search for: dev_kfree_skb_any

...ed-off-by: Jason Wang <jasowang at redhat.com> > > Not sure how much it's worth but since Eric suggested it ... Probably depends on the actual cost of u64_stats_update_begin/end against the likely extra saving of the tx_bytes and tx_packets values onto the stack across the call to dev_kfree_skb_any(). (Which depends on the number of caller saved registers.) > Acked-by: Michael S. Tsirkin <mst at redhat.com> > > > --- > > drivers/net/virtio_net.c | 12 ++++++++---- > > 1 files changed, 8 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/n...

...ed-off-by: Jason Wang <jasowang at redhat.com> > > Not sure how much it's worth but since Eric suggested it ... Probably depends on the actual cost of u64_stats_update_begin/end against the likely extra saving of the tx_bytes and tx_packets values onto the stack across the call to dev_kfree_skb_any(). (Which depends on the number of caller saved registers.) > Acked-by: Michael S. Tsirkin <mst at redhat.com> > > > --- > > drivers/net/virtio_net.c | 12 ++++++++---- > > 1 files changed, 8 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/n...

...> > > > > Not sure how much it's worth but since Eric suggested it ... > > > > Probably depends on the actual cost of u64_stats_update_begin/end > > against the likely extra saving of the tx_bytes and tx_packets > > values onto the stack across the call to dev_kfree_skb_any(). > > (Which depends on the number of caller saved registers.) > > Yea, some benchmark results would be nice to see. I there are likely to be multiple skb on the queue the fastest code would probably do one 'virtqueue_get_all()' that returned a linked list of buffers, then fo...

...> > > > > Not sure how much it's worth but since Eric suggested it ... > > > > Probably depends on the actual cost of u64_stats_update_begin/end > > against the likely extra saving of the tx_bytes and tx_packets > > values onto the stack across the call to dev_kfree_skb_any(). > > (Which depends on the number of caller saved registers.) > > Yea, some benchmark results would be nice to see. I there are likely to be multiple skb on the queue the fastest code would probably do one 'virtqueue_get_all()' that returned a linked list of buffers, then fo...

...uf(sq->vq, &len)) != NULL) { pr_debug("Sent skb %p\n", skb); - u64_stats_update_begin(&stats->tx_syncp); - stats->tx_bytes += skb->len; - stats->tx_packets++; - u64_stats_update_end(&stats->tx_syncp); + tx_bytes += skb->len; + tx_packets++; dev_kfree_skb_any(skb); } + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += tx_bytes; + stats->tx_packets =+ tx_packets; + u64_stats_update_end(&stats->tx_syncp); } static int xmit_skb(struct send_queue *sq, struct sk_buff *skb) -- 1.7.1

...uf(sq->vq, &len)) != NULL) { pr_debug("Sent skb %p\n", skb); - u64_stats_update_begin(&stats->tx_syncp); - stats->tx_bytes += skb->len; - stats->tx_packets++; - u64_stats_update_end(&stats->tx_syncp); + tx_bytes += skb->len; + tx_packets++; dev_kfree_skb_any(skb); } + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += tx_bytes; + stats->tx_packets =+ tx_packets; + u64_stats_update_end(&stats->tx_syncp); } static int xmit_skb(struct send_queue *sq, struct sk_buff *skb) -- 1.7.1

...uct virtnet_stats *stats = this_cpu_ptr(vi->stats); + u64 tx_bytes = 0, tx_packets = 0; + + while (tx_packets < budget && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + tx_bytes += skb->len; + tx_packets++; + + dev_kfree_skb_any(skb); + } + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += tx_bytes; + stats->tx_packets =+ tx_packets; + u64_stats_update_end(&stats->tx_syncp); + + return tx_packets; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->...

...uct virtnet_stats *stats = this_cpu_ptr(vi->stats); + u64 tx_bytes = 0, tx_packets = 0; + + while (tx_packets < budget && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + tx_bytes += skb->len; + tx_packets++; + + dev_kfree_skb_any(skb); + } + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += tx_bytes; + stats->tx_packets =+ tx_packets; + u64_stats_update_end(&stats->tx_syncp); + + return tx_packets; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->...

...le (tx_packets < budget && >> > + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { >> > + pr_debug("Sent skb %p\n", skb); >> > + >> > + tx_bytes += skb->len; >> > + tx_packets++; >> > + >> > + dev_kfree_skb_any(skb); >> > + } >> > + >> > + u64_stats_update_begin(&stats->tx_syncp); >> > + stats->tx_bytes += tx_bytes; >> > + stats->tx_packets =+ tx_packets; >> > + u64_stats_update_end(&stats->tx_syncp); >> > + >> > +...

