Re: [PATCH V3 3/3] tcm_vhost: Use vq->private_data to indicate if the endpoint is setup

[Asias He <asias@redhat.com>]

On Mon, Mar 18, 2013 at 11:30:57AM +0200, Michael S. Tsirkin wrote:
> On Mon, Mar 18, 2013 at 05:14:33PM +0800, Asias He wrote:
> > On Mon, Mar 18, 2013 at 10:19:00AM +0200, Michael S. Tsirkin wrote:
> > > On Fri, Mar 15, 2013 at 09:14:07AM +0800, Asias He wrote:
> > > > Currently, vs->vs_endpoint is used indicate if the endpoint is setup or
> > > > not. It is set or cleared in vhost_scsi_set_endpoint() or
> > > > vhost_scsi_clear_endpoint() under the vs->dev.mutex lock. However, when
> > > > we check it in vhost_scsi_handle_vq(), we ignored the lock, this is
> > > > wrong.
> > > 
> > > This one, I don't get. Why is it wrong? Could you please describe the
> > > race codition you are trying to prevent?
> > 
> > Why is it safe to access vs->vs_endpoint without any lock?
> 
> For the same reason it's safe with the pointer: either readers
> see the old or the new value, and we flush before relying on
> the new value.

vs_endpoint is a bool not a pointer. Here it is even more implicit to
understand the whole story. Why not make it more consistent with the other
user of vhost. Using vq->private_data for backend related data.

We have enough special tricks (vhost rcu, vhost work queue).

> RCU macros also include barriers that are irrelevant if you are not
> going to access any data through the pointer.
> Nowdays they also including lockdep-like checks, which you override.

vhost-net is also overriding, no? And I am not seeing any effect to make
the '1' gonna.

  sock = rcu_dereference_check(vq->private_data, 1);

> > > > Instead of using the vs->vs_endpoint and the vs->dev.mutex lock to
> > > > indicate the status of the endpoint, we use per virtqueue
> > > > vq->private_data to indicate it. In this way, we can only take the
> > > > vq->mutex lock which is per queue and make the concurrent multiqueue
> > > > process having less lock contention. Further, in the read side of
> > > > vq->private_data, we can even do not take only lock if it is accessed in
> > > > the vhost worker thread, because it is protected by "vhost rcu".
> > > 
> > > But (unlike with -net) you never actually need the pointer. So why all
> > > the complexity?
> > 
> > It works as a flag, NULL or !NULL.
> > 
> > This is from your other mail:
> > 
> > '''
> > This takes dev mutex on data path which will introduce
> > contention esp for multiqueue.
> > How about storing the endpoint as part of vq
> > private data and protecting with vq mutex?
> > '''
> 
> Yes this is better than taking the mutex but I don't see
> a problem as is, either. For patch to go into 3.9 it needs
> to fix a bug, not just be a refactoring.

Well, if it is not fix a real bug. Let's skip it for 3.9. 

