), rq->tag);
-			return ret;
+		if (unlikely(ret < 0)) {
+			goto msg_control;
 		}
 		if (queue->data_digest)
 			nvme_tcp_ddgst_calc(req, &queue->rcv_crc, ret);
@@ -1002,6 +1009,29 @@ static int nvme_tcp_recvmsg_data(struct nvme_tcp_queue *queue)
 		}
 	}
 
+	return 0;
+
+msg_control:
+	/*
+	 * If MSG_CTRUNC is set, it's a control message,
+	 * so let's read the control message.
+	 */
+	if (msg.msg_flags & MSG_CTRUNC) {
+		memset(&msg, 0, sizeof(msg));
+		msg.msg_flags = MSG_DONTWAIT;
+		msg.msg_control = cbuf;
+		msg.msg_controllen = sizeof(cbuf);
+
+		ret = sock_recvmsg(queue->sock, &msg, msg.msg_flags);
+	}
+
+	if (ret < 0) {
+		dev_err(queue->ctrl->ctrl.device,
+			"queue %d failed to receive request %#x data, %d",
+			nvme_tcp_queue_id(queue), rq->tag, ret);
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -1384,15 +1414,37 @@ static int nvme_tcp_try_recvmsg(struct nvme_tcp_queue *queue)
 		}
 	} while (result >= 0);
 
-	if (result < 0 && result != -EAGAIN) {
-		dev_err(queue->ctrl->ctrl.device,
-			"receive failed:  %d\n", result);
-		queue->rd_enabled = false;
-		nvme_tcp_error_recovery(&queue->ctrl->ctrl);
-	} else if (result == -EAGAIN)
-		result = 0;
+	if (result < 0) {
+		if (result != -EKEYEXPIRED && result != -EAGAIN) {
+			dev_err(queue->ctrl->ctrl.device,
+				"receive failed:  %d\n", result);
+			queue->rd_enabled = false;
+			nvme_tcp_error_recovery(&queue->ctrl->ctrl);
+		}
+		return result;
+	}
 
-	return result < 0 ? result : (queue->nr_cqe = nr_cqe);
+	queue->nr_cqe = nr_cqe;
+	return nr_cqe;
+}
+
+static void update_tls_keys(struct nvme_tcp_queue *queue)
+{
+	int qid = nvme_tcp_queue_id(queue);
+	int ret;
+
+	dev_dbg(queue->ctrl->ctrl.device,
+		"updating key for queue %d\n", qid);
+
+	ret = nvme_tcp_start_tls(&(queue->ctrl->ctrl),
+				 queue, queue->ctrl->ctrl.tls_pskid,
+				 HANDSHAKE_KEY_UPDATE_TYPE_RECEIVED);
+
+	if (ret < 0) {
+		dev_err(queue->ctrl->ctrl.device,
+			"failed to update the keys %d\n", ret);
+		nvme_tcp_fail_request(queue->request);
+	}
 }
 
 static void nvme_tcp_io_work(struct work_struct *w)
@@ -1417,8 +1469,11 @@ static void nvme_tcp_io_work(struct work_struct *w)
 		result = nvme_tcp_try_recvmsg(queue);
 		if (result > 0)
 			pending = true;
-		else if (unlikely(result < 0))
-			return;
+		else if (unlikely(result < 0)) {
+			if (result == -EKEYEXPIRED)
+				update_tls_keys(queue);
+			break;
+		}
 
 		/* did we get some space after spending time in recv? */
 		if (nvme_tcp_queue_has_pending(queue) &&
@@ -1726,6 +1781,7 @@ static void nvme_tcp_tls_done(void *data, int status, key_serial_t pskid,
 			ctrl->ctrl.tls_pskid = key_serial(tls_key);
 		key_put(tls_key);
 		queue->tls_err = 0;
+		queue->handshake_session_id = handshake_session_id;
 	}
 
 out_complete:
@@ -1755,6 +1811,7 @@ static int nvme_tcp_start_tls(struct nvme_ctrl *nctrl,
 		keyring = key_serial(nctrl->opts->keyring);
 	args.ta_keyring = keyring;
 	args.ta_timeout_ms = tls_handshake_timeout * 1000;
+	args.ta_handshake_session_id = queue->handshake_session_id;
 	queue->tls_err = -EOPNOTSUPP;
 	init_completion(&queue->tls_complete);
 	if (keyupdate == HANDSHAKE_KEY_UPDATE_TYPE_UNSPEC)
-- 
2.51.1

