Re: [PATCH net 2/2] bnxt_en: Linearize TX SKB if the fragments exceed the max

On Mon, May 5, 2025 at 8:37 AM Eric Dumazet [off-list ref] wrote:

On Fri, Mar 21, 2025 at 2:17 PM Michael Chan [off-list ref] wrote:

quoted

 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 11 +++++++++++
 drivers/net/ethernet/broadcom/bnxt/bnxt.h |  4 ++++
 2 files changed, 15 insertions(+)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 158e9789c1f4..2cd79b59cf00 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c

@@ -485,6 +485,17 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
        txr = &bp->tx_ring[bp->tx_ring_map[i]];
        prod = txr->tx_prod;

+#if (MAX_SKB_FRAGS > TX_MAX_FRAGS)
+       if (skb_shinfo(skb)->nr_frags > TX_MAX_FRAGS) {
+               netdev_warn_once(dev, "SKB has too many (%d) fragments, max supported is %d.  SKB will be linearized.\n",
+                                skb_shinfo(skb)->nr_frags, TX_MAX_FRAGS);
+               if (skb_linearize(skb)) {
+                       dev_kfree_skb_any(skb);
+                       dev_core_stats_tx_dropped_inc(dev);
+                       return NETDEV_TX_OK;
+               }
+       }
+#endif

Hi Michael

Sorry for the delay. I just saw your patch.

GSO would be more efficient and less likely to fail under memory pressure.

Could you test the following patch for me ?

Thanks !

The patch works.  I forced a smaller TX_MAX_FRAGS to easily test it.
But we now skip the warn_once() above since we intercept it earlier.
 Should we add a warn_once() here also?  Thanks.