Re: [PATCH v4 2/2] tcm_vhost: Wait for pending requests in vhost_scsi_flush()

["Michael S. Tsirkin" <mst@redhat.com>]

On Sun, Apr 14, 2013 at 08:27:14PM +0800, Asias He wrote:
> On Sun, Apr 14, 2013 at 12:58:03PM +0300, Michael S. Tsirkin wrote:
> > On Sat, Apr 13, 2013 at 11:29:14AM +0800, Asias He wrote:
> > > This patch makes vhost_scsi_flush() wait for all the pending requests
> > > issued before the flush operation to be finished.
> > > 
> > > Changes in v4:
> > > - Introduce vhost_scsi_inflight
> > > - Drop array to track flush
> > > - Use RCU to protect vs_inflight explicitly
> > > 
> > > Changes in v3:
> > > - Rebase
> > > - Drop 'tcm_vhost: Wait for pending requests in
> > >   vhost_scsi_clear_endpoint()' in this series, we already did that in
> > >   'tcm_vhost: Use vq->private_data to indicate if the endpoint is setup'
> > > 
> > > Changes in v2:
> > > - Increase/Decrease inflight requests in
> > >   vhost_scsi_{allocate,free}_cmd and tcm_vhost_{allocate,free}_evt
> > > 
> > > Signed-off-by: Asias He <asias@redhat.com>
> > > ---
> > >  drivers/vhost/tcm_vhost.c | 73 +++++++++++++++++++++++++++++++++++++++++++++++
> > >  drivers/vhost/tcm_vhost.h |  5 ++++
> > >  2 files changed, 78 insertions(+)
> > > 
> > > diff --git a/drivers/vhost/tcm_vhost.c b/drivers/vhost/tcm_vhost.c
> > > index e09f0fe..5dde525 100644
> > > --- a/drivers/vhost/tcm_vhost.c
> > > +++ b/drivers/vhost/tcm_vhost.c
> > > @@ -74,6 +74,11 @@ enum {
> > >  #define VHOST_SCSI_MAX_VQ	128
> > >  #define VHOST_SCSI_MAX_EVENT	128
> > >  
> > > +struct vhost_scsi_inflight {
> > > +	wait_queue_head_t wait;
> > > +	atomic_t count;
> > 
> > Okay now let's switch to kref + completion, make it more
> > descriptive.
> 
> I still do not see why kref is better.

It makes the fact you are doing reference counting, explicit.

> Completion sounds good.
> 
> > > +};
> > > +
> > >  struct vhost_scsi {
> > >  	/* Protected by vhost_scsi->dev.mutex */
> > >  	struct tcm_vhost_tpg **vs_tpg;
> > > @@ -91,6 +96,8 @@ struct vhost_scsi {
> > >  	struct mutex vs_events_lock; /* protect vs_events_dropped,events_nr */
> > >  	bool vs_events_dropped; /* any missed events */
> > >  	int vs_events_nr; /* num of pending events */
> > > +
> > > +	struct vhost_scsi_inflight __rcu *vs_inflight; /* track inflight req */
> > >  };
> > >  
> > >  /* Local pointer to allocated TCM configfs fabric module */
> > > @@ -108,6 +115,51 @@ static int iov_num_pages(struct iovec *iov)
> > >  	       ((unsigned long)iov->iov_base & PAGE_MASK)) >> PAGE_SHIFT;
> > >  }
> > >  
> > > +static struct vhost_scsi_inflight *
> > > +tcm_vhost_alloc_inflight(struct vhost_scsi *vs)
> > > +{
> > > +	struct vhost_scsi_inflight *inflight;
> > > +
> > > +	inflight = kzalloc(sizeof(*inflight), GFP_KERNEL);
> > > +	if (inflight) {
> > 
> > This is used in set_features, so let's make it int
> > and return error to user if not. No need to corrupt kernel
> > memory silently like this.
> 
> What do you mean by used in set_features? The return value of
> tcm_vhost_alloc_inflight was used at some point, but now it is not used.
> So I will return int.
> 
> Why is it corrupted?

You skip flushes so something can be in flight, our code
assumes flush actually flushes things.

> > 
> > 
> > > +		atomic_set(&inflight->count, 0);
> > 
> > 
> > Ugh. So once all requests finish, refcount is 0
> > and then inflight is freed, and then the next request will
> > get a freed inflight value and dereference. Looks pretty bad,
> > but maybe there's an increment somewhere that fixes it.
> 
> How can the next request get a freed inflight? It can not happen.  The
> old inflight is freed only after all the requests which reference it are
> finished. See the last few lines in vhost_scsi_flush.
> 
> > But let's not go there.  That's why I said above we should use kref +
> > completion. That makes is very clear how to use it correctly.
> > So:
> > 	- initialize to 1
> > 	- swap pointer with RCU
> > 	- decrement
> > 	- wait_for_completion
> 
> We can not go there.