[PATCH v6 5/5] nvmet-tcp: Support KeyUpdate

[alistair23]

From: Alistair Francis <alistair.francis@wdc.com>

If the nvmet_tcp_try_recv() function return EKEYEXPIRED or if we receive
a KeyUpdate handshake type then the underlying TLS keys need to be
updated.

If the NVMe Host (TLS client) initiates a KeyUpdate this patch will
allow the NVMe layer to process the KeyUpdate request and forward the
request to userspace. Userspace must then update the key to keep the
connection alive.

This patch allows us to handle the NVMe host sending a KeyUpdate
request without aborting the connection. At this time we don't support
initiating a KeyUpdate.

Link: https://datatracker.ietf.org/doc/html/rfc8446#section-4.6.3
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
---
v6:
 - Simplify the nvmet_tls_key_expired() check
v5:
 - No change
v4:
 - Restructure code to avoid #ifdefs and forward declarations
 - Use a helper function for checking -EKEYEXPIRED
 - Remove all support for initiating KeyUpdate
 - Use helper function for restoring callbacks
v3:
 - Use a write lock for sk_user_data
 - Fix build with CONFIG_NVME_TARGET_TCP_TLS disabled
 - Remove unused variable
v2:
 - Use a helper function for KeyUpdates
 - Ensure keep alive timer is stopped
 - Wait for TLS KeyUpdate to complete

 drivers/nvme/target/tcp.c | 200 ++++++++++++++++++++++++++------------
 1 file changed, 139 insertions(+), 61 deletions(-)

diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 818efdeccef1..0458a9691cbc 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -175,6 +175,7 @@ struct nvmet_tcp_queue {
 
 	/* TLS state */
 	key_serial_t		tls_pskid;
+	key_serial_t		handshake_session_id;
 	struct delayed_work	tls_handshake_tmo_work;
 
 	unsigned long           poll_end;
@@ -186,6 +187,8 @@ struct nvmet_tcp_queue {
 	struct sockaddr_storage	sockaddr_peer;
 	struct work_struct	release_work;
 
+	struct completion       tls_complete;
+
 	int			idx;
 	struct list_head	queue_list;
 
@@ -214,6 +217,10 @@ static struct workqueue_struct *nvmet_tcp_wq;
 static const struct nvmet_fabrics_ops nvmet_tcp_ops;
 static void nvmet_tcp_free_cmd(struct nvmet_tcp_cmd *c);
 static void nvmet_tcp_free_cmd_buffers(struct nvmet_tcp_cmd *cmd);
+#ifdef CONFIG_NVME_TARGET_TCP_TLS
+static int nvmet_tcp_tls_handshake(struct nvmet_tcp_queue *queue,
+				   enum handshake_key_update_type keyupdate);
+#endif
 
 static inline u16 nvmet_tcp_cmd_tag(struct nvmet_tcp_queue *queue,
 		struct nvmet_tcp_cmd *cmd)
@@ -832,6 +839,20 @@ static int nvmet_tcp_try_send_one(struct nvmet_tcp_queue *queue,
 	return 1;
 }
 
