From: Paul E. McKenney > Sent: 03 August 2023 14:54 .... > > > If both are READ_ONCE(), you should not need the barrier(). Unless there > > > is some other code not shown in your example that requires it, that is. > > > > And unless the compiler has a bug. :) > > > > So, the barrier() in hlist_nulls_for_each_entry_rcu() is a workaround for a compiler bug. > > Fair enough!!! ;-) Except that it is likely that the compiler bug is avoided by the implementation of READ_ONCE() rather than ACCESS_ONCE(). Also the code that looped forever (UDP receive socket lookup) no longer has the retry - which is a different bug. If a socket rehash hits the lookup then an erroneous ICMP 'port unreachable' is sent rather than doing a rescan. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
