Merge branch 'cxgb4-add-mirror-action-support-for-TC-MATCHALL'

Rahul Lakkireddy says:

====================
cxgb4: add mirror action support for TC-MATCHALL

This series of patches add support to mirror all ingress traffic
for TC-MATCHALL ingress offload.

Patch 1 adds support to dynamically create a mirror Virtual Interface
(VI) that accepts all mirror ingress traffic when mirror action is
set in TC-MATCHALL offload.

Patch 2 adds support to allocate mirror Rxqs and setup RSS for the
mirror VI.

Patch 3 adds support to replicate all the main VI configuration to
mirror VI. This includes replicating MTU, promiscuous mode,
all-multicast mode, and enabled netdev Rx feature offloads.

v3:
- Replace mirror VI refcount_t with normal u32 variable in all patches.
- Add back calling cxgb4_port_mirror_start() in cxgb_open(), which
  was there in v1, but got missed in v2 during refactoring, in patch
  3.

v2:
- Add mutex to protect all mirror VI data, instead of just
  mirror Rxqs, in patch 1 and 2.
- Remove the un-needed mirror Rxq mutex in patch 2.
- Simplify the replication code by refactoring t4_set_rxmode()
  to handle mirror VI, instead of duplicating the t4_set_rxmode()
  calls in multiple places in patch 3.
====================

Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 files changed, 606 insertions, 49 deletions

diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index d811df1b93d9..ff53c78307c5 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -679,6 +679,12 @@ struct port_info {
 	u8 rx_cchan;
 
 	bool tc_block_shared;
+
+	/* Mirror VI information */
+	u16 viid_mirror;
+	u16 nmirrorqsets;
+	u32 vi_mirror_count;
+	struct mutex vi_mirror_mutex; /* Sync access to Mirror VI info */
 };
 
 struct dentry;
@@ -705,6 +711,13 @@ enum {
 	ULP_CRYPTO_KTLS_INLINE  = 1 << 3,
 };
 
+#define CXGB4_MIRROR_RXQ_DEFAULT_DESC_NUM 1024
+#define CXGB4_MIRROR_RXQ_DEFAULT_DESC_SIZE 64
+#define CXGB4_MIRROR_RXQ_DEFAULT_INTR_USEC 5
+#define CXGB4_MIRROR_RXQ_DEFAULT_PKT_CNT 8
+
+#define CXGB4_MIRROR_FLQ_DEFAULT_DESC_NUM 72
+
 struct rx_sw_desc;
 
 struct sge_fl {                     /* SGE free-buffer queue state */
@@ -954,12 +967,15 @@ struct sge {
 	struct sge_eohw_txq *eohw_txq;
 	struct sge_ofld_rxq *eohw_rxq;
 
+	struct sge_eth_rxq *mirror_rxq[NCHAN];
+
 	u16 max_ethqsets;           /* # of available Ethernet queue sets */
 	u16 ethqsets;               /* # of active Ethernet queue sets */
 	u16 ethtxq_rover;           /* Tx queue to clean up next */
 	u16 ofldqsets;              /* # of active ofld queue sets */
 	u16 nqs_per_uld;	    /* # of Rx queues per ULD */
 	u16 eoqsets;                /* # of ETHOFLD queues */
+	u16 mirrorqsets;            /* # of Mirror queues */
 
 	u16 timer_val[SGE_NTIMERS];
 	u8 counter_val[SGE_NCOUNTERS];
@@ -1857,6 +1873,8 @@ int t4_init_rss_mode(struct adapter *adap, int mbox);
 int t4_init_portinfo(struct port_info *pi, int mbox,
 		     int port, int pf, int vf, u8 mac[]);
 int t4_port_init(struct adapter *adap, int mbox, int pf, int vf);
+int t4_init_port_mirror(struct port_info *pi, u8 mbox, u8 port, u8 pf, u8 vf,
+			u16 *mirror_viid);
 void t4_fatal_err(struct adapter *adapter);
 unsigned int t4_chip_rss_size(struct adapter *adapter);
 int t4_config_rss_range(struct adapter *adapter, int mbox, unsigned int viid,
@@ -1966,8 +1984,8 @@ int t4_free_vi(struct adapter *adap, unsigned int mbox,
 	       unsigned int pf, unsigned int vf,
 	       unsigned int viid);
 int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
-		int mtu, int promisc, int all_multi, int bcast, int vlanex,
-		bool sleep_ok);
+		  unsigned int viid_mirror, int mtu, int promisc, int all_multi,
+		  int bcast, int vlanex, bool sleep_ok);
 int t4_free_raw_mac_filt(struct adapter *adap, unsigned int viid,
 			 const u8 *addr, const u8 *mask, unsigned int idx,
 			 u8 lookup_type, u8 port_id, bool sleep_ok);
@@ -2141,6 +2159,8 @@ int cxgb_open(struct net_device *dev);
 int cxgb_close(struct net_device *dev);
 void cxgb4_enable_rx(struct adapter *adap, struct sge_rspq *q);
 void cxgb4_quiesce_rx(struct sge_rspq *q);
+int cxgb4_port_mirror_alloc(struct net_device *dev);
+void cxgb4_port_mirror_free(struct net_device *dev);
 #ifdef CONFIG_CHELSIO_TLS_DEVICE
 int cxgb4_set_ktls_feature(struct adapter *adap, bool enable);
 #endif
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
index b477b8842905..05f33b7e3677 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
@@ -2742,6 +2742,58 @@ do { \
 	}
 
