[PATCH net-next v3 1/2] net: mana: Handle SKB if TX SGEs exceed hardware limit

[Aditya Garg <gargaditya@linux.microsoft.com>]

The MANA hardware supports a maximum of 30 scatter-gather entries (SGEs)
per TX WQE. Exceeding this limit can cause TX failures.
Add ndo_features_check() callback to validate SKB layout before
transmission. For GSO SKBs that would exceed the hardware SGE limit, clear
NETIF_F_GSO_MASK to enforce software segmentation in the stack.
Add a fallback in mana_start_xmit() to linearize non-GSO SKBs that still
exceed the SGE limit.

Also, Add ethtool counter for SKBs linearized

Co-developed-by: Dipayaan Roy <dipayanroy@linux.microsoft.com>
Signed-off-by: Dipayaan Roy <dipayanroy@linux.microsoft.com>
Signed-off-by: Aditya Garg <gargaditya@linux.microsoft.com>
---
 drivers/net/ethernet/microsoft/mana/mana_en.c | 37 ++++++++++++++++++-
 .../ethernet/microsoft/mana/mana_ethtool.c    |  2 +
 include/net/mana/gdma.h                       |  6 ++-
 include/net/mana/mana.h                       |  1 +
 4 files changed, 43 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/microsoft/mana/mana_en.c b/drivers/net/ethernet/microsoft/mana/mana_en.c
index cccd5b63cee6..67ae5421f9ee 100644
--- a/drivers/net/ethernet/microsoft/mana/mana_en.c
+++ b/drivers/net/ethernet/microsoft/mana/mana_en.c
@@ -11,6 +11,7 @@
 #include <linux/mm.h>
 #include <linux/pci.h>
 #include <linux/export.h>
+#include <linux/skbuff.h>
 
 #include <net/checksum.h>
 #include <net/ip6_checksum.h>
@@ -329,6 +330,20 @@ netdev_tx_t mana_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 	cq = &apc->tx_qp[txq_idx].tx_cq;
 	tx_stats = &txq->stats;
 
+	if (MAX_SKB_FRAGS + 2 > MAX_TX_WQE_SGL_ENTRIES &&
+	    skb_shinfo(skb)->nr_frags + 2 > MAX_TX_WQE_SGL_ENTRIES) {
+		/* GSO skb with Hardware SGE limit exceeded is not expected here
+		 * as they are handled in mana_features_check() callback
+		 */
+		if (skb_linearize(skb)) {
+			netdev_warn_once(ndev, "Failed to linearize skb with nr_frags=%d and is_gso=%d\n",
+					 skb_shinfo(skb)->nr_frags,
+					 skb_is_gso(skb));
+			goto tx_drop_count;
+		}
+		apc->eth_stats.linear_pkt_tx_cnt++;
+	}
+
 	pkg.tx_oob.s_oob.vcq_num = cq->gdma_id;
 	pkg.tx_oob.s_oob.vsq_frame = txq->vsq_frame;
 
@@ -442,8 +457,6 @@ netdev_tx_t mana_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 		}
 	}
 
-	WARN_ON_ONCE(pkg.wqe_req.num_sge > MAX_TX_WQE_SGL_ENTRIES);
-
 	if (pkg.wqe_req.num_sge <= ARRAY_SIZE(pkg.sgl_array)) {
 		pkg.wqe_req.sgl = pkg.sgl_array;
 	} else {
@@ -518,6 +531,25 @@ netdev_tx_t mana_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 	return NETDEV_TX_OK;
 }
 