+#ifdef CONFIG_NVME_TARGET_TCP_TLS
+static bool nvmet_tls_key_expired(struct nvmet_tcp_queue *queue, int ret)
+{
+	return ret == -EKEYEXPIRED &&
+		queue->state != NVMET_TCP_Q_DISCONNECTING &&
+		queue->state != NVMET_TCP_Q_TLS_HANDSHAKE;
+}
+#else
+static bool nvmet_tls_key_expired(struct nvmet_tcp_queue *queue, int ret)
+{
+	return false;
+}
+#endif
+
 static int nvmet_tcp_try_send(struct nvmet_tcp_queue *queue,
 		int budget, int *sends)
 {
@@ -1106,6 +1127,103 @@ static inline bool nvmet_tcp_pdu_valid(u8 type)
 	return false;
 }
 
+static void nvmet_tcp_release_queue(struct kref *kref)
+{
+	struct nvmet_tcp_queue *queue =
+		container_of(kref, struct nvmet_tcp_queue, kref);
+
+	WARN_ON(queue->state != NVMET_TCP_Q_DISCONNECTING);
+	queue_work(nvmet_wq, &queue->release_work);
+}
+
+static void nvmet_tcp_schedule_release_queue(struct nvmet_tcp_queue *queue)
+{
+	spin_lock_bh(&queue->state_lock);
+	if (queue->state == NVMET_TCP_Q_TLS_HANDSHAKE) {
+		/* Socket closed during handshake */
+		tls_handshake_cancel(queue->sock->sk);
+	}
+	if (queue->state != NVMET_TCP_Q_DISCONNECTING) {
+		queue->state = NVMET_TCP_Q_DISCONNECTING;
+		kref_put(&queue->kref, nvmet_tcp_release_queue);
+	}
+	spin_unlock_bh(&queue->state_lock);
+}
+
+static void nvmet_tcp_restore_socket_callbacks(struct nvmet_tcp_queue *queue)
+{
+	struct socket *sock = queue->sock;
+
+	if (!queue->state_change)
+		return;
+
+	write_lock_bh(&sock->sk->sk_callback_lock);
+	sock->sk->sk_data_ready =  queue->data_ready;
+	sock->sk->sk_state_change = queue->state_change;
+	sock->sk->sk_write_space = queue->write_space;
+	sock->sk->sk_user_data = NULL;
+	write_unlock_bh(&sock->sk->sk_callback_lock);
+}
+
+#ifdef CONFIG_NVME_TARGET_TCP_TLS
+static void nvmet_tcp_tls_handshake_timeout(struct work_struct *w)
+{
+	struct nvmet_tcp_queue *queue = container_of(to_delayed_work(w),
+			struct nvmet_tcp_queue, tls_handshake_tmo_work);
+
+	pr_warn("queue %d: TLS handshake timeout\n", queue->idx);
+	/*
+	 * If tls_handshake_cancel() fails we've lost the race with
+	 * nvmet_tcp_tls_handshake_done() */
+	if (!tls_handshake_cancel(queue->sock->sk))
+		return;
+	spin_lock_bh(&queue->state_lock);
+	if (WARN_ON(queue->state != NVMET_TCP_Q_TLS_HANDSHAKE)) {
+		spin_unlock_bh(&queue->state_lock);
+		return;
+	}
+	queue->state = NVMET_TCP_Q_FAILED;
+	spin_unlock_bh(&queue->state_lock);
+	nvmet_tcp_schedule_release_queue(queue);
+	kref_put(&queue->kref, nvmet_tcp_release_queue);
+}
+
+static int update_tls_keys(struct nvmet_tcp_queue *queue)
+{
+	int ret;
+
+	cancel_work(&queue->io_work);
+	queue->state = NVMET_TCP_Q_TLS_HANDSHAKE;
+
+	nvmet_tcp_restore_socket_callbacks(queue);
+
+	INIT_DELAYED_WORK(&queue->tls_handshake_tmo_work,
+			  nvmet_tcp_tls_handshake_timeout);
+
+	ret = nvmet_tcp_tls_handshake(queue, HANDSHAKE_KEY_UPDATE_TYPE_RECEIVED);
+
+	if (ret < 0)
+		return ret;
+
+	ret = wait_for_completion_interruptible_timeout(&queue->tls_complete,
+							10 * HZ);
+
+	if (ret <= 0) {
+		tls_handshake_cancel(queue->sock->sk);
+		return ret;
+	}
+
+	queue->state = NVMET_TCP_Q_LIVE;
+
+	return 0;
+}
+#else
+static int update_tls_keys(struct nvmet_tcp_queue *queue)
+{
+	return -EPFNOSUPPORT;
+}
+#endif
+
 static int nvmet_tcp_tls_record_ok(struct nvmet_tcp_queue *queue,
 		struct msghdr *msg, char *cbuf)
 {
@@ -1131,6 +1249,9 @@ static int nvmet_tcp_tls_record_ok(struct nvmet_tcp_queue *queue,
 			ret = -EAGAIN;
 		}
 		break;
+	case TLS_RECORD_TYPE_HANDSHAKE:
+		ret = -EAGAIN;
+		break;
 	default:
 		/* discard this record type */
 		pr_err("queue %d: TLS record %d unhandled\n",
@@ -1340,6 +1461,8 @@ static int nvmet_tcp_try_recv(struct nvmet_tcp_queue *queue,
 	for (i = 0; i < budget; i++) {
 		ret = nvmet_tcp_try_recv_one(queue);
 		if (unlikely(ret < 0)) {
+			if (nvmet_tls_key_expired(queue, ret))
+					goto done;
 			nvmet_tcp_socket_error(queue, ret);
 			goto done;
 		} else if (ret == 0) {
@@ -1351,29 +1474,6 @@ static int nvmet_tcp_try_recv(struct nvmet_tcp_queue *queue,
 	return ret;
 }
 