 	r -= eth_entries;
+	for_each_port(adap, j) {
+		struct port_info *pi = adap2pinfo(adap, j);
+		const struct sge_eth_rxq *rx;
+
+		mutex_lock(&pi->vi_mirror_mutex);
+		if (!pi->vi_mirror_count) {
+			mutex_unlock(&pi->vi_mirror_mutex);
+			continue;
+		}
+
+		if (r >= DIV_ROUND_UP(pi->nmirrorqsets, 4)) {
+			r -= DIV_ROUND_UP(pi->nmirrorqsets, 4);
+			mutex_unlock(&pi->vi_mirror_mutex);
+			continue;
+		}
+
+		rx = &s->mirror_rxq[j][r * 4];
+		n = min(4, pi->nmirrorqsets - 4 * r);
+
+		S("QType:", "Mirror-Rxq");
+		S("Interface:",
+		  rx[i].rspq.netdev ? rx[i].rspq.netdev->name : "N/A");
+		R("RspQ ID:", rspq.abs_id);
+		R("RspQ size:", rspq.size);
+		R("RspQE size:", rspq.iqe_len);
+		R("RspQ CIDX:", rspq.cidx);
+		R("RspQ Gen:", rspq.gen);
+		S3("u", "Intr delay:", qtimer_val(adap, &rx[i].rspq));
+		S3("u", "Intr pktcnt:", s->counter_val[rx[i].rspq.pktcnt_idx]);
+		R("FL ID:", fl.cntxt_id);
+		R("FL size:", fl.size - 8);
+		R("FL pend:", fl.pend_cred);
+		R("FL avail:", fl.avail);
+		R("FL PIDX:", fl.pidx);
+		R("FL CIDX:", fl.cidx);
+		RL("RxPackets:", stats.pkts);
+		RL("RxCSO:", stats.rx_cso);
+		RL("VLANxtract:", stats.vlan_ex);
+		RL("LROmerged:", stats.lro_merged);
+		RL("LROpackets:", stats.lro_pkts);
+		RL("RxDrops:", stats.rx_drops);
+		RL("RxBadPkts:", stats.bad_rx_pkts);
+		RL("FLAllocErr:", fl.alloc_failed);
+		RL("FLLrgAlcErr:", fl.large_alloc_failed);
+		RL("FLMapErr:", fl.mapping_err);
+		RL("FLLow:", fl.low);
+		RL("FLStarving:", fl.starving);
+
+		mutex_unlock(&pi->vi_mirror_mutex);
+		goto out;
+	}
+
 	if (!adap->tc_mqprio)
 		goto skip_mqprio;
 
@@ -3098,9 +3150,10 @@ unlock:
 	return 0;
 }
 
-static int sge_queue_entries(const struct adapter *adap)
+static int sge_queue_entries(struct adapter *adap)
 {
 	int i, tot_uld_entries = 0, eohw_entries = 0, eosw_entries = 0;
+	int mirror_rxq_entries = 0;
 
 	if (adap->tc_mqprio) {
 		struct cxgb4_tc_port_mqprio *port_mqprio;
@@ -3123,6 +3176,15 @@ static int sge_queue_entries(const struct adapter *adap)
 		mutex_unlock(&adap->tc_mqprio->mqprio_mutex);
 	}
 
+	for_each_port(adap, i) {
+		struct port_info *pi = adap2pinfo(adap, i);
+
+		mutex_lock(&pi->vi_mirror_mutex);
+		if (pi->vi_mirror_count)
+			mirror_rxq_entries += DIV_ROUND_UP(pi->nmirrorqsets, 4);
+		mutex_unlock(&pi->vi_mirror_mutex);
+	}
+
 	if (!is_uld(adap))
 		goto lld_only;
 
@@ -3137,7 +3199,7 @@ static int sge_queue_entries(const struct adapter *adap)
 	mutex_unlock(&uld_mutex);
 
 lld_only:
-	return DIV_ROUND_UP(adap->sge.ethqsets, 4) +
+	return DIV_ROUND_UP(adap->sge.ethqsets, 4) + mirror_rxq_entries +
 	       eohw_entries + eosw_entries + tot_uld_entries +
 	       DIV_ROUND_UP(MAX_CTRL_QUEUES, 4) + 1;
 }
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 234f01389cb1..0991631f3a91 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -435,8 +435,8 @@ static int set_rxmode(struct net_device *dev, int mtu, bool sleep_ok)
 	__dev_uc_sync(dev, cxgb4_mac_sync, cxgb4_mac_unsync);
 	__dev_mc_sync(dev, cxgb4_mac_sync, cxgb4_mac_unsync);
 
-	return t4_set_rxmode(adapter, adapter->mbox, pi->viid, mtu,
-			     (dev->flags & IFF_PROMISC) ? 1 : 0,
+	return t4_set_rxmode(adapter, adapter->mbox, pi->viid, pi->viid_mirror,
+			     mtu, (dev->flags & IFF_PROMISC) ? 1 : 0,
 			     (dev->flags & IFF_ALLMULTI) ? 1 : 0, 1, -1,
 			     sleep_ok);
 }
@@ -503,15 +503,16 @@ set_hash:
  */
 static int link_start(struct net_device *dev)
 {
-	int ret;
 	struct port_info *pi = netdev_priv(dev);
-	unsigned int mb = pi->adapter->pf;
+	unsigned int mb = pi->adapter->mbox;
+	int ret;
 
 	/*
 	 * We do not set address filters and promiscuity here, the stack does
 	 * that step explicitly.
 	 */
-	ret = t4_set_rxmode(pi->adapter, mb, pi->viid, dev->mtu, -1, -1, -1,
+	ret = t4_set_rxmode(pi->adapter, mb, pi->viid, pi->viid_mirror,
+			    dev->mtu, -1, -1, -1,
 			    !!(dev->features & NETIF_F_HW_VLAN_CTAG_RX), true);
 	if (ret == 0)
 		ret = cxgb4_update_mac_filt(pi, pi->viid, &pi->xact_addr_filt,
@@ -822,6 +823,31 @@ static void adap_config_hpfilter(struct adapter *adapter)
 			"HP filter region isn't supported by FW\n");
 }
 
+static int cxgb4_config_rss(const struct port_info *pi, u16 *rss,
+			    u16 rss_size, u16 viid)
+{
+	struct adapter *adap = pi->adapter;
+	int ret;
+
+	ret = t4_config_rss_range(adap, adap->mbox, viid, 0, rss_size, rss,
+				  rss_size);
+	if (ret)
+		return ret;
+
+	/* If Tunnel All Lookup isn't specified in the global RSS
+	 * Configuration, then we need to specify a default Ingress
+	 * Queue for any ingress packets which aren't hashed.  We'll
+	 * use our first ingress queue ...
+	 */
+	return t4_config_vi_rss(adap, adap->mbox, viid,
+				FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN_F |
+				FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN_F |
+				FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN_F |
+				FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN_F |
+				FW_RSS_VI_CONFIG_CMD_UDPEN_F,
+				rss[0]);
+}
+
 /**
  *	cxgb4_write_rss - write the RSS table for a given port
  *	@pi: the port
@@ -833,10 +859,10 @@ static void adap_config_hpfilter(struct adapter *adapter)
  */
 int cxgb4_write_rss(const struct port_info *pi, const u16 *queues)
 {
-	u16 *rss;
-	int i, err;
 	struct adapter *adapter = pi->adapter;
 	const struct sge_eth_rxq *rxq;
+	int i, err;
+	u16 *rss;
 
 	rxq = &adapter->sge.ethrxq[pi->first_qset];
 	rss = kmalloc_array(pi->rss_size, sizeof(u16), GFP_KERNEL);
@@ -847,21 +873,7 @@ int cxgb4_write_rss(const struct port_info *pi, const u16 *queues)
 	for (i = 0; i < pi->rss_size; i++, queues++)
 		rss[i] = rxq[*queues].rspq.abs_id;
 
-	err = t4_config_rss_range(adapter, adapter->pf, pi->viid, 0,
-				  pi->rss_size, rss, pi->rss_size);
-	/* If Tunnel All Lookup isn't specified in the global RSS
-	 * Configuration, then we need to specify a default Ingress
-	 * Queue for any ingress packets which aren't hashed.  We'll
-	 * use our first ingress queue ...
-	 */
-	if (!err)
-		err = t4_config_vi_rss(adapter, adapter->mbox, pi->viid,
-				       FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN_F |
-				       FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN_F |
-				       FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN_F |
-				       FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN_F |
-				       FW_RSS_VI_CONFIG_CMD_UDPEN_F,
-				       rss[0]);
+	err = cxgb4_config_rss(pi, rss, pi->rss_size, pi->viid);
 	kfree(rss);
 	return err;
 }
@@ -1259,15 +1271,15 @@ int cxgb4_set_rspq_intr_params(struct sge_rspq *q,
 
 static int cxgb_set_features(struct net_device *dev, netdev_features_t features)
 {
-	const struct port_info *pi = netdev_priv(dev);
 	netdev_features_t changed = dev->features ^ features;
+	const struct port_info *pi = netdev_priv(dev);
 	int err;
 
 	if (!(changed & NETIF_F_HW_VLAN_CTAG_RX))
 		return 0;
 
-	err = t4_set_rxmode(pi->adapter, pi->adapter->pf, pi->viid, -1,
-			    -1, -1, -1,
+	err = t4_set_rxmode(pi->adapter, pi->adapter->mbox, pi->viid,
+			    pi->viid_mirror, -1, -1, -1, -1,
 			    !!(features & NETIF_F_HW_VLAN_CTAG_RX), true);
 	if (unlikely(err))
 		dev->features = features ^ NETIF_F_HW_VLAN_CTAG_RX;
@@ -1285,6 +1297,292 @@ static int setup_debugfs(struct adapter *adap)
 	return 0;
 }
 
