diff mbox series

[net] nfp: fix issue of skb segments exceeds descriptor limitation

Message ID 20220708100718.791158-1-simon.horman@corigine.com (mailing list archive)
State Accepted
Commit 9c840d5f9aaef87e65db900bae21c70b059aba5f
Delegated to: Netdev Maintainers
Headers show
Series [net] nfp: fix issue of skb segments exceeds descriptor limitation | expand

Checks

Context Check Description
netdev/tree_selection success Clearly marked for net
netdev/fixes_present success Fixes tag present in non-next series
netdev/subject_prefix success Link
netdev/cover_letter success Single patches do not need cover letters
netdev/patch_count success Link
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 0 this patch: 0
netdev/cc_maintainers fail 3 blamed authors not CCed: fei.qin@corigine.com xingfeng.hu@corigine.com dianchao.wang@corigine.com; 5 maintainers not CCed: fei.qin@corigine.com xingfeng.hu@corigine.com yinjun.zhang@corigine.com dianchao.wang@corigine.com edumazet@google.com
netdev/build_clang success Errors and warnings before: 0 this patch: 0
netdev/module_param success Was 0 now: 0
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/check_selftest success No net selftest shell script
netdev/verify_fixes success Fixes tag looks correct
netdev/build_allmodconfig_warn success Errors and warnings before: 0 this patch: 0
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 80 lines checked
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

Simon Horman July 8, 2022, 10:07 a.m. UTC 
From: Baowen Zheng <baowen.zheng@corigine.com>

TCP packets will be dropped if the segments number in the tx skb
exceeds limitation when sending iperf3 traffic with --zerocopy option.

we make the following changes:

Get nr_frags in nfp_nfdk_tx_maybe_close_block instead of passing from
outside because it will be changed after skb_linearize operation.

Fill maximum dma_len in first tx descriptor to make sure the whole
head is included in the first descriptor.

Fixes: c10d12e3dce8 ("nfp: add support for NFDK data path")
Signed-off-by: Baowen Zheng <baowen.zheng@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/ethernet/netronome/nfp/nfdk/dp.c | 33 +++++++++++++++-----
 1 file changed, 25 insertions(+), 8 deletions(-)

Comments

patchwork-bot+netdevbpf@kernel.org July 9, 2022, 11:30 a.m. UTC | #1 
Hello:

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

On Fri,  8 Jul 2022 11:07:18 +0100 you wrote:
> From: Baowen Zheng <baowen.zheng@corigine.com>
> 
> TCP packets will be dropped if the segments number in the tx skb
> exceeds limitation when sending iperf3 traffic with --zerocopy option.
> 
> we make the following changes:
> 
> [...]

Here is the summary with links:
  - [net] nfp: fix issue of skb segments exceeds descriptor limitation
    https://git.kernel.org/netdev/net/c/9c840d5f9aae

You are awesome, thank you!
diff mbox series

Patch

diff --git a/drivers/net/ethernet/netronome/nfp/nfdk/dp.c b/drivers/net/ethernet/netronome/nfp/nfdk/dp.c
index e509d6dcba5c..805071d64a20 100644
--- a/drivers/net/ethernet/netronome/nfp/nfdk/dp.c
+++ b/drivers/net/ethernet/netronome/nfp/nfdk/dp.c
@@ -125,17 +125,18 @@  nfp_nfdk_tx_csum(struct nfp_net_dp *dp, struct nfp_net_r_vector *r_vec,
 
 static int
 nfp_nfdk_tx_maybe_close_block(struct nfp_net_tx_ring *tx_ring,
-			      unsigned int nr_frags, struct sk_buff *skb)
+			      struct sk_buff *skb)
 {
 	unsigned int n_descs, wr_p, nop_slots;
 	const skb_frag_t *frag, *fend;
 	struct nfp_nfdk_tx_desc *txd;
+	unsigned int nr_frags;
 	unsigned int wr_idx;
 	int err;
 
 recount_descs:
 	n_descs = nfp_nfdk_headlen_to_segs(skb_headlen(skb));
-
+	nr_frags = skb_shinfo(skb)->nr_frags;
 	frag = skb_shinfo(skb)->frags;
 	fend = frag + nr_frags;
 	for (; frag < fend; frag++)
@@ -281,10 +282,13 @@  netdev_tx_t nfp_nfdk_tx(struct sk_buff *skb, struct net_device *netdev)
 	if (unlikely((int)metadata < 0))
 		goto err_flush;
 
