[PATCH net-next v2 0/2] net: enetc: add support for software TSO

[PATCH net-next v2 0/2] net: enetc: add support for software TSO

[Ioana Ciornei]

This series adds support for driver level TSO in the enetc driver.

Ever since the ENETC MDIO erratum workaround is in place, the Tx path is
incurring a penalty (enetc_lock_mdio/enetc_unlock_mdio) for each skb to
be sent out. On top of this, ENETC does not support Tx checksum
offloading. This means that software TSO would help performance just by
the fact that one single mdio lock/unlock sequence would cover multiple
packets sent. On the other hand, checksum needs to be computed in
software since the controller cannot handle it.

This is why, beside using the usual tso_build_hdr()/tso_build_data()
this specific implementation also has to compute the checksum, both IP
and L4, for each resulted segment.

Even with that, the performance improvement of a TCP flow running on a
single A72@1.3GHz of the LS1028A SoC (2.5Gbit/s port) is the following:

before: 1.63 Gbits/sec
after:  2.34 Gbits/sec

Changes in v2:
 - declare NETIF_F_HW_CSUM instead of NETIF_F_IP_CSUM in 1/2
 - add support for TSO over IPv6 (NETIF_F_TSO6 and csum compute) in 2/2

Ioana Ciornei (2):
  net: enetc: declare NETIF_F_HW_CSUM and do it in software
  net: enetc: add support for software TSO

 drivers/net/ethernet/freescale/enetc/enetc.c  | 317 +++++++++++++++++-
 drivers/net/ethernet/freescale/enetc/enetc.h  |   4 +
 .../net/ethernet/freescale/enetc/enetc_pf.c   |   8 +-
 .../net/ethernet/freescale/enetc/enetc_vf.c   |   8 +-
 4 files changed, 317 insertions(+), 20 deletions(-)

-- 
2.31.1

[PATCH net-next v2 1/2] net: enetc: declare NETIF_F_HW_CSUM and do it in software

[Ioana Ciornei]

This is just a preparation patch for software TSO in the enetc driver.
Unfortunately, ENETC does not support Tx checksum offload which would
normally render TSO, even software, impossible.

Declare NETIF_F_HW_CSUM as part of the feature set and do it at driver
level using skb_csum_hwoffload_help() so that we can move forward and
also add support for TSO in the next patch.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
---
Changes in v2:
 - declare NETIF_F_HW_CSUM instead of NETIF_F_IP_CSUM

 drivers/net/ethernet/freescale/enetc/enetc.c    | 8 +++++++-
 drivers/net/ethernet/freescale/enetc/enetc_pf.c | 6 ++++--
 drivers/net/ethernet/freescale/enetc/enetc_vf.c | 6 ++++--
 3 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/freescale/enetc/enetc.c b/drivers/net/ethernet/freescale/enetc/enetc.c
index 3cbfa8b4e265..17d8e04c10a8 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc.c
@@ -319,7 +319,7 @@ static netdev_tx_t enetc_start_xmit(struct sk_buff *skb,
 {
 	struct enetc_ndev_priv *priv = netdev_priv(ndev);
 	struct enetc_bdr *tx_ring;
-	int count;
+	int count, err;
 
 	/* Queue one-step Sync packet if already locked */
 	if (skb->cb[0] & ENETC_F_TX_ONESTEP_SYNC_TSTAMP) {
@@ -342,6 +342,12 @@ static netdev_tx_t enetc_start_xmit(struct sk_buff *skb,
 		return NETDEV_TX_BUSY;
 	}
 
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		err = skb_checksum_help(skb);
+		if (err)
+			goto drop_packet_err;
+	}
+
 	enetc_lock_mdio();
 	count = enetc_map_tx_buffs(tx_ring, skb);
 	enetc_unlock_mdio();
diff --git a/drivers/net/ethernet/freescale/enetc/enetc_pf.c b/drivers/net/ethernet/freescale/enetc/enetc_pf.c
index 628a74ba630c..1e3b2e191562 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc_pf.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc_pf.c
@@ -759,10 +759,12 @@ static void enetc_pf_netdev_setup(struct enetc_si *si, struct net_device *ndev,
 
