On Tue, Jul 15, 2025 at 07:14:21PM +0530, G Thomas, Rohan wrote:
> Hi Simon,
>
> Thanks for reviewing the patch.
>
> On 7/14/2025 7:10 PM, Simon Horman wrote:
> > On Mon, Jul 14, 2025 at 03:59:19PM +0800, Rohan G Thomas via B4 Relay wrote:
> > > From: Rohan G Thomas <rohan.g.thomas at altera.com>
> > >
> > > Currently, in the AF_XDP transmit paths, the CIC bit of
> > > TX Desc3 is set for all packets. Setting this bit for
> > > packets transmitting through queues that don't support
> > > checksum offloading causes the TX DMA to get stuck after
> > > transmitting some packets. This patch ensures the CIC bit
> > > of TX Desc3 is set only if the TX queue supports checksum
> > > offloading.
> > >
> > > Signed-off-by: Rohan G Thomas <rohan.g.thomas at altera.com>
> > > Reviewed-by: Matthew Gerlach <matthew.gerlach at altera.com>
> >
> > Hi Rohan,
> >
> > I notice that stmmac_xmit() handles a few other cases where
> > checksum offload should not be requested via stmmac_prepare_tx_desc:
> >
> > csum_insertion = (skb->ip_summed == CHECKSUM_PARTIAL);
> > /* DWMAC IPs can be synthesized to support tx coe only for a few tx
> > * queues. In that case, checksum offloading for those queues that don't
> > * support tx coe needs to fallback to software checksum calculation.
> > *
> > * Packets that won't trigger the COE e.g. most DSA-tagged packets will
> > * also have to be checksummed in software.
> > */
> > if (csum_insertion &&
> > (priv->plat->tx_queues_cfg[queue].coe_unsupported ||
> > !stmmac_has_ip_ethertype(skb))) {
> > if (unlikely(skb_checksum_help(skb)))
> > goto dma_map_err;
> > csum_insertion = !csum_insertion;
> > }
> >
> > Do we need to care about them in stmmac_xdp_xmit_zc()
> > and stmmac_xdp_xmit_xdpf() too?
>
> This patch only addresses avoiding the TX DMA hang by ensuring the CIC
> bit is only set when the queue supports checksum offload. For DSA tagged
> packets checksum offloading is not supported by the DWMAC IPs but no TX
> DMA hang. AFAIK, currently AF_XDP paths don't have equivalent handling
> like skb_checksum_help(), since they operate on xdp buffers. So this
> patch doesn't attempt to implement a sw fallback but just avoids DMA
> stall.
Ok, fair enough.
As per Andrew's advice elsewhere in this thread.
This patch also looks like it should be a fix for net,
and should have a Fixes tag.