-	nr_frags = skb_shinfo(skb)->nr_frags;
-	if (nfp_nfdk_tx_maybe_close_block(tx_ring, nr_frags, skb))
+	if (nfp_nfdk_tx_maybe_close_block(tx_ring, skb))
 		goto err_flush;
 
+	/* nr_frags will change after skb_linearize so we get nr_frags after
+	 * nfp_nfdk_tx_maybe_close_block function
+	 */
+	nr_frags = skb_shinfo(skb)->nr_frags;
 	/* DMA map all */
 	wr_idx = D_IDX(tx_ring, tx_ring->wr_p);
 	txd = &tx_ring->ktxds[wr_idx];
@@ -310,7 +314,16 @@  netdev_tx_t nfp_nfdk_tx(struct sk_buff *skb, struct net_device *netdev)
 
 	/* FIELD_PREP() implicitly truncates to chunk */
 	dma_len -= 1;
-	dlen_type = FIELD_PREP(NFDK_DESC_TX_DMA_LEN_HEAD, dma_len) |
+
+	/* We will do our best to pass as much data as we can in descriptor
+	 * and we need to make sure the first descriptor includes whole head
+	 * since there is limitation in firmware side. Sometimes the value of
+	 * dma_len bitwise and NFDK_DESC_TX_DMA_LEN_HEAD will less than
+	 * headlen.
+	 */
+	dlen_type = FIELD_PREP(NFDK_DESC_TX_DMA_LEN_HEAD,
+			       dma_len > NFDK_DESC_TX_DMA_LEN_HEAD ?
+			       NFDK_DESC_TX_DMA_LEN_HEAD : dma_len) |
 		    FIELD_PREP(NFDK_DESC_TX_TYPE_HEAD, type);
 
 	txd->dma_len_type = cpu_to_le16(dlen_type);
@@ -925,7 +938,9 @@  nfp_nfdk_tx_xdp_buf(struct nfp_net_dp *dp, struct nfp_net_rx_ring *rx_ring,
 
 	/* FIELD_PREP() implicitly truncates to chunk */
 	dma_len -= 1;
-	dlen_type = FIELD_PREP(NFDK_DESC_TX_DMA_LEN_HEAD, dma_len) |
+	dlen_type = FIELD_PREP(NFDK_DESC_TX_DMA_LEN_HEAD,
+			       dma_len > NFDK_DESC_TX_DMA_LEN_HEAD ?
+			       NFDK_DESC_TX_DMA_LEN_HEAD : dma_len) |
 		    FIELD_PREP(NFDK_DESC_TX_TYPE_HEAD, type);
 
 	txd->dma_len_type = cpu_to_le16(dlen_type);
@@ -1303,7 +1318,7 @@  nfp_nfdk_ctrl_tx_one(struct nfp_net *nn, struct nfp_net_r_vector *r_vec,
 				   skb_push(skb, 4));
 	}
 
-	if (nfp_nfdk_tx_maybe_close_block(tx_ring, 0, skb))
+	if (nfp_nfdk_tx_maybe_close_block(tx_ring, skb))
 		goto err_free;
 
 	/* DMA map all */
@@ -1328,7 +1343,9 @@  nfp_nfdk_ctrl_tx_one(struct nfp_net *nn, struct nfp_net_r_vector *r_vec,
 	txbuf++;
 
 	dma_len -= 1;
-	dlen_type = FIELD_PREP(NFDK_DESC_TX_DMA_LEN_HEAD, dma_len) |
+	dlen_type = FIELD_PREP(NFDK_DESC_TX_DMA_LEN_HEAD,
+			       dma_len > NFDK_DESC_TX_DMA_LEN_HEAD ?
+			       NFDK_DESC_TX_DMA_LEN_HEAD : dma_len) |
 		    FIELD_PREP(NFDK_DESC_TX_TYPE_HEAD, type);
 
 	txd->dma_len_type = cpu_to_le16(dlen_type);