]> git.hungrycats.org Git - linux/commit
SUNRPC: timeout and cancel TLS handshake with -ETIMEDOUT
authorBenjamin Coddington <bcodding@redhat.com>
Fri, 15 Nov 2024 13:59:36 +0000 (08:59 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 9 Dec 2024 09:33:00 +0000 (10:33 +0100)
commit931be6b73830bc3d358fe8e896221a22f882fdf7
tree76babf0e54991e57f27a965c8c457c3923cd3faf
parent3811172e8c98ceebd12fe526ca6cb37a1263c964
SUNRPC: timeout and cancel TLS handshake with -ETIMEDOUT

[ Upstream commit d7bdd849ef1b681da03ac05ca0957b2cbe2d24b6 ]

We've noticed a situation where an unstable TCP connection can cause the
TLS handshake to timeout waiting for userspace to complete it.  When this
happens, we don't want to return from xs_tls_handshake_sync() with zero, as
this will cause the upper xprt to be set CONNECTED, and subsequent attempts
to transmit will be returned with -EPIPE.  The sunrpc machine does not
recover from this situation and will spin attempting to transmit.

The return value of tls_handshake_cancel() can be used to detect a race
with completion:

 * tls_handshake_cancel - cancel a pending handshake
 * Return values:
 *   %true - Uncompleted handshake request was canceled
 *   %false - Handshake request already completed or not found

If true, we do not want the upper xprt to be connected, so return
-ETIMEDOUT.  If false, its possible the handshake request was lost and
that may be the reason for our timeout.  Again we do not want the upper
xprt to be connected, so return -ETIMEDOUT.

Ensure that we alway return an error from xs_tls_handshake_sync() if we
call tls_handshake_cancel().

Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Fixes: 75eb6af7acdf ("SUNRPC: Add a TCP-with-TLS RPC transport class")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
net/sunrpc/xprtsock.c