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
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <1725435002.9733856-1-xuanzhuo@linux.alibaba.com>
Date: Wed, 4 Sep 2024 15:30:02 +0800
From: Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
To: 文博李 <liwenbo.martin@...edance.com>
Cc: virtualization@...ts.linux.dev,
 netdev@...r.kernel.org,
 linux-kernel@...r.kernel.org,
 Jiahui Cen <cenjiahui@...edance.com>,
 Ying Fang <fangying.tommy@...edance.com>,
 mst@...hat.com,
 jasowang@...hat.com,
 eperezma@...hat.com,
 davem@...emloft.net,
 edumazet@...gle.com,
 kuba@...nel.org,
 pabeni@...hat.com
Subject: Re: [External] Re: [PATCH v2] virtio_net: Fix mismatched buf address when unmapping for small packets

On Wed, 4 Sep 2024 15:21:28 +0800, =?utf-8?b?5paH5Y2a5p2O?= <liwenbo.martin@...edance.com> wrote:
> When SWIOTLB is enabled, a DMA map will allocate a bounce buffer for real
> DMA operations,
> and when unmapping, SWIOTLB copies the content in the bounce buffer to the
> driver-allocated
> buffer (the `buf` variable). Such copy only synchronizes data in the buffer
> range, not the whole page.
> So we should give the correct offset for DMA unmapping.

I see.

But I think we should pass the "correct" buf to virtio core as the "data" by
virtqueue_add_inbuf_ctx().

In the merge mode, we pass the pointer that points to the virtnet header.
In the small mode, we pass the pointer that points to the virtnet header - offset.

But this is not the only problem, we try to get the virtnet header by the buf
inside receive_buf(before receive_small).

	flags = ((struct virtio_net_common_hdr *)buf)->hdr.flags;

So I think it is time to unify the buf that passed to the virtio core into a
pointer pointed to the virtnet header.

Thanks.


>
> Thanks.
>
> On Wed, Sep 4, 2024 at 2:46 PM Xuan Zhuo <xuanzhuo@...ux.alibaba.com> wrote:
> >
> > On Wed,  4 Sep 2024 14:10:09 +0800, Wenbo Li <liwenbo.martin@...edance.com>
> wrote:
> > > Currently, the virtio-net driver will perform a pre-dma-mapping for
> > > small or mergeable RX buffer. But for small packets, a mismatched
> address
> > > without VIRTNET_RX_PAD and xdp_headroom is used for unmapping.
> >
> > Will used virt_to_head_page(), so could you say more about it?
> >
> >         struct page *page = virt_to_head_page(buf);
> >
> > Thanks.
> >
> > >
> > > That will result in unsynchronized buffers when SWIOTLB is enabled, for
> > > example, when running as a TDX guest.
> > >
> > > This patch handles small and mergeable packets separately and fixes
> > > the mismatched buffer address.
> > >
> > > Changes from v1: Use ctx to get xdp_headroom.
> > >
> > > Fixes: 295525e29a5b ("virtio_net: merge dma operations when filling
> mergeable buffers")
> > > Signed-off-by: Wenbo Li <liwenbo.martin@...edance.com>
> > > Signed-off-by: Jiahui Cen <cenjiahui@...edance.com>
> > > Signed-off-by: Ying Fang <fangying.tommy@...edance.com>
> > > ---
> > >  drivers/net/virtio_net.c | 29 ++++++++++++++++++++++++++++-
> > >  1 file changed, 28 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > > index c6af18948..cbc3c0ae4 100644
> > > --- a/drivers/net/virtio_net.c
> > > +++ b/drivers/net/virtio_net.c
> > > @@ -891,6 +891,23 @@ static void *virtnet_rq_get_buf(struct
> receive_queue *rq, u32 *len, void **ctx)
> > >       return buf;
> > >  }
> > >
> > > +static void *virtnet_rq_get_buf_small(struct receive_queue *rq,
> > > +                                   u32 *len,
> > > +                                   void **ctx,
> > > +                                   unsigned int header_offset)
> > > +{
> > > +     void *buf;
> > > +     unsigned int xdp_headroom;
> > > +
> > > +     buf = virtqueue_get_buf_ctx(rq->vq, len, ctx);
> > > +     if (buf) {
> > > +             xdp_headroom = (unsigned long)*ctx;
> > > +             virtnet_rq_unmap(rq, buf + VIRTNET_RX_PAD + xdp_headroom,
> *len);
> > > +     }
> > > +
> > > +     return buf;
> > > +}
> > > +
> > >  static void virtnet_rq_init_one_sg(struct receive_queue *rq, void
> *buf, u32 len)
> > >  {
> > >       struct virtnet_rq_dma *dma;
> > > @@ -2692,13 +2709,23 @@ static int virtnet_receive_packets(struct
> virtnet_info *vi,
> > >       int packets = 0;
> > >       void *buf;
> > >
> > > -     if (!vi->big_packets || vi->mergeable_rx_bufs) {
> > > +     if (vi->mergeable_rx_bufs) {
> > >               void *ctx;
> > >               while (packets < budget &&
> > >                      (buf = virtnet_rq_get_buf(rq, &len, &ctx))) {
> > >                       receive_buf(vi, rq, buf, len, ctx, xdp_xmit,
> stats);
> > >                       packets++;
> > >               }
> > > +     } else if (!vi->big_packets) {
> > > +             void *ctx;
> > > +             unsigned int xdp_headroom = virtnet_get_headroom(vi);
> > > +             unsigned int header_offset = VIRTNET_RX_PAD +
> xdp_headroom;
> > > +
> > > +             while (packets < budget &&
> > > +                    (buf = virtnet_rq_get_buf_small(rq, &len, &ctx,
> header_offset))) {
> > > +                     receive_buf(vi, rq, buf, len, ctx, xdp_xmit,
> stats);
> > > +                     packets++;
> > > +             }
> > >       } else {
> > >               while (packets < budget &&
> > >                      (buf = virtqueue_get_buf(rq->vq, &len)) != NULL) {
> > > --
> > > 2.20.1
> > >
>

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