On 13.04.2023 21:22, Andrew Cooper wrote: > Despite rcu_read_unlock() being fully inlineable, the optimiser cannot fold > these exit paths, because of the various compiler barriers providing RCU > safety. Help the compiler out.
Mind me asking about "cannot"? Iirc the compiler is permitted to fold even volatile asm()s, as long as certain guarantees aren't violated. With "... often won't ..." or alike I'd be happy to give my ack, albeit doing so with another maintainer's questions pending (and imo wrongly taken by you as possibly insulting) it's never really clear to me whether doing so is appropriate. I would specifically argue that ... > This compiles to marginally better code in all cases. No functional change. ... even minor improvements are worth it. > Signed-off-by: Andrew Cooper <[email protected]> Preferably (unless proven wrong) with said adjustment and on the basis that Julien signals his question as addressed (once it really is, of course): Acked-by: Jan Beulich <[email protected]> Jan