Right. But it's confusing, and also adds overhead on data path
(wakeup each time last request is completed).
Let's do standard ref counting: init to 1, before flush - decrement
and wait for completion.

> > 
> > 
> > > +		init_waitqueue_head(&inflight->wait);
> > > +	}
> > > +	rcu_assign_pointer(vs->vs_inflight, inflight);
> > > +	synchronize_rcu();
> > > +
> > > +	return inflight;
> > > +}
> > > +
> > 
> > This looks like it will overwrite inflight without
> > freeing the old one. In fact it won't because caller
> > has saved the pointer but this interface is
> > just too tricky. Please just opencode this function.
> > 
> 
> Did you see the old inflight was freed in vhost_scsi_flush().
> It was code in a helper function because it is used in too places.
> One is in vhost_scsi_open and the other is in vhost_scsi_scsi?

The name is still confusing.
alloc should simply allocate and return pointer.
Have callers do assign and flush as appropriate.
In particular open does not need synchronize_rcu,
and not checking old inflight value.

> > 
> > > +static struct vhost_scsi_inflight *
> > > +tcm_vhost_inc_inflight(struct vhost_scsi *vs)
> > > +{
> > > +	struct vhost_scsi_inflight *inflight;
> > > +
> > > +	rcu_read_lock();
> > > +	inflight = rcu_dereference(vs->vs_inflight);
> > > +	if (inflight)
> > 
> > How can it be NULL?
> 
> When tcm_vhost_alloc_inflight failed to allocate inflight.

Then we won't flush which is a wrong way to handle such
and error. Instead, fail the command.

> > > +		atomic_inc(&inflight->count);
> > > +	rcu_read_unlock();
> > > +
> > > +	return inflight;
> > > +}
> > > +
> > > +static void tcm_vhost_dec_inflight(struct vhost_scsi_inflight *inflight)
> > > +{
> > > +	/*
> > > +	 * Wakeup the waiter when all the requests issued before the flush
> > > +	 * operation are finished.
> > > +	 */
> > > +	if (inflight && !atomic_dec_return(&inflight->count))
> > > +		wake_up(&inflight->wait);
> > > +}
> > > +
> > > +static bool tcm_vhost_done_inflight(struct vhost_scsi_inflight *inflight)
> > > +{
> > > +	return atomic_read(&inflight->count) == 0;
> > > +}
> > > +
> > >  static bool tcm_vhost_check_feature(struct vhost_scsi *vs, int feature)
> > >  {
> > >  	bool ret = false;
> > > @@ -402,6 +454,7 @@ static int tcm_vhost_queue_tm_rsp(struct se_cmd *se_cmd)
> > >  static void tcm_vhost_free_evt(struct vhost_scsi *vs, struct tcm_vhost_evt *evt)
> > >  {
> > >  	mutex_lock(&vs->vs_events_lock);
> > > +	tcm_vhost_dec_inflight(evt->inflight);
> > >  	vs->vs_events_nr--;
> > >  	kfree(evt);
> > >  	mutex_unlock(&vs->vs_events_lock);
> > > @@ -423,6 +476,7 @@ static struct tcm_vhost_evt *tcm_vhost_allocate_evt(struct vhost_scsi *vs,
> > >  	if (evt) {
> > >  		evt->event.event = event;
> > >  		evt->event.reason = reason;
> > > +		evt->inflight = tcm_vhost_inc_inflight(vs);
> > >  		vs->vs_events_nr++;
> > >  	}
> > >  	mutex_unlock(&vs->vs_events_lock);
> > > @@ -445,6 +499,8 @@ static void vhost_scsi_free_cmd(struct tcm_vhost_cmd *tv_cmd)
> > >  		kfree(tv_cmd->tvc_sgl);
> > >  	}
> > >  
> > > +	tcm_vhost_dec_inflight(tv_cmd->inflight);
> > > +
> > >  	kfree(tv_cmd);
> > >  }
> > >  
> > > @@ -595,6 +651,7 @@ static struct tcm_vhost_cmd *vhost_scsi_allocate_cmd(
> > >  	tv_cmd->tvc_data_direction = data_direction;
> > >  	tv_cmd->tvc_nexus = tv_nexus;
> > >  	tv_cmd->tvc_vhost = vs;
> > > +	tv_cmd->inflight = tcm_vhost_inc_inflight(vs);
> > >  
> > >  	return tv_cmd;
> > >  }
> > > @@ -983,10 +1040,22 @@ static void vhost_scsi_flush_vq(struct vhost_scsi *vs, int index)
> > >  static void vhost_scsi_flush(struct vhost_scsi *vs)
> > >  {
> > >  	int i;
> > > +	struct vhost_scsi_inflight *inflight;
> > > +
> > > +	inflight = ACCESS_ONCE(vs->vs_inflight);
> > 
> > rcu_dereference_protected ? This ACCESS_ONCE looks bogus.
> 
> okay.
> 
> > > +	if (!tcm_vhost_alloc_inflight(vs))
> > > +		return;
> > >  
> > >  	for (i = 0; i < VHOST_SCSI_MAX_VQ; i++)
> > >  		vhost_scsi_flush_vq(vs, i);
> > >  	vhost_work_flush(&vs->dev, &vs->vs_completion_work);
> > > +	vhost_work_flush(&vs->dev, &vs->vs_event_work);
> > > +
> > > +	/* Wait until all requests issued before the flush to be finished */
> > 
> > s/until/for/
> 
> okay.
> 
> > > +	if (inflight) {
> > 
> > How can this be NULL?
> 
> When tcm_vhost_alloc_inflight failed to allocate inflight.