-static void nvmet_tcp_release_queue(struct kref *kref)
-{
-	struct nvmet_tcp_queue *queue =
-		container_of(kref, struct nvmet_tcp_queue, kref);
-
-	WARN_ON(queue->state != NVMET_TCP_Q_DISCONNECTING);
-	queue_work(nvmet_wq, &queue->release_work);
-}
-
-static void nvmet_tcp_schedule_release_queue(struct nvmet_tcp_queue *queue)
-{
-	spin_lock_bh(&queue->state_lock);
-	if (queue->state == NVMET_TCP_Q_TLS_HANDSHAKE) {
-		/* Socket closed during handshake */
-		tls_handshake_cancel(queue->sock->sk);
-	}
-	if (queue->state != NVMET_TCP_Q_DISCONNECTING) {
-		queue->state = NVMET_TCP_Q_DISCONNECTING;
-		kref_put(&queue->kref, nvmet_tcp_release_queue);
-	}
-	spin_unlock_bh(&queue->state_lock);
-}
-
 static inline void nvmet_tcp_arm_queue_deadline(struct nvmet_tcp_queue *queue)
 {
 	queue->poll_end = jiffies + usecs_to_jiffies(idle_poll_period_usecs);
@@ -1404,8 +1504,12 @@ static void nvmet_tcp_io_work(struct work_struct *w)
 		ret = nvmet_tcp_try_recv(queue, NVMET_TCP_RECV_BUDGET, &ops);
 		if (ret > 0)
 			pending = true;
-		else if (ret < 0)
+		else if (ret < 0) {
+			if (ret == -EKEYEXPIRED)
+				break;
+
 			return;
+		}
 
 		ret = nvmet_tcp_try_send(queue, NVMET_TCP_SEND_BUDGET, &ops);
 		if (ret > 0)
@@ -1415,6 +1519,11 @@ static void nvmet_tcp_io_work(struct work_struct *w)
 
 	} while (pending && ops < NVMET_TCP_IO_WORK_BUDGET);
 
+	if (ret == -EKEYEXPIRED) {
+		update_tls_keys(queue);
+		pending = true;
+	}
+
 	/*
 	 * Requeue the worker if idle deadline period is in progress or any
 	 * ops activity was recorded during the do-while loop above.
@@ -1517,21 +1626,6 @@ static void nvmet_tcp_free_cmds(struct nvmet_tcp_queue *queue)
 	kfree(cmds);
 }
 
