On Fri, Mar 22, 2019 at 09:13:51AM +0800, Ming Lei wrote: > Inside sbitmap_queue_clear(), once the clear bit is set, it will be > visiable to allocation path immediately. Meantime READ/WRITE on old > associated instance(such as request in case of blk-mq) may be > out-of-order with the setting clear bit, so race with re-allocation > may be triggered. > > Adds one memory barrier for ordering READ/WRITE of the freed associated > instance with setting clear bit for avoiding race with re-allocation. Just to make sure I understand, the race is something like this? Thread 1 Thread 2 sbitmap_queue_get() ... write to request _____ sbitmap_queue_clear() \ \ sbitmap_queue_get() \______ observes write If so, you can add Reviewed-by: Omar Sandoval <osandov@xxxxxx>
