| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Tue, 4 Apr 2023 16:03:49 +0800
From: Jason Wang <jasowang@...hat.com>
To: Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
Cc: netdev@...r.kernel.org, "Michael S. Tsirkin" <mst@...hat.com>,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
Alexei Starovoitov <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>,
Jesper Dangaard Brouer <hawk@...nel.org>,
John Fastabend <john.fastabend@...il.com>,
virtualization@...ts.linux-foundation.org, bpf@...r.kernel.org
Subject: Re: [PATCH net-next 3/8] virtio_net: introduce virtnet_xdp_handler()
to seprate the logic of run xdp
On Tue, Apr 4, 2023 at 3:12 PM Xuan Zhuo <xuanzhuo@...ux.alibaba.com> wrote:
>
> On Tue, 4 Apr 2023 15:01:36 +0800, Jason Wang <jasowang@...hat.com> wrote:
> > On Tue, Apr 4, 2023 at 2:55 PM Xuan Zhuo <xuanzhuo@...ux.alibaba.com> wrote:
> > >
> > > On Tue, 4 Apr 2023 14:35:05 +0800, Jason Wang <jasowang@...hat.com> wrote:
> > > > On Tue, Apr 4, 2023 at 2:22 PM Xuan Zhuo <xuanzhuo@...ux.alibaba.com> wrote:
> > > > >
> > > > > On Tue, 4 Apr 2023 13:04:02 +0800, Jason Wang <jasowang@...hat.com> wrote:
> > > > > > On Mon, Apr 3, 2023 at 12:17 PM Xuan Zhuo <xuanzhuo@...ux.alibaba.com> wrote:
> > > > > > >
> > > > > > > On Mon, 3 Apr 2023 10:43:03 +0800, Jason Wang <jasowang@...hat.com> wrote:
> > > > > > > > On Tue, Mar 28, 2023 at 8:04 PM Xuan Zhuo <xuanzhuo@...ux.alibaba.com> wrote:
> > > > > > > > >
> > > > > > > > > At present, we have two similar logic to perform the XDP prog.
> > > > > > > > >
> > > > > > > > > Therefore, this PATCH separates the code of executing XDP, which is
> > > > > > > > > conducive to later maintenance.
> > > > > > > > >
> > > > > > > > > Signed-off-by: Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
> > > > > > > > > ---
> > > > > > > > > drivers/net/virtio_net.c | 142 +++++++++++++++++++++------------------
> > > > > > > > > 1 file changed, 75 insertions(+), 67 deletions(-)
> > > > > > > > >
> > > > > > > > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > > > > > > > > index bb426958cdd4..72b9d6ee4024 100644
> > > > > > > > > --- a/drivers/net/virtio_net.c
> > > > > > > > > +++ b/drivers/net/virtio_net.c
> > > > > > > > > @@ -301,6 +301,15 @@ struct padded_vnet_hdr {
> > > > > > > > > char padding[12];
> > > > > > > > > };
> > > > > > > > >
> > > > > > > > > +enum {
> > > > > > > > > + /* xdp pass */
> > > > > > > > > + VIRTNET_XDP_RES_PASS,
> > > > > > > > > + /* drop packet. the caller needs to release the page. */
> > > > > > > > > + VIRTNET_XDP_RES_DROP,
> > > > > > > > > + /* packet is consumed by xdp. the caller needs to do nothing. */
> > > > > > > > > + VIRTNET_XDP_RES_CONSUMED,
> > > > > > > > > +};
> > > > > > > >
> > > > > > > > I'd prefer this to be done on top unless it is a must. But I don't see
> > > > > > > > any advantage of introducing this, it's partial mapping of XDP action
> > > > > > > > and it needs to be extended when XDP action is extended. (And we've
> > > > > > > > already had: VIRTIO_XDP_REDIR and VIRTIO_XDP_TX ...)
> > > > > > >
> > > > > > > No, these are the three states of buffer after XDP processing.
> > > > > > >
> > > > > > > * PASS: goto make skb
> > > > > >
> > > > > > XDP_PASS goes for this.
> > > > > >
> > > > > > > * DROP: we should release buffer
> > > > > >
> > > > > > XDP_DROP and error conditions go with this.
> > > > > >
> > > > > > > * CUNSUMED: xdp prog used the buffer, we do nothing
> > > > > >
> > > > > > XDP_TX/XDP_REDIRECTION goes for this.