...le (tx_packets < budget && >> > + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { >> > + pr_debug("Sent skb %p\n", skb); >> > + >> > + tx_bytes += skb->len; >> > + tx_packets++; >> > + >> > + dev_kfree_skb_any(skb); >> > + } >> > + >> > + u64_stats_update_begin(&stats->tx_syncp); >> > + stats->tx_bytes += tx_bytes; >> > + stats->tx_packets =+ tx_packets; >> > + u64_stats_update_end(&stats->tx_syncp); >> > + >> > +...

...wang at redhat.com> > > > > Not sure how much it's worth but since Eric suggested it ... > > Probably depends on the actual cost of u64_stats_update_begin/end > against the likely extra saving of the tx_bytes and tx_packets > values onto the stack across the call to dev_kfree_skb_any(). > (Which depends on the number of caller saved registers.) Yea, some benchmark results would be nice to see. > > Acked-by: Michael S. Tsirkin <mst at redhat.com> > > > > > --- > > > drivers/net/virtio_net.c | 12 ++++++++---- > > > 1 files c...

...et && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += skb->len; + stats->tx_packets++; + u64_stats_update_end(&stats->tx_syncp); + + dev_kfree_skb_any(skb); + sent++; + } + + return sent; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->vdev->priv; + struct send_queue *sq = &vi->sq[vq2txq(vq)]; - /* Suppress further interrupts. */ - virtqueue_disable_cb(vq); - - /* We were probably waiting f...

...et && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += skb->len; + stats->tx_packets++; + u64_stats_update_end(&stats->tx_syncp); + + dev_kfree_skb_any(skb); + sent++; + } + + return sent; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->vdev->priv; + struct send_queue *sq = &vi->sq[vq2txq(vq)]; - /* Suppress further interrupts. */ - virtqueue_disable_cb(vq); - - /* We were probably waiting f...

...et && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += skb->len; + stats->tx_packets++; + u64_stats_update_end(&stats->tx_syncp); + + dev_kfree_skb_any(skb); + packets++; + } + + return packets; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->vdev->priv; + struct send_queue *sq = &vi->sq[vq2txq(vq)]; - /* Suppress further interrupts. */ - virtqueue_disable_cb(vq); - - /* We were probably wai...

...et && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += skb->len; + stats->tx_packets++; + u64_stats_update_end(&stats->tx_syncp); + + dev_kfree_skb_any(skb); + packets++; + } + + return packets; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->vdev->priv; + struct send_queue *sq = &vi->sq[vq2txq(vq)]; - /* Suppress further interrupts. */ - virtqueue_disable_cb(vq); - - /* We were probably wai...

According to David, proper accounting and queueing (at all levels, not just TCP sockets) is more important than trying to skim a bunch of cycles by avoiding TX interrupts. Having an event to free the SKB is absolutely essential for the stack to operate correctly. This series tries to enable tx interrupt for virtio-net. The idea is simple: enable tx interrupt and schedule a tx napi to free old

According to David, proper accounting and queueing (at all levels, not just TCP sockets) is more important than trying to skim a bunch of cycles by avoiding TX interrupts. Having an event to free the SKB is absolutely essential for the stack to operate correctly. This series tries to enable tx interrupt for virtio-net. The idea is simple: enable tx interrupt and schedule a tx napi to free old

...et && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += skb->len; + stats->tx_packets++; + u64_stats_update_end(&stats->tx_syncp); + + dev_kfree_skb_any(skb); + sent++; + } + + return sent; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->vdev->priv; + struct send_queue *sq = &vi->sq[vq2txq(vq)]; - /* Suppress further interrupts. */ - virtqueue_disable_cb(vq); - - /* We were probably waiting f...

...et && + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { + pr_debug("Sent skb %p\n", skb); + + u64_stats_update_begin(&stats->tx_syncp); + stats->tx_bytes += skb->len; + stats->tx_packets++; + u64_stats_update_end(&stats->tx_syncp); + + dev_kfree_skb_any(skb); + sent++; + } + + return sent; +} + static void skb_xmit_done(struct virtqueue *vq) { struct virtnet_info *vi = vq->vdev->priv; + struct send_queue *sq = &vi->sq[vq2txq(vq)]; - /* Suppress further interrupts. */ - virtqueue_disable_cb(vq); - - /* We were probably waiting f...

...p; > >> > + (skb = virtqueue_get_buf(sq->vq, &len)) != NULL) { > >> > + pr_debug("Sent skb %p\n", skb); > >> > + > >> > + tx_bytes += skb->len; > >> > + tx_packets++; > >> > + > >> > + dev_kfree_skb_any(skb); > >> > + } > >> > + > >> > + u64_stats_update_begin(&stats->tx_syncp); > >> > + stats->tx_bytes += tx_bytes; > >> > + stats->tx_packets =+ tx_packets; > >> > + u64_stats_update_end(&stats->tx_syncp); &...