]> git.baikalelectronics.ru Git - kernel.git/commitdiff
cxgb4: add ethtool n-tuple filter insertion
authorVishal Kulkarni <vishal@chelsio.com>
Fri, 19 Jun 2020 14:21:36 +0000 (19:51 +0530)
committerDavid S. Miller <davem@davemloft.net>
Fri, 19 Jun 2020 20:17:31 +0000 (13:17 -0700)
Add support to parse and insert ethtool n-tuple filters.
Translate n-tuple spec to flow spec and use the existing tc-flower
offload infra to insert ethtool n-tuple filters.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Vishal Kulkarni <vishal@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c
drivers/net/ethernet/chelsio/cxgb4/cxgb4_filter.c
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h

index 51f1d5f87bc3431626a97bb9ed8bb60fd86f6d78..82fc09b6dc8ed703df04eb54bf80390a0696e1eb 100644 (file)
@@ -11,6 +11,7 @@
 #include "t4fw_api.h"
 #include "cxgb4_cudbg.h"
 #include "cxgb4_filter.h"
+#include "cxgb4_tc_flower.h"
 
 #define EEPROM_MAGIC 0x38E2F10C
 
@@ -1635,6 +1636,79 @@ static int get_rxnfc(struct net_device *dev, struct ethtool_rxnfc *info,
        return -EOPNOTSUPP;
 }
 
+/* Add Ethtool n-tuple filters. */
+static int cxgb4_ntuple_set_filter(struct net_device *netdev,
+                                  struct ethtool_rxnfc *cmd)
+{
+       struct ethtool_rx_flow_spec_input input = {};
+       struct cxgb4_ethtool_filter_info *filter_info;
+       struct adapter *adapter = netdev2adap(netdev);
+       struct port_info *pi = netdev_priv(netdev);
+       struct ch_filter_specification fs;
+       struct ethtool_rx_flow_rule *flow;
+       u32 tid;
+       int ret;
+
+       if (!(adapter->flags & CXGB4_FULL_INIT_DONE))
+               return -EAGAIN;  /* can still change nfilters */
+
+       if (!adapter->ethtool_filters)
+               return -EOPNOTSUPP;
+
+       if (cmd->fs.location >= adapter->ethtool_filters->nentries) {
+               dev_err(adapter->pdev_dev,
+                       "Location must be < %u",
+                       adapter->ethtool_filters->nentries);
+               return -ERANGE;
+       }
+
+       if (test_bit(cmd->fs.location,
+                    adapter->ethtool_filters->port[pi->port_id].bmap))
+               return -EEXIST;
+
+       memset(&fs, 0, sizeof(fs));
+
+       input.fs = &cmd->fs;
+       flow = ethtool_rx_flow_rule_create(&input);
+       if (IS_ERR(flow)) {
+               ret = PTR_ERR(flow);
+               goto exit;
+       }
+
+       fs.hitcnts = 1;
+
+       ret = cxgb4_flow_rule_replace(netdev, flow->rule, cmd->fs.location,
+                                     NULL, &fs, &tid);
+       if (ret)
+               goto free;
+
+       filter_info = &adapter->ethtool_filters->port[pi->port_id];
+
+       filter_info->loc_array[cmd->fs.location] = tid;
+       set_bit(cmd->fs.location, filter_info->bmap);
+       filter_info->in_use++;
+
+free:
+       ethtool_rx_flow_rule_destroy(flow);
+exit:
+       return ret;
+}
+
+static int set_rxnfc(struct net_device *dev, struct ethtool_rxnfc *cmd)
+{
+       int ret = -EOPNOTSUPP;
+
+       switch (cmd->cmd) {
+       case ETHTOOL_SRXCLSRLINS:
+               ret = cxgb4_ntuple_set_filter(dev, cmd);
+               break;
+       default:
+               break;
+       }
+
+       return ret;
+}
+
 static int set_dump(struct net_device *dev, struct ethtool_dump *eth_dump)
 {
        struct adapter *adapter = netdev2adap(dev);
@@ -1840,6 +1914,7 @@ static const struct ethtool_ops cxgb_ethtool_ops = {
        .get_regs_len      = get_regs_len,
        .get_regs          = get_regs,
        .get_rxnfc         = get_rxnfc,
+       .set_rxnfc         = set_rxnfc,
        .get_rxfh_indir_size = get_rss_table_size,
        .get_rxfh          = get_rss_table,
        .set_rxfh          = set_rss_table,
index 796555255207c6511f1e58429f792458bdc634a6..a12df792d832223323d52553b65e5db5c815998b 100644 (file)
@@ -1152,6 +1152,11 @@ bool is_filter_exact_match(struct adapter *adap,
        if (!is_hashfilter(adap))
                return false;
 