+static void cxgb4_port_mirror_free_rxq(struct adapter *adap,
+				       struct sge_eth_rxq *mirror_rxq)
+{
+	if ((adap->flags & CXGB4_FULL_INIT_DONE) &&
+	    !(adap->flags & CXGB4_SHUTTING_DOWN))
+		cxgb4_quiesce_rx(&mirror_rxq->rspq);
+
+	if (adap->flags & CXGB4_USING_MSIX) {
+		cxgb4_clear_msix_aff(mirror_rxq->msix->vec,
+				     mirror_rxq->msix->aff_mask);
+		free_irq(mirror_rxq->msix->vec, &mirror_rxq->rspq);
+		cxgb4_free_msix_idx_in_bmap(adap, mirror_rxq->msix->idx);
+	}
+
+	free_rspq_fl(adap, &mirror_rxq->rspq, &mirror_rxq->fl);
+}
+
+static int cxgb4_port_mirror_alloc_queues(struct net_device *dev)
+{
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+	struct sge_eth_rxq *mirror_rxq;
+	struct sge *s = &adap->sge;
+	int ret = 0, msix = 0;
+	u16 i, rxqid;
+	u16 *rss;
+
+	if (!pi->vi_mirror_count)
+		return 0;
+
+	if (s->mirror_rxq[pi->port_id])
+		return 0;
+
+	mirror_rxq = kcalloc(pi->nmirrorqsets, sizeof(*mirror_rxq), GFP_KERNEL);
+	if (!mirror_rxq)
+		return -ENOMEM;
+
+	s->mirror_rxq[pi->port_id] = mirror_rxq;
+
+	if (!(adap->flags & CXGB4_USING_MSIX))
+		msix = -((int)adap->sge.intrq.abs_id + 1);
+
+	for (i = 0, rxqid = 0; i < pi->nmirrorqsets; i++, rxqid++) {
+		mirror_rxq = &s->mirror_rxq[pi->port_id][i];
+
+		/* Allocate Mirror Rxqs */
+		if (msix >= 0) {
+			msix = cxgb4_get_msix_idx_from_bmap(adap);
+			if (msix < 0) {
+				ret = msix;
+				goto out_free_queues;
+			}
+
+			mirror_rxq->msix = &adap->msix_info[msix];
+			snprintf(mirror_rxq->msix->desc,
+				 sizeof(mirror_rxq->msix->desc),
+				 "%s-mirrorrxq%d", dev->name, i);
+		}
+
+		init_rspq(adap, &mirror_rxq->rspq,
+			  CXGB4_MIRROR_RXQ_DEFAULT_INTR_USEC,
+			  CXGB4_MIRROR_RXQ_DEFAULT_PKT_CNT,
+			  CXGB4_MIRROR_RXQ_DEFAULT_DESC_NUM,
+			  CXGB4_MIRROR_RXQ_DEFAULT_DESC_SIZE);
+
+		mirror_rxq->fl.size = CXGB4_MIRROR_FLQ_DEFAULT_DESC_NUM;
+
+		ret = t4_sge_alloc_rxq(adap, &mirror_rxq->rspq, false,
+				       dev, msix, &mirror_rxq->fl,
+				       t4_ethrx_handler, NULL, 0);
+		if (ret)
+			goto out_free_msix_idx;
+
+		/* Setup MSI-X vectors for Mirror Rxqs */
+		if (adap->flags & CXGB4_USING_MSIX) {
+			ret = request_irq(mirror_rxq->msix->vec,
+					  t4_sge_intr_msix, 0,
+					  mirror_rxq->msix->desc,
+					  &mirror_rxq->rspq);
+			if (ret)
+				goto out_free_rxq;
+
+			cxgb4_set_msix_aff(adap, mirror_rxq->msix->vec,
+					   &mirror_rxq->msix->aff_mask, i);
+		}
+
+		/* Start NAPI for Mirror Rxqs */
+		cxgb4_enable_rx(adap, &mirror_rxq->rspq);
+	}
+
+	/* Setup RSS for Mirror Rxqs */
+	rss = kcalloc(pi->rss_size, sizeof(u16), GFP_KERNEL);
+	if (!rss) {
+		ret = -ENOMEM;
+		goto out_free_queues;
+	}
+
+	mirror_rxq = &s->mirror_rxq[pi->port_id][0];
+	for (i = 0; i < pi->rss_size; i++)
+		rss[i] = mirror_rxq[i % pi->nmirrorqsets].rspq.abs_id;
+
+	ret = cxgb4_config_rss(pi, rss, pi->rss_size, pi->viid_mirror);
+	kfree(rss);
+	if (ret)
+		goto out_free_queues;
+
+	return 0;
+
+out_free_rxq:
+	free_rspq_fl(adap, &mirror_rxq->rspq, &mirror_rxq->fl);
+
+out_free_msix_idx:
+	cxgb4_free_msix_idx_in_bmap(adap, mirror_rxq->msix->idx);
+
+out_free_queues:
+	while (rxqid-- > 0)
+		cxgb4_port_mirror_free_rxq(adap,
+					   &s->mirror_rxq[pi->port_id][rxqid]);
+
+	kfree(s->mirror_rxq[pi->port_id]);
+	s->mirror_rxq[pi->port_id] = NULL;
+	return ret;
+}
+
+static void cxgb4_port_mirror_free_queues(struct net_device *dev)
+{
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+	struct sge *s = &adap->sge;
+	u16 i;
+
+	if (!pi->vi_mirror_count)
+		return;
+
+	if (!s->mirror_rxq[pi->port_id])
+		return;
+
+	for (i = 0; i < pi->nmirrorqsets; i++)
+		cxgb4_port_mirror_free_rxq(adap,
+					   &s->mirror_rxq[pi->port_id][i]);
+
+	kfree(s->mirror_rxq[pi->port_id]);
+	s->mirror_rxq[pi->port_id] = NULL;
+}
+
+static int cxgb4_port_mirror_start(struct net_device *dev)
+{
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+	int ret, idx = -1;
+
+	if (!pi->vi_mirror_count)
+		return 0;
+
+	/* Mirror VIs can be created dynamically after stack had
+	 * already setup Rx modes like MTU, promisc, allmulti, etc.
+	 * on main VI. So, parse what the stack had setup on the
+	 * main VI and update the same on the mirror VI.
+	 */
+	ret = t4_set_rxmode(adap, adap->mbox, pi->viid, pi->viid_mirror,
+			    dev->mtu, (dev->flags & IFF_PROMISC) ? 1 : 0,
+			    (dev->flags & IFF_ALLMULTI) ? 1 : 0, 1,
+			    !!(dev->features & NETIF_F_HW_VLAN_CTAG_RX), true);
+	if (ret) {
+		dev_err(adap->pdev_dev,
+			"Failed start up Rx mode for Mirror VI 0x%x, ret: %d\n",
+			pi->viid_mirror, ret);
+		return ret;
+	}
+
+	/* Enable replication bit for the device's MAC address
+	 * in MPS TCAM, so that the packets for the main VI are
+	 * replicated to mirror VI.
+	 */
+	ret = cxgb4_update_mac_filt(pi, pi->viid_mirror, &idx,
+				    dev->dev_addr, true, NULL);
+	if (ret) {
+		dev_err(adap->pdev_dev,
+			"Failed updating MAC filter for Mirror VI 0x%x, ret: %d\n",
+			pi->viid_mirror, ret);
+		return ret;
+	}
+
+	/* Enabling a Virtual Interface can result in an interrupt
+	 * during the processing of the VI Enable command and, in some
+	 * paths, result in an attempt to issue another command in the
+	 * interrupt context. Thus, we disable interrupts during the
+	 * course of the VI Enable command ...
+	 */
+	local_bh_disable();
+	ret = t4_enable_vi_params(adap, adap->mbox, pi->viid_mirror, true, true,
+				  false);
+	local_bh_enable();
+	if (ret)
+		dev_err(adap->pdev_dev,
+			"Failed starting Mirror VI 0x%x, ret: %d\n",
+			pi->viid_mirror, ret);
+
+	return ret;
+}
+
+static void cxgb4_port_mirror_stop(struct net_device *dev)
+{
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+
+	if (!pi->vi_mirror_count)
+		return;
+
+	t4_enable_vi_params(adap, adap->mbox, pi->viid_mirror, false, false,
+			    false);
+}
+
+int cxgb4_port_mirror_alloc(struct net_device *dev)
+{
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+	int ret = 0;
+
+	if (!pi->nmirrorqsets)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&pi->vi_mirror_mutex);
+	if (pi->viid_mirror) {
+		pi->vi_mirror_count++;
+		goto out_unlock;
+	}
+
+	ret = t4_init_port_mirror(pi, adap->mbox, pi->port_id, adap->pf, 0,
+				  &pi->viid_mirror);
+	if (ret)
+		goto out_unlock;
+
+	pi->vi_mirror_count = 1;
+
+	if (adap->flags & CXGB4_FULL_INIT_DONE) {
+		ret = cxgb4_port_mirror_alloc_queues(dev);
+		if (ret)
+			goto out_free_vi;
+
+		ret = cxgb4_port_mirror_start(dev);
+		if (ret)
+			goto out_free_queues;
+	}
+
+	mutex_unlock(&pi->vi_mirror_mutex);
+	return 0;
+
+out_free_queues:
+	cxgb4_port_mirror_free_queues(dev);
+
+out_free_vi:
+	pi->vi_mirror_count = 0;
+	t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
+	pi->viid_mirror = 0;
+
+out_unlock:
+	mutex_unlock(&pi->vi_mirror_mutex);
+	return ret;
+}
+
+void cxgb4_port_mirror_free(struct net_device *dev)
+{
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+
+	mutex_lock(&pi->vi_mirror_mutex);
+	if (!pi->viid_mirror)
+		goto out_unlock;
+
+	if (pi->vi_mirror_count > 1) {
+		pi->vi_mirror_count--;
+		goto out_unlock;
+	}
+
+	cxgb4_port_mirror_stop(dev);
+	cxgb4_port_mirror_free_queues(dev);
+
+	pi->vi_mirror_count = 0;
+	t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
+	pi->viid_mirror = 0;
+
+out_unlock:
+	mutex_unlock(&pi->vi_mirror_mutex);
+}
+
 /*
  * upper-layer driver support
  */
@@ -2557,8 +2855,29 @@ int cxgb_open(struct net_device *dev)
 		return err;
 
 	err = link_start(dev);
-	if (!err)
-		netif_tx_start_all_queues(dev);
+	if (err)
+		return err;
+
+	if (pi->nmirrorqsets) {
+		mutex_lock(&pi->vi_mirror_mutex);
+		err = cxgb4_port_mirror_alloc_queues(dev);
+		if (err)
+			goto out_unlock;
+
+		err = cxgb4_port_mirror_start(dev);
+		if (err)
+			goto out_free_queues;
+		mutex_unlock(&pi->vi_mirror_mutex);
+	}
+
+	netif_tx_start_all_queues(dev);
+	return 0;
+
+out_free_queues:
+	cxgb4_port_mirror_free_queues(dev);
+
+out_unlock:
+	mutex_unlock(&pi->vi_mirror_mutex);
 	return err;
 }
 
@@ -2576,7 +2895,17 @@ int cxgb_close(struct net_device *dev)
 	cxgb4_dcb_reset(dev);
 	dcb_tx_queue_prio_enable(dev, false);
 #endif
-	return ret;
+	if (ret)
+		return ret;
+
+	if (pi->nmirrorqsets) {
+		mutex_lock(&pi->vi_mirror_mutex);
+		cxgb4_port_mirror_stop(dev);
+		cxgb4_port_mirror_free_queues(dev);
+		mutex_unlock(&pi->vi_mirror_mutex);
+	}
+
+	return 0;
 }
 
 int cxgb4_create_server_filter(const struct net_device *dev, unsigned int stid,
@@ -2842,11 +3171,11 @@ static void cxgb_set_rxmode(struct net_device *dev)
 
 static int cxgb_change_mtu(struct net_device *dev, int new_mtu)
 {
-	int ret;
 	struct port_info *pi = netdev_priv(dev);
+	int ret;
 
-	ret = t4_set_rxmode(pi->adapter, pi->adapter->pf, pi->viid, new_mtu, -1,
-			    -1, -1, -1, true);
+	ret = t4_set_rxmode(pi->adapter, pi->adapter->mbox, pi->viid,
+			    pi->viid_mirror, new_mtu, -1, -1, -1, -1, true);
 	if (!ret)
 		dev->mtu = new_mtu;
 	return ret;
@@ -5504,6 +5833,19 @@ static int cfg_queues(struct adapter *adap)
 		avail_qsets -= s->eoqsets;
 	}
 
+	/* Mirror queues must follow same scheme as normal Ethernet
+	 * Queues, when there are enough queues available. Otherwise,
+	 * allocate at least 1 queue per port. If even 1 queue is not
+	 * available, then disable mirror queues support.
+	 */
+	if (avail_qsets >= s->max_ethqsets)
+		s->mirrorqsets = s->max_ethqsets;
+	else if (avail_qsets >= adap->params.nports)
+		s->mirrorqsets = adap->params.nports;
+	else
+		s->mirrorqsets = 0;
+	avail_qsets -= s->mirrorqsets;
+
 	for (i = 0; i < ARRAY_SIZE(s->ethrxq); i++) {
 		struct sge_eth_rxq *r = &s->ethrxq[i];
 
@@ -5617,8 +5959,8 @@ void cxgb4_free_msix_idx_in_bmap(struct adapter *adap,
 
 static int enable_msix(struct adapter *adap)
 {
-	u32 eth_need, uld_need = 0, ethofld_need = 0;
-	u32 ethqsets = 0, ofldqsets = 0, eoqsets = 0;
+	u32 eth_need, uld_need = 0, ethofld_need = 0, mirror_need = 0;
+	u32 ethqsets = 0, ofldqsets = 0, eoqsets = 0, mirrorqsets = 0;
 	u8 num_uld = 0, nchan = adap->params.nports;
 	u32 i, want, need, num_vec;
 	struct sge *s = &adap->sge;
@@ -5649,6 +5991,12 @@ static int enable_msix(struct adapter *adap)
 		need += ethofld_need;
 	}
 
+	if (s->mirrorqsets) {
+		want += s->mirrorqsets;
+		mirror_need = nchan;
+		need += mirror_need;
+	}
+
 	want += EXTRA_VECS;
 	need += EXTRA_VECS;
 
@@ -5682,8 +6030,10 @@ static int enable_msix(struct adapter *adap)
 		adap->params.ethofld = 0;
 		s->ofldqsets = 0;
 		s->eoqsets = 0;
+		s->mirrorqsets = 0;
 		uld_need = 0;
 		ethofld_need = 0;
+		mirror_need = 0;
 	}
 
 	num_vec = allocated;
@@ -5697,6 +6047,8 @@ static int enable_msix(struct adapter *adap)
 			ofldqsets = nchan;
 		if (is_ethofld(adap))
 			eoqsets = ethofld_need;
+		if (s->mirrorqsets)
+			mirrorqsets = mirror_need;
 
 		num_vec -= need;
 		while (num_vec) {
@@ -5728,12 +6080,25 @@ static int enable_msix(struct adapter *adap)
 				num_vec -= uld_need;
 			}
 		}
+
+		if (s->mirrorqsets) {
+			while (num_vec) {
+				if (num_vec < mirror_need ||
+				    mirrorqsets > s->mirrorqsets)
+					break;
+
+				mirrorqsets++;
+				num_vec -= mirror_need;
+			}
+		}
 	} else {
 		ethqsets = s->max_ethqsets;
 		if (is_uld(adap))
 			ofldqsets = s->ofldqsets;
 		if (is_ethofld(adap))
 			eoqsets = s->eoqsets;
+		if (s->mirrorqsets)
+			mirrorqsets = s->mirrorqsets;
 	}
 
 	if (ethqsets < s->max_ethqsets) {
@@ -5749,6 +6114,15 @@ static int enable_msix(struct adapter *adap)
 	if (is_ethofld(adap))
 		s->eoqsets = eoqsets;
 
