]> git.hungrycats.org Git - linux/commit
SUNRPC: Reject a TLS alert record that is not two octets
authorChuck Lever <cel@kernel.org>
Sat, 8 Aug 2026 15:40:10 +0000 (11:40 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 6 Nov 2026 17:06:08 +0000 (12:06 -0500)
commit560956c8d2479bbf3a215b94391d9ada2063206e
treec9a98ca367577eb970bf7034349f787a08bf6791
parentf44daa50489fede01012c575d9f5e7b11a212fe4
SUNRPC: Reject a TLS alert record that is not two octets

commit d16bebe45272ced4f962d27b9efd8b32a3876bd9 upstream.

tls_alert_recv() reads two octets from the kvec it is handed and does
not check the length (net/handshake/alert.c). svc_tcp_sock_recv_cmsg()
calls it for any positive receive, and the alert[] buffer it supplies
carries no initializer. A one-octet alert body leaves the description
read from uninitialized stack and reported through
trace_tls_alert_recv().

The peer controls that length. Neither tls_rx_msg_size() nor
tls_rx_one_record() enforces the two-octet Alert payload. A TLS 1.3
record carrying only the inner content-type octet decrypts to a
zero-length payload. RFC 8446 Section 5.1 requires a record with an
Alert type to carry exactly one message, so any other length is
malformed.

Require exactly two octets before parsing and return -EBADMSG
otherwise. That closes the transport rather than acting on a partly
uninitialized alert. Gate the path on a control message rather than a
positive count so that a zero-length record reaches the check.

Fixes: bee47cb026e7 ("sunrpc: fix handling of server side tls alerts")
Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/20260808-svcsock-cmsg-fixes-v3-2-62d9a631c880@kernel.org
Signed-off-by: Chuck Lever <cel@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/sunrpc/svcsock.c