| Message ID | 08f8656e906ff69bd30915a6a37a01d5f0422194.1597842004.git.lorenzo@kernel.org |
|---|---|
| State | Changes Requested |
| Delegated to: | David Miller |
| Headers | show |
| Series | mvneta: introduce XDP multi-buffer support | expand |
On Wed, 19 Aug 2020 15:13:48 +0200 Lorenzo Bianconi <lorenzo@kernel.org> wrote: > Update multi-buffer bit (mb) in xdp_buff to notify XDP/eBPF layer and > XDP remote drivers if this is a "non-linear" XDP buffer > > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> > --- > drivers/net/ethernet/marvell/mvneta.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c > index 832bbb8b05c8..36a3defa63fa 100644 > --- a/drivers/net/ethernet/marvell/mvneta.c > +++ b/drivers/net/ethernet/marvell/mvneta.c > @@ -2170,11 +2170,14 @@ mvneta_run_xdp(struct mvneta_port *pp, struct mvneta_rx_queue *rxq, > struct bpf_prog *prog, struct xdp_buff *xdp, > u32 frame_sz, struct mvneta_stats *stats) > { > + struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); > unsigned int len, data_len, sync; > u32 ret, act; > > len = xdp->data_end - xdp->data_hard_start - pp->rx_offset_correction; > data_len = xdp->data_end - xdp->data; > + > + xdp->mb = !!sinfo->nr_frags; > act = bpf_prog_run_xdp(prog, xdp); Reading the memory sinfo->nr_frags could be a performance issue for our baseline case of no-multi-buffer. As you are reading a cache-line that you don't need to (and driver have not touch yet).
> On Wed, 19 Aug 2020 15:13:48 +0200 > Lorenzo Bianconi <lorenzo@kernel.org> wrote: > > > Update multi-buffer bit (mb) in xdp_buff to notify XDP/eBPF layer and > > XDP remote drivers if this is a "non-linear" XDP buffer > > > > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> > > --- > > drivers/net/ethernet/marvell/mvneta.c | 3 +++ > > 1 file changed, 3 insertions(+) > > > > diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c > > index 832bbb8b05c8..36a3defa63fa 100644 > > --- a/drivers/net/ethernet/marvell/mvneta.c > > +++ b/drivers/net/ethernet/marvell/mvneta.c > > @@ -2170,11 +2170,14 @@ mvneta_run_xdp(struct mvneta_port *pp, struct mvneta_rx_queue *rxq, > > struct bpf_prog *prog, struct xdp_buff *xdp, > > u32 frame_sz, struct mvneta_stats *stats) > > { > > + struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); > > unsigned int len, data_len, sync; > > u32 ret, act; > > > > len = xdp->data_end - xdp->data_hard_start - pp->rx_offset_correction; > > data_len = xdp->data_end - xdp->data; > > + > > + xdp->mb = !!sinfo->nr_frags; > > act = bpf_prog_run_xdp(prog, xdp); > > Reading the memory sinfo->nr_frags could be a performance issue for our > baseline case of no-multi-buffer. As you are reading a cache-line that > you don't need to (and driver have not touch yet). ack, I will rework it in v2 to remove this access. Regards, Lorenzo > > -- > Best regards, > Jesper Dangaard Brouer > MSc.CS, Principal Kernel Engineer at Red Hat > LinkedIn: http://www.linkedin.com/in/brouer >
