From: Eric Dumazet <edumazet@google.com>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Xuan Zhuo <xuanzhuo@linux.alibaba.com>,
Eric Dumazet <eric.dumazet@gmail.com>,
netdev <netdev@vger.kernel.org>,
Alexander Duyck <alexanderduyck@fb.com>,
Paolo Abeni <pabeni@redhat.com>, Greg Thelen <gthelen@google.com>,
"David S . Miller" <davem@davemloft.net>,
Jakub Kicinski <kuba@kernel.org>,
su-lifan@linux.alibaba.com, "dust.li" <dust.li@linux.alibaba.com>,
Jason Wang <jasowang@redhat.com>
Subject: Re: [PATCH net] net: avoid 32 x truesize under-estimation for tiny skbs
Date: Wed, 31 Mar 2021 10:46:41 +0200 [thread overview]
Message-ID: <CANn89i+E=Bu4zPwtQGNZzoPdTzsb+9rWx0d73ZHm_nWQfHRHMA@mail.gmail.com> (raw)
In-Reply-To: <CANn89iJN3SQDctZxaPdZMSPGRbjLrsYGM7=Y2POv-3Ysw-EZ_w@mail.gmail.com>
On Wed, Mar 31, 2021 at 10:36 AM Eric Dumazet <edumazet@google.com> wrote:
>
> On Wed, Mar 31, 2021 at 10:11 AM Michael S. Tsirkin <mst@redhat.com> wrote:
> >
> > On Mon, Mar 29, 2021 at 11:06:09AM +0200, Eric Dumazet wrote:
> > > On Mon, Mar 29, 2021 at 10:52 AM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
> > > >
> > > > On Wed, 13 Jan 2021 08:18:19 -0800, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> > > > > From: Eric Dumazet <edumazet@google.com>
> > > > >
> > > > > Both virtio net and napi_get_frags() allocate skbs
> > > > > with a very small skb->head
> > > > >
> > > > > While using page fragments instead of a kmalloc backed skb->head might give
> > > > > a small performance improvement in some cases, there is a huge risk of
> > > > > under estimating memory usage.
> > > > >
> > > > > For both GOOD_COPY_LEN and GRO_MAX_HEAD, we can fit at least 32 allocations
> > > > > per page (order-3 page in x86), or even 64 on PowerPC
> > > > >
> > > > > We have been tracking OOM issues on GKE hosts hitting tcp_mem limits
> > > > > but consuming far more memory for TCP buffers than instructed in tcp_mem[2]
> > > > >
> > > > > Even if we force napi_alloc_skb() to only use order-0 pages, the issue
> > > > > would still be there on arches with PAGE_SIZE >= 32768
> > > > >
> > > > > This patch makes sure that small skb head are kmalloc backed, so that
> > > > > other objects in the slab page can be reused instead of being held as long
> > > > > as skbs are sitting in socket queues.
> > > > >
> > > > > Note that we might in the future use the sk_buff napi cache,
> > > > > instead of going through a more expensive __alloc_skb()
> > > > >
> > > > > Another idea would be to use separate page sizes depending
> > > > > on the allocated length (to never have more than 4 frags per page)
> > > > >
> > > > > I would like to thank Greg Thelen for his precious help on this matter,
> > > > > analysing crash dumps is always a time consuming task.
> > > >
> > > >
> > > > This patch causes a performance degradation of about 10% in the scenario of
> > > > virtio-net + GRO.
> > > >
> > > > For GRO, there is no way to merge skbs based on frags with this patch, only
> > > > frag_list can be used to link skbs. The problem that this cause are that compared
> > > > to the GRO package merged into the frags way, the current skb needs to call
> > > > kfree_skb_list to release each skb, resulting in performance degradation.
> > > >
> > > > virtio-net will store some data onto the linear space after receiving it. In
> > > > addition to the header, there are also some payloads, so "headlen <= offset"
> > > > fails. And skb->head_frag is failing when use kmalloc() for skb->head allocation.
> > > >
> > >
> > > Thanks for the report.
> > >
> > > There is no way we can make things both fast for existing strategies
> > > used by _insert_your_driver
> > > and malicious usages of data that can sit for seconds/minutes in socket queues.
> > >
> > > I think that if you want to gain this 10% back, you have to change
> > > virtio_net to meet optimal behavior.
> > >
> > > Normal drivers make sure to not pull payload in skb->head, only headers.
> >
> > Hmm we do have hdr_len field, but seem to ignore it on RX.
