[dpdk-dev,v4,1/6] vhost: fix windows vm hang
Commit Message
This patch fixes a Windows VM compatibility issue in DPDK 16.07 vhost code,
which causes the guest to hang once any packets are enqueued when mrg_rxbuf
is turned on.
How to test?
1. Start testpmd in the host with a vhost port.
2. Start a Windows VM image with qemu and connect to the vhost port.
3. Start io forwarding with tx_first in host testpmd.
For 16.07 code, the Windows VM will hang once any packets are enqueued.
Cc: <stable@dpdk.org>
Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
---
lib/librte_vhost/vhost_rxtx.c | 17 ++++++++++++-----
1 file changed, 12 insertions(+), 5 deletions(-)
Comments
On Mon, Aug 29, 2016 at 11:35:59PM -0400, Zhihong Wang wrote:
> This patch fixes a Windows VM compatibility issue in DPDK 16.07 vhost code,
> which causes the guest to hang once any packets are enqueued when mrg_rxbuf
> is turned on.
This commit log lacks two important pieces: why does the hang happen and
how does your patch fix it.
> How to test?
>
> 1. Start testpmd in the host with a vhost port.
>
> 2. Start a Windows VM image with qemu and connect to the vhost port.
>
> 3. Start io forwarding with tx_first in host testpmd.
>
> For 16.07 code, the Windows VM will hang once any packets are enqueued.
>
> Cc: <stable@dpdk.org>
> Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
> ---
> lib/librte_vhost/vhost_rxtx.c | 17 ++++++++++++-----
> 1 file changed, 12 insertions(+), 5 deletions(-)
>
> diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
> index 08a73fd..5806f99 100644
> --- a/lib/librte_vhost/vhost_rxtx.c
> +++ b/lib/librte_vhost/vhost_rxtx.c
> @@ -384,6 +384,8 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
> uint16_t start_idx = vq->last_used_idx;
> uint16_t cur_idx = start_idx;
> uint64_t desc_addr;
> + uint32_t desc_chain_head;
> + uint32_t desc_chain_len;
What's the point of introducing "desc_chain_len"? It has the same value
of desc_offset.
--yliu
> -----Original Message-----
> From: Yuanhan Liu [mailto:yuanhan.liu@linux.intel.com]
> Sent: Monday, September 5, 2016 1:25 PM
> To: Wang, Zhihong <zhihong.wang@intel.com>
> Cc: dev@dpdk.org; maxime.coquelin@redhat.com;
> yuanhan.liu@linux.intel.com; thomas.monjalon@6wind.com;
> stable@dpdk.org
> Subject: Re: [dpdk-stable] [PATCH v4 1/6] vhost: fix windows vm hang
>
> On Mon, Aug 29, 2016 at 11:35:59PM -0400, Zhihong Wang wrote:
> > This patch fixes a Windows VM compatibility issue in DPDK 16.07 vhost
> code,
> > which causes the guest to hang once any packets are enqueued when
> mrg_rxbuf
> > is turned on.
>
> This commit log lacks two important pieces: why does the hang happen and
> how does your patch fix it.
Okay, I'll add it in v5.
>
> > How to test?
> >
> > 1. Start testpmd in the host with a vhost port.
> >
> > 2. Start a Windows VM image with qemu and connect to the vhost port.
> >
> > 3. Start io forwarding with tx_first in host testpmd.
> >
> > For 16.07 code, the Windows VM will hang once any packets are enqueued.
> >
> > Cc: <stable@dpdk.org>
> > Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
> > ---
> > lib/librte_vhost/vhost_rxtx.c | 17 ++++++++++++-----
> > 1 file changed, 12 insertions(+), 5 deletions(-)
> >
> > diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
> > index 08a73fd..5806f99 100644
> > --- a/lib/librte_vhost/vhost_rxtx.c
> > +++ b/lib/librte_vhost/vhost_rxtx.c
> > @@ -384,6 +384,8 @@ copy_mbuf_to_desc_mergeable(struct virtio_net
> *dev, struct vhost_virtqueue *vq,
> > uint16_t start_idx = vq->last_used_idx;
> > uint16_t cur_idx = start_idx;
> > uint64_t desc_addr;
> > + uint32_t desc_chain_head;
> > + uint32_t desc_chain_len;
>
> What's the point of introducing "desc_chain_len"? It has the same value
> of desc_offset.
No it's not, desc_offset is the offset of the current desc only.
That's where the old code goes wrong.
