data_len in skbuff represents bytes resident in fragment lists or
unmapped page buffers. For such packets, when data_len is non-zero,
skb_put cannot be used - this will throw a kernel bug. Hence do not
use macb_pad_and_fcs for such fragments.
Signed-off-by: Harini Katakam <[email protected]>
Signed-off-by: Michal Simek <[email protected]>
Signed-off-by: Radhey Shyam Pandey <[email protected]>
---
drivers/net/ethernet/cadence/macb_main.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
index 6434e74c04f1..0b03305ad6a0 100644
--- a/drivers/net/ethernet/cadence/macb_main.c
+++ b/drivers/net/ethernet/cadence/macb_main.c
@@ -1995,7 +1995,8 @@ static unsigned int macb_tx_map(struct macb *bp,
ctrl |= MACB_BF(TX_LSO, lso_ctrl);
ctrl |= MACB_BF(TX_TCP_SEQ_SRC, seq_ctrl);
if ((bp->dev->features & NETIF_F_HW_CSUM) &&
- skb->ip_summed != CHECKSUM_PARTIAL && !lso_ctrl)
+ skb->ip_summed != CHECKSUM_PARTIAL && !lso_ctrl &&
+ (skb->data_len == 0))
ctrl |= MACB_BIT(TX_NOCRC);
} else
/* Only set MSS/MFS on payload descriptors
@@ -2091,9 +2092,11 @@ static int macb_pad_and_fcs(struct sk_buff **skb, struct net_device *ndev)
struct sk_buff *nskb;
u32 fcs;
+ /* Not available for GSO and fragments */
if (!(ndev->features & NETIF_F_HW_CSUM) ||
!((*skb)->ip_summed != CHECKSUM_PARTIAL) ||
- skb_shinfo(*skb)->gso_size) /* Not available for GSO */
+ skb_shinfo(*skb)->gso_size ||
+ ((*skb)->data_len > 0))
return 0;
if (padlen <= 0) {
--
2.17.1
On Mon, 2022-05-09 at 17:45 +0530, Harini Katakam wrote:
> data_len in skbuff represents bytes resident in fragment lists or
> unmapped page buffers. For such packets, when data_len is non-zero,
> skb_put cannot be used - this will throw a kernel bug. Hence do not
> use macb_pad_and_fcs for such fragments.
>
> Signed-off-by: Harini Katakam <[email protected]>
> Signed-off-by: Michal Simek <[email protected]>
> Signed-off-by: Radhey Shyam Pandey <[email protected]>
This looks like a fix suitable for the net tree. Please add a relevant
'Fixes' tag.
> ---
> drivers/net/ethernet/cadence/macb_main.c | 7 +++++--
> 1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
> index 6434e74c04f1..0b03305ad6a0 100644
> --- a/drivers/net/ethernet/cadence/macb_main.c
> +++ b/drivers/net/ethernet/cadence/macb_main.c
> @@ -1995,7 +1995,8 @@ static unsigned int macb_tx_map(struct macb *bp,
> ctrl |= MACB_BF(TX_LSO, lso_ctrl);
> ctrl |= MACB_BF(TX_TCP_SEQ_SRC, seq_ctrl);
> if ((bp->dev->features & NETIF_F_HW_CSUM) &&
> - skb->ip_summed != CHECKSUM_PARTIAL && !lso_ctrl)
> + skb->ip_summed != CHECKSUM_PARTIAL && !lso_ctrl &&
> + (skb->data_len == 0))
> ctrl |= MACB_BIT(TX_NOCRC);
> } else
> /* Only set MSS/MFS on payload descriptors
This chunk looks unrelated to the commit message ?!? only the next one
looks relevant.
Thanks.
Paolo
Hi Paolo,
On Tue, May 10, 2022 at 6:48 PM Paolo Abeni <[email protected]> wrote:
>
> On Mon, 2022-05-09 at 17:45 +0530, Harini Katakam wrote:
> > data_len in skbuff represents bytes resident in fragment lists or
> > unmapped page buffers. For such packets, when data_len is non-zero,
> > skb_put cannot be used - this will throw a kernel bug. Hence do not
> > use macb_pad_and_fcs for such fragments.
> >
> > Signed-off-by: Harini Katakam <[email protected]>
> > Signed-off-by: Michal Simek <[email protected]>
> > Signed-off-by: Radhey Shyam Pandey <[email protected]>
>
> This looks like a fix suitable for the net tree. Please add a relevant
> 'Fixes' tag.
>
> > ---
> > drivers/net/ethernet/cadence/macb_main.c | 7 +++++--
> > 1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
> > index 6434e74c04f1..0b03305ad6a0 100644
> > --- a/drivers/net/ethernet/cadence/macb_main.c
> > +++ b/drivers/net/ethernet/cadence/macb_main.c
> > @@ -1995,7 +1995,8 @@ static unsigned int macb_tx_map(struct macb *bp,
> > ctrl |= MACB_BF(TX_LSO, lso_ctrl);
> > ctrl |= MACB_BF(TX_TCP_SEQ_SRC, seq_ctrl);
> > if ((bp->dev->features & NETIF_F_HW_CSUM) &&
> > - skb->ip_summed != CHECKSUM_PARTIAL && !lso_ctrl)
> > + skb->ip_summed != CHECKSUM_PARTIAL && !lso_ctrl &&
> > + (skb->data_len == 0))
> > ctrl |= MACB_BIT(TX_NOCRC);
> > } else
> > /* Only set MSS/MFS on payload descriptors
>
> This chunk looks unrelated to the commit message ?!? only the next one
> looks relevant.
Thanks for the review. This code is related to the commit message.
macb_pad_and_fcs is performed on the same packets on which
TX_NOCRC is set (instruct HW not to perform CRC and perform
CRC in SW). This is the patch which added both:
https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/commit/drivers/net/ethernet/cadence?id=653e92a9175ea7ed67efe209c725222051a3713d
I'll mention the same in Fixes tag and also add Claudiu's review tag.
Regards,
Harini