wexu at redhat.com
2017-Nov-28 17:17 UTC
[PATCH net,stable] vhost: fix skb leak in handle_rx()
From: Wei Xu <wexu at redhat.com>
Matthew found a roughly 40% tcp throughput regression with commit
c67df11f(vhost_net: try batch dequing from skb array) as discussed
in the following thread:
https://www.mail-archive.com/netdev at vger.kernel.org/msg187936.html
Eventually we figured out that it was a skb leak in handle_rx()
when sending packets to the VM. This usually happens when a guest
can not drain out vq as fast as vhost fills in, afterwards it sets
off the traffic jam and leaks skb(s) which occurs as no headcount
to send on the vq from vhost side.
This can be avoided by making sure we have got enough headcount
before actually consuming a skb from the batched rx array while
transmitting, which is simply done by deferring it a moment later
in this patch.
Signed-off-by: Wei Xu <wexu at redhat.com>
---
drivers/vhost/net.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
index 8d626d7..e76535e 100644
--- a/drivers/vhost/net.c
+++ b/drivers/vhost/net.c
@@ -778,8 +778,6 @@ static void handle_rx(struct vhost_net *net)
/* On error, stop handling until the next kick. */
if (unlikely(headcount < 0))
goto out;
- if (nvq->rx_array)
- msg.msg_control = vhost_net_buf_consume(&nvq->rxq);
/* On overrun, truncate and discard */
if (unlikely(headcount > UIO_MAXIOV)) {
iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1);
@@ -809,6 +807,8 @@ static void handle_rx(struct vhost_net *net)
*/
iov_iter_advance(&msg.msg_iter, vhost_hlen);
}
+ if (nvq->rx_array)
+ msg.msg_control = vhost_net_buf_consume(&nvq->rxq);
err = sock->ops->recvmsg(sock, &msg,
sock_len, MSG_DONTWAIT | MSG_TRUNC);
/* Userspace might have consumed the packet meanwhile:
--
1.8.3.1
Michael S. Tsirkin
2017-Nov-28 17:50 UTC
[PATCH net,stable] vhost: fix skb leak in handle_rx()
On Tue, Nov 28, 2017 at 12:17:16PM -0500, wexu at redhat.com wrote:> From: Wei Xu <wexu at redhat.com> > > Matthew found a roughly 40% tcp throughput regression with commit > c67df11f(vhost_net: try batch dequing from skb array) as discussed > in the following thread: > https://www.mail-archive.com/netdev at vger.kernel.org/msg187936.html > > Eventually we figured out that it was a skb leak in handle_rx() > when sending packets to the VM. This usually happens when a guest > can not drain out vq as fast as vhost fills in, afterwards it sets > off the traffic jam and leaks skb(s) which occurs as no headcount > to send on the vq from vhost side. > > This can be avoided by making sure we have got enough headcount > before actually consuming a skb from the batched rx array while > transmitting, which is simply done by deferring it a moment later > in this patch. > > Signed-off-by: Wei Xu <wexu at redhat.com>Looks like a good way to fix it, but it will still leak if recvmsg returns a wrong length (I'm not sure this can happen in practice, but best to keep it simple). Also, we need to add this before each recvmsg, including overrun, and discard on error.> --- > drivers/vhost/net.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c > index 8d626d7..e76535e 100644 > --- a/drivers/vhost/net.c > +++ b/drivers/vhost/net.c > @@ -778,8 +778,6 @@ static void handle_rx(struct vhost_net *net) > /* On error, stop handling until the next kick. */ > if (unlikely(headcount < 0)) > goto out; > - if (nvq->rx_array) > - msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > /* On overrun, truncate and discard */ > if (unlikely(headcount > UIO_MAXIOV)) { > iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1); > @@ -809,6 +807,8 @@ static void handle_rx(struct vhost_net *net) > */ > iov_iter_advance(&msg.msg_iter, vhost_hlen); > } > + if (nvq->rx_array) > + msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > err = sock->ops->recvmsg(sock, &msg, > sock_len, MSG_DONTWAIT | MSG_TRUNC); > /* Userspace might have consumed the packet meanwhile: > -- > 1.8.3.1
Michael S. Tsirkin
2017-Nov-28 17:53 UTC
[PATCH net,stable] vhost: fix skb leak in handle_rx()
On Tue, Nov 28, 2017 at 12:17:16PM -0500, wexu at redhat.com wrote:> From: Wei Xu <wexu at redhat.com> > > Matthew found a roughly 40% tcp throughput regression with commit > c67df11f(vhost_net: try batch dequing from skb array) as discussed > in the following thread: > https://www.mail-archive.com/netdev at vger.kernel.org/msg187936.html > > Eventually we figured out that it was a skb leak in handle_rx() > when sending packets to the VM. This usually happens when a guest > can not drain out vq as fast as vhost fills in, afterwards it sets > off the traffic jam and leaks skb(s) which occurs as no headcount > to send on the vq from vhost side. > > This can be avoided by making sure we have got enough headcount > before actually consuming a skb from the batched rx array while > transmitting, which is simply done by deferring it a moment later > in this patch. > > Signed-off-by: Wei Xu <wexu at redhat.com>Given the amount of effort Matthew has put into this, you definitely want Reported-by: Matthew Rosato <mjrosato at linux.vnet.ibm.com> here. Let's give credit where credit is due. Thanks a lot Matthew!