On Saturday 19 January 2013 09:01 AM, Al Viro wrote: > On Fri, Jan 18, 2013 at 05:54:27PM +0530, Vineet Gupta wrote: > >> + ; --- (Slow Path #3) notify_resume --- >> +.Lchk_notify_resume: >> + btst r9, TIF_NOTIFY_RESUME >> + blnz @do_notify_resume >> + b resume_user_mode_begin ; unconditionally back to U mode ret >> chks >> + ; for single exit point from this block > Umm... Can we even get there without NOTIFY_RESUME? Again, there's > future-proofing and there's laying minefields - think what will happen > if we *do* get there with some bit in _TIF_WORK_MASK that isn't recognized > by any of these cases. Looping forever?
IMHO, for future safe-ing, the test for TIF_NOTIFY_RESUME is correct (as we will need to add that check the moment a new bit is introduced in _TIF_WORK_MASK). Regarding the infinite loop, I would assume that _TIF_WORK_MASK is golden (fixed by your prior comment) so anyone touching it needs to add corresponding code here - IMHO we don't need to handle that scenario (maybe add a comment in thread_info.h). With that assumption, the unconditional branch would go back to start and the re-test for TIF_WORK_MASK will break the loop even if any stray bit was set. So essentially we don't need any code change ! Am I overlooking something here ? -Vineet -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