+	if (s->mirrorqsets) {
+		s->mirrorqsets = mirrorqsets;
+		for_each_port(adap, i) {
+			pi = adap2pinfo(adap, i);
+			pi->nmirrorqsets = s->mirrorqsets / nchan;
+			mutex_init(&pi->vi_mirror_mutex);
+		}
+	}
+
 	/* map for msix */
 	ret = alloc_msix_info(adap, allocated);
 	if (ret)
@@ -5760,8 +6134,9 @@ static int enable_msix(struct adapter *adap)
 	}
 
 	dev_info(adap->pdev_dev,
-		 "%d MSI-X vectors allocated, nic %d eoqsets %d per uld %d\n",
-		 allocated, s->max_ethqsets, s->eoqsets, s->nqs_per_uld);
+		 "%d MSI-X vectors allocated, nic %d eoqsets %d per uld %d mirrorqsets %d\n",
+		 allocated, s->max_ethqsets, s->eoqsets, s->nqs_per_uld,
+		 s->mirrorqsets);
 
 	kfree(entries);
 	return 0;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
index ae0e998d9338..6251444ffa58 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.c
@@ -372,6 +372,7 @@ void cxgb4_process_flow_actions(struct net_device *in,
 		case FLOW_ACTION_DROP:
 			fs->action = FILTER_DROP;
 			break;
+		case FLOW_ACTION_MIRRED:
 		case FLOW_ACTION_REDIRECT: {
 			struct net_device *out = act->dev;
 			struct port_info *pi = netdev_priv(out);
@@ -529,7 +530,8 @@ static bool valid_pedit_action(struct net_device *dev,
 
 int cxgb4_validate_flow_actions(struct net_device *dev,
 				struct flow_action *actions,
-				struct netlink_ext_ack *extack)
+				struct netlink_ext_ack *extack,
+				u8 matchall_filter)
 {
 	struct flow_action_entry *act;
 	bool act_redir = false;
@@ -546,11 +548,19 @@ int cxgb4_validate_flow_actions(struct net_device *dev,
 		case FLOW_ACTION_DROP:
 			/* Do nothing */
 			break;
+		case FLOW_ACTION_MIRRED:
 		case FLOW_ACTION_REDIRECT: {
 			struct adapter *adap = netdev2adap(dev);
 			struct net_device *n_dev, *target_dev;
-			unsigned int i;
 			bool found = false;
+			unsigned int i;
+
+			if (act->id == FLOW_ACTION_MIRRED &&
+			    !matchall_filter) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "Egress mirror action is only supported for tc-matchall");
+				return -EOPNOTSUPP;
+			}
 
 			target_dev = act->dev;
 			for_each_port(adap, i) {
@@ -689,7 +699,7 @@ int cxgb4_flow_rule_replace(struct net_device *dev, struct flow_rule *rule,
 	u8 inet_family;
 	int fidx, ret;
 
-	if (cxgb4_validate_flow_actions(dev, &rule->action, extack))
+	if (cxgb4_validate_flow_actions(dev, &rule->action, extack, 0))
 		return -EOPNOTSUPP;
 
 	if (cxgb4_validate_flow_match(dev, rule))
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
index befa459324fb..6296e1d5a12b 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_flower.h
@@ -113,7 +113,8 @@ void cxgb4_process_flow_actions(struct net_device *in,
 				struct ch_filter_specification *fs);
 int cxgb4_validate_flow_actions(struct net_device *dev,
 				struct flow_action *actions,
-				struct netlink_ext_ack *extack);
+				struct netlink_ext_ack *extack,
+				u8 matchall_filter);
 
 int cxgb4_tc_flower_replace(struct net_device *dev,
 			    struct flow_cls_offload *cls);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
index c439b5bce9c9..e377e50c2492 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.c
@@ -188,6 +188,49 @@ static void cxgb4_matchall_free_tc(struct net_device *dev)
 	tc_port_matchall->egress.state = CXGB4_MATCHALL_STATE_DISABLED;
 }
 
+static int cxgb4_matchall_mirror_alloc(struct net_device *dev,
+				       struct tc_cls_matchall_offload *cls)
+{
+	struct netlink_ext_ack *extack = cls->common.extack;
+	struct cxgb4_tc_port_matchall *tc_port_matchall;
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+	struct flow_action_entry *act;
+	int ret;
+	u32 i;
+
+	tc_port_matchall = &adap->tc_matchall->port_matchall[pi->port_id];
+	flow_action_for_each(i, act, &cls->rule->action) {
+		if (act->id == FLOW_ACTION_MIRRED) {
+			ret = cxgb4_port_mirror_alloc(dev);
+			if (ret) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "Couldn't allocate mirror");
+				return ret;
+			}
+
+			tc_port_matchall->ingress.viid_mirror = pi->viid_mirror;
+			break;
+		}
+	}
+
+	return 0;
+}
+
+static void cxgb4_matchall_mirror_free(struct net_device *dev)
+{
+	struct cxgb4_tc_port_matchall *tc_port_matchall;
+	struct port_info *pi = netdev2pinfo(dev);
+	struct adapter *adap = netdev2adap(dev);
+
+	tc_port_matchall = &adap->tc_matchall->port_matchall[pi->port_id];
+	if (!tc_port_matchall->ingress.viid_mirror)
+		return;
+
+	cxgb4_port_mirror_free(dev);
+	tc_port_matchall->ingress.viid_mirror = 0;
+}
+
 static int cxgb4_matchall_alloc_filter(struct net_device *dev,
 				       struct tc_cls_matchall_offload *cls)
 {
@@ -211,6 +254,10 @@ static int cxgb4_matchall_alloc_filter(struct net_device *dev,
 		return -ENOMEM;
 	}
 
