]> asedeno.scripts.mit.edu Git - linux.git/commitdiff
net/tls: fix poll ignoring partially copied records
authorJakub Kicinski <jakub.kicinski@netronome.com>
Thu, 4 Jul 2019 21:50:36 +0000 (14:50 -0700)
committerDavid S. Miller <davem@davemloft.net>
Sun, 7 Jul 2019 21:11:44 +0000 (14:11 -0700)
David reports that RPC applications which use epoll() occasionally
get stuck, and that TLS ULP causes the kernel to not wake applications,
even though read() will return data.

This is indeed true. The ctx->rx_list which holds partially copied
records is not consulted when deciding whether socket is readable.

Note that SO_RCVLOWAT with epoll() is and has always been broken for
kernel TLS. We'd need to parse all records from the TCP layer, instead
of just the first one.

Fixes: 692d7b5d1f91 ("tls: Fix recvmsg() to be able to peek across multiple records")
Reported-by: David Beckett <david.beckett@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/tls/tls_sw.c

index 455a782c7658f6de2d23a7f8978c63eda819d552..e2385183526e460586cdef4bf59c0fc1329d1942 100644 (file)
@@ -1958,7 +1958,8 @@ bool tls_sw_stream_read(const struct sock *sk)
                ingress_empty = list_empty(&psock->ingress_msg);
        rcu_read_unlock();
 
-       return !ingress_empty || ctx->recv_pkt;
+       return !ingress_empty || ctx->recv_pkt ||
+               !skb_queue_empty(&ctx->rx_list);
 }
 
 static int tls_read_size(struct strparser *strp, struct sk_buff *skb)