qede: Validate unsupported configurations
authorManish Chopra <manish.chopra@cavium.com>
Thu, 24 May 2018 16:54:50 +0000 (09:54 -0700)
committerDavid S. Miller <davem@davemloft.net>
Fri, 25 May 2018 20:10:42 +0000 (16:10 -0400)
Validate and prevent some of the configurations for
unsupported [by firmware] inputs [for example - mac ext,
vlans, masks/prefix, tos/tclass] via ethtool -N/-U.

Signed-off-by: Manish Chopra <manish.chopra@cavium.com>
Signed-off-by: Shahed Shaikh <shahed.shaikh@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/qlogic/qede/qede_filter.c

index bd5b4e4c1d479354de514488ae1dcb0996059fe7..43ed420900c1279019cb61e77b34aa8cd2083be6 100644 (file)
@@ -1560,10 +1560,63 @@ static void qede_flow_build_ipv6_hdr(struct qede_arfs_tuple *t,
        ports[1] = t->dst_port;
 }
 
+/* Validate fields which are set and not accepted by the driver */
+static int qede_flow_spec_validate_unused(struct qede_dev *edev,
+                                         struct ethtool_rx_flow_spec *fs)
+{
+       if (fs->flow_type & FLOW_MAC_EXT) {
+               DP_INFO(edev, "Don't support MAC extensions\n");
+               return -EOPNOTSUPP;
+       }
+
+       if ((fs->flow_type & FLOW_EXT) &&
+           (fs->h_ext.vlan_etype || fs->h_ext.vlan_tci)) {
+               DP_INFO(edev, "Don't support vlan-based classification\n");
+               return -EOPNOTSUPP;
+       }
+
+       if ((fs->flow_type & FLOW_EXT) &&
+           (fs->h_ext.data[0] || fs->h_ext.data[1])) {
+               DP_INFO(edev, "Don't support user defined data\n");
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
 static int qede_flow_spec_to_tuple_ipv4_common(struct qede_dev *edev,
                                               struct qede_arfs_tuple *t,
                                               struct ethtool_rx_flow_spec *fs)
 {
+       if ((fs->h_u.tcp_ip4_spec.ip4src &
+            fs->m_u.tcp_ip4_spec.ip4src) != fs->h_u.tcp_ip4_spec.ip4src) {
+               DP_INFO(edev, "Don't support IP-masks\n");
+               return -EOPNOTSUPP;
+       }
+
+       if ((fs->h_u.tcp_ip4_spec.ip4dst &
+            fs->m_u.tcp_ip4_spec.ip4dst) != fs->h_u.tcp_ip4_spec.ip4dst) {
+               DP_INFO(edev, "Don't support IP-masks\n");
+               return -EOPNOTSUPP;
+       }
+
+       if ((fs->h_u.tcp_ip4_spec.psrc &
+            fs->m_u.tcp_ip4_spec.psrc) != fs->h_u.tcp_ip4_spec.psrc) {
+               DP_INFO(edev, "Don't support port-masks\n");
+               return -EOPNOTSUPP;
+       }
+
+       if ((fs->h_u.tcp_ip4_spec.pdst &
+            fs->m_u.tcp_ip4_spec.pdst) != fs->h_u.tcp_ip4_spec.pdst) {
+               DP_INFO(edev, "Don't support port-masks\n");
+               return -EOPNOTSUPP;
+       }
+
+       if (fs->h_u.tcp_ip4_spec.tos) {
+               DP_INFO(edev, "Don't support tos\n");
+               return -EOPNOTSUPP;
+       }
+
        t->eth_proto = htons(ETH_P_IP);
        t->src_ipv4 = fs->h_u.tcp_ip4_spec.ip4src;
        t->dst_ipv4 = fs->h_u.tcp_ip4_spec.ip4dst;
@@ -1619,6 +1672,23 @@ static int qede_flow_spec_to_tuple_ipv6_common(struct qede_dev *edev,
        p = &zero_addr;
        memset(p, 0, sizeof(zero_addr));
 
+       if ((fs->h_u.tcp_ip6_spec.psrc &
+            fs->m_u.tcp_ip6_spec.psrc) != fs->h_u.tcp_ip6_spec.psrc) {
+               DP_INFO(edev, "Don't support port-masks\n");
+               return -EOPNOTSUPP;
+       }
+
+       if ((fs->h_u.tcp_ip6_spec.pdst &
+            fs->m_u.tcp_ip6_spec.pdst) != fs->h_u.tcp_ip6_spec.pdst) {
+               DP_INFO(edev, "Don't support port-masks\n");
+               return -EOPNOTSUPP;
+       }
+
+       if (fs->h_u.tcp_ip6_spec.tclass) {
+               DP_INFO(edev, "Don't support tclass\n");
+               return -EOPNOTSUPP;
+       }
+
        t->eth_proto = htons(ETH_P_IPV6);
        memcpy(&t->src_ipv6, &fs->h_u.tcp_ip6_spec.ip6src,
               sizeof(struct in6_addr));
@@ -1673,6 +1743,9 @@ static int qede_flow_spec_to_tuple(struct qede_dev *edev,
 {
        memset(t, 0, sizeof(*t));
 
+       if (qede_flow_spec_validate_unused(edev, fs))
+               return -EOPNOTSUPP;
+
        switch ((fs->flow_type & ~FLOW_EXT)) {
        case TCP_V4_FLOW:
                return qede_flow_spec_to_tuple_tcpv4(edev, t, fs);