This patch standardizes the inconsistent return values for unsuccessful
XDP transmits by using standardized error codes (-EBUSY or -ENOMEM).
Fixes: 26312c685ae0 ("net: fec: correct the counting of XDP sent frames")
Signed-off-by: Shenwei Wang <[email protected]>
---
v2:
- focusing on code clean up per Simon's feedback.
drivers/net/ethernet/freescale/fec_main.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 42ec6ca3bf03..6a021fe24dfe 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -3798,8 +3798,7 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
entries_free = fec_enet_get_free_txdesc_num(txq);
if (entries_free < MAX_SKB_FRAGS + 1) {
netdev_err(fep->netdev, "NOT enough BD for SG!\n");
- xdp_return_frame(frame);
- return NETDEV_TX_BUSY;
+ return -EBUSY;
}
/* Fill in a Tx ring entry */
@@ -3813,7 +3812,7 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
dma_addr = dma_map_single(&fep->pdev->dev, frame->data,
frame->len, DMA_TO_DEVICE);
if (dma_mapping_error(&fep->pdev->dev, dma_addr))
- return FEC_ENET_XDP_CONSUMED;
+ return -ENOMEM;
status |= (BD_ENET_TX_INTR | BD_ENET_TX_LAST);
if (fep->bufdesc_ex)
@@ -3869,7 +3868,7 @@ static int fec_enet_xdp_xmit(struct net_device *dev,
__netif_tx_lock(nq, cpu);
for (i = 0; i < num_frames; i++) {
- if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) != 0)
+ if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) < 0)
break;
sent_frames++;
}
--
2.34.1
The 05/10/2023 15:05, Shenwei Wang wrote:
>
> This patch standardizes the inconsistent return values for unsuccessful
> XDP transmits by using standardized error codes (-EBUSY or -ENOMEM).
Shouldn't this patch target net-next instead of net? As Simon suggested
here [1], or maybe is just me who misunderstood that part.
Also it is nice to CC people who comment at your previous patches in all
the next versions.
Just a small thing, if there is only 1 patch in the series, you don't
need to add 1/1 in the subject.
[1] https://lore.kernel.org/netdev/[email protected]/T/#m4b6b21c75512391496294fc78db2fbdf687f1381
>
> Fixes: 26312c685ae0 ("net: fec: correct the counting of XDP sent frames")
> Signed-off-by: Shenwei Wang <[email protected]>
> ---
> v2:
> - focusing on code clean up per Simon's feedback.
>
> drivers/net/ethernet/freescale/fec_main.c | 7 +++----
> 1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
> index 42ec6ca3bf03..6a021fe24dfe 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -3798,8 +3798,7 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
> entries_free = fec_enet_get_free_txdesc_num(txq);
> if (entries_free < MAX_SKB_FRAGS + 1) {
> netdev_err(fep->netdev, "NOT enough BD for SG!\n");
> - xdp_return_frame(frame);
> - return NETDEV_TX_BUSY;
> + return -EBUSY;
> }
>
> /* Fill in a Tx ring entry */
> @@ -3813,7 +3812,7 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
> dma_addr = dma_map_single(&fep->pdev->dev, frame->data,
> frame->len, DMA_TO_DEVICE);
> if (dma_mapping_error(&fep->pdev->dev, dma_addr))
> - return FEC_ENET_XDP_CONSUMED;
> + return -ENOMEM;
>
> status |= (BD_ENET_TX_INTR | BD_ENET_TX_LAST);
> if (fep->bufdesc_ex)
> @@ -3869,7 +3868,7 @@ static int fec_enet_xdp_xmit(struct net_device *dev,
> __netif_tx_lock(nq, cpu);
>
> for (i = 0; i < num_frames; i++) {
> - if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) != 0)
> + if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) < 0)
> break;
> sent_frames++;
> }
> --
> 2.34.1
>
>
--
/Horatiu
On Thu, May 11, 2023 at 09:24:52AM +0200, Horatiu Vultur wrote:
> The 05/10/2023 15:05, Shenwei Wang wrote:
> >
> > This patch standardizes the inconsistent return values for unsuccessful
> > XDP transmits by using standardized error codes (-EBUSY or -ENOMEM).
>
> Shouldn't this patch target net-next instead of net? As Simon suggested
> here [1], or maybe is just me who misunderstood that part.
> Also it is nice to CC people who comment at your previous patches in all
> the next versions.
>
> Just a small thing, if there is only 1 patch in the series, you don't
> need to add 1/1 in the subject.
>
> [1] https://lore.kernel.org/netdev/[email protected]/T/#m4b6b21c75512391496294fc78db2fbdf687f1381
>
> >
> > Fixes: 26312c685ae0 ("net: fec: correct the counting of XDP sent frames")
Hi Shenwei,
I agree with Horatiu.
Also, this is not a fix. So it should not have a Fixes tag.
