Stefano Garzarella
2023-Jun-26 15:20 UTC
[RFC PATCH v4 01/17] vsock/virtio: read data from non-linear skb
On Sat, Jun 03, 2023 at 11:49:23PM +0300, Arseniy Krasnov wrote:>This is preparation patch for non-linear skbuff handling. It replaces >direct calls of 'memcpy_to_msg()' with 'skb_copy_datagram_iter()'. Main >advantage of the second one is that is can handle paged part of the skb >by using 'kmap()' on each page, but if there are no pages in the skb, >it behaves like simple copying to iov iterator. This patch also adds >new field to the control block of skb - this value shows current offset >in the skb to read next portion of data (it doesn't matter linear it or >not). Idea is that 'skb_copy_datagram_iter()' handles both types of >skb internally - it just needs an offset from which to copy data from >the given skb. This offset is incremented on each read from skb. This >approach allows to avoid special handling of non-linear skbs: >1) We can't call 'skb_pull()' on it, because it updates 'data' pointer. >2) We need to update 'data_len' also on each read from this skb. > >Signed-off-by: Arseniy Krasnov <AVKrasnov at sberdevices.ru> >--- > include/linux/virtio_vsock.h | 1 + > net/vmw_vsock/virtio_transport_common.c | 26 +++++++++++++++++-------- > 2 files changed, 19 insertions(+), 8 deletions(-) > >diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h >index c58453699ee9..17dbb7176e37 100644 >--- a/include/linux/virtio_vsock.h >+++ b/include/linux/virtio_vsock.h >@@ -12,6 +12,7 @@ > struct virtio_vsock_skb_cb { > bool reply; > bool tap_delivered; >+ u32 frag_off; > }; > > #define VIRTIO_VSOCK_SKB_CB(skb) ((struct virtio_vsock_skb_cb *)((skb)->cb)) >diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c >index b769fc258931..5819a9cd4515 100644 >--- a/net/vmw_vsock/virtio_transport_common.c >+++ b/net/vmw_vsock/virtio_transport_common.c >@@ -355,7 +355,7 @@ virtio_transport_stream_do_peek(struct vsock_sock *vsk, > spin_lock_bh(&vvs->rx_lock); > > skb_queue_walk_safe(&vvs->rx_queue, skb, tmp) { >- off = 0; >+ off = VIRTIO_VSOCK_SKB_CB(skb)->frag_off; > > if (total == len) > break; >@@ -370,7 +370,10 @@ virtio_transport_stream_do_peek(struct vsock_sock *vsk, > */ > spin_unlock_bh(&vvs->rx_lock); > >- err = memcpy_to_msg(msg, skb->data + off, bytes); >+ err = skb_copy_datagram_iter(skb, off, >+ &msg->msg_iter, >+ bytes); >+ > if (err) > goto out; > >@@ -414,24 +417,28 @@ virtio_transport_stream_do_dequeue(struct vsock_sock *vsk, > skb = skb_peek(&vvs->rx_queue); > > bytes = len - total; >- if (bytes > skb->len) >- bytes = skb->len; >+ if (bytes > skb->len - VIRTIO_VSOCK_SKB_CB(skb)->frag_off) >+ bytes = skb->len - VIRTIO_VSOCK_SKB_CB(skb)->frag_off;What about storing `VIRTIO_VSOCK_SKB_CB(skb)->frag_off` in a variable? More for readability than optimization, which I hope the compiler already does on its own. The rest LGTM. Stefano> > /* sk_lock is held by caller so no one else can dequeue. > * Unlock rx_lock since memcpy_to_msg() may sleep. > */ > spin_unlock_bh(&vvs->rx_lock); > >- err = memcpy_to_msg(msg, skb->data, bytes); >+ err = skb_copy_datagram_iter(skb, >+ VIRTIO_VSOCK_SKB_CB(skb)->frag_off, >+ &msg->msg_iter, bytes); >+ > if (err) > goto out; > > spin_lock_bh(&vvs->rx_lock); > > total += bytes; >- skb_pull(skb, bytes); > >- if (skb->len == 0) { >+ VIRTIO_VSOCK_SKB_CB(skb)->frag_off += bytes; >+ >+ if (skb->len == VIRTIO_VSOCK_SKB_CB(skb)->frag_off) { > u32 pkt_len = le32_to_cpu(virtio_vsock_hdr(skb)->len); > > virtio_transport_dec_rx_pkt(vvs, pkt_len); >@@ -503,7 +510,10 @@ static int virtio_transport_seqpacket_do_dequeue(struct vsock_sock *vsk, > */ > spin_unlock_bh(&vvs->rx_lock); > >- err = memcpy_to_msg(msg, skb->data, bytes_to_copy); >+ err = skb_copy_datagram_iter(skb, 0, >+ &msg->msg_iter, >+ bytes_to_copy); >+ > if (err) { > /* Copy of message failed. Rest of > * fragments will be freed without copy. >-- >2.25.1 >