Re: [PATCH bpf-next v3 6/7] ice: xsk: improve AF_XDP ZC Tx and use batching API

[Maciej Fijalkowski <maciej.fijalkowski@intel.com>]

On Fri, Jan 21, 2022 at 02:17:42PM +0100, Alexander Lobakin wrote:
> From: Alexander Lobakin <alexandr.lobakin@intel.com>
> Date: Fri, 21 Jan 2022 13:54:47 +0100
> 
> > From: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> > Date: Fri, 21 Jan 2022 13:00:10 +0100
> > 
> > > Apply the logic that was done for regular XDP from commit 9610bd988df9
> > > ("ice: optimize XDP_TX workloads") to the ZC side of the driver. On top
> > > of that, introduce batching to Tx that is inspired by i40e's
> > > implementation with adjustments to the cleaning logic - take into the
> > > account NAPI budget in ice_clean_xdp_irq_zc().
> > > 
> > > Separating the stats structs onto separate cache lines seemed to improve
> > > the performance.
> > > 
> > > Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> > > ---
> > >  drivers/net/ethernet/intel/ice/ice_txrx.c |   2 +-
> > >  drivers/net/ethernet/intel/ice/ice_txrx.h |   2 +-
> > >  drivers/net/ethernet/intel/ice/ice_xsk.c  | 256 ++++++++++++++--------
> > >  drivers/net/ethernet/intel/ice/ice_xsk.h  |  27 ++-
> > >  4 files changed, 186 insertions(+), 101 deletions(-)
> > > 
> > > +/**
> > > + * ice_fill_tx_hw_ring - produce the number of Tx descriptors onto ring
> > > + * @xdp_ring: XDP ring to produce the HW Tx descriptors on
> > > + * @descs: AF_XDP descriptors to pull the DMA addresses and lengths from
> > > + * @nb_pkts: count of packets to be send
> > > + * @total_bytes: bytes accumulator that will be used for stats update
> > > + */
> > > +static void ice_fill_tx_hw_ring(struct ice_tx_ring *xdp_ring, struct xdp_desc *descs,
> > > +				u32 nb_pkts, unsigned int *total_bytes)
> > > +{
> > > +	u16 tx_thresh = xdp_ring->tx_thresh;
> > > +	struct ice_tx_desc *tx_desc;
> > 
> > And @tx_desc as well.
> > 
> > > +	u32 batched, leftover, i;
> > > +
> > > +	batched = nb_pkts & ~(PKTS_PER_BATCH - 1);
> > > +	leftover = nb_pkts & (PKTS_PER_BATCH - 1);
> > > +	for (i = 0; i < batched; i += PKTS_PER_BATCH)
> > > +		ice_xmit_pkt_batch(xdp_ring, &descs[i], total_bytes);
> > > +	for (i = batched; i < batched + leftover; i++)
> 
> Breh, I overlooked that. @i will equal @batched after exiting the
> first loop, so the assignment here is redundant (probably harmless
> tho if the compilers are smart enough).

I can drop this and scope variables properly, thanks!

> 
> > > +		ice_xmit_pkt(xdp_ring, &descs[i], total_bytes);
> > > +
> > > +	if (xdp_ring->next_to_use > xdp_ring->next_rs) {
> > > +		tx_desc = ICE_TX_DESC(xdp_ring, xdp_ring->next_rs);
> > > +		tx_desc->cmd_type_offset_bsz |=
> > > +			cpu_to_le64(ICE_TX_DESC_CMD_RS << ICE_TXD_QW1_CMD_S);
> > > +		xdp_ring->next_rs += tx_thresh;
> > > +	}
> > > +}
> > >  
> > > -		prefetch(tx_desc);
> > > +/**