Re: [PATCH] vhost/crypto: fix locking

2025-03-05 Thread Thomas Monjalon
05/03/2025 10:52, Maxime Coquelin: > > On 3/5/25 10:12 AM, David Marchand wrote: > > vc_req_out->vq->iotlb_lock is taken twice on the same thread: > > vc_req_out->vq refers to vc_req->vq (after a memcpy), which itself > > is a reference to the vq. > > > > clang probably does not detect that the s

[PATCH] vhost/crypto: fix locking

2025-03-05 Thread David Marchand
vc_req_out->vq->iotlb_lock is taken twice on the same thread: vc_req_out->vq refers to vc_req->vq (after a memcpy), which itself is a reference to the vq. clang probably does not detect that the same object is already locked as it does not track object referencies. Finish the incomplete and incor

Re: [PATCH] vhost/crypto: fix locking

2025-03-05 Thread Maxime Coquelin
On 3/5/25 10:12 AM, David Marchand wrote: vc_req_out->vq->iotlb_lock is taken twice on the same thread: vc_req_out->vq refers to vc_req->vq (after a memcpy), which itself is a reference to the vq. clang probably does not detect that the same object is already locked as it does not track objec

RE: [EXTERNAL] [PATCH] vhost/crypto: fix locking

2025-03-05 Thread Gowrishankar Muthukrishnan
> > clang probably does not detect that the same object is already locked as it > does > not track object referencies. > > Finish the incomplete and incorrect cleanup and only refer to the > &vq->iotlb_lock > capability (required by vhost_iova_to_vva). > > Fixes: 88c73b5434e6 ("vhost/crypto: f