> > Jason do you see any issues with using it for the head len?
> >
>
> I was looking at this code (page_to_skb()) a few minutes ago ;)
>
> pulling payload would make sense only if can pull of of it (to free the page)
> (This is what some drivers implement and call copybreak)
>
> Even if we do not have an accurate knowledge of header sizes,
> it would be better to pull only the Ethernet header and let GRO do the
> rest during its dissection.
>
> Once fixed, virtio_net will reduce by 2x number of frags per skb,
> compared to the situation before "net: avoid 32 x truesize
> under-estimation for tiny skbs"
Ie I suspect the simple way to fix this would be :
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index bb4ea9dbc16bcb19c5969fc8247478aa66c63fce..a5500bf6ac01051be949edf9fead934a90335f4f
100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -409,9 +409,7 @@ static struct sk_buff *page_to_skb(struct virtnet_info *vi,
offset += hdr_padded_len;
p += hdr_padded_len;
- copy = len;
- if (copy > skb_tailroom(skb))
- copy = skb_tailroom(skb);
+ copy = min_t(int, len, ETH_HLEN);
skb_put_data(skb, p, copy);
if (metasize) {
next prev parent reply other threads:[~2021-03-31 8:47 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-13 16:18 [PATCH net] net: avoid 32 x truesize under-estimation for tiny skbs Eric Dumazet
2021-01-13 18:00 ` Alexander Duyck
2021-01-13 19:19 ` Michael S. Tsirkin
2021-01-13 22:23 ` David Laight
2021-01-14 5:16 ` Eric Dumazet
2021-01-14 9:29 ` David Laight
2021-01-14 19:00 ` patchwork-bot+netdevbpf
[not found] ` <1617007696.5731978-1-xuanzhuo@linux.alibaba.com>
2021-03-29 9:06 ` Eric Dumazet
2021-03-31 8:11 ` Michael S. Tsirkin
2021-03-31 8:36 ` Eric Dumazet
2021-03-31 8:46 ` Eric Dumazet [this message]
2021-03-31 8:49 ` Eric Dumazet
2021-03-31 8:54 ` Eric Dumazet
[not found] ` <1617248264.4993114-2-xuanzhuo@linux.alibaba.com>
2021-04-01 5:06 ` Eric Dumazet
[not found] ` <1617357110.3822439-1-xuanzhuo@linux.alibaba.com>
2021-04-02 12:52 ` Eric Dumazet
2021-04-01 13:51 ` Michael S. Tsirkin
2021-04-01 14:08 ` Eric Dumazet
2021-04-01 7:14 ` Jason Wang
[not found] ` <1617267183.5697193-1-xuanzhuo@linux.alibaba.com>
2021-04-01 9:58 ` Eric Dumazet
2021-04-02 2:52 ` Jason Wang
[not found] ` <1617361253.1788838-2-xuanzhuo@linux.alibaba.com>
2021-04-02 12:53 ` Eric Dumazet
2021-04-06 2:04 ` Jason Wang
[not found] ` <1617190239.1035674-1-xuanzhuo@linux.alibaba.com>
2021-03-31 12:08 ` Eric Dumazet
2021-04-01 13:36 ` Michael S. Tsirkin
2022-09-07 20:19 ` Paolo Abeni
2022-09-07 20:40 ` Eric Dumazet
2022-09-08 10:48 ` Paolo Abeni
2022-09-08 12:20 ` Eric Dumazet
2022-09-08 14:26 ` Paolo Abeni
2022-09-08 16:00 ` Eric Dumazet
2022-09-07 21:36 ` Alexander H Duyck
2022-09-08 11:00 ` Paolo Abeni
2022-09-08 14:53 ` Alexander H Duyck
2022-09-08 18:01 ` Paolo Abeni
2022-09-08 19:26 ` Alexander Duyck
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CANn89i+E=Bu4zPwtQGNZzoPdTzsb+9rWx0d73ZHm_nWQfHRHMA@mail.gmail.com' \
--to=edumazet@google.com \
--cc=alexanderduyck@fb.com \
--cc=davem@davemloft.net \
--cc=dust.li@linux.alibaba.com \
--cc=eric.dumazet@gmail.com \
--cc=gthelen@google.com \
--cc=jasowang@redhat.com \
--cc=kuba@kernel.org \
--cc=mst@redhat.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=su-lifan@linux.alibaba.com \
--cc=xuanzhuo@linux.alibaba.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).