> > > > Signed-off-by: Asias He <asias@redhat.com>
> > > > ---
> > > >  drivers/vhost/tcm_vhost.c | 46 ++++++++++++++++++++++++++++++++++++++++------
> > > >  1 file changed, 40 insertions(+), 6 deletions(-)
> > > > 
> > > > diff --git a/drivers/vhost/tcm_vhost.c b/drivers/vhost/tcm_vhost.c
> > > > index 43fb11e..099feef 100644
> > > > --- a/drivers/vhost/tcm_vhost.c
> > > > +++ b/drivers/vhost/tcm_vhost.c
> > > > @@ -67,7 +67,6 @@ struct vhost_scsi {
> > > >  	/* Protected by vhost_scsi->dev.mutex */
> > > >  	struct tcm_vhost_tpg *vs_tpg[VHOST_SCSI_MAX_TARGET];
> > > >  	char vs_vhost_wwpn[TRANSPORT_IQN_LEN];
> > > > -	bool vs_endpoint;
> > > >  
> > > >  	struct vhost_dev dev;
> > > >  	struct vhost_virtqueue vqs[VHOST_SCSI_MAX_VQ];
> > > > @@ -91,6 +90,24 @@ static int iov_num_pages(struct iovec *iov)
> > > >  	       ((unsigned long)iov->iov_base & PAGE_MASK)) >> PAGE_SHIFT;
> > > >  }
> > > >  
> > > > +static bool tcm_vhost_check_endpoint(struct vhost_virtqueue *vq)
> > > > +{
> > > > +	bool ret = false;
> > > > +
> > > > +	/*
> > > > +	 * We can handle the vq only after the endpoint is setup by calling the
> > > > +	 * VHOST_SCSI_SET_ENDPOINT ioctl.
> > > > +	 *
> > > > +	 * TODO: Check that we are running from vhost_worker which acts
> > > > +	 * as read-side critical section for vhost kind of RCU.
> > > > +	 * See the comments in struct vhost_virtqueue in drivers/vhost/vhost.h
> > > > +	 */
> > > > +	if (rcu_dereference_check(vq->private_data, 1))
> > > > +		ret = true;
> > > > +
> > > > +	return ret;
> > > > +}
> > > > +
> > > >  static int tcm_vhost_check_true(struct se_portal_group *se_tpg)
> > > >  {
> > > >  	return 1;
> > > > @@ -581,8 +598,7 @@ static void vhost_scsi_handle_vq(struct vhost_scsi *vs,
> > > >  	int head, ret;
> > > >  	u8 target;
> > > >  
> > > > -	/* Must use ioctl VHOST_SCSI_SET_ENDPOINT */
> > > > -	if (unlikely(!vs->vs_endpoint))
> > > > +	if (!tcm_vhost_check_endpoint(vq))
> > > >  		return;
> > > >  
> > > >  	mutex_lock(&vq->mutex);
> > > > @@ -781,8 +797,9 @@ static int vhost_scsi_set_endpoint(
> > > >  {
> > > >  	struct tcm_vhost_tport *tv_tport;
> > > >  	struct tcm_vhost_tpg *tv_tpg;
> > > > +	struct vhost_virtqueue *vq;
> > > >  	bool match = false;
> > > > -	int index, ret;
> > > > +	int index, ret, i;
> > > >  
> > > >  	mutex_lock(&vs->dev.mutex);
> > > >  	/* Verify that ring has been setup correctly. */
> > > > @@ -826,7 +843,13 @@ static int vhost_scsi_set_endpoint(
> > > >  	if (match) {
> > > >  		memcpy(vs->vs_vhost_wwpn, t->vhost_wwpn,
> > > >  		       sizeof(vs->vs_vhost_wwpn));
> > > > -		vs->vs_endpoint = true;
> > > > +		for (i = 0; i < VHOST_SCSI_MAX_VQ; i++) {
> > > > +			vq = &vs->vqs[i];
> > > > +			/* Flushing the vhost_work acts as synchronize_rcu */
> > > > +			mutex_lock(&vq->mutex);
> > > > +			rcu_assign_pointer(vq->private_data, vs);
> > > > +			mutex_unlock(&vq->mutex);
> > > > +		}
> > > >  		ret = 0;
> > > >  	} else {
> > > >  		ret = -EEXIST;
> > > > @@ -842,6 +865,8 @@ static int vhost_scsi_clear_endpoint(
> > > >  {
> > > >  	struct tcm_vhost_tport *tv_tport;
> > > >  	struct tcm_vhost_tpg *tv_tpg;
> > > > +	struct vhost_virtqueue *vq;
> > > > +	bool match = false;
> > > >  	int index, ret, i;
> > > >  	u8 target;
> > > >  
> > > > @@ -877,9 +902,18 @@ static int vhost_scsi_clear_endpoint(
> > > >  		}
> > > >  		tv_tpg->tv_tpg_vhost_count--;
> > > >  		vs->vs_tpg[target] = NULL;
> > > > -		vs->vs_endpoint = false;
> > > > +		match = true;
> > > >  		mutex_unlock(&tv_tpg->tv_tpg_mutex);
> > > >  	}
> > > > +	if (match) {
> > > > +		for (i = 0; i < VHOST_SCSI_MAX_VQ; i++) {
> > > > +			vq = &vs->vqs[i];
> > > > +			/* Flushing the vhost_work acts as synchronize_rcu */
> > > > +			mutex_lock(&vq->mutex);
> > > > +			rcu_assign_pointer(vq->private_data, NULL);
> > > > +			mutex_unlock(&vq->mutex);
> > > > +		}
> > > > +	}
> > > >  	mutex_unlock(&vs->dev.mutex);
> > > >  	return 0;
> > > >  
> > > > -- 
> > > > 1.8.1.4
> > 
> > -- 
> > Asias

-- 
Asias