On Wed, Aug 19, 2020 at 03:13:48PM +0200, Lorenzo Bianconi wrote: > Update multi-buffer bit (mb) in xdp_buff to notify XDP/eBPF layer and > XDP remote drivers if this is a "non-linear" XDP buffer > > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> > --- > drivers/net/ethernet/marvell/mvneta.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c > index 832bbb8b05c8..36a3defa63fa 100644 > --- a/drivers/net/ethernet/marvell/mvneta.c > +++ b/drivers/net/ethernet/marvell/mvneta.c > @@ -2170,11 +2170,14 @@ mvneta_run_xdp(struct mvneta_port *pp, struct mvneta_rx_queue *rxq, > struct bpf_prog *prog, struct xdp_buff *xdp, > u32 frame_sz, struct mvneta_stats *stats) > { > + struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); > unsigned int len, data_len, sync; > u32 ret, act; > > len = xdp->data_end - xdp->data_hard_start - pp->rx_offset_correction; > data_len = xdp->data_end - xdp->data; > + > + xdp->mb = !!sinfo->nr_frags; But this set is not utilizing it from BPF side in any way. Personally I would like to see this as a part of work where BPF program would actually be taught how to rely on xdp->mb. Especially after John's comment in other patch. > act = bpf_prog_run_xdp(prog, xdp); > > /* Due xdp_adjust_tail: DMA sync for_device cover max len CPU touch */ > -- > 2.26.2 >
On Aug 20, Maciej Fijalkowski wrote: > On Wed, Aug 19, 2020 at 03:13:48PM +0200, Lorenzo Bianconi wrote: > > Update multi-buffer bit (mb) in xdp_buff to notify XDP/eBPF layer and > > XDP remote drivers if this is a "non-linear" XDP buffer > > > > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> > > --- > > drivers/net/ethernet/marvell/mvneta.c | 3 +++ > > 1 file changed, 3 insertions(+) > > > > diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c > > index 832bbb8b05c8..36a3defa63fa 100644 > > --- a/drivers/net/ethernet/marvell/mvneta.c > > +++ b/drivers/net/ethernet/marvell/mvneta.c > > @@ -2170,11 +2170,14 @@ mvneta_run_xdp(struct mvneta_port *pp, struct mvneta_rx_queue *rxq, > > struct bpf_prog *prog, struct xdp_buff *xdp, > > u32 frame_sz, struct mvneta_stats *stats) > > { > > + struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); > > unsigned int len, data_len, sync; > > u32 ret, act; > > > > len = xdp->data_end - xdp->data_hard_start - pp->rx_offset_correction; > > data_len = xdp->data_end - xdp->data; > > + > > + xdp->mb = !!sinfo->nr_frags; > > But this set is not utilizing it from BPF side in any way. Personally I > would like to see this as a part of work where BPF program would actually > be taught how to rely on xdp->mb. Especially after John's comment in other > patch. > Sameeh is working on them. I did not include the patches since IMO they are not strictly related to this series. I will include Sameeh's patches in v2. Regards, Lorenzo > > act = bpf_prog_run_xdp(prog, xdp); > > > > /* Due xdp_adjust_tail: DMA sync for_device cover max len CPU touch */ > > -- > > 2.26.2 > > >
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c index 832bbb8b05c8..36a3defa63fa 100644 --- a/drivers/net/ethernet/marvell/mvneta.c +++ b/drivers/net/ethernet/marvell/mvneta.c @@ -2170,11 +2170,14 @@ mvneta_run_xdp(struct mvneta_port *pp, struct mvneta_rx_queue *rxq, struct bpf_prog *prog, struct xdp_buff *xdp, u32 frame_sz, struct mvneta_stats *stats) { + struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(xdp); unsigned int len, data_len, sync; u32 ret, act; len = xdp->data_end - xdp->data_hard_start - pp->rx_offset_correction; data_len = xdp->data_end - xdp->data; + + xdp->mb = !!sinfo->nr_frags; act = bpf_prog_run_xdp(prog, xdp); /* Due xdp_adjust_tail: DMA sync for_device cover max len CPU touch */
Update multi-buffer bit (mb) in xdp_buff to notify XDP/eBPF layer and XDP remote drivers if this is a "non-linear" XDP buffer Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> --- drivers/net/ethernet/marvell/mvneta.c | 3 +++ 1 file changed, 3 insertions(+)