| Message ID | 62165fcacf47521edae67ae739827aa5f751fb8b.1603185591.git.lorenzo@kernel.org |
|---|---|
| State | Not Applicable |
| Delegated to: | BPF Maintainers |
| Headers | show |
| Series | xdp: introduce bulking for page_pool tx return path | expand |
| Context | Check | Description |
|---|---|---|
| jkicinski/cover_letter | success | Link |
| jkicinski/fixes_present | success | Link |
| jkicinski/patch_count | success | Link |
| jkicinski/tree_selection | success | Guessed tree name to be net-next |
| jkicinski/subject_prefix | warning | Target tree name not specified in the subject |
| jkicinski/source_inline | success | Was 0 now: 0 |
| jkicinski/verify_signedoff | success | Link |
| jkicinski/module_param | success | Was 0 now: 0 |
| jkicinski/build_32bit | success | Errors and warnings before: 7921 this patch: 7921 |
| jkicinski/kdoc | success | Errors and warnings before: 0 this patch: 0 |
| jkicinski/verify_fixes | success | Link |
| jkicinski/checkpatch | success | total: 0 errors, 0 warnings, 0 checks, 101 lines checked |
| jkicinski/build_allmodconfig_warn | success | Errors and warnings before: 8419 this patch: 8419 |
| jkicinski/header_inline | success | Link |
| jkicinski/stable | success | Stable not CCed |
On Tue, 20 Oct 2020 11:33:37 +0200 Lorenzo Bianconi <lorenzo@kernel.org> wrote: > diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c > index 54b0bf574c05..af33cc62ed4c 100644 > --- a/drivers/net/ethernet/marvell/mvneta.c > +++ b/drivers/net/ethernet/marvell/mvneta.c > @@ -663,6 +663,8 @@ struct mvneta_tx_queue { > > /* Affinity mask for CPUs*/ > cpumask_t affinity_mask; > + > + struct xdp_frame_bulk bq; > }; > > struct mvneta_rx_queue { > @@ -1854,12 +1856,10 @@ static void mvneta_txq_bufs_free(struct mvneta_port *pp, > dev_kfree_skb_any(buf->skb); > } else if (buf->type == MVNETA_TYPE_XDP_TX || > buf->type == MVNETA_TYPE_XDP_NDO) { > - if (napi && buf->type == MVNETA_TYPE_XDP_TX) > - xdp_return_frame_rx_napi(buf->xdpf); > - else > - xdp_return_frame(buf->xdpf); > + xdp_return_frame_bulk(buf->xdpf, &txq->bq, napi); Hmm, I don't think you can use 'napi' directly here. You are circumventing check (buf->type == MVNETA_TYPE_XDP_TX), and will now also allow XDP_NDO (XDP_REDIRECT) to basically use xdp_return_frame_rx_napi(). > } > } > + xdp_flush_frame_bulk(&txq->bq, napi); > > netdev_tx_completed_queue(nq, pkts_compl, bytes_compl); > }
> On Tue, 20 Oct 2020 11:33:37 +0200 > Lorenzo Bianconi <lorenzo@kernel.org> wrote: > > > diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c > > index 54b0bf574c05..af33cc62ed4c 100644 > > --- a/drivers/net/ethernet/marvell/mvneta.c > > +++ b/drivers/net/ethernet/marvell/mvneta.c > > @@ -663,6 +663,8 @@ struct mvneta_tx_queue { > > > > /* Affinity mask for CPUs*/ > > cpumask_t affinity_mask; > > + > > + struct xdp_frame_bulk bq; > > }; > > > > struct mvneta_rx_queue { > > @@ -1854,12 +1856,10 @@ static void mvneta_txq_bufs_free(struct mvneta_port *pp, > > dev_kfree_skb_any(buf->skb); > > } else if (buf->type == MVNETA_TYPE_XDP_TX || > > buf->type == MVNETA_TYPE_XDP_NDO) { > > - if (napi && buf->type == MVNETA_TYPE_XDP_TX) > > - xdp_return_frame_rx_napi(buf->xdpf); > > - else > > - xdp_return_frame(buf->xdpf); > > + xdp_return_frame_bulk(buf->xdpf, &txq->bq, napi); > > Hmm, I don't think you can use 'napi' directly here. > > You are circumventing check (buf->type == MVNETA_TYPE_XDP_TX), and will > now also allow XDP_NDO (XDP_REDIRECT) to basically use xdp_return_frame_rx_napi(). ack, right. I will fix it. Regards, Lorenzo > > > > } > > } > > + xdp_flush_frame_bulk(&txq->bq, napi); > > > > netdev_tx_completed_queue(nq, pkts_compl, bytes_compl); > > } > > > > -- > Best regards, > Jesper Dangaard Brouer > MSc.CS, Principal Kernel Engineer at Red Hat > LinkedIn: http://www.linkedin.com/in/brouer >
