Re: [PATCH V3 2/2] vhost: handle polling errors

From: Jason Wang
Date: Mon Jan 07 2013 - 10:04:27 EST


On 01/07/2013 10:55 PM, Michael S. Tsirkin wrote:
> On Mon, Jan 07, 2013 at 12:38:17PM +0800, Jason Wang wrote:
>> On 01/06/2013 09:22 PM, Michael S. Tsirkin wrote:
>>> On Sun, Jan 06, 2013 at 03:18:38PM +0800, Jason Wang wrote:
>>>> Polling errors were ignored by vhost/vhost_net, this may lead to crash when
>>>> trying to remove vhost from waitqueue when after the polling is failed. Solve
>>>> this problem by:
>>>>
>>>> - checking the poll->wqh before trying to remove from waitqueue
>>>> - report an error when poll() returns a POLLERR in vhost_start_poll()
>>>> - report an error when vhost_start_poll() fails in
>>>> vhost_vring_ioctl()/vhost_net_set_backend() which is used to notify the
>>>> failure to userspace.
>>>> - report an error in the data path in vhost_net when meet polling errors.
>>>>
>>>> After those changes, we can safely drop the tx polling state in vhost_net since
>>>> it was replaced by the checking of poll->wqh.
>>>>
>>>> Signed-off-by: Jason Wang <jasowang@xxxxxxxxxx>
>>>> ---
>>>> drivers/vhost/net.c | 74 ++++++++++++++++--------------------------------
>>>> drivers/vhost/vhost.c | 31 +++++++++++++++-----
>>>> drivers/vhost/vhost.h | 2 +-
>>>> 3 files changed, 49 insertions(+), 58 deletions(-)
>>>>
>>>> diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
>>>> index d10ad6f..125c1e5 100644
>>>> --- a/drivers/vhost/net.c
>>>> +++ b/drivers/vhost/net.c
>>>> @@ -64,20 +64,10 @@ enum {
>>>> VHOST_NET_VQ_MAX = 2,
>>>> };
>>>>
>>>> -enum vhost_net_poll_state {
>>>> - VHOST_NET_POLL_DISABLED = 0,
>>>> - VHOST_NET_POLL_STARTED = 1,
>>>> - VHOST_NET_POLL_STOPPED = 2,
>>>> -};
>>>> -
>>>> struct vhost_net {
>>>> struct vhost_dev dev;
>>>> struct vhost_virtqueue vqs[VHOST_NET_VQ_MAX];
>>>> struct vhost_poll poll[VHOST_NET_VQ_MAX];
>>>> - /* Tells us whether we are polling a socket for TX.
>>>> - * We only do this when socket buffer fills up.
>>>> - * Protected by tx vq lock. */
>>>> - enum vhost_net_poll_state tx_poll_state;
>>>> /* Number of TX recently submitted.
>>>> * Protected by tx vq lock. */
>>>> unsigned tx_packets;
>>>> @@ -155,24 +145,6 @@ static void copy_iovec_hdr(const struct iovec *from, struct iovec *to,
>>>> }
>>>> }
>>>>
>>>> -/* Caller must have TX VQ lock */
>>>> -static void tx_poll_stop(struct vhost_net *net)
>>>> -{
>>>> - if (likely(net->tx_poll_state != VHOST_NET_POLL_STARTED))
>>>> - return;
>>>> - vhost_poll_stop(net->poll + VHOST_NET_VQ_TX);
>>>> - net->tx_poll_state = VHOST_NET_POLL_STOPPED;
>>>> -}
>>>> -
>>>> -/* Caller must have TX VQ lock */
>>>> -static void tx_poll_start(struct vhost_net *net, struct socket *sock)
>>>> -{
>>>> - if (unlikely(net->tx_poll_state != VHOST_NET_POLL_STOPPED))
>>>> - return;
>>>> - vhost_poll_start(net->poll + VHOST_NET_VQ_TX, sock->file);
>>>> - net->tx_poll_state = VHOST_NET_POLL_STARTED;
>>>> -}
>>>> -
>>>> /* In case of DMA done not in order in lower device driver for some reason.
>>>> * upend_idx is used to track end of used idx, done_idx is used to track head
>>>> * of used idx. Once lower device DMA done contiguously, we will signal KVM
>>>> @@ -227,6 +199,7 @@ static void vhost_zerocopy_callback(struct ubuf_info *ubuf, bool success)
>>>> static void handle_tx(struct vhost_net *net)
>>>> {
>>>> struct vhost_virtqueue *vq = &net->dev.vqs[VHOST_NET_VQ_TX];
>>>> + struct vhost_poll *poll = net->poll + VHOST_NET_VQ_TX;
>>>> unsigned out, in, s;
>>>> int head;
>>>> struct msghdr msg = {
>>>> @@ -252,7 +225,8 @@ static void handle_tx(struct vhost_net *net)
>>>> wmem = atomic_read(&sock->sk->sk_wmem_alloc);
>>>> if (wmem >= sock->sk->sk_sndbuf) {
>>>> mutex_lock(&vq->mutex);
>>>> - tx_poll_start(net, sock);
>>>> + if (vhost_poll_start(poll, sock->file))
>>>> + vq_err(vq, "Fail to start TX polling\n");
>>> s/Fail/Failed/
>>>
>>> A question though: how can this happen? Could you clarify please?