 	ndev->hw_features = NETIF_F_SG | NETIF_F_RXCSUM |
 			    NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
-			    NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_LOOPBACK;
+			    NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_LOOPBACK |
+			    NETIF_F_HW_CSUM;
 	ndev->features = NETIF_F_HIGHDMA | NETIF_F_SG | NETIF_F_RXCSUM |
 			 NETIF_F_HW_VLAN_CTAG_TX |
-			 NETIF_F_HW_VLAN_CTAG_RX;
+			 NETIF_F_HW_VLAN_CTAG_RX |
+			 NETIF_F_HW_CSUM;
 
 	if (si->num_rss)
 		ndev->hw_features |= NETIF_F_RXHASH;
diff --git a/drivers/net/ethernet/freescale/enetc/enetc_vf.c b/drivers/net/ethernet/freescale/enetc/enetc_vf.c
index 0704f6bf12fd..f2a0c0f9fe1d 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc_vf.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc_vf.c
@@ -122,10 +122,12 @@ static void enetc_vf_netdev_setup(struct enetc_si *si, struct net_device *ndev,
 
 	ndev->hw_features = NETIF_F_SG | NETIF_F_RXCSUM |
 			    NETIF_F_HW_VLAN_CTAG_TX |
-			    NETIF_F_HW_VLAN_CTAG_RX;
+			    NETIF_F_HW_VLAN_CTAG_RX |
+			    NETIF_F_HW_CSUM;
 	ndev->features = NETIF_F_HIGHDMA | NETIF_F_SG | NETIF_F_RXCSUM |
 			 NETIF_F_HW_VLAN_CTAG_TX |
-			 NETIF_F_HW_VLAN_CTAG_RX;
+			 NETIF_F_HW_VLAN_CTAG_RX |
+			 NETIF_F_HW_CSUM;
 
 	if (si->num_rss)
 		ndev->hw_features |= NETIF_F_RXHASH;
-- 
2.31.1

[PATCH net-next v2 2/2] net: enetc: add support for software TSO

[Ioana Ciornei]

This patch adds support for driver level TSO in the enetc driver using
the TSO API.

Beside using the usual tso_build_hdr(), tso_build_data() this specific
implementation also has to compute the checksum, both IP and L4, for
each resulted segment. This is because the ENETC controller does not
support Tx checksum offload which is needed in order to take advantage
of TSO.

With the workaround for the ENETC MDIO erratum in place the Tx path of
the driver is forced to lock/unlock for each skb sent. This is why, even
though we are computing the checksum by hand we see the following
improvement in TCP termination on the LS1028A SoC, on a single A72 core
running at 1.3GHz:

before: 1.63 Gbits/sec
after:  2.34 Gbits/sec

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
---
Changes in v2:
 - add support for TSO over IPv6 (NETIF_F_TSO6 and csum compute)

 drivers/net/ethernet/freescale/enetc/enetc.c  | 319 ++++++++++++++++--
 drivers/net/ethernet/freescale/enetc/enetc.h  |   4 +
 .../net/ethernet/freescale/enetc/enetc_pf.c   |   6 +-
 .../net/ethernet/freescale/enetc/enetc_vf.c   |   6 +-
 4 files changed, 311 insertions(+), 24 deletions(-)

diff --git a/drivers/net/ethernet/freescale/enetc/enetc.c b/drivers/net/ethernet/freescale/enetc/enetc.c
index 17d8e04c10a8..09193b478ab3 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc.c
@@ -8,6 +8,7 @@
 #include <linux/vmalloc.h>
 #include <linux/ptp_classify.h>
 #include <net/pkt_sched.h>
+#include <net/tso.h>
 
 static int enetc_num_stack_tx_queues(struct enetc_ndev_priv *priv)
 {
@@ -314,6 +315,255 @@ static int enetc_map_tx_buffs(struct enetc_bdr *tx_ring, struct sk_buff *skb)
 	return 0;
 }
 