After waiting for further review please send a v3 with these updates.
pw-bot: cr
> -----Original Message-----
> From: Horatiu Vultur <[email protected]>
> Sent: Thursday, May 11, 2023 2:25 AM
> To: Shenwei Wang <[email protected]>
> Cc: Wei Fang <[email protected]>; David S. Miller <[email protected]>;
> Eric Dumazet <[email protected]>; Jakub Kicinski <[email protected]>;
> Paolo Abeni <[email protected]>; Clark Wang <[email protected]>; dl-
> linux-imx <[email protected]>; Alexei Starovoitov <[email protected]>; Daniel
> Borkmann <[email protected]>; Jesper Dangaard Brouer
> <[email protected]>; John Fastabend <[email protected]>; Alexander
> Lobakin <[email protected]>; [email protected]; linux-
> [email protected]; [email protected]
> Subject: [EXT] Re: [PATCH v2 net 1/1] net: fec: using the standard return codes
> when xdp xmit errors
>
> Caution: This is an external email. Please take care when clicking links or
> opening attachments. When in doubt, report the message using the 'Report this
> email' button
>
>
> The 05/10/2023 15:05, Shenwei Wang wrote:
> >
> > This patch standardizes the inconsistent return values for
> > unsuccessful XDP transmits by using standardized error codes (-EBUSY or -
> ENOMEM).
>
> Shouldn't this patch target net-next instead of net? As Simon suggested here [1],
> or maybe is just me who misunderstood that part.
- xdp_return_frame(frame);
- return NETDEV_TX_BUSY;
+ return -EBUSY;
As the "xdp_return_frame(frame)" should not be called when error branch, the patch
fix it too.
> Also it is nice to CC people who comment at your previous patches in all the next
> versions.
>
> Just a small thing, if there is only 1 patch in the series, you don't need to add 1/1
> in the subject.
>
The patch was generated by using "git format-patch -n1". The strange thing is that
it appends "1/1" on some machines but not on others. The output seems inconsistent.
However, using "git format-patch -N1" can fix this.
Thanks,
Shenwei
> [1]
> https://lore.kern/
> el.org%2Fnetdev%2F20230509193845.1090040-1-
> shenwei.wang%40nxp.com%2FT%2F%23m4b6b21c75512391496294fc78db2fbd
> f687f1381&data=05%7C01%7Cshenwei.wang%40nxp.com%7Ca92a1dbe84f348
> 42e79e08db51f0d24f%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C
> 638193866998878533%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMD
> AiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&s
> data=MO7RbDtCUU82clB2CyltzpvEdc7%2FjwLWvcCzgGN8mDc%3D&reserved=0
>
> >
> > Fixes: 26312c685ae0 ("net: fec: correct the counting of XDP sent
> > frames")
> > Signed-off-by: Shenwei Wang <[email protected]>
> > ---
> > v2:
> > - focusing on code clean up per Simon's feedback.
> >
> > drivers/net/ethernet/freescale/fec_main.c | 7 +++----
> > 1 file changed, 3 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/freescale/fec_main.c
> > b/drivers/net/ethernet/freescale/fec_main.c
> > index 42ec6ca3bf03..6a021fe24dfe 100644
> > --- a/drivers/net/ethernet/freescale/fec_main.c
> > +++ b/drivers/net/ethernet/freescale/fec_main.c
> > @@ -3798,8 +3798,7 @@ static int fec_enet_txq_xmit_frame(struct
> fec_enet_private *fep,
> > entries_free = fec_enet_get_free_txdesc_num(txq);
> > if (entries_free < MAX_SKB_FRAGS + 1) {
> > netdev_err(fep->netdev, "NOT enough BD for SG!\n");
> > - xdp_return_frame(frame);
> > - return NETDEV_TX_BUSY;
> > + return -EBUSY;
> > }
> >
> > /* Fill in a Tx ring entry */
> > @@ -3813,7 +3812,7 @@ static int fec_enet_txq_xmit_frame(struct
> fec_enet_private *fep,
> > dma_addr = dma_map_single(&fep->pdev->dev, frame->data,
> > frame->len, DMA_TO_DEVICE);
> > if (dma_mapping_error(&fep->pdev->dev, dma_addr))
> > - return FEC_ENET_XDP_CONSUMED;
> > + return -ENOMEM;
> >
> > status |= (BD_ENET_TX_INTR | BD_ENET_TX_LAST);
> > if (fep->bufdesc_ex)
> > @@ -3869,7 +3868,7 @@ static int fec_enet_xdp_xmit(struct net_device *dev,
> > __netif_tx_lock(nq, cpu);
> >
> > for (i = 0; i < num_frames; i++) {
> > - if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) != 0)
> > + if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) < 0)
> > break;
> > sent_frames++;
> > }
> > --
> > 2.34.1
> >
> >
>
> --
> /Horatiu
On Wed, 10 May 2023 15:05:23 -0500 Shenwei Wang wrote:
> - xdp_return_frame(frame);
This line is a bug fix (double free).
I'm going to apply v2, it's good enough.
On Thu, 11 May 2023 17:55:42 -0700 Jakub Kicinski wrote:
> On Wed, 10 May 2023 15:05:23 -0500 Shenwei Wang wrote:
> > - xdp_return_frame(frame);
>
> This line is a bug fix (double free).
>
> I'm going to apply v2, it's good enough.
Let me take that back, I'll reply to v2.