>>> Maybe we can find a way to prevent this error?
>> Two conditions I think this can happen:
>>
>> 1) a buggy userspace disable a queue through TUNSETQUEUE
>> 2) the net device were gone
>>
>> For 1, looks like we can delay the disabling until the refcnt goes to
>> zero. For 2 may needs more changes.
> I'd expect keeping a socket reference would prevent both issues.
> Doesn't it?

Doesn't work for 2 I think, the socket didn't hold a refcnt of the
device, so the device can go away at anytime. Although we can change
this, but it's the behaviour before multiqueue support.
>
>> Not sure it's worth to do this work,
>> maybe a warning is enough just like other failure.
> With other failures, you normally can correct the error then
> kick to have it restart. This is soomething thagt would not
> work here.

If userspace is wrote correctly, (e.g passing a fd with correct state)
it can also be corrected.
>
>>>> mutex_unlock(&vq->mutex);
>>>> return;
>>>> }
>>>> @@ -261,7 +235,7 @@ static void handle_tx(struct vhost_net *net)
>>>> vhost_disable_notify(&net->dev, vq);
>>>>
>>>> if (wmem < sock->sk->sk_sndbuf / 2)
>>>> - tx_poll_stop(net);
>>>> + vhost_poll_stop(poll);
>>>> hdr_size = vq->vhost_hlen;
>>>> zcopy = vq->ubufs;
>>>>
>>>> @@ -283,8 +257,10 @@ static void handle_tx(struct vhost_net *net)
>>>>
>>>> wmem = atomic_read(&sock->sk->sk_wmem_alloc);
>>>> if (wmem >= sock->sk->sk_sndbuf * 3 / 4) {
>>>> - tx_poll_start(net, sock);
>>>> - set_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
>>>> + if (vhost_poll_start(poll, sock->file))
>>>> + vq_err(vq, "Fail to start TX polling\n");
>>>> + else
>>>> + set_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
>>>> break;
>>>> }
>>>> /* If more outstanding DMAs, queue the work.
>>>> @@ -294,8 +270,10 @@ static void handle_tx(struct vhost_net *net)
>>>> (vq->upend_idx - vq->done_idx) :
>>>> (vq->upend_idx + UIO_MAXIOV - vq->done_idx);
>>>> if (unlikely(num_pends > VHOST_MAX_PEND)) {
>>>> - tx_poll_start(net, sock);
>>>> - set_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
>>>> + if (vhost_poll_start(poll, sock->file))
>>>> + vq_err(vq, "Fail to start TX polling\n");
>>>> + else
>>>> + set_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
>>>> break;
>>>> }
>>>> if (unlikely(vhost_enable_notify(&net->dev, vq))) {
>>>> @@ -360,7 +338,8 @@ static void handle_tx(struct vhost_net *net)
>>>> }
>>>> vhost_discard_vq_desc(vq, 1);
>>>> if (err == -EAGAIN || err == -ENOBUFS)
>>>> - tx_poll_start(net, sock);
>>>> + if (vhost_poll_start(poll, sock->file))
>>>> + vq_err(vq, "Fail to start TX polling\n");
>>>> break;
>>>> }
>>>> if (err != len)
>>>> @@ -623,7 +602,6 @@ static int vhost_net_open(struct inode *inode, struct file *f)
>>>>
>>>> vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, POLLOUT, dev);
>>>> vhost_poll_init(n->poll + VHOST_NET_VQ_RX, handle_rx_net, POLLIN, dev);
>>>> - n->tx_poll_state = VHOST_NET_POLL_DISABLED;
>>>>
>>>> f->private_data = n;
>>>>
>>>> @@ -633,29 +611,25 @@ static int vhost_net_open(struct inode *inode, struct file *f)
>>>> static void vhost_net_disable_vq(struct vhost_net *n,
>>>> struct vhost_virtqueue *vq)
>>>> {
>>>> + struct vhost_poll *poll = n->poll + (vq - n->vqs);
>>>> +
>>>> if (!vq->private_data)
>>>> return;
>>>> - if (vq == n->vqs + VHOST_NET_VQ_TX) {
>>>> - tx_poll_stop(n);
>>>> - n->tx_poll_state = VHOST_NET_POLL_DISABLED;
>>>> - } else
>>>> - vhost_poll_stop(n->poll + VHOST_NET_VQ_RX);
>>>> + vhost_poll_stop(poll);
>>>> }
>>>>
>>>> -static void vhost_net_enable_vq(struct vhost_net *n,
>>>> +static int vhost_net_enable_vq(struct vhost_net *n,
>>>> struct vhost_virtqueue *vq)
>>>> {
>>>> struct socket *sock;
>>>> + struct vhost_poll *poll = n->poll + (vq - n->vqs);
>>>>
>>>> sock = rcu_dereference_protected(vq->private_data,