If you take a look at the virtio spec:
/* le32 is used here for ids for padding reasons. */
struct vring_used_elem {
/* Index of start of used descriptor chain. */
le32 id;
/* Total length of the descriptor chain which was written to. */
le32 len;
};
>
> --yliu
On Mon, Sep 05, 2016 at 05:25:31AM +0000, Wang, Zhihong wrote:
>
>
> > -----Original Message-----
> > From: Yuanhan Liu [mailto:yuanhan.liu@linux.intel.com]
> > Sent: Monday, September 5, 2016 1:25 PM
> > To: Wang, Zhihong <zhihong.wang@intel.com>
> > Cc: dev@dpdk.org; maxime.coquelin@redhat.com;
> > yuanhan.liu@linux.intel.com; thomas.monjalon@6wind.com;
> > stable@dpdk.org
> > Subject: Re: [dpdk-stable] [PATCH v4 1/6] vhost: fix windows vm hang
> >
> > On Mon, Aug 29, 2016 at 11:35:59PM -0400, Zhihong Wang wrote:
> > > This patch fixes a Windows VM compatibility issue in DPDK 16.07 vhost
> > code,
> > > which causes the guest to hang once any packets are enqueued when
> > mrg_rxbuf
> > > is turned on.
> >
> > This commit log lacks two important pieces: why does the hang happen and
> > how does your patch fix it.
>
> Okay, I'll add it in v5.
>
> >
> > > How to test?
> > >
> > > 1. Start testpmd in the host with a vhost port.
> > >
> > > 2. Start a Windows VM image with qemu and connect to the vhost port.
> > >
> > > 3. Start io forwarding with tx_first in host testpmd.
> > >
> > > For 16.07 code, the Windows VM will hang once any packets are enqueued.
> > >
> > > Cc: <stable@dpdk.org>
> > > Signed-off-by: Zhihong Wang <zhihong.wang@intel.com>
> > > ---
> > > lib/librte_vhost/vhost_rxtx.c | 17 ++++++++++++-----
> > > 1 file changed, 12 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
> > > index 08a73fd..5806f99 100644
> > > --- a/lib/librte_vhost/vhost_rxtx.c
> > > +++ b/lib/librte_vhost/vhost_rxtx.c
> > > @@ -384,6 +384,8 @@ copy_mbuf_to_desc_mergeable(struct virtio_net
> > *dev, struct vhost_virtqueue *vq,
> > > uint16_t start_idx = vq->last_used_idx;
> > > uint16_t cur_idx = start_idx;
> > > uint64_t desc_addr;
> > > + uint32_t desc_chain_head;
> > > + uint32_t desc_chain_len;
> >
> > What's the point of introducing "desc_chain_len"? It has the same value
> > of desc_offset.
>
> No it's not, desc_offset is the offset of the current desc only.
> That's where the old code goes wrong.
Oh, right.
--yliu
@@ -384,6 +384,8 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
uint16_t start_idx = vq->last_used_idx;
uint16_t cur_idx = start_idx;
uint64_t desc_addr;
+ uint32_t desc_chain_head;
+ uint32_t desc_chain_len;
uint32_t mbuf_offset, mbuf_avail;
uint32_t desc_offset, desc_avail;
uint32_t cpy_len;
@@ -412,6 +414,8 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
desc_avail = buf_vec[vec_idx].buf_len - dev->vhost_hlen;
desc_offset = dev->vhost_hlen;
+ desc_chain_head = buf_vec[vec_idx].desc_idx;
+ desc_chain_len = desc_offset;
mbuf_avail = rte_pktmbuf_data_len(m);
mbuf_offset = 0;
@@ -419,19 +423,21 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
/* done with current desc buf, get the next one */
if (desc_avail == 0) {
desc_idx = buf_vec[vec_idx].desc_idx;
+ vec_idx++;
if (!(vq->desc[desc_idx].flags & VRING_DESC_F_NEXT)) {
/* Update used ring with desc information */
used_idx = cur_idx++ & (vq->size - 1);
- vq->used->ring[used_idx].id = desc_idx;
- vq->used->ring[used_idx].len = desc_offset;
+ vq->used->ring[used_idx].id = desc_chain_head;
+ vq->used->ring[used_idx].len = desc_chain_len;
vhost_log_used_vring(dev, vq,
offsetof(struct vring_used,
ring[used_idx]),
sizeof(vq->used->ring[used_idx]));
+ desc_chain_head = buf_vec[vec_idx].desc_idx;
+ desc_chain_len = 0;
}
- vec_idx++;
desc_addr = gpa_to_vva(dev, buf_vec[vec_idx].buf_addr);
if (unlikely(!desc_addr))
return 0;
@@ -463,11 +469,12 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
mbuf_offset += cpy_len;
desc_avail -= cpy_len;
desc_offset += cpy_len;
+ desc_chain_len += cpy_len;
}
used_idx = cur_idx & (vq->size - 1);
- vq->used->ring[used_idx].id = buf_vec[vec_idx].desc_idx;
- vq->used->ring[used_idx].len = desc_offset;
+ vq->used->ring[used_idx].id = desc_chain_head;
+ vq->used->ring[used_idx].len = desc_chain_len;
vhost_log_used_vring(dev, vq,
offsetof(struct vring_used, ring[used_idx]),
sizeof(vq->used->ring[used_idx]));