Again, wrong way to handle it.

> > > +		wait_event(inflight->wait, tcm_vhost_done_inflight(inflight));
> > > +		kfree(inflight);
> > > +	}
> > >  }
> > >  
> > >  /*
> > > @@ -1195,6 +1264,9 @@ static int vhost_scsi_open(struct inode *inode, struct file *f)
> > >  	s->vs_events_dropped = false;
> > >  	mutex_init(&s->vs_events_lock);
> > >  
> > > +	if (!tcm_vhost_alloc_inflight(s))
> > > +		return -ENOMEM;
> > > +
> > >  	s->vqs[VHOST_SCSI_VQ_CTL].handle_kick = vhost_scsi_ctl_handle_kick;
> > >  	s->vqs[VHOST_SCSI_VQ_EVT].handle_kick = vhost_scsi_evt_handle_kick;
> > >  	for (i = VHOST_SCSI_VQ_IO; i < VHOST_SCSI_MAX_VQ; i++)
> > > @@ -1220,6 +1292,7 @@ static int vhost_scsi_release(struct inode *inode, struct file *f)
> > >  	vhost_scsi_clear_endpoint(s, &t);
> > >  	vhost_dev_stop(&s->dev);
> > >  	vhost_dev_cleanup(&s->dev, false);
> > > +	kfree(s->vs_inflight);
> > >  	kfree(s);
> > >  	return 0;
> > >  }
> > > diff --git a/drivers/vhost/tcm_vhost.h b/drivers/vhost/tcm_vhost.h
> > > index 94e9ee53..7567767 100644
> > > --- a/drivers/vhost/tcm_vhost.h
> > > +++ b/drivers/vhost/tcm_vhost.h
> > > @@ -2,6 +2,7 @@
> > >  #define TCM_VHOST_NAMELEN 256
> > >  #define TCM_VHOST_MAX_CDB_SIZE 32
> > >  
> > > +struct vhost_scsi_inflight;
> > >  struct tcm_vhost_cmd {
> > >  	/* Descriptor from vhost_get_vq_desc() for virt_queue segment */
> > >  	int tvc_vq_desc;
> > > @@ -37,6 +38,8 @@ struct tcm_vhost_cmd {
> > >  	unsigned char tvc_sense_buf[TRANSPORT_SENSE_BUFFER];
> > >  	/* Completed commands list, serviced from vhost worker thread */
> > >  	struct llist_node tvc_completion_list;
> > > +	/* Used to track inflight req */
> > > +	struct vhost_scsi_inflight *inflight;
> > >  };
> > >  
> > >  struct tcm_vhost_nexus {
> > > @@ -91,6 +94,8 @@ struct tcm_vhost_evt {
> > >  	struct virtio_scsi_event event;
> > >  	/* virtio_scsi event list, serviced from vhost worker thread */
> > >  	struct llist_node list;
> > > +	/* Used to track inflight req */
> > > +	struct vhost_scsi_inflight *inflight;
> > >  };
> > >  
> > >  /*
> > > -- 
> > > 1.8.1.4
> 
> -- 
> Asias