>>>> lockdep_is_held(&vq->mutex));
>>>> if (!sock)
>>>> - return;
>>>> - if (vq == n->vqs + VHOST_NET_VQ_TX) {
>>>> - n->tx_poll_state = VHOST_NET_POLL_STOPPED;
>>>> - tx_poll_start(n, sock);
>>>> - } else
>>>> - vhost_poll_start(n->poll + VHOST_NET_VQ_RX, sock->file);
>>>> + return 0;
>>>> +
>>>> + return vhost_poll_start(poll, sock->file);
>>>> }
>>>>
>>>> static struct socket *vhost_net_stop_vq(struct vhost_net *n,
>>>> @@ -833,7 +807,9 @@ static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd)
>>>> r = vhost_init_used(vq);
>>>> if (r)
>>>> goto err_used;
>>>> - vhost_net_enable_vq(n, vq);
>>>> + r = vhost_net_enable_vq(n, vq);
>>>> + if (r)
>>>> + goto err_used;
>>>>
>>>> oldubufs = vq->ubufs;
>>>> vq->ubufs = ubufs;
>>>> diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
>>>> index 34389f7..5c7a466 100644
>>>> --- a/drivers/vhost/vhost.c
>>>> +++ b/drivers/vhost/vhost.c
>>>> @@ -77,26 +77,41 @@ void vhost_poll_init(struct vhost_poll *poll, vhost_work_fn_t fn,
>>>> init_poll_funcptr(&poll->table, vhost_poll_func);
>>>> poll->mask = mask;
>>>> poll->dev = dev;
>>>> + poll->wqh = NULL;
>>>>
>>>> vhost_work_init(&poll->work, fn);
>>>> }
>>>>
>>>> +/* Stop polling a file. After this function returns, it becomes safe to drop the
>>>> + * file reference. You must also flush afterwards. */
>>>> +void vhost_poll_stop(struct vhost_poll *poll)
>>>> +{
>>>> + if (poll->wqh) {
>>>> + remove_wait_queue(poll->wqh, &poll->wait);
>>>> + poll->wqh = NULL;
>>>> + }
>>>> +}
>>>> +
>>>> /* Start polling a file. We add ourselves to file's wait queue. The caller must
>>>> * keep a reference to a file until after vhost_poll_stop is called. */
>>>> -void vhost_poll_start(struct vhost_poll *poll, struct file *file)
>>>> +int vhost_poll_start(struct vhost_poll *poll, struct file *file)
>>>> {
>>>> unsigned long mask;
>>>> + int ret = 0;
>>>> +
>>>> + if (poll->wqh)
>>>> + return -EBUSY;
>>>>
>>> I think this should return success: we are already polling.
>>> Otherwise this would trigger a bug below I think.
>> Ok.
>>>> mask = file->f_op->poll(file, &poll->table);
>>>> if (mask)
>>>> vhost_poll_wakeup(&poll->wait, 0, 0, (void *)mask);
>>>> -}
>>>>
>>>> -/* Stop polling a file. After this function returns, it becomes safe to drop the
>>>> - * file reference. You must also flush afterwards. */
>>>> -void vhost_poll_stop(struct vhost_poll *poll)
>>>> -{
>>>> - remove_wait_queue(poll->wqh, &poll->wait);
>>>> + if (mask & POLLERR) {
>>>> + ret = -EINVAL;
>>>> + vhost_poll_stop(poll);
>>>> + }
>>>> +
>>>> + return ret;
>>>> }
>>>>
>>>> static bool vhost_work_seq_done(struct vhost_dev *dev, struct vhost_work *work,
>>>> @@ -792,7 +807,7 @@ long vhost_vring_ioctl(struct vhost_dev *d, int ioctl, void __user *argp)
>>>> fput(filep);
>>>>
>>>> if (pollstart && vq->handle_kick)
>>>> - vhost_poll_start(&vq->poll, vq->kick);
>>>> + r = vhost_poll_start(&vq->poll, vq->kick);
>>>>
>>>> mutex_unlock(&vq->mutex);
>>>>
>>>> diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
>>>> index 2639c58..17261e2 100644
>>>> --- a/drivers/vhost/vhost.h
>>>> +++ b/drivers/vhost/vhost.h
>>>> @@ -42,7 +42,7 @@ void vhost_work_queue(struct vhost_dev *dev, struct vhost_work *work);
>>>>
>>>> void vhost_poll_init(struct vhost_poll *poll, vhost_work_fn_t fn,
>>>> unsigned long mask, struct vhost_dev *dev);
>>>> -void vhost_poll_start(struct vhost_poll *poll, struct file *file);
>>>> +int vhost_poll_start(struct vhost_poll *poll, struct file *file);
>>>> void vhost_poll_stop(struct vhost_poll *poll);
>>>> void vhost_poll_flush(struct vhost_poll *poll);
>>>> void vhost_poll_queue(struct vhost_poll *poll);
>>>> --
>>>> 1.7.1
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe netdev" in
>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/