+static netdev_features_t mana_features_check(struct sk_buff *skb,
+					     struct net_device *ndev,
+					     netdev_features_t features)
+{
+	if (MAX_SKB_FRAGS + 2 > MAX_TX_WQE_SGL_ENTRIES &&
+	    skb_shinfo(skb)->nr_frags + 2 > MAX_TX_WQE_SGL_ENTRIES) {
+		/* Exceeds HW SGE limit.
+		 * GSO case:
+		 *   Disable GSO so the stack will software-segment the skb
+		 *   into smaller skbs that fit the SGE budget.
+		 * Non-GSO case:
+		 *   The xmit path will attempt skb_linearize() as a fallback.
+		 */
+		if (skb_is_gso(skb))
+			features &= ~NETIF_F_GSO_MASK;
+	}
+	return features;
+}
+
 static void mana_get_stats64(struct net_device *ndev,
 			     struct rtnl_link_stats64 *st)
 {
@@ -878,6 +910,7 @@ static const struct net_device_ops mana_devops = {
 	.ndo_open		= mana_open,
 	.ndo_stop		= mana_close,
 	.ndo_select_queue	= mana_select_queue,
+	.ndo_features_check	= mana_features_check,
 	.ndo_start_xmit		= mana_start_xmit,
 	.ndo_validate_addr	= eth_validate_addr,
 	.ndo_get_stats64	= mana_get_stats64,
diff --git a/drivers/net/ethernet/microsoft/mana/mana_ethtool.c b/drivers/net/ethernet/microsoft/mana/mana_ethtool.c
index a1afa75a9463..fa5e1a2f06a9 100644
--- a/drivers/net/ethernet/microsoft/mana/mana_ethtool.c
+++ b/drivers/net/ethernet/microsoft/mana/mana_ethtool.c
@@ -71,6 +71,8 @@ static const struct mana_stats_desc mana_eth_stats[] = {
 	{"tx_cq_err", offsetof(struct mana_ethtool_stats, tx_cqe_err)},
 	{"tx_cqe_unknown_type", offsetof(struct mana_ethtool_stats,
 					tx_cqe_unknown_type)},
+	{"linear_pkt_tx_cnt", offsetof(struct mana_ethtool_stats,
+					linear_pkt_tx_cnt)},
 	{"rx_coalesced_err", offsetof(struct mana_ethtool_stats,
 					rx_coalesced_err)},
 	{"rx_cqe_unknown_type", offsetof(struct mana_ethtool_stats,
diff --git a/include/net/mana/gdma.h b/include/net/mana/gdma.h
index 637f42485dba..84614ebe0f4c 100644
--- a/include/net/mana/gdma.h
+++ b/include/net/mana/gdma.h
@@ -592,6 +592,9 @@ enum {
 #define GDMA_DRV_CAP_FLAG_1_HANDLE_RECONFIG_EQE BIT(17)
 #define GDMA_DRV_CAP_FLAG_1_HW_VPORT_LINK_AWARE BIT(6)
 
+/* Driver supports linearizing the skb when num_sge exceeds hardware limit */
+#define GDMA_DRV_CAP_FLAG_1_SKB_LINEARIZE BIT(20)
+
 #define GDMA_DRV_CAP_FLAGS1 \
 	(GDMA_DRV_CAP_FLAG_1_EQ_SHARING_MULTI_VPORT | \
 	 GDMA_DRV_CAP_FLAG_1_NAPI_WKDONE_FIX | \
@@ -601,7 +604,8 @@ enum {
 	 GDMA_DRV_CAP_FLAG_1_DYNAMIC_IRQ_ALLOC_SUPPORT | \
 	 GDMA_DRV_CAP_FLAG_1_SELF_RESET_ON_EQE | \
 	 GDMA_DRV_CAP_FLAG_1_HANDLE_RECONFIG_EQE | \
-	 GDMA_DRV_CAP_FLAG_1_HW_VPORT_LINK_AWARE)
+	 GDMA_DRV_CAP_FLAG_1_HW_VPORT_LINK_AWARE | \
+	 GDMA_DRV_CAP_FLAG_1_SKB_LINEARIZE)
 
 #define GDMA_DRV_CAP_FLAGS2 0
 
diff --git a/include/net/mana/mana.h b/include/net/mana/mana.h
index 8906901535f5..50a532fb30d6 100644
--- a/include/net/mana/mana.h
+++ b/include/net/mana/mana.h
@@ -404,6 +404,7 @@ struct mana_ethtool_stats {
 	u64 hc_tx_err_gdma;
 	u64 tx_cqe_err;
 	u64 tx_cqe_unknown_type;
+	u64 linear_pkt_tx_cnt;
 	u64 rx_coalesced_err;
 	u64 rx_cqe_unknown_type;
 };
-- 
2.43.0