+static void enetc_map_tx_tso_hdr(struct enetc_bdr *tx_ring, struct sk_buff *skb,
+				 struct enetc_tx_swbd *tx_swbd,
+				 union enetc_tx_bd *txbd, int *i, int hdr_len,
+				 int data_len)
+{
+	union enetc_tx_bd txbd_tmp;
+	u8 flags = 0, e_flags = 0;
+	dma_addr_t addr;
+
+	enetc_clear_tx_bd(&txbd_tmp);
+	addr = tx_ring->tso_headers_dma + *i * TSO_HEADER_SIZE;
+
+	if (skb_vlan_tag_present(skb))
+		flags |= ENETC_TXBD_FLAGS_EX;
+
+	txbd_tmp.addr = cpu_to_le64(addr);
+	txbd_tmp.buf_len = cpu_to_le16(hdr_len);
+
+	/* first BD needs frm_len and offload flags set */
+	txbd_tmp.frm_len = cpu_to_le16(hdr_len + data_len);
+	txbd_tmp.flags = flags;
+
+	/* For the TSO header we do not set the dma address since we do not
+	 * want it unmapped when we do cleanup. We still set len so that we
+	 * count the bytes sent.
+	 */
+	tx_swbd->len = hdr_len;
+	tx_swbd->do_twostep_tstamp = false;
+	tx_swbd->check_wb = false;
+
+	/* Actually write the header in the BD */
+	*txbd = txbd_tmp;
+
+	/* Add extension BD for VLAN */
+	if (flags & ENETC_TXBD_FLAGS_EX) {
+		/* Get the next BD */
+		enetc_bdr_idx_inc(tx_ring, i);
+		txbd = ENETC_TXBD(*tx_ring, *i);
+		tx_swbd = &tx_ring->tx_swbd[*i];
+		prefetchw(txbd);
+
+		/* Setup the VLAN fields */
+		enetc_clear_tx_bd(&txbd_tmp);
+		txbd_tmp.ext.vid = cpu_to_le16(skb_vlan_tag_get(skb));
+		txbd_tmp.ext.tpid = 0; /* < C-TAG */
+		e_flags |= ENETC_TXBD_E_FLAGS_VLAN_INS;
+
+		/* Write the BD */
+		txbd_tmp.ext.e_flags = e_flags;
+		*txbd = txbd_tmp;
+	}
+}
+
+static int enetc_map_tx_tso_data(struct enetc_bdr *tx_ring, struct sk_buff *skb,
+				 struct enetc_tx_swbd *tx_swbd,
+				 union enetc_tx_bd *txbd, char *data,
+				 int size, bool last_bd)
+{
+	union enetc_tx_bd txbd_tmp;
+	dma_addr_t addr;
+	u8 flags = 0;
+
+	enetc_clear_tx_bd(&txbd_tmp);
+
+	addr = dma_map_single(tx_ring->dev, data, size, DMA_TO_DEVICE);
+	if (unlikely(dma_mapping_error(tx_ring->dev, addr))) {
+		netdev_err(tx_ring->ndev, "DMA map error\n");
+		return -ENOMEM;
+	}
+
+	if (last_bd) {
+		flags |= ENETC_TXBD_FLAGS_F;
+		tx_swbd->is_eof = 1;
+	}
+
+	txbd_tmp.addr = cpu_to_le64(addr);
+	txbd_tmp.buf_len = cpu_to_le16(size);
+	txbd_tmp.flags = flags;
+
+	tx_swbd->dma = addr;
+	tx_swbd->len = size;
+	tx_swbd->dir = DMA_TO_DEVICE;
+
+	*txbd = txbd_tmp;
+
+	return 0;
+}
+
+static __wsum enetc_tso_hdr_csum(struct tso_t *tso, struct sk_buff *skb,
+				 char *hdr, int hdr_len, int *l4_hdr_len)
+{
+	char *l4_hdr = hdr + skb_transport_offset(skb);
+	int mac_hdr_len = skb_network_offset(skb);
+
+	if (tso->tlen != sizeof(struct udphdr)) {
+		struct tcphdr *tcph = (struct tcphdr *)(l4_hdr);
+
+		tcph->check = 0;
+	} else {
+		struct udphdr *udph = (struct udphdr *)(l4_hdr);
+
+		udph->check = 0;
+	}
+
+	/* Compute the IP checksum. This is necessary since tso_build_hdr()
+	 * already incremented the IP ID field.
+	 */
+	if (!tso->ipv6) {
+		struct iphdr *iph = (void *)(hdr + mac_hdr_len);
+
+		iph->check = 0;
+		iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
+	}
+
+	/* Compute the checksum over the L4 header. */
+	*l4_hdr_len = hdr_len - skb_transport_offset(skb);
+	return csum_partial(l4_hdr, *l4_hdr_len, 0);
+}
+
+static void enetc_tso_complete_csum(struct enetc_bdr *tx_ring, struct tso_t *tso,
+				    struct sk_buff *skb, char *hdr, int len,
+				    __wsum sum)
+{
+	char *l4_hdr = hdr + skb_transport_offset(skb);