> > > > > >
> > > > > > So t virtnet_xdp_handler() just maps XDP ACTION plus the error
> > > > > > conditions to the above three states.
> > > > > >
> > > > > > We can simply map error to XDP_DROP like:
> > > > > >
> > > > > > case XDP_TX:
> > > > > > stats->xdp_tx++;
> > > > > > xdpf = xdp_convert_buff_to_frame(xdp);
> > > > > > if (unlikely(!xdpf))
> > > > > > return XDP_DROP;
> > > > > >
> > > > > > A good side effect is to avoid the xdp_xmit pointer to be passed to
> > > > > > the function.
> > > > >
> > > > >
> > > > > So, I guess you mean this:
> > > > >
> > > > > switch (act) {
> > > > > case XDP_PASS:
> > > > > /* handle pass */
> > > > > return skb;
> > > > >
> > > > > case XDP_TX:
> > > > > *xdp_xmit |= VIRTIO_XDP_TX;
> > > > > goto xmit;
> > > > >
> > > > > case XDP_REDIRECT:
> > > > > *xdp_xmit |= VIRTIO_XDP_REDIR;
> > > > > goto xmit;
> > > > >
> > > > > case XDP_DROP:
> > > > > default:
> > > > > goto err_xdp;
> > > > > }
> > > > >
> > > > > I have to say there is no problem from the perspective of code implementation.
> > > >
> > > > Note that this is the current logic where it is determined in
> > > > receive_small() and receive_mergeable().
> > >
> > > Yes, but the purpose of this patches is to simplify the call.
> >
> > You mean simplify the receive_small()/mergeable()?
>
> YES.
>
>
> >
> > >
> > > >
> > > > >
> > > > > But if the a new ACTION liking XDP_TX,XDP_REDIRECT is added in the future, then
> > > > > we must modify all the callers.
> > > >
> > > > This is fine since we only use a single type for XDP action.
> > >
> > > a single type?
> >
> > Instead of (partial) duplicating XDP actions in the new enums.
>
>
> I think it's really misunderstand here. So your thought is these?
>
> VIRTNET_XDP_RES_PASS,
> VIRTNET_XDP_RES_TX_REDIRECT,
> VIRTNET_XDP_RES_DROP,
No, I meant the enum you introduced.
>
>
>
> >
> > >
> > > >
> > > > > This is the benefit of using CUNSUMED.
> > > >
> > > > It's very hard to say, e.g if we want to support cloning in the future.
> > >
> > > cloning? You mean clone one new buffer.
> > >
> > > It is true that no matter what realization, the logic must be modified.
> >
> > Yes.
> >
> > >
> > > >
> > > > >
> > > > > I think it is a good advantage to put xdp_xmit in virtnet_xdp_handler(),
> > > > > which makes the caller not care too much about these details.
> > > >
> > > > This part I don't understand, having xdp_xmit means the caller need to
> > > > know whether it is xmited or redirected. The point of the enum is to
> > > > hide the XDP actions, but it's conflict with what xdp_xmit who want to
> > > > expose (part of) the XDP actions.
> > >
> > > I mean, no matter what virtnet_xdp_handler () returns? XDP_ACTION or some one I
> > > defined, I want to hide the modification of xdp_xmit to virtnet_xdp_handler().
> > >
> > > Even if virtnet_xdp_handler() returns XDP_TX, we can also complete the
> > > modification of XDP_XMIT within Virtnet_xdp_handler().
> > >
> > >
> > > >
> > > > > If you take into
> > > > > account the problem of increasing the number of parameters, I advise to put it
> > > > > in rq.
> > > >
> > > > I don't have strong opinion to introduce the enum,
> > >
> > > OK, I will drop these new enums.
> >
> > Just to make sure we are at the same page. I mean, if there is no
> > objection from others, I'm ok to have an enum, but we need to use a
> > separate patch to do that.
>
> Do you refer to introduce enums alone without virtnet_xdp_handler()?
I meant, having two patches
1) split out virtnet_xdp_handler() without introducing any new enums
2) introduce the new enum to simplify the codes
Thanks
>
> >
> > >
> > > > what I want to say
> > > > is, use a separated patch to do that.
> > >
> > > Does this part refer to putting xdp_xmit in rq?
> >
> > I mean it's better to be done separately. But I don't see the
> > advantage of this other than reducing the parameters.
>
> I think so also.
>
> Thanks.
>
>
> >
> > Thanks
> >
> > >
> > > Thanks.
