On 2022-01-28 16:33:44 [+0100], Jason A. Donenfeld wrote: > From: Andy Lutomirski <luto@xxxxxxxxxx> > > We don't need spinlocks to protect batched entropy -- all we need > is a little bit of care. This should fix up the following splat that > Jonathan received with a PROVE_LOCKING=y/PROVE_RAW_LOCK_NESTING=y > kernel: NO. Could we please look at my RANDOM patches first? This affects PREEMPT_RT. There is no need to stuff this in and tag it stable. I can repost my rebased patched if there no objection. This patch invokes extract_crng() with disabled interrupts so we didn't gain anything IMHO. Sebastian
