| Message ID | 20230701063947.3422088-3-AVKrasnov@sberdevices.ru (mailing list archive) |
|---|---|
| State | RFC |
| Headers | show |
| Series | vsock: MSG_ZEROCOPY flag support | expand |
On Sat, Jul 01, 2023 at 09:39:32AM +0300, Arseniy Krasnov wrote: >This adds copying to guest's virtio buffers from non-linear skbs. Such >skbs are created by protocol layer when MSG_ZEROCOPY flags is used. It >replaces call of 'copy_to_iter()' to 'skb_copy_datagram_iter()'- second >function can read data from non-linear skb. Also this patch uses field >'frag_off' from skb control block. This field shows current offset to >read data from skb which could be both linear or not. > >Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> >--- > Changelog: > v4 -> v5: > * Use local variable for 'frag_off'. > * Update commit message by adding some details about 'frag_off' field. > * R-b from Bobby Eshleman removed due to patch update. I think we should merge this patch with the previous one, since vhost-vsock for example uses virtio_transport_stream_do_dequeue() that we change in the previous commit, so we will break the bisection. The patch LGTM! Stefano > > drivers/vhost/vsock.c | 14 +++++++++----- > 1 file changed, 9 insertions(+), 5 deletions(-) > >diff --git a/drivers/vhost/vsock.c b/drivers/vhost/vsock.c >index 6578db78f0ae..cb00e0e059e4 100644 >--- a/drivers/vhost/vsock.c >+++ b/drivers/vhost/vsock.c >@@ -114,6 +114,7 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, > struct sk_buff *skb; > unsigned out, in; > size_t nbytes; >+ u32 frag_off; > int head; > > skb = virtio_vsock_skb_dequeue(&vsock->send_pkt_queue); >@@ -156,7 +157,8 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, > } > > iov_iter_init(&iov_iter, ITER_DEST, &vq->iov[out], in, iov_len); >- payload_len = skb->len; >+ frag_off = VIRTIO_VSOCK_SKB_CB(skb)->frag_off; >+ payload_len = skb->len - frag_off; > hdr = virtio_vsock_hdr(skb); > > /* If the packet is greater than the space available in the >@@ -197,8 +199,10 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, > break; > } > >- nbytes = copy_to_iter(skb->data, payload_len, &iov_iter); >- if (nbytes != payload_len) { >+ if (skb_copy_datagram_iter(skb, >+ frag_off, >+ &iov_iter, >+ payload_len)) { > kfree_skb(skb); > vq_err(vq, "Faulted on copying pkt buf\n"); > break; >@@ -212,13 +216,13 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, > vhost_add_used(vq, head, sizeof(*hdr) + payload_len); > added = true; > >- skb_pull(skb, payload_len); >+ VIRTIO_VSOCK_SKB_CB(skb)->frag_off += payload_len; > total_len += payload_len; > > /* If we didn't send all the payload we can requeue the packet > * to send it with the next available buffer. > */ >- if (skb->len > 0) { >+ if (VIRTIO_VSOCK_SKB_CB(skb)->frag_off < skb->len) { > hdr->flags |= cpu_to_le32(flags_to_restore); > > /* We are queueing the same skb to handle >-- >2.25.1 >
diff --git a/drivers/vhost/vsock.c b/drivers/vhost/vsock.c index 6578db78f0ae..cb00e0e059e4 100644 --- a/drivers/vhost/vsock.c +++ b/drivers/vhost/vsock.c @@ -114,6 +114,7 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, struct sk_buff *skb; unsigned out, in; size_t nbytes; + u32 frag_off; int head; skb = virtio_vsock_skb_dequeue(&vsock->send_pkt_queue); @@ -156,7 +157,8 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, } iov_iter_init(&iov_iter, ITER_DEST, &vq->iov[out], in, iov_len); - payload_len = skb->len; + frag_off = VIRTIO_VSOCK_SKB_CB(skb)->frag_off; + payload_len = skb->len - frag_off; hdr = virtio_vsock_hdr(skb); /* If the packet is greater than the space available in the @@ -197,8 +199,10 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, break; } - nbytes = copy_to_iter(skb->data, payload_len, &iov_iter); - if (nbytes != payload_len) { + if (skb_copy_datagram_iter(skb, + frag_off, + &iov_iter, + payload_len)) { kfree_skb(skb); vq_err(vq, "Faulted on copying pkt buf\n"); break; @@ -212,13 +216,13 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock, vhost_add_used(vq, head, sizeof(*hdr) + payload_len); added = true; - skb_pull(skb, payload_len); + VIRTIO_VSOCK_SKB_CB(skb)->frag_off += payload_len; total_len += payload_len; /* If we didn't send all the payload we can requeue the packet * to send it with the next available buffer. */ - if (skb->len > 0) { + if (VIRTIO_VSOCK_SKB_CB(skb)->frag_off < skb->len) { hdr->flags |= cpu_to_le32(flags_to_restore); /* We are queueing the same skb to handle
This adds copying to guest's virtio buffers from non-linear skbs. Such skbs are created by protocol layer when MSG_ZEROCOPY flags is used. It replaces call of 'copy_to_iter()' to 'skb_copy_datagram_iter()'- second function can read data from non-linear skb. Also this patch uses field 'frag_off' from skb control block. This field shows current offset to read data from skb which could be both linear or not. Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru> --- Changelog: v4 -> v5: * Use local variable for 'frag_off'. * Update commit message by adding some details about 'frag_off' field. * R-b from Bobby Eshleman removed due to patch update. drivers/vhost/vsock.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)