+	__sum16 csum_final;
+
+	/* Complete the L4 checksum by appending the pseudo-header to the
+	 * already computed checksum.
+	 */
+	if (!tso->ipv6)
+		csum_final = csum_tcpudp_magic(ip_hdr(skb)->saddr,
+					       ip_hdr(skb)->daddr,
+					       len, ip_hdr(skb)->protocol, sum);
+	else
+		csum_final = csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
+					     &ipv6_hdr(skb)->daddr,
+					     len, ipv6_hdr(skb)->nexthdr, sum);
+
+	if (tso->tlen != sizeof(struct udphdr)) {
+		struct tcphdr *tcph = (struct tcphdr *)(l4_hdr);
+
+		tcph->check = csum_final;
+	} else {
+		struct udphdr *udph = (struct udphdr *)(l4_hdr);
+
+		udph->check = csum_final;
+	}
+}
+
+static int enetc_map_tx_tso_buffs(struct enetc_bdr *tx_ring, struct sk_buff *skb)
+{
+	int hdr_len, total_len, data_len;
+	struct enetc_tx_swbd *tx_swbd;
+	union enetc_tx_bd *txbd;
+	struct tso_t tso;
+	__wsum csum, csum2;
+	int count = 0, pos;
+	int err, i, bd_data_num;
+
+	/* Initialize the TSO handler, and prepare the first payload */
+	hdr_len = tso_start(skb, &tso);
+	total_len = skb->len - hdr_len;
+	i = tx_ring->next_to_use;
+
+	while (total_len > 0) {
+		char *hdr;
+
+		/* Get the BD */
+		txbd = ENETC_TXBD(*tx_ring, i);
+		tx_swbd = &tx_ring->tx_swbd[i];
+		prefetchw(txbd);
+
+		/* Determine the length of this packet */
+		data_len = min_t(int, skb_shinfo(skb)->gso_size, total_len);
+		total_len -= data_len;
+
+		/* prepare packet headers: MAC + IP + TCP */
+		hdr = tx_ring->tso_headers + i * TSO_HEADER_SIZE;
+		tso_build_hdr(skb, hdr, &tso, data_len, total_len == 0);
+
+		/* compute the csum over the L4 header */
+		csum = enetc_tso_hdr_csum(&tso, skb, hdr, hdr_len, &pos);
+		enetc_map_tx_tso_hdr(tx_ring, skb, tx_swbd, txbd, &i, hdr_len, data_len);
+		bd_data_num = 0;
+		count++;
+
+		while (data_len > 0) {
+			int size;
+
+			size = min_t(int, tso.size, data_len);
+
+			/* Advance the index in the BDR */
+			enetc_bdr_idx_inc(tx_ring, &i);
+			txbd = ENETC_TXBD(*tx_ring, i);
+			tx_swbd = &tx_ring->tx_swbd[i];
+			prefetchw(txbd);
+
+			/* Compute the checksum over this segment of data and
+			 * add it to the csum already computed (over the L4
+			 * header and possible other data segments).
+			 */
+			csum2 = csum_partial(tso.data, size, 0);
+			csum = csum_block_add(csum, csum2, pos);
+			pos += size;
+
+			err = enetc_map_tx_tso_data(tx_ring, skb, tx_swbd, txbd,
+						    tso.data, size,
+						    size == data_len);
+			if (err)
+				goto err_map_data;
+
+			data_len -= size;
+			count++;
+			bd_data_num++;
+			tso_build_data(skb, &tso, size);
+
+			if (unlikely(bd_data_num >= ENETC_MAX_SKB_FRAGS && data_len))
+				goto err_chained_bd;
+		}
+
+		enetc_tso_complete_csum(tx_ring, &tso, skb, hdr, pos, csum);
+
+		if (total_len == 0)
+			tx_swbd->skb = skb;
+
+		/* Go to the next BD */
+		enetc_bdr_idx_inc(tx_ring, &i);
+	}
+
+	tx_ring->next_to_use = i;
+	enetc_update_tx_ring_tail(tx_ring);
+
+	return count;
+
+err_map_data:
+	dev_err(tx_ring->dev, "DMA map error");
+
+err_chained_bd:
+	do {
+		tx_swbd = &tx_ring->tx_swbd[i];
+		enetc_free_tx_frame(tx_ring, tx_swbd);
+		if (i == 0)
+			i = tx_ring->bd_count;
+		i--;
+	} while (count--);
+
+	return 0;
+}
+
 static netdev_tx_t enetc_start_xmit(struct sk_buff *skb,
 				    struct net_device *ndev)
 {
@@ -332,26 +582,36 @@ static netdev_tx_t enetc_start_xmit(struct sk_buff *skb,
 
