On 2022/3/2 18:19, Vitaly Wool wrote: > On Wed, Mar 2, 2022 at 10:12 AM David Laight <David.Laight@xxxxxxxxxx> wrote: >> >>>> Atomic operations aren't magic. >>>> Atomic operations are (at best) one slow locked bus cycle. >>>> Acquiring a lock is the same. >>>> Releasing a lock might be cheaper, but is probably a locked bus cycle. >>>> >>>> So if you use state_lock to protect pages_nr then you lose an atomic >>>> operation for the decrement and gain one (for the unlock) in the increment. >>>> That is even or maybe a slight gain. >>>> OTOH a 64bit atomic is a PITA on some 32bit systems. >>>> (In fact any atomic is a PITA on sparc32.) >>> >>> It's actually *stale_lock* and it's very misleading to use it for this. >>> I would actually like to keep atomics but I have no problem with >>> making it 32-bit for 32-bit systems. Would that work for you guys? >> >> It would be better to rename the lock. > > No it would not because that lock is protecting the list of entries > that could not be immediately freed. > Or could we use pool->lock to do this ? > ~Vitaly Vitaly, is the patch itself worth a Reviewed-by tag and go to the mm-tree ? Could this enhance discussed here be sent as another separate patch or am I supposed to make this change into the current patch? Many thanks for comment. > . >
