On Mon 14-12-15 12:03:20, Michal Hocko wrote: > JFYI: Andrey Ryabinin has noticed that this might be related to > http://lkml.kernel.org/r/CAPAsAGzrOQAABhOta_o-MzocnikjPtwJLfEKQJ3n5mbBm0T7Bw@xxxxxxxxxxxxxx > > and indeed if somebody with pending signals would do wait_on_page_locked > then it could race AFAIU. No, a simple lock_page would fail to lock the page (thanks Andrey for the clarification in the parallel email thread). I have missed that __lock_page uses bit_wait_io and that would return EINTR when using signal_pending. The fix has changed that to singnal_pending_state so this will not happen anymore. I think the mystery is solved... -- Michal Hocko SUSE Labs -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>