 	tx_ring = priv->tx_ring[skb->queue_mapping];
 
-	if (unlikely(skb_shinfo(skb)->nr_frags > ENETC_MAX_SKB_FRAGS))
-		if (unlikely(skb_linearize(skb)))
-			goto drop_packet_err;
+	if (skb_is_gso(skb)) {
+		if (enetc_bd_unused(tx_ring) < tso_count_descs(skb)) {
+			netif_stop_subqueue(ndev, tx_ring->index);
+			return NETDEV_TX_BUSY;
+		}
 
-	count = skb_shinfo(skb)->nr_frags + 1; /* fragments + head */
-	if (enetc_bd_unused(tx_ring) < ENETC_TXBDS_NEEDED(count)) {
-		netif_stop_subqueue(ndev, tx_ring->index);
-		return NETDEV_TX_BUSY;
-	}
+		enetc_lock_mdio();
+		count = enetc_map_tx_tso_buffs(tx_ring, skb);
+		enetc_unlock_mdio();
+	} else {
+		if (unlikely(skb_shinfo(skb)->nr_frags > ENETC_MAX_SKB_FRAGS))
+			if (unlikely(skb_linearize(skb)))
+				goto drop_packet_err;
+
+		count = skb_shinfo(skb)->nr_frags + 1; /* fragments + head */
+		if (enetc_bd_unused(tx_ring) < ENETC_TXBDS_NEEDED(count)) {
+			netif_stop_subqueue(ndev, tx_ring->index);
+			return NETDEV_TX_BUSY;
+		}
 
-	if (skb->ip_summed == CHECKSUM_PARTIAL) {
-		err = skb_checksum_help(skb);
-		if (err)
-			goto drop_packet_err;
+		if (skb->ip_summed == CHECKSUM_PARTIAL) {
+			err = skb_checksum_help(skb);
+			if (err)
+				goto drop_packet_err;
+		}
+		enetc_lock_mdio();
+		count = enetc_map_tx_buffs(tx_ring, skb);
+		enetc_unlock_mdio();
 	}
 
-	enetc_lock_mdio();
-	count = enetc_map_tx_buffs(tx_ring, skb);
-	enetc_unlock_mdio();
-
 	if (unlikely(!count))
 		goto drop_packet_err;
 
@@ -1499,15 +1759,30 @@ static int enetc_alloc_txbdr(struct enetc_bdr *txr)
 		return -ENOMEM;
 
 	err = enetc_dma_alloc_bdr(txr, sizeof(union enetc_tx_bd));
-	if (err) {
-		vfree(txr->tx_swbd);
-		return err;
-	}
+	if (err)
+		goto err_alloc_bdr;
+
+	txr->tso_headers = dma_alloc_coherent(txr->dev,
+					      txr->bd_count * TSO_HEADER_SIZE,
+					      &txr->tso_headers_dma,
+					      GFP_KERNEL);
+	if (err)
+		goto err_alloc_tso;
 
 	txr->next_to_clean = 0;
 	txr->next_to_use = 0;
 
 	return 0;
+
+err_alloc_tso:
+	dma_free_coherent(txr->dev, txr->bd_count * sizeof(union enetc_tx_bd),
+			  txr->bd_base, txr->bd_dma_base);
+	txr->bd_base = NULL;
+err_alloc_bdr:
+	vfree(txr->tx_swbd);
+	txr->tx_swbd = NULL;
+
+	return err;
 }
 
 static void enetc_free_txbdr(struct enetc_bdr *txr)
@@ -1519,6 +1794,10 @@ static void enetc_free_txbdr(struct enetc_bdr *txr)
 
 	size = txr->bd_count * sizeof(union enetc_tx_bd);
 