> > >
> > >
> > > >
> > > > Thanks
> > > >
> > > > >
> > > > > Thanks.
> > > > >
> > > > >
> > > > >
> > > > > >
> > > > > > >
> > > > > > > The latter two are not particularly related to XDP ACTION. And it does not need
> > > > > > > to extend when XDP action is extended. At least I have not thought of this
> > > > > > > situation.
> > > > > >
> > > > > > What's the advantages of such indirection compared to using XDP action directly?
> > > > > >
> > > > > > Thanks
> > > > > >
> > > > > > >
> > > > > > >
> > > > > > > >
> > > > > > > > > +
> > > > > > > > > static void virtnet_rq_free_unused_buf(struct virtqueue *vq, void *buf);
> > > > > > > > > static void virtnet_sq_free_unused_buf(struct virtqueue *vq, void *buf);
> > > > > > > > >
> > > > > > > > > @@ -789,6 +798,59 @@ static int virtnet_xdp_xmit(struct net_device *dev,
> > > > > > > > > return ret;
> > > > > > > > > }
> > > > > > > > >
> > > > > > > > > +static int virtnet_xdp_handler(struct bpf_prog *xdp_prog, struct xdp_buff *xdp,
> > > > > > > > > + struct net_device *dev,
> > > > > > > > > + unsigned int *xdp_xmit,
> > > > > > > > > + struct virtnet_rq_stats *stats)
> > > > > > > > > +{
> > > > > > > > > + struct xdp_frame *xdpf;
> > > > > > > > > + int err;
> > > > > > > > > + u32 act;
> > > > > > > > > +
> > > > > > > > > + act = bpf_prog_run_xdp(xdp_prog, xdp);
> > > > > > > > > + stats->xdp_packets++;
> > > > > > > > > +
> > > > > > > > > + switch (act) {
> > > > > > > > > + case XDP_PASS:
> > > > > > > > > + return VIRTNET_XDP_RES_PASS;
> > > > > > > > > +
> > > > > > > > > + case XDP_TX:
> > > > > > > > > + stats->xdp_tx++;
> > > > > > > > > + xdpf = xdp_convert_buff_to_frame(xdp);
> > > > > > > > > + if (unlikely(!xdpf))
> > > > > > > > > + return VIRTNET_XDP_RES_DROP;
> > > > > > > > > +
> > > > > > > > > + err = virtnet_xdp_xmit(dev, 1, &xdpf, 0);
> > > > > > > > > + if (unlikely(!err)) {
> > > > > > > > > + xdp_return_frame_rx_napi(xdpf);
> > > > > > > > > + } else if (unlikely(err < 0)) {
> > > > > > > > > + trace_xdp_exception(dev, xdp_prog, act);
> > > > > > > > > + return VIRTNET_XDP_RES_DROP;
> > > > > > > > > + }
> > > > > > > > > +
> > > > > > > > > + *xdp_xmit |= VIRTIO_XDP_TX;
> > > > > > > > > + return VIRTNET_XDP_RES_CONSUMED;
> > > > > > > > > +
> > > > > > > > > + case XDP_REDIRECT:
> > > > > > > > > + stats->xdp_redirects++;
> > > > > > > > > + err = xdp_do_redirect(dev, xdp, xdp_prog);
> > > > > > > > > + if (err)
> > > > > > > > > + return VIRTNET_XDP_RES_DROP;
> > > > > > > > > +
> > > > > > > > > + *xdp_xmit |= VIRTIO_XDP_REDIR;
> > > > > > > > > + return VIRTNET_XDP_RES_CONSUMED;
> > > > > > > > > +
> > > > > > > > > + default:
> > > > > > > > > + bpf_warn_invalid_xdp_action(dev, xdp_prog, act);
> > > > > > > > > + fallthrough;
> > > > > > > > > + case XDP_ABORTED:
> > > > > > > > > + trace_xdp_exception(dev, xdp_prog, act);
> > > > > > > > > + fallthrough;
> > > > > > > > > + case XDP_DROP:
> > > > > > > > > + return VIRTNET_XDP_RES_DROP;
> > > > > > > > > + }
> > > > > > > > > +}
> > > > > > > > > +
> > > > > > > > > static unsigned int virtnet_get_headroom(struct virtnet_info *vi)
> > > > > > > > > {
> > > > > > > > > return vi->xdp_enabled ? VIRTIO_XDP_HEADROOM : 0;