> --- > drivers/vhost/net.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c > index 8d626d7..e76535e 100644 > --- a/drivers/vhost/net.c > +++ b/drivers/vhost/net.c > @@ -778,8 +778,6 @@ static void handle_rx(struct vhost_net *net) > /* On error, stop handling until the next kick. */ > if (unlikely(headcount < 0)) > goto out; > - if (nvq->rx_array) > - msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > /* On overrun, truncate and discard */ > if (unlikely(headcount > UIO_MAXIOV)) { > iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1); > @@ -809,6 +807,8 @@ static void handle_rx(struct vhost_net *net) > */ > iov_iter_advance(&msg.msg_iter, vhost_hlen); > } > + if (nvq->rx_array) > + msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > err = sock->ops->recvmsg(sock, &msg, > sock_len, MSG_DONTWAIT | MSG_TRUNC); > /* Userspace might have consumed the packet meanwhile: > -- > 1.8.3.1
On 2017?11?29? 01:17, wexu at redhat.com wrote:> From: Wei Xu <wexu at redhat.com> > > Matthew found a roughly 40% tcp throughput regression with commit > c67df11f(vhost_net: try batch dequing from skb array) as discussed > in the following thread: > https://www.mail-archive.com/netdev at vger.kernel.org/msg187936.html > > Eventually we figured out that it was a skb leak in handle_rx() > when sending packets to the VM. This usually happens when a guest > can not drain out vq as fast as vhost fills in, afterwards it sets > off the traffic jam and leaks skb(s) which occurs as no headcount > to send on the vq from vhost side. > > This can be avoided by making sure we have got enough headcount > before actually consuming a skb from the batched rx array while > transmitting, which is simply done by deferring it a moment later > in this patch. > > Signed-off-by: Wei Xu <wexu at redhat.com> > --- > drivers/vhost/net.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c > index 8d626d7..e76535e 100644 > --- a/drivers/vhost/net.c > +++ b/drivers/vhost/net.c > @@ -778,8 +778,6 @@ static void handle_rx(struct vhost_net *net) > /* On error, stop handling until the next kick. */ > if (unlikely(headcount < 0)) > goto out; > - if (nvq->rx_array) > - msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > /* On overrun, truncate and discard */ > if (unlikely(headcount > UIO_MAXIOV)) {You need do msg.msg_control = vhost_net_buf_consume() here too, otherwise we may still get it leaked. Thanks> iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1); > @@ -809,6 +807,8 @@ static void handle_rx(struct vhost_net *net) > */ > iov_iter_advance(&msg.msg_iter, vhost_hlen); > } > + if (nvq->rx_array) > + msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > err = sock->ops->recvmsg(sock, &msg, > sock_len, MSG_DONTWAIT | MSG_TRUNC); > /* Userspace might have consumed the packet meanwhile:
On 2017?11?29? 09:53, Jason Wang wrote:> > > On 2017???11???29??? 01:17, wexu at redhat.com wrote: >> From: Wei Xu <wexu at redhat.com> >> >> Matthew found a roughly 40% tcp throughput regression with commit >> c67df11f(vhost_net: try batch dequing from skb array) as discussed >> in the following thread: >> https://www.mail-archive.com/netdev at vger.kernel.org/msg187936.html >> >> Eventually we figured out that it was a skb leak in handle_rx() >> when sending packets to the VM. This usually happens when a guest >> can not drain out vq as fast as vhost fills in, afterwards it sets >> off the traffic jam and leaks skb(s) which occurs as no headcount >> to send on the vq from vhost side. >> >> This can be avoided by making sure we have got enough headcount >> before actually consuming a skb from the batched rx array while >> transmitting, which is simply done by deferring it a moment later >> in this patch. >> >> Signed-off-by: Wei Xu <wexu at redhat.com> >> --- >> ? drivers/vhost/net.c | 4 ++-- >> ? 1 file changed, 2 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c >> index 8d626d7..e76535e 100644 >> --- a/drivers/vhost/net.c >> +++ b/drivers/vhost/net.c >> @@ -778,8 +778,6 @@ static void handle_rx(struct vhost_net *net) >> ????????? /* On error, stop handling until the next kick. */ >> ????????? if (unlikely(headcount < 0)) >> ????????????? goto out; >> -??????? if (nvq->rx_array) >> -??????????? msg.msg_control = vhost_net_buf_consume(&nvq->rxq); >> ????????? /* On overrun, truncate and discard */ >> ????????? if (unlikely(headcount > UIO_MAXIOV)) { > > You need do msg.msg_control = vhost_net_buf_consume() here too, > otherwise we may still get it leaked. > > ThanksNot a leak actually, but the packet won't be consumed and we will hit UIO_MAXIOV forever in this case. Thanks