+	dma_free_coherent(txr->dev, txr->bd_count * TSO_HEADER_SIZE,
+			  txr->tso_headers, txr->tso_headers_dma);
+	txr->tso_headers = NULL;
+
 	dma_free_coherent(txr->dev, size, txr->bd_base, txr->bd_dma_base);
 	txr->bd_base = NULL;
 
diff --git a/drivers/net/ethernet/freescale/enetc/enetc.h b/drivers/net/ethernet/freescale/enetc/enetc.h
index 08b283347d9c..fb39e406b7fc 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc.h
+++ b/drivers/net/ethernet/freescale/enetc/enetc.h
@@ -112,6 +112,10 @@ struct enetc_bdr {
 	dma_addr_t bd_dma_base;
 	u8 tsd_enable; /* Time specific departure */
 	bool ext_en; /* enable h/w descriptor extensions */
+
+	/* DMA buffer for TSO headers */
+	char *tso_headers;
+	dma_addr_t tso_headers_dma;
 } ____cacheline_aligned_in_smp;
 
 static inline void enetc_bdr_idx_inc(struct enetc_bdr *bdr, int *i)
diff --git a/drivers/net/ethernet/freescale/enetc/enetc_pf.c b/drivers/net/ethernet/freescale/enetc/enetc_pf.c
index 1e3b2e191562..8281dd664f4e 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc_pf.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc_pf.c
@@ -760,11 +760,13 @@ static void enetc_pf_netdev_setup(struct enetc_si *si, struct net_device *ndev,
 	ndev->hw_features = NETIF_F_SG | NETIF_F_RXCSUM |
 			    NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
 			    NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_LOOPBACK |
-			    NETIF_F_HW_CSUM;
+			    NETIF_F_HW_CSUM | NETIF_F_TSO | NETIF_F_TSO6;
 	ndev->features = NETIF_F_HIGHDMA | NETIF_F_SG | NETIF_F_RXCSUM |
 			 NETIF_F_HW_VLAN_CTAG_TX |
 			 NETIF_F_HW_VLAN_CTAG_RX |
-			 NETIF_F_HW_CSUM;
+			 NETIF_F_HW_CSUM | NETIF_F_TSO | NETIF_F_TSO6;
+	ndev->vlan_features = NETIF_F_SG | NETIF_F_HW_CSUM |
+			      NETIF_F_TSO | NETIF_F_TSO6;
 
 	if (si->num_rss)
 		ndev->hw_features |= NETIF_F_RXHASH;
diff --git a/drivers/net/ethernet/freescale/enetc/enetc_vf.c b/drivers/net/ethernet/freescale/enetc/enetc_vf.c
index f2a0c0f9fe1d..df312c9f8243 100644
--- a/drivers/net/ethernet/freescale/enetc/enetc_vf.c
+++ b/drivers/net/ethernet/freescale/enetc/enetc_vf.c
@@ -123,11 +123,13 @@ static void enetc_vf_netdev_setup(struct enetc_si *si, struct net_device *ndev,
 	ndev->hw_features = NETIF_F_SG | NETIF_F_RXCSUM |
 			    NETIF_F_HW_VLAN_CTAG_TX |
 			    NETIF_F_HW_VLAN_CTAG_RX |
-			    NETIF_F_HW_CSUM;
+			    NETIF_F_HW_CSUM | NETIF_F_TSO | NETIF_F_TSO6;
 	ndev->features = NETIF_F_HIGHDMA | NETIF_F_SG | NETIF_F_RXCSUM |
 			 NETIF_F_HW_VLAN_CTAG_TX |
 			 NETIF_F_HW_VLAN_CTAG_RX |
-			 NETIF_F_HW_CSUM;
+			 NETIF_F_HW_CSUM | NETIF_F_TSO | NETIF_F_TSO6;
+	ndev->vlan_features = NETIF_F_SG | NETIF_F_HW_CSUM |
+			      NETIF_F_TSO | NETIF_F_TSO6;
 
 	if (si->num_rss)
 		ndev->hw_features |= NETIF_F_RXHASH;
-- 
2.31.1

RE: [PATCH net-next v2 2/2] net: enetc: add support for software TSO

[Claudiu Manoil]