-static void nvmet_tcp_restore_socket_callbacks(struct nvmet_tcp_queue *queue)
-{
-	struct socket *sock = queue->sock;
-
-	if (!queue->state_change)
-		return;
-
-	write_lock_bh(&sock->sk->sk_callback_lock);
-	sock->sk->sk_data_ready =  queue->data_ready;
-	sock->sk->sk_state_change = queue->state_change;
-	sock->sk->sk_write_space = queue->write_space;
-	sock->sk->sk_user_data = NULL;
-	write_unlock_bh(&sock->sk->sk_callback_lock);
-}
-
 static void nvmet_tcp_uninit_data_in_cmds(struct nvmet_tcp_queue *queue)
 {
 	struct nvmet_tcp_cmd *cmd = queue->cmds;
@@ -1794,6 +1888,7 @@ static void nvmet_tcp_tls_handshake_done(void *data, int status,
 	}
 	if (!status) {
 		queue->tls_pskid = peerid;
+		queue->handshake_session_id = handshake_session_id;
 		queue->state = NVMET_TCP_Q_CONNECTING;
 	} else
 		queue->state = NVMET_TCP_Q_FAILED;
@@ -1809,28 +1904,7 @@ static void nvmet_tcp_tls_handshake_done(void *data, int status,
 	else
 		nvmet_tcp_set_queue_sock(queue);
 	kref_put(&queue->kref, nvmet_tcp_release_queue);
-}
-
-static void nvmet_tcp_tls_handshake_timeout(struct work_struct *w)
-{
-	struct nvmet_tcp_queue *queue = container_of(to_delayed_work(w),
-			struct nvmet_tcp_queue, tls_handshake_tmo_work);
-
-	pr_warn("queue %d: TLS handshake timeout\n", queue->idx);
-	/*
-	 * If tls_handshake_cancel() fails we've lost the race with
-	 * nvmet_tcp_tls_handshake_done() */
-	if (!tls_handshake_cancel(queue->sock->sk))
-		return;
-	spin_lock_bh(&queue->state_lock);
-	if (WARN_ON(queue->state != NVMET_TCP_Q_TLS_HANDSHAKE)) {
-		spin_unlock_bh(&queue->state_lock);
-		return;
-	}
-	queue->state = NVMET_TCP_Q_FAILED;
-	spin_unlock_bh(&queue->state_lock);
-	nvmet_tcp_schedule_release_queue(queue);
-	kref_put(&queue->kref, nvmet_tcp_release_queue);
+	complete(&queue->tls_complete);
 }
 
 static int nvmet_tcp_tls_handshake(struct nvmet_tcp_queue *queue,
@@ -1852,11 +1926,15 @@ static int nvmet_tcp_tls_handshake(struct nvmet_tcp_queue *queue,
 	args.ta_data = queue;
 	args.ta_keyring = key_serial(queue->port->nport->keyring);
 	args.ta_timeout_ms = tls_handshake_timeout * 1000;
+	args.ta_handshake_session_id = queue->handshake_session_id;
+
+	init_completion(&queue->tls_complete);
 
 	if (keyupdate == HANDSHAKE_KEY_UPDATE_TYPE_UNSPEC)
 		ret = tls_server_hello_psk(&args, GFP_KERNEL);
 	else
 		ret = tls_server_keyupdate_psk(&args, GFP_KERNEL, keyupdate);
+
 	if (ret) {
 		kref_put(&queue->kref, nvmet_tcp_release_queue);
 		pr_err("failed to start TLS, err=%d\n", ret);
-- 
2.51.1

Re: [PATCH v6 4/5] nvme-tcp: Support KeyUpdate

[Christoph Hellwig]

On Tue, Dec 02, 2025 at 11:34:28AM +1000, alistair23@gmail.com wrote:
> +		if (unlikely(ret < 0)) {
> +			goto msg_control;
>  		}

This should drop the superfluous braces.