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c index 54b0bf574c05..af33cc62ed4c 100644 --- a/drivers/net/ethernet/marvell/mvneta.c +++ b/drivers/net/ethernet/marvell/mvneta.c @@ -663,6 +663,8 @@ struct mvneta_tx_queue { /* Affinity mask for CPUs*/ cpumask_t affinity_mask; + + struct xdp_frame_bulk bq; }; struct mvneta_rx_queue { @@ -1854,12 +1856,10 @@ static void mvneta_txq_bufs_free(struct mvneta_port *pp, dev_kfree_skb_any(buf->skb); } else if (buf->type == MVNETA_TYPE_XDP_TX || buf->type == MVNETA_TYPE_XDP_NDO) { - if (napi && buf->type == MVNETA_TYPE_XDP_TX) - xdp_return_frame_rx_napi(buf->xdpf); - else - xdp_return_frame(buf->xdpf); + xdp_return_frame_bulk(buf->xdpf, &txq->bq, napi); } } + xdp_flush_frame_bulk(&txq->bq, napi); netdev_tx_completed_queue(nq, pkts_compl, bytes_compl); } diff --git a/include/net/xdp.h b/include/net/xdp.h index 3814fb631d52..4b79d50afe36 100644 --- a/include/net/xdp.h +++ b/include/net/xdp.h @@ -104,6 +104,12 @@ struct xdp_frame { struct net_device *dev_rx; /* used by cpumap */ }; +#define XDP_BULK_QUEUE_SIZE 16 +struct xdp_frame_bulk { + void *q[XDP_BULK_QUEUE_SIZE]; + int count; + void *xa; +}; static inline struct skb_shared_info * xdp_get_shared_info_from_frame(struct xdp_frame *frame) @@ -194,6 +200,11 @@ struct xdp_frame *xdp_convert_buff_to_frame(struct xdp_buff *xdp) void xdp_return_frame(struct xdp_frame *xdpf); void xdp_return_frame_rx_napi(struct xdp_frame *xdpf); void xdp_return_buff(struct xdp_buff *xdp); +void xdp_flush_frame_bulk(struct xdp_frame_bulk *bq, + bool napi_direct); +void xdp_return_frame_bulk(struct xdp_frame *xdpf, + struct xdp_frame_bulk *bq, + bool napi_direct); /* When sending xdp_frame into the network stack, then there is no * return point callback, which is needed to release e.g. DMA-mapping diff --git a/net/core/xdp.c b/net/core/xdp.c index 48aba933a5a8..b05467a916b4 100644 --- a/net/core/xdp.c +++ b/net/core/xdp.c @@ -380,6 +380,56 @@ void xdp_return_frame_rx_napi(struct xdp_frame *xdpf) } EXPORT_SYMBOL_GPL(xdp_return_frame_rx_napi); +void xdp_flush_frame_bulk(struct xdp_frame_bulk *bq, + bool napi_direct) +{ + struct xdp_mem_allocator *xa = bq->xa; + int i; + + for (i = 0; i < bq->count; i++) { + napi_direct &= !xdp_return_frame_no_direct(); + page_pool_put_full_page(xa->page_pool, + virt_to_head_page(bq->q[i]), + napi_direct); + } + bq->count = 0; +} +EXPORT_SYMBOL_GPL(xdp_flush_frame_bulk); + +void xdp_return_frame_bulk(struct xdp_frame *xdpf, + struct xdp_frame_bulk *bq, + bool napi_direct) +{ + struct xdp_mem_info *mem = &xdpf->mem; + struct xdp_mem_allocator *xa, *nxa; + + if (mem->type != MEM_TYPE_PAGE_POOL) { + __xdp_return(xdpf->data, &xdpf->mem, napi_direct); + return; + } + + rcu_read_lock(); + + xa = bq->xa; + if (unlikely(!xa || mem->id != xa->mem.id)) { + nxa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params); + if (unlikely(!xa)) { + bq->xa = nxa; + xa = nxa; + } + } + + if (mem->id != xa->mem.id || bq->count == XDP_BULK_QUEUE_SIZE) + xdp_flush_frame_bulk(bq, napi_direct); + + bq->q[bq->count++] = xdpf->data; + if (mem->id != xa->mem.id) + bq->xa = nxa; + + rcu_read_unlock(); +} +EXPORT_SYMBOL_GPL(xdp_return_frame_bulk); + void xdp_return_buff(struct xdp_buff *xdp) { __xdp_return(xdp->data, &xdp->rxq->mem, true);
Introduce bulking capability in xdp tx return path (XDP_TX and XDP_REDIRECT). xdp_return_frame and xdp_return_frame_napi are usually run inside the driver NAPI tx completion loop so it is possible batch them. Current implementation considers only page_pool memory model. Convert mvneta driver to xdp_return_frame_bulk APIs. Suggested-by: Jesper Dangaard Brouer <brouer@redhat.com> Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> --- drivers/net/ethernet/marvell/mvneta.c | 8 ++--- include/net/xdp.h | 11 ++++++ net/core/xdp.c | 50 +++++++++++++++++++++++++++ 3 files changed, 65 insertions(+), 4 deletions(-)