libcxgb: Don't accidentally set RTO_ONLINK in cxgb_find_route()
authorGuillaume Nault <gnault@redhat.com>
Mon, 10 Jan 2022 13:43:11 +0000 (14:43 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 27 Jan 2022 08:01:00 +0000 (09:01 +0100)
commit a915deaa9abe4fb3a440312c954253a6a733608e upstream.

Mask the ECN bits before calling ip_route_output_ports(). The tos
variable might be passed directly from an IPv4 header, so it may have
the last ECN bit set. This interferes with the route lookup process as
ip_route_output_key_hash() interpretes this bit specially (to restrict
the route scope).

Found by code inspection, compile tested only.

Fixes: 804c2f3e36ef ("libcxgb,iw_cxgb4,cxgbit: add cxgb_find_route()")
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/net/ethernet/chelsio/libcxgb/libcxgb_cm.c

index d04a6c1634452034217e9dc4ab8771bf02899b86..da8d10475a08e6e23c466b6f62637f7c840e86bd 100644 (file)
@@ -32,6 +32,7 @@
 
 #include <linux/tcp.h>
 #include <linux/ipv6.h>
+#include <net/inet_ecn.h>
 #include <net/route.h>
 #include <net/ip6_route.h>
 
@@ -99,7 +100,7 @@ cxgb_find_route(struct cxgb4_lld_info *lldi,
 
        rt = ip_route_output_ports(&init_net, &fl4, NULL, peer_ip, local_ip,
                                   peer_port, local_port, IPPROTO_TCP,
-                                  tos, 0);
+                                  tos & ~INET_ECN_MASK, 0);
        if (IS_ERR(rt))
                return NULL;
        n = dst_neigh_lookup(&rt->dst, &peer_ip);