> -----Original Message-----
> From: Ioana Ciornei <ioana.ciornei@nxp.com>
> Sent: Thursday, October 7, 2021 6:31 PM
[...]
> Subject: [PATCH net-next v2 2/2] net: enetc: add support for software TSO
> 
> This patch adds support for driver level TSO in the enetc driver using
> the TSO API.
> 
> Beside using the usual tso_build_hdr(), tso_build_data() this specific
> implementation also has to compute the checksum, both IP and L4, for
> each resulted segment. This is because the ENETC controller does not
> support Tx checksum offload which is needed in order to take advantage
> of TSO.
> 
> With the workaround for the ENETC MDIO erratum in place the Tx path of
> the driver is forced to lock/unlock for each skb sent. This is why, even
> though we are computing the checksum by hand we see the following
> improvement in TCP termination on the LS1028A SoC, on a single A72 core
> running at 1.3GHz:
> 
> before: 1.63 Gbits/sec
> after:  2.34 Gbits/sec
> 
> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
> ---
> Changes in v2:
>  - add support for TSO over IPv6 (NETIF_F_TSO6 and csum compute)
> 

Reviewed-by: Claudiu Manoil <claudiu.manoil@nxp.com>

Re: [PATCH net-next v2 1/2] net: enetc: declare NETIF_F_HW_CSUM and do it in software

[Vladimir Oltean]

On Thu, Oct 07, 2021 at 06:30:42PM +0300, Ioana Ciornei wrote:
> This is just a preparation patch for software TSO in the enetc driver.
> Unfortunately, ENETC does not support Tx checksum offload which would
> normally render TSO, even software, impossible.
> 
> Declare NETIF_F_HW_CSUM as part of the feature set and do it at driver
> level using skb_csum_hwoffload_help() so that we can move forward and

The comment is still referencing the other function you were calling in v1.
Anyway, not a big deal, in the end everything ends up in skb_checksum_help().

> also add support for TSO in the next patch.
> 
> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
> ---

Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>

Re: [PATCH net-next v2 2/2] net: enetc: add support for software TSO

[Vladimir Oltean]

On Thu, Oct 07, 2021 at 06:30:43PM +0300, Ioana Ciornei wrote:
> This patch adds support for driver level TSO in the enetc driver using
> the TSO API.
> 
> Beside using the usual tso_build_hdr(), tso_build_data() this specific
> implementation also has to compute the checksum, both IP and L4, for
> each resulted segment. This is because the ENETC controller does not
> support Tx checksum offload which is needed in order to take advantage
> of TSO.
> 
> With the workaround for the ENETC MDIO erratum in place the Tx path of
> the driver is forced to lock/unlock for each skb sent. This is why, even
> though we are computing the checksum by hand we see the following
> improvement in TCP termination on the LS1028A SoC, on a single A72 core
> running at 1.3GHz:
> 
> before: 1.63 Gbits/sec
> after:  2.34 Gbits/sec
> 
> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
> ---

Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>

Re: [PATCH net-next v2 0/2] net: enetc: add support for software TSO

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (master)
by David S. Miller <davem@davemloft.net>:

On Thu,  7 Oct 2021 18:30:41 +0300 you wrote:
> This series adds support for driver level TSO in the enetc driver.
> 
> Ever since the ENETC MDIO erratum workaround is in place, the Tx path is
> incurring a penalty (enetc_lock_mdio/enetc_unlock_mdio) for each skb to
> be sent out. On top of this, ENETC does not support Tx checksum
> offloading. This means that software TSO would help performance just by
> the fact that one single mdio lock/unlock sequence would cover multiple
> packets sent. On the other hand, checksum needs to be computed in
> software since the controller cannot handle it.
> 
> [...]

Here is the summary with links:
  - [net-next,v2,1/2] net: enetc: declare NETIF_F_HW_CSUM and do it in software
    https://git.kernel.org/netdev/net-next/c/acede3c5dad5
  - [net-next,v2,2/2] net: enetc: add support for software TSO
    https://git.kernel.org/netdev/net-next/c/fb8629e2cbfc

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html