Otherwise looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v6 5/5] nvmet-tcp: Support KeyUpdate

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v6 2/5] net/handshake: Define handshake_req_keyupdate

[Chuck Lever]

On Mon, Dec 1, 2025, at 8:34 PM, alistair23@gmail.com wrote:
> From: Alistair Francis <alistair.francis@wdc.com>
>
> Add a new handshake_req_keyupdate() function which is similar to the
> existing handshake_req_submit().
>
> The new handshake_req_keyupdate() does not add the request to the hash
> table (unlike handshake_req_submit()) but instead uses the existing
> request from the initial handshake.
>
> During the initial handshake handshake_req_submit() will add the request
> to the hash table. The request will not be removed from the hash table
> unless the socket is closed (reference count hits zero).
>
> After the initial handshake handshake_req_keyupdate() can be used to re-use
> the existing request in the hash table to trigger a KeyUpdate with
> userspace.
>
> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
> ---
> v6:
>  - New patch
>
>  net/handshake/handshake.h |  2 +
>  net/handshake/request.c   | 95 +++++++++++++++++++++++++++++++++++++++
>  2 files changed, 97 insertions(+)
>
> diff --git a/net/handshake/handshake.h b/net/handshake/handshake.h
> index a48163765a7a..04feacd1e21d 100644
> --- a/net/handshake/handshake.h
> +++ b/net/handshake/handshake.h
> @@ -84,6 +84,8 @@ void handshake_req_hash_destroy(void);
>  void *handshake_req_private(struct handshake_req *req);
>  struct handshake_req *handshake_req_hash_lookup(struct sock *sk);
>  struct handshake_req *handshake_req_next(struct handshake_net *hn, int 
> class);
> +int handshake_req_keyupdate(struct socket *sock, struct handshake_req 
> *req,
> +			 gfp_t flags);
>  int handshake_req_submit(struct socket *sock, struct handshake_req 
> *req,
>  			 gfp_t flags);
>  void handshake_complete(struct handshake_req *req, unsigned int status,
> diff --git a/net/handshake/request.c b/net/handshake/request.c
> index 274d2c89b6b2..916caab88fe0 100644
> --- a/net/handshake/request.c
> +++ b/net/handshake/request.c
> @@ -196,6 +196,101 @@ struct handshake_req *handshake_req_next(struct 
> handshake_net *hn, int class)
>  }
>  EXPORT_SYMBOL_IF_KUNIT(handshake_req_next);
> 
> +/**
> + * handshake_req_keyupdate - Submit a KeyUpdate request
> + * @sock: open socket on which to perform the handshake
> + * @req: handshake arguments, this must already be allocated and exist
> + * in the hash table, which happens as part of handshake_req_submit()
> + * @flags: memory allocation flags
> + *
> + * Return values:
> + *   %0: Request queued
> + *   %-EINVAL: Invalid argument
> + *   %-EBUSY: A handshake is already under way for this socket

Or, an initial handshake hasn't been done yet. Perhaps that
deserves its own errno value...

Thanks for considering this idea!


