Re: [PATCH net-next] net: fec: implement TSO descriptor cleanup

[Dheeraj Reddy Jonnalagadda]

On Fri, Jan 17, 2025 at 02:47:33AM +0000, Wei Fang wrote:
> > Implement the TODO in fec_enet_txq_submit_tso() error path to properly
> > release buffer descriptors that were allocated during a failed TSO
> > operation. This prevents descriptor leaks when TSO operations fail
> > partway through.
> > 
> > The cleanup iterates from the starting descriptor to where the error
> > occurred, resetting the status and buffer address fields of each
> > descriptor.
> > 
> > Signed-off-by: Dheeraj Reddy Jonnalagadda <dheeraj.linuxdev@xxxxxxxxx>
> > ---
> >  drivers/net/ethernet/freescale/fec_main.c | 13 ++++++++++++-
> >  1 file changed, 12 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/net/ethernet/freescale/fec_main.c
> > b/drivers/net/ethernet/freescale/fec_main.c
> > index b2daed55bf6c..eff065010c9e 100644
> > --- a/drivers/net/ethernet/freescale/fec_main.c
> > +++ b/drivers/net/ethernet/freescale/fec_main.c
> > @@ -913,7 +913,18 @@ static int fec_enet_txq_submit_tso(struct
> > fec_enet_priv_tx_q *txq,
> >  	return 0;
> > 
> >  err_release:
> > -	/* TODO: Release all used data descriptors for TSO */
> > +	/* Release all used data descriptors for TSO */
> > +	struct bufdesc *tmp_bdp = txq->bd.cur;
> > +
> > +	while (tmp_bdp != bdp) {
> > +		tmp_bdp->cbd_sc = 0;
> > +		tmp_bdp->cbd_bufaddr = 0;
> > +		tmp_bdp->cbd_datlen = 0;
> > +		tmp_bdp = fec_enet_get_nextdesc(tmp_bdp, &txq->bd);
> > +	}
> 
> There is still some cleanup to do.
> 1. If bufdesc_ex is used, we also need to clear it, such as ebdp->cbd_esc.
> 2. The data buffers have been mapped in fec_enet_txq_put_data_tso(),
> I think we need to unmap them in the error path. But do not unmap
> the TSO header buff, which is a DMA memory. Actually it is not necessary
> for fec_enet_txq_put_hdr_tso() to call dma_map_single(). If you are
> interested, you can add a separate patch to remove dma_map_single()
> in fec_enet_txq_put_hdr_tso().
> 
> > +
> > +	dev_kfree_skb_any(skb);
> > +
> >  	return ret;
> >  }
> > 
> > --
> > 2.34.1
> 
Hello Simon and Fang,

Thank you you valuable feedback. I will incorporate the suggestions and 
send the updated patch.

-Dheeraj