On 2016年12月23日 23:57, John Fastabend wrote:
On 16-12-23 06:37 AM, Jason Wang wrote:When XDP_PASS were determined for linearized packets, we try to get new buffers in the virtqueue and build skbs from them. This is wrong, we should create skbs based on existed buffers instead. Fixing them by creating skb based on xdp_page. With this patch "ping 192.168.100.4 -s 3900 -M do" works for XDP_PASS. Cc: John Fastabend <john.r.fastabend@xxxxxxxxx> Signed-off-by: Jason Wang <jasowang@xxxxxxxxxx> --- drivers/net/virtio_net.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index 58ad40e..470293e 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -578,8 +578,14 @@ static struct sk_buff *receive_mergeable(struct net_device *dev, act = do_xdp_prog(vi, rq, xdp_prog, xdp_page, offset, len); switch (act) { case XDP_PASS: - if (unlikely(xdp_page != page)) - __free_pages(xdp_page, 0); + /* We can only create skb based on xdp_page. */ + if (unlikely(xdp_page != page)) { + rcu_read_unlock(); + put_page(page); + head_skb = page_to_skb(vi, rq, xdp_page, + 0, len, PAGE_SIZE); + return head_skb; + } break; case XDP_TX: if (unlikely(xdp_page != page))Great thanks. This was likely working before because of the memory leak fixed in 3/9.
Looks not, without this and 3/9 the code will try to get buffers and build skb for a new packet instead of existed buffers.
Thanks
Acked-by: John Fastabend <john.r.fastabend@xxxxxxxxx>
_______________________________________________ Virtualization mailing list Virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linuxfoundation.org/mailman/listinfo/virtualization
