similar to: [PATCH 6/6] virtio_net: reduce alignment for buffers

Sorry, forgot to Cc Michael and the virt list - patch reproduced below in full. johannes >From 22500fbcf722748fe3471b2e4c6156db47aade15 Mon Sep 17 00:00:00 2001 From: Johannes Berg <johannes.berg at intel.com> Date: Wed, 19 Aug 2015 09:25:18 +0200 Subject: [PATCH] virtio_net: use DECLARE_EWMA Instead of using the out-of-line EWMA calculation, use DECLARE_EWMA() to create static

Sorry, forgot to Cc Michael and the virt list - patch reproduced below in full. johannes >From 22500fbcf722748fe3471b2e4c6156db47aade15 Mon Sep 17 00:00:00 2001 From: Johannes Berg <johannes.berg at intel.com> Date: Wed, 19 Aug 2015 09:25:18 +0200 Subject: [PATCH] virtio_net: use DECLARE_EWMA Instead of using the out-of-line EWMA calculation, use DECLARE_EWMA() to create static

On Fri, Mar 02, 2018 at 05:29:14PM +0800, Jason Wang wrote: > XDP_REDIRECT support for mergeable buffer was removed since commit > 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() > case"). This is because we don't reserve enough tailroom for struct > skb_shared_info which breaks XDP assumption. So this patch fixes this > by reserving enough

XDP_REDIRECT support for mergeable buffer was removed since commit 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() case"). This is because we don't reserve enough tailroom for struct skb_shared_info which breaks XDP assumption. So this patch fixes this by reserving enough tailroom and using fixed size of rx buffer. Signed-off-by: Jason Wang <jasowang at

XDP_REDIRECT support for mergeable buffer was removed since commit 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() case"). This is because we don't reserve enough tailroom for struct skb_shared_info which breaks XDP assumption. So this patch fixes this by reserving enough tailroom and using fixed size of rx buffer. Signed-off-by: Jason Wang <jasowang at

When ring size is small (<32 entries) making buffers smaller means a full ring might not be able to hold enough buffers to fit a single large packet. Make sure a ring full of buffers is large enough to allow at least one packet of max size. Fixes: 2613af0ed18a ("virtio_net: migrate mergeable rx buffers to page frag allocators") Signed-off-by: Michael S. Tsirkin <mst at

When ring size is small (<32 entries) making buffers smaller means a full ring might not be able to hold enough buffers to fit a single large packet. Make sure a ring full of buffers is large enough to allow at least one packet of max size. Fixes: 2613af0ed18a ("virtio_net: migrate mergeable rx buffers to page frag allocators") Signed-off-by: Michael S. Tsirkin <mst at

XDP_REDIRECT support for mergeable buffer was removed since commit 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() case"). This is because we don't reserve enough tailroom for struct skb_shared_info which breaks XDP assumption. Other complaints are, the complex linearize logic and EWMA estimation may increase the possibility of linearizing. Signed-off-by:

When ring size is small (<32 entries) making buffers smaller means a full ring might not be able to hold enough buffers to fit a single large packet. Make sure a ring full of buffers is large enough to allow at least one packet of max size. Fixes: 2613af0ed18a ("virtio_net: migrate mergeable rx buffers to page frag allocators") Signed-off-by: Michael S. Tsirkin <mst at

When ring size is small (<32 entries) making buffers smaller means a full ring might not be able to hold enough buffers to fit a single large packet. Make sure a ring full of buffers is large enough to allow at least one packet of max size. Fixes: 2613af0ed18a ("virtio_net: migrate mergeable rx buffers to page frag allocators") Signed-off-by: Michael S. Tsirkin <mst at

For small packets data copy was observed to take up about 15% CPU time. Switch to build_skb and avoid the copy when using mergeable rx buffers. As a bonus, medium-size skbs that fit in a page will be completely linear. Of course, we now need to lower the lower bound on packet size, to make sure a sane number of skbs fits in rx socket buffer. By how much? I don't know yet. It might also be

For small packets data copy was observed to take up about 15% CPU time. Switch to build_skb and avoid the copy when using mergeable rx buffers. As a bonus, medium-size skbs that fit in a page will be completely linear. Of course, we now need to lower the lower bound on packet size, to make sure a sane number of skbs fits in rx socket buffer. By how much? I don't know yet. It might also be

On Thu, Mar 01, 2018 at 11:19:04AM +0800, Jason Wang wrote: > XDP_REDIRECT support for mergeable buffer was removed since commit > 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() > case"). This is because we don't reserve enough tailroom for struct > skb_shared_info which breaks XDP assumption. Other complaints are, the > complex linearize logic

XDP_REDIRECT support for mergeable buffer was removed since commit 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() case"). This is because we don't reserve enough tailroom for struct skb_shared_info which breaks XDP assumption. So this patch fixes this by reserving enough tailroom and using fixed size of rx buffer. Signed-off-by: Jason Wang <jasowang at

XDP_REDIRECT support for mergeable buffer was removed since commit 7324f5399b06 ("virtio_net: disable XDP_REDIRECT in receive_mergeable() case"). This is because we don't reserve enough tailroom for struct skb_shared_info which breaks XDP assumption. So this patch fixes this by reserving enough tailroom and using fixed size of rx buffer. Signed-off-by: Jason Wang <jasowang at

I don't have any guests with PAGE_SIZE > 64k but the code seems to be clearly broken in that case as PAGE_SIZE / MERGEABLE_BUFFER_ALIGN will need more than 8 bit and so the code in mergeable_ctx_to_buf_address does not give us the actual true size. Cc: John Fastabend <john.fastabend at gmail.com> Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- changes from v1: fix

I don't have any guests with PAGE_SIZE > 64k but the code seems to be clearly broken in that case as PAGE_SIZE / MERGEABLE_BUFFER_ALIGN will need more than 8 bit and so the code in mergeable_ctx_to_buf_address does not give us the actual true size. Cc: John Fastabend <john.fastabend at gmail.com> Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- changes from v1: fix

I don't have any guests with PAGE_SIZE > 64k but the code seems to be clearly broken in that case as PAGE_SIZE / MERGEABLE_BUFFER_ALIGN will need more than 8 bit and so the code in mergeable_ctx_to_buf_address does not give us the actual true size. Cc: John Fastabend <john.fastabend at gmail.com> Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- Lightly tested on x86

I don't have any guests with PAGE_SIZE > 64k but the code seems to be clearly broken in that case as PAGE_SIZE / MERGEABLE_BUFFER_ALIGN will need more than 8 bit and so the code in mergeable_ctx_to_buf_address does not give us the actual true size. Cc: John Fastabend <john.fastabend at gmail.com> Signed-off-by: Michael S. Tsirkin <mst at redhat.com> --- Lightly tested on x86

Hi Michael, Here's a quick sketch of some code that enforces a minimum buffer alignment of only 64, and has a maximum theoretical buffer size of aligned GOOD_PACKET_LEN + (BUF_ALIGN - 1) * BUF_ALIGN, which is at least 1536 + 63 * 64 = 5568. On x86, we already use a 64 byte alignment, and this code supports all current buffer sizes, from 1536 to PAGE_SIZE. #if L1_CACHE_BYTES < 64 #define