+       if ((atomic_read(&adap->tids.hash_tids_in_use) +
+            atomic_read(&adap->tids.tids_in_use)) >=
+           (adap->tids.nhash + (adap->tids.stid_base - adap->tids.tid_base)))
+               return false;
+
         /* Keep tunnel VNI match disabled for hash-filters for now */
        if (fs->mask.encap_vld)
                return false;
index 501917751b7f3bbedc2068beae8cc9f5c3885861..7423980bc49a0d34811d47473fcc9d63fc949485 100644 (file)
@@ -6371,7 +6371,7 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
                        NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
                        NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_GRO |
                        NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
-                       NETIF_F_HW_TC;
+                       NETIF_F_HW_TC | NETIF_F_NTUPLE;
 
                if (chip_ver > CHELSIO_T5) {
                        netdev->hw_enc_features |= NETIF_F_IP_CSUM |
index 030de20a5d2754d451bb778d79e4c636932c8c99..260d1219b723b65f3ed01abb11fbaf3027ea0fa4 100644 (file)
@@ -81,19 +81,9 @@ static struct ch_tc_flower_entry *ch_flower_lookup(struct adapter *adap,
 }
 
 static void cxgb4_process_flow_match(struct net_device *dev,
-                                    struct flow_cls_offload *cls,
+                                    struct flow_rule *rule,
                                     struct ch_filter_specification *fs)
 {
-       struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
-       u16 addr_type = 0;
-
-       if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CONTROL)) {
-               struct flow_match_control match;
-
-               flow_rule_match_control(rule, &match);
-               addr_type = match.key->addr_type;
-       }
-
        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
                struct flow_match_basic match;
                u16 ethtype_key, ethtype_mask;
@@ -116,7 +106,7 @@ static void cxgb4_process_flow_match(struct net_device *dev,
                fs->mask.proto = match.mask->ip_proto;
        }
 
-       if (addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
+       if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
                struct flow_match_ipv4_addrs match;
 
                flow_rule_match_ipv4_addrs(rule, &match);
@@ -131,7 +121,7 @@ static void cxgb4_process_flow_match(struct net_device *dev,
                memcpy(&fs->nat_fip[0], &match.key->src, sizeof(match.key->src));
        }
 
-       if (addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
+       if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
                struct flow_match_ipv6_addrs match;
 
                flow_rule_match_ipv6_addrs(rule, &match);
@@ -224,9 +214,8 @@ static void cxgb4_process_flow_match(struct net_device *dev,
 }
 
 static int cxgb4_validate_flow_match(struct net_device *dev,
-                                    struct flow_cls_offload *cls)
+                                    struct flow_rule *rule)
 {
-       struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
        struct flow_dissector *dissector = rule->match.dissector;
        u16 ethtype_mask = 0;
        u16 ethtype_key = 0;
@@ -693,14 +682,11 @@ out_unlock:
        spin_unlock_bh(&t->ftid_lock);
 }
 
-int cxgb4_tc_flower_replace(struct net_device *dev,
-                           struct flow_cls_offload *cls)
+int cxgb4_flow_rule_replace(struct net_device *dev, struct flow_rule *rule,
+                           u32 tc_prio, struct netlink_ext_ack *extack,
+                           struct ch_filter_specification *fs, u32 *tid)
 {
-       struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
-       struct netlink_ext_ack *extack = cls->common.extack;
        struct adapter *adap = netdev2adap(dev);
-       struct ch_tc_flower_entry *ch_flower;
-       struct ch_filter_specification *fs;
        struct filter_ctx ctx;
        u8 inet_family;
        int fidx, ret;
@@ -708,18 +694,10 @@ int cxgb4_tc_flower_replace(struct net_device *dev,
        if (cxgb4_validate_flow_actions(dev, &rule->action, extack))
                return -EOPNOTSUPP;
 
-       if (cxgb4_validate_flow_match(dev, cls))
+       if (cxgb4_validate_flow_match(dev, rule))
                return -EOPNOTSUPP;
 
