On 1/26/2022 3:30 AM, Michael S. Tsirkin wrote:

On Tue, Jan 25, 2022 at 05:17:43PM +0800, Zhu Lingshan wrote:

quoted

When irq number is negative(e.g., -EINVAL), the virtqueue
may be disabled or the virtqueues are sharing a device irq.
In such case, we should not setup irq offloading for a virtqueue.

Signed-off-by: Zhu Lingshan <redacted>
---
  drivers/vhost/vdpa.c | 3 +++
  1 file changed, 3 insertions(+)

diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c
index 851539807bc9..909891d518e8 100644
--- a/drivers/vhost/vdpa.c
+++ b/drivers/vhost/vdpa.c

@@ -96,6 +96,9 @@ static void vhost_vdpa_setup_vq_irq(struct vhost_vdpa *v, u16 qid)
  	if (!ops->get_vq_irq)
  		return;
  
+	if (irq < 0)
+		return;
+
  	irq = ops->get_vq_irq(vdpa, qid);

So it's used before it's initialized. Ugh.
How was this patchset tested?

Sorry, my bad, it is not rebased properly, V3 can fix this for sure.

Thanks

quoted

  	irq_bypass_unregister_producer(&vq->call_ctx.producer);
  	if (!vq->call_ctx.ctx || irq < 0)
-- 
2.27.0