On Mon, Mar 28, 2016 at 06:20:41AM -0600, Lin Ma wrote: > We know that RX_BUF_SIZE is assigned the value 1522 in > src/drivers/net/virtio-net.c. > > I'd like to know that is it ok if I write a patch to change the value from > 1522 to IEEE80211_MAX_FRAME_LEN(2352)? > If the answer is negative, Would you please tell me why? > > I got an bug report about this, During a pxe guest installation, after > loading kernel, if sending jumbo frame packets, say > 'ping -i0.5 -M do -s 8972 $GUEST_IP' from host, The guest installation will > hang, So I want to increase the rx buffer > size of virtio-net to avoid this overflow. That's why I ask this question.
The virtio-net specification says: If the VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6 or VIRTIO_NET_F_GUEST_UFO features are used, the maximum incoming packet will be to 65550 bytes long (the maximum size of a TCP or UDP packet, plus the 14 byte ethernet header), otherwise 1514 bytes. The 12-byte struct virtio_net_hdr is prepended to this, making for 65562 or 1526 bytes. iPXE doesn't use the segmentation/fragmentation offload features so it sticks to the small 1526 byte size. (I haven't checked if the 1522 vs 1526 difference is legitimate or a bug.) Although it probably wouldn't hurt to increase the rx buffer size to around IEEE80211_MAX_FRAME_LEN, I do wonder why this particular constant would be the true maximum value. 802.11 can use up to 7935 byte frames. Michael Tsirkin: Can you comment on virtio-net jumbo frame usage? Is it common? Which number is a popular frame size limit? Stefan
signature.asc
Description: PGP signature
_______________________________________________ ipxe-devel mailing list ipxe-devel@lists.ipxe.org https://lists.ipxe.org/mailman/listinfo.cgi/ipxe-devel