> > > > > > > > > @@ -876,7 +938,6 @@ static struct sk_buff *receive_small(struct net_device *dev,
> > > > > > > > > struct page *page = virt_to_head_page(buf);
> > > > > > > > > unsigned int delta = 0;
> > > > > > > > > struct page *xdp_page;
> > > > > > > > > - int err;
> > > > > > > > > unsigned int metasize = 0;
> > > > > > > > >
> > > > > > > > > len -= vi->hdr_len;
> > > > > > > > > @@ -898,7 +959,6 @@ static struct sk_buff *receive_small(struct net_device *dev,
> > > > > > > > > xdp_prog = rcu_dereference(rq->xdp_prog);
> > > > > > > > > if (xdp_prog) {
> > > > > > > > > struct virtio_net_hdr_mrg_rxbuf *hdr = buf + header_offset;
> > > > > > > > > - struct xdp_frame *xdpf;
> > > > > > > > > struct xdp_buff xdp;
> > > > > > > > > void *orig_data;
> > > > > > > > > u32 act;
> > > > > > > > > @@ -931,46 +991,22 @@ static struct sk_buff *receive_small(struct net_device *dev,
> > > > > > > > > xdp_prepare_buff(&xdp, buf + VIRTNET_RX_PAD + vi->hdr_len,
> > > > > > > > > xdp_headroom, len, true);
> > > > > > > > > orig_data = xdp.data;
> > > > > > > > > - act = bpf_prog_run_xdp(xdp_prog, &xdp);
> > > > > > > > > - stats->xdp_packets++;
> > > > > > > > > +
> > > > > > > > > + act = virtnet_xdp_handler(xdp_prog, &xdp, dev, xdp_xmit, stats);
> > > > > > > > >
> > > > > > > > > switch (act) {
> > > > > > > > > - case XDP_PASS:
> > > > > > > > > + case VIRTNET_XDP_RES_PASS:
> > > > > > > > > /* Recalculate length in case bpf program changed it */
> > > > > > > > > delta = orig_data - xdp.data;
> > > > > > > > > len = xdp.data_end - xdp.data;
> > > > > > > > > metasize = xdp.data - xdp.data_meta;
> > > > > > > > > break;
> > > > > > > > > - case XDP_TX:
> > > > > > > > > - stats->xdp_tx++;
> > > > > > > > > - xdpf = xdp_convert_buff_to_frame(&xdp);
> > > > > > > > > - if (unlikely(!xdpf))
> > > > > > > > > - goto err_xdp;
> > > > > > > > > - err = virtnet_xdp_xmit(dev, 1, &xdpf, 0);
> > > > > > > > > - if (unlikely(!err)) {
> > > > > > > > > - xdp_return_frame_rx_napi(xdpf);
> > > > > > > > > - } else if (unlikely(err < 0)) {
> > > > > > > > > - trace_xdp_exception(vi->dev, xdp_prog, act);
> > > > > > > > > - goto err_xdp;
> > > > > > > > > - }
> > > > > > > > > - *xdp_xmit |= VIRTIO_XDP_TX;
> > > > > > > > > - rcu_read_unlock();
> > > > > > > > > - goto xdp_xmit;
> > > > > > > > > - case XDP_REDIRECT:
> > > > > > > > > - stats->xdp_redirects++;
> > > > > > > > > - err = xdp_do_redirect(dev, &xdp, xdp_prog);
> > > > > > > > > - if (err)
> > > > > > > > > - goto err_xdp;
> > > > > > > > > - *xdp_xmit |= VIRTIO_XDP_REDIR;
> > > > > > > > > +
> > > > > > > > > + case VIRTNET_XDP_RES_CONSUMED:
> > > > > > > > > rcu_read_unlock();
> > > > > > > > > goto xdp_xmit;
> > > > > > > > > - default:
> > > > > > > > > - bpf_warn_invalid_xdp_action(vi->dev, xdp_prog, act);
> > > > > > > > > - fallthrough;
> > > > > > > > > - case XDP_ABORTED:
> > > > > > > > > - trace_xdp_exception(vi->dev, xdp_prog, act);
> > > > > > > > > - goto err_xdp;
> > > > > > > > > - case XDP_DROP:
> > > > > > > > > +
> > > > > > > > > + case VIRTNET_XDP_RES_DROP:
> > > > > > > > > goto err_xdp;
> > > > > > > > > }
> > > > > > > > > }