-       ch_flower = allocate_flower_entry();
-       if (!ch_flower) {
-               netdev_err(dev, "%s: ch_flower alloc failed.\n", __func__);
-               return -ENOMEM;
-       }
-
-       fs = &ch_flower->fs;
-       fs->hitcnts = 1;
-       cxgb4_process_flow_match(dev, cls, fs);
+       cxgb4_process_flow_match(dev, rule, fs);
        cxgb4_process_flow_actions(dev, &rule->action, fs);
 
        fs->hash = is_filter_exact_match(adap, fs);
@@ -730,12 +708,11 @@ int cxgb4_tc_flower_replace(struct net_device *dev,
         * existing rules.
         */
        fidx = cxgb4_get_free_ftid(dev, inet_family, fs->hash,
-                                  cls->common.prio);
+                                  tc_prio);
        if (fidx < 0) {
                NL_SET_ERR_MSG_MOD(extack,
                                   "No free LETCAM index available");
-               ret = -ENOMEM;
-               goto free_entry;
+               return -ENOMEM;
        }
 
        if (fidx < adap->tids.nhpftids) {
@@ -749,42 +726,70 @@ int cxgb4_tc_flower_replace(struct net_device *dev,
        if (fs->hash)
                fidx = 0;
 
-       fs->tc_prio = cls->common.prio;
-       fs->tc_cookie = cls->cookie;
+       fs->tc_prio = tc_prio;
 
        init_completion(&ctx.completion);
        ret = __cxgb4_set_filter(dev, fidx, fs, &ctx);
        if (ret) {
                netdev_err(dev, "%s: filter creation err %d\n",
                           __func__, ret);
-               goto free_entry;
+               return ret;
        }
 
        /* Wait for reply */
        ret = wait_for_completion_timeout(&ctx.completion, 10 * HZ);
-       if (!ret) {
-               ret = -ETIMEDOUT;
-               goto free_entry;
-       }
+       if (!ret)
+               return -ETIMEDOUT;
 
-       ret = ctx.result;
        /* Check if hw returned error for filter creation */
+       if (ctx.result)
+               return ctx.result;
+
+       *tid = ctx.tid;
+
+       if (fs->hash)
+               cxgb4_tc_flower_hash_prio_add(adap, tc_prio);
+
+       return 0;
+}
+
+int cxgb4_tc_flower_replace(struct net_device *dev,
+                           struct flow_cls_offload *cls)
+{
+       struct flow_rule *rule = flow_cls_offload_flow_rule(cls);
+       struct netlink_ext_ack *extack = cls->common.extack;
+       struct adapter *adap = netdev2adap(dev);
+       struct ch_tc_flower_entry *ch_flower;
+       struct ch_filter_specification *fs;
+       int ret;
+
+       ch_flower = allocate_flower_entry();
+       if (!ch_flower) {
+               netdev_err(dev, "%s: ch_flower alloc failed.\n", __func__);
+               return -ENOMEM;
+       }
+
+       fs = &ch_flower->fs;
+       fs->hitcnts = 1;
+       fs->tc_cookie = cls->cookie;
+
+       ret = cxgb4_flow_rule_replace(dev, rule, cls->common.prio, extack, fs,
+                                     &ch_flower->filter_id);
        if (ret)
                goto free_entry;
 
        ch_flower->tc_flower_cookie = cls->cookie;
-       ch_flower->filter_id = ctx.tid;
        ret = rhashtable_insert_fast(&adap->flower_tbl, &ch_flower->node,
                                     adap->flower_ht_params);
        if (ret)
                goto del_filter;
 
-       if (fs->hash)
-               cxgb4_tc_flower_hash_prio_add(adap, cls->common.prio);
-
        return 0;
 
 del_filter:
+       if (fs->hash)
+               cxgb4_tc_flower_hash_prio_del(adap, cls->common.prio);
+
        cxgb4_del_filter(dev, ch_flower->filter_id, &ch_flower->fs);
 
 free_entry:
index 0a30c96b81ffa2f66a89435811cb8787f51d50ea..7fa3797495004e818d2f1d73a95fab5beaeceb18 100644 (file)
@@ -121,6 +121,9 @@ int cxgb4_tc_flower_destroy(struct net_device *dev,
                            struct flow_cls_offload *cls);
 int cxgb4_tc_flower_stats(struct net_device *dev,
                          struct flow_cls_offload *cls);
+int cxgb4_flow_rule_replace(struct net_device *dev, struct flow_rule *rule,
+                           u32 tc_prio, struct netlink_ext_ack *extack,
+                           struct ch_filter_specification *fs, u32 *tid);
 
 int cxgb4_init_tc_flower(struct adapter *adap);
 void cxgb4_cleanup_tc_flower(struct adapter *adap);