On Tue, Nov 28, 2017 at 07:50:58PM +0200, Michael S. Tsirkin wrote:> On Tue, Nov 28, 2017 at 12:17:16PM -0500, wexu at redhat.com wrote: > > From: Wei Xu <wexu at redhat.com> > > > > Matthew found a roughly 40% tcp throughput regression with commit > > c67df11f(vhost_net: try batch dequing from skb array) as discussed > > in the following thread: > > https://www.mail-archive.com/netdev at vger.kernel.org/msg187936.html > > > > Eventually we figured out that it was a skb leak in handle_rx() > > when sending packets to the VM. This usually happens when a guest > > can not drain out vq as fast as vhost fills in, afterwards it sets > > off the traffic jam and leaks skb(s) which occurs as no headcount > > to send on the vq from vhost side. > > > > This can be avoided by making sure we have got enough headcount > > before actually consuming a skb from the batched rx array while > > transmitting, which is simply done by deferring it a moment later > > in this patch. > > > > Signed-off-by: Wei Xu <wexu at redhat.com> > > Looks like a good way to fix it, but it will still leak if recvmsg > returns a wrong length (I'm not sure this can happen in practice, but > best to keep it simple).Right, it is better to defend this case, will include it in v2.> > Also, we need to add this before each recvmsg, including overrun, > and discard on error.OK. Wei> > > --- > > drivers/vhost/net.c | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c > > index 8d626d7..e76535e 100644 > > --- a/drivers/vhost/net.c > > +++ b/drivers/vhost/net.c > > @@ -778,8 +778,6 @@ static void handle_rx(struct vhost_net *net) > > /* On error, stop handling until the next kick. */ > > if (unlikely(headcount < 0)) > > goto out; > > - if (nvq->rx_array) > > - msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > > /* On overrun, truncate and discard */ > > if (unlikely(headcount > UIO_MAXIOV)) { > > iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1); > > @@ -809,6 +807,8 @@ static void handle_rx(struct vhost_net *net) > > */ > > iov_iter_advance(&msg.msg_iter, vhost_hlen); > > } > > + if (nvq->rx_array) > > + msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > > err = sock->ops->recvmsg(sock, &msg, > > sock_len, MSG_DONTWAIT | MSG_TRUNC); > > /* Userspace might have consumed the packet meanwhile: > > -- > > 1.8.3.1
On Tue, Nov 28, 2017 at 07:53:33PM +0200, Michael S. Tsirkin wrote:> On Tue, Nov 28, 2017 at 12:17:16PM -0500, wexu at redhat.com wrote: > > From: Wei Xu <wexu at redhat.com> > > > > Matthew found a roughly 40% tcp throughput regression with commit > > c67df11f(vhost_net: try batch dequing from skb array) as discussed > > in the following thread: > > https://www.mail-archive.com/netdev at vger.kernel.org/msg187936.html > > > > Eventually we figured out that it was a skb leak in handle_rx() > > when sending packets to the VM. This usually happens when a guest > > can not drain out vq as fast as vhost fills in, afterwards it sets > > off the traffic jam and leaks skb(s) which occurs as no headcount > > to send on the vq from vhost side. > > > > This can be avoided by making sure we have got enough headcount > > before actually consuming a skb from the batched rx array while > > transmitting, which is simply done by deferring it a moment later > > in this patch. > > > > Signed-off-by: Wei Xu <wexu at redhat.com> > > Given the amount of effort Matthew has put into this, > you definitely want > > Reported-by: Matthew Rosato <mjrosato at linux.vnet.ibm.com> > > here. >Absolutely we want that, sorry for missed you here, Matthew. It is more like you have figured out this issue independently all by yourself with a wide assortment of extremely quick tests(tools, throughput, slub leak, etc), and I am happy that I have the opportunity to do the paperwork on behalf of you.:) Thanks a lot Matthew. Wei> Let's give credit where credit is due. > > Thanks a lot Matthew! > > > --- > > drivers/vhost/net.c | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c > > index 8d626d7..e76535e 100644 > > --- a/drivers/vhost/net.c > > +++ b/drivers/vhost/net.c > > @@ -778,8 +778,6 @@ static void handle_rx(struct vhost_net *net) > > /* On error, stop handling until the next kick. */ > > if (unlikely(headcount < 0)) > > goto out; > > - if (nvq->rx_array) > > - msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > > /* On overrun, truncate and discard */ > > if (unlikely(headcount > UIO_MAXIOV)) { > > iov_iter_init(&msg.msg_iter, READ, vq->iov, 1, 1); > > @@ -809,6 +807,8 @@ static void handle_rx(struct vhost_net *net) > > */ > > iov_iter_advance(&msg.msg_iter, vhost_hlen); > > } > > + if (nvq->rx_array) > > + msg.msg_control = vhost_net_buf_consume(&nvq->rxq); > > err = sock->ops->recvmsg(sock, &msg, > > sock_len, MSG_DONTWAIT | MSG_TRUNC); > > /* Userspace might have consumed the packet meanwhile: > > -- > > 1.8.3.1