> + *   %-ESRCH: No handshake agent is available
> + *   %-EAGAIN: Too many pending handshake requests
> + *   %-ENOMEM: Failed to allocate memory
> + *   %-EMSGSIZE: Failed to construct notification message
> + *   %-EOPNOTSUPP: Handshake module not initialized
> + *
> + * A zero return value from handshake_req_submit() means that
> + * exactly one subsequent completion callback is guaranteed.
> + *
> + * A negative return value from handshake_req_submit() means that
> + * no completion callback will be done and that @req has been
> + * destroyed.
> + */
> +int handshake_req_keyupdate(struct socket *sock, struct handshake_req *req,
> +			    gfp_t flags)
> +{
> +	struct handshake_net *hn;
> +	struct net *net;
> +	struct handshake_req *req_lookup;
> +	int ret;
> +
> +	if (!sock || !req || !sock->file) {
> +		kfree(req);
> +		return -EINVAL;
> +	}
> +
> +	req->hr_sk = sock->sk;
> +	if (!req->hr_sk) {
> +		kfree(req);
> +		return -EINVAL;
> +	}
> +	req->hr_odestruct = req->hr_sk->sk_destruct;
> +	req->hr_sk->sk_destruct = handshake_sk_destruct;
> +
> +	ret = -EOPNOTSUPP;
> +	net = sock_net(req->hr_sk);
> +	hn = handshake_pernet(net);
> +	if (!hn)
> +		goto out_err;
> +
> +	ret = -EAGAIN;
> +	if (READ_ONCE(hn->hn_pending) >= hn->hn_pending_max)
> +		goto out_err;
> +
> +	spin_lock(&hn->hn_lock);
> +	ret = -EOPNOTSUPP;
> +	if (test_bit(HANDSHAKE_F_NET_DRAINING, &hn->hn_flags))
> +		goto out_unlock;
> +	ret = -EBUSY;
> +
> +	req_lookup = handshake_req_hash_lookup(sock->sk);
> +	if (!req_lookup)
> +		goto out_unlock;
> +
> +	if (req_lookup != req)
> +		goto out_unlock;
> +	if (!__add_pending_locked(hn, req))
> +		goto out_unlock;
> +	spin_unlock(&hn->hn_lock);
> +
> +	test_and_clear_bit(HANDSHAKE_F_REQ_COMPLETED, &req->hr_flags);
> +
> +	ret = handshake_genl_notify(net, req->hr_proto, flags);
> +	if (ret) {
> +		trace_handshake_notify_err(net, req, req->hr_sk, ret);
> +		if (remove_pending(hn, req))
> +			goto out_err;
> +	}
> +
> +	/* Prevent socket release while a handshake request is pending */
> +	sock_hold(req->hr_sk);
> +
> +	trace_handshake_submit(net, req, req->hr_sk);
> +	return 0;
> +
> +out_unlock:
> +	spin_unlock(&hn->hn_lock);
> +out_err:
> +	trace_handshake_submit_err(net, req, req->hr_sk, ret);
> +	handshake_req_destroy(req);
> +	return ret;
> +}
> +EXPORT_SYMBOL(handshake_req_keyupdate);
> +
>  /**
>   * handshake_req_submit - Submit a handshake request
>   * @sock: open socket on which to perform the handshake
> -- 
> 2.51.1

-- 
Chuck Lever

Re: [PATCH v6 2/5] net/handshake: Define handshake_req_keyupdate

[Hannes Reinecke]

On 12/2/25 02:34, alistair23@gmail.com wrote:
> From: Alistair Francis <alistair.francis@wdc.com>
> 
> Add a new handshake_req_keyupdate() function which is similar to the
> existing handshake_req_submit().
> 
> The new handshake_req_keyupdate() does not add the request to the hash
> table (unlike handshake_req_submit()) but instead uses the existing
> request from the initial handshake.
> 
> During the initial handshake handshake_req_submit() will add the request
> to the hash table. The request will not be removed from the hash table
> unless the socket is closed (reference count hits zero).
> 
> After the initial handshake handshake_req_keyupdate() can be used to re-use
> the existing request in the hash table to trigger a KeyUpdate with
> userspace.
> 
> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
> ---
> v6:
>   - New patch
> 
>   net/handshake/handshake.h |  2 +
>   net/handshake/request.c   | 95 +++++++++++++++++++++++++++++++++++++++
>   2 files changed, 97 insertions(+)
> 
Reviewed-by: Hannes Reinecke <hare@suse.de>

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                  Kernel Storage Architect
hare@suse.de                                +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich

Re: [PATCH v6 1/5] net/handshake: Store the key serial number on completion

[Jakub Kicinski]

On Tue,  2 Dec 2025 11:34:25 +1000 alistair23@gmail.com wrote:
>  net/handshake/genl.c                       |  5 +++--
>  net/handshake/tlshd.c                      | 15 +++++++++++++--

We don't process -next patches during the merge window in networking.
If someone thinks this is urgent do what you gotta do, otherwise please
repost this once merge window is over.