Re: [Patch bpf-next] unix_bpf: fix a potential deadlock in unix_dgram_bpf_recvmsg()

On Tue, Jul 27, 2021 at 9:12 AM John Fastabend [off-list ref] wrote:

Is there a reason to grab the mutex_lock(u->iolock) above the
skb_queue_emptyaand sk_psock_queue_empty checks?

Could it be move here just above the msg_bytes_ready label?

The check of the receive queue is more accurate with lock.

Thanks.