+	ret = cxgb4_matchall_mirror_alloc(dev, cls);
+	if (ret)
+		return ret;
+
 	tc_port_matchall = &adap->tc_matchall->port_matchall[pi->port_id];
 	fs = &tc_port_matchall->ingress.fs;
 	memset(fs, 0, sizeof(*fs));
@@ -229,11 +276,15 @@ static int cxgb4_matchall_alloc_filter(struct net_device *dev,
 
 	ret = cxgb4_set_filter(dev, fidx, fs);
 	if (ret)
-		return ret;
+		goto out_free;
 
 	tc_port_matchall->ingress.tid = fidx;
 	tc_port_matchall->ingress.state = CXGB4_MATCHALL_STATE_ENABLED;
 	return 0;
+
+out_free:
+	cxgb4_matchall_mirror_free(dev);
+	return ret;
 }
 
 static int cxgb4_matchall_free_filter(struct net_device *dev)
@@ -250,6 +301,8 @@ static int cxgb4_matchall_free_filter(struct net_device *dev)
 	if (ret)
 		return ret;
 
+	cxgb4_matchall_mirror_free(dev);
+
 	tc_port_matchall->ingress.packets = 0;
 	tc_port_matchall->ingress.bytes = 0;
 	tc_port_matchall->ingress.last_used = 0;
@@ -279,7 +332,7 @@ int cxgb4_tc_matchall_replace(struct net_device *dev,
 
 		ret = cxgb4_validate_flow_actions(dev,
 						  &cls_matchall->rule->action,
-						  extack);
+						  extack, 1);
 		if (ret)
 			return ret;
 
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
index ab6b5683dfd3..e264b6e606c4 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_tc_matchall.h
@@ -21,6 +21,7 @@ struct cxgb4_matchall_ingress_entry {
 	enum cxgb4_matchall_state state; /* Current MATCHALL offload state */
 	u32 tid; /* Index to hardware filter entry */
 	struct ch_filter_specification fs; /* Filter entry */
+	u16 viid_mirror; /* Identifier for allocated Mirror VI */
 	u64 bytes; /* # of bytes hitting the filter */
 	u64 packets; /* # of packets hitting the filter */
 	u64 last_used; /* Last updated jiffies time */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
index 70fe189202be..0af5ee9975df 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
@@ -7711,6 +7711,7 @@ int t4_free_vi(struct adapter *adap, unsigned int mbox, unsigned int pf,
  *	@adap: the adapter
  *	@mbox: mailbox to use for the FW command
  *	@viid: the VI id
+ *	@viid_mirror: the mirror VI id
  *	@mtu: the new MTU or -1
  *	@promisc: 1 to enable promiscuous mode, 0 to disable it, -1 no change
  *	@all_multi: 1 to enable all-multi mode, 0 to disable it, -1 no change
@@ -7721,10 +7722,11 @@ int t4_free_vi(struct adapter *adap, unsigned int mbox, unsigned int pf,
  *	Sets Rx properties of a virtual interface.
  */
 int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
-		  int mtu, int promisc, int all_multi, int bcast, int vlanex,
-		  bool sleep_ok)
+		  unsigned int viid_mirror, int mtu, int promisc, int all_multi,
+		  int bcast, int vlanex, bool sleep_ok)
 {
-	struct fw_vi_rxmode_cmd c;
+	struct fw_vi_rxmode_cmd c, c_mirror;
+	int ret;
 
 	/* convert to FW values */
 	if (mtu < 0)
@@ -7749,7 +7751,24 @@ int t4_set_rxmode(struct adapter *adap, unsigned int mbox, unsigned int viid,
 			    FW_VI_RXMODE_CMD_ALLMULTIEN_V(all_multi) |
 			    FW_VI_RXMODE_CMD_BROADCASTEN_V(bcast) |
 			    FW_VI_RXMODE_CMD_VLANEXEN_V(vlanex));
-	return t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), NULL, sleep_ok);
+
+	if (viid_mirror) {
+		memcpy(&c_mirror, &c, sizeof(c_mirror));
+		c_mirror.op_to_viid =
+			cpu_to_be32(FW_CMD_OP_V(FW_VI_RXMODE_CMD) |
+				    FW_CMD_REQUEST_F | FW_CMD_WRITE_F |
+				    FW_VI_RXMODE_CMD_VIID_V(viid_mirror));
+	}
+
+	ret = t4_wr_mbox_meat(adap, mbox, &c, sizeof(c), NULL, sleep_ok);
+	if (ret)
+		return ret;
+
+	if (viid_mirror)
+		ret = t4_wr_mbox_meat(adap, mbox, &c_mirror, sizeof(c_mirror),
+				      NULL, sleep_ok);
+
+	return ret;
 }
 
 /**
@@ -9711,6 +9730,22 @@ int t4_port_init(struct adapter *adap, int mbox, int pf, int vf)
 	return 0;
 }
 
+int t4_init_port_mirror(struct port_info *pi, u8 mbox, u8 port, u8 pf, u8 vf,
+			u16 *mirror_viid)
+{
+	int ret;
+
+	ret = t4_alloc_vi(pi->adapter, mbox, port, pf, vf, 1, NULL, NULL,
+			  NULL, NULL);
+	if (ret < 0)
+		return ret;
+
+	if (mirror_viid)
+		*mirror_viid = ret;
+
+	return 0;
+}
+
 /**
  *	t4_read_cimq_cfg - read CIM queue configuration
  *	@adap: the adapter