> > > > > > > > > @@ -1277,7 +1313,6 @@ static struct sk_buff *receive_mergeable(struct net_device *dev,
> > > > > > > > > if (xdp_prog) {
> > > > > > > > > unsigned int xdp_frags_truesz = 0;
> > > > > > > > > struct skb_shared_info *shinfo;
> > > > > > > > > - struct xdp_frame *xdpf;
> > > > > > > > > struct page *xdp_page;
> > > > > > > > > struct xdp_buff xdp;
> > > > > > > > > void *data;
> > > > > > > > > @@ -1294,49 +1329,22 @@ static struct sk_buff *receive_mergeable(struct net_device *dev,
> > > > > > > > > if (unlikely(err))
> > > > > > > > > goto err_xdp_frags;
> > > > > > > > >
> > > > > > > > > - act = bpf_prog_run_xdp(xdp_prog, &xdp);
> > > > > > > > > - stats->xdp_packets++;
> > > > > > > > > + act = virtnet_xdp_handler(xdp_prog, &xdp, dev, xdp_xmit, stats);
> > > > > > > > >
> > > > > > > > > switch (act) {
> > > > > > > > > - case XDP_PASS:
> > > > > > > > > + case VIRTNET_XDP_RES_PASS:
> > > > > > > > > head_skb = build_skb_from_xdp_buff(dev, vi, &xdp, xdp_frags_truesz);
> > > > > > > > > if (unlikely(!head_skb))
> > > > > > > > > goto err_xdp_frags;
> > > > > > > > >
> > > > > > > > > rcu_read_unlock();
> > > > > > > > > return head_skb;
> > > > > > > > > - case XDP_TX:
> > > > > > > > > - stats->xdp_tx++;
> > > > > > > > > - xdpf = xdp_convert_buff_to_frame(&xdp);
> > > > > > > > > - if (unlikely(!xdpf)) {
> > > > > > > > > - netdev_dbg(dev, "convert buff to frame failed for xdp\n");
> > > > > > > >
> > > > > > > > Nit: This debug is lost after the conversion.
> > > > > > >
> > > > > > > Will fix.
> > > > > > >
> > > > > > > Thanks.
> > > > > > >
> > > > > > > >
> > > > > > > > Thanks
> > > > > > > >
> > > > > > > > > - goto err_xdp_frags;
> > > > > > > > > - }
> > > > > > > > > - err = virtnet_xdp_xmit(dev, 1, &xdpf, 0);
> > > > > > > > > - if (unlikely(!err)) {
> > > > > > > > > - xdp_return_frame_rx_napi(xdpf);
> > > > > > > > > - } else if (unlikely(err < 0)) {
> > > > > > > > > - trace_xdp_exception(vi->dev, xdp_prog, act);
> > > > > > > > > - goto err_xdp_frags;
> > > > > > > > > - }
> > > > > > > > > - *xdp_xmit |= VIRTIO_XDP_TX;
> > > > > > > > > - rcu_read_unlock();
> > > > > > > > > - goto xdp_xmit;
> > > > > > > > > - case XDP_REDIRECT:
> > > > > > > > > - stats->xdp_redirects++;
> > > > > > > > > - err = xdp_do_redirect(dev, &xdp, xdp_prog);
> > > > > > > > > - if (err)
> > > > > > > > > - goto err_xdp_frags;
> > > > > > > > > - *xdp_xmit |= VIRTIO_XDP_REDIR;
> > > > > > > > > +
> > > > > > > > > + case VIRTNET_XDP_RES_CONSUMED:
> > > > > > > > > rcu_read_unlock();
> > > > > > > > > goto xdp_xmit;
> > > > > > > > > - default:
> > > > > > > > > - bpf_warn_invalid_xdp_action(vi->dev, xdp_prog, act);
> > > > > > > > > - fallthrough;
> > > > > > > > > - case XDP_ABORTED:
> > > > > > > > > - trace_xdp_exception(vi->dev, xdp_prog, act);
> > > > > > > > > - fallthrough;
> > > > > > > > > - case XDP_DROP:
> > > > > > > > > +
> > > > > > > > > + case VIRTNET_XDP_RES_DROP:
> > > > > > > > > goto err_xdp_frags;
> > > > > > > > > }
> > > > > > > > > err_xdp_frags:
> > > > > > > > > --
> > > > > > > > > 2.32.0.3.g01195cf9f
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>
Powered by blists - more mailing lists