On Fri, Dec 01, 2023 at 08:34:12PM +0800, Edward Adam Davis wrote: > > According to splat, after a page fault occurred, the attempt to retrieve > rcu_read_lock() failed, resulting in a timeout of 143s. This is my speculation. Oh I see what's going on. The reproducer is mapping /dev/hwrng, so the write to user-space is then triggering another read which then dead-locks. Let me think about this. Thanks, -- Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt