> On Apr 8, 2016, at 17:37 , Segher Boessenkool <seg...@kernel.crashing.org> > wrote: > > On Fri, Apr 08, 2016 at 10:24:50AM +0200, Olivier Hainque wrote: >>> But I expect for stage4, the best solution is to strengthen the stack_tie >>> pattern to block all memory. Early scheduling of the stack frame >>> deallocation (a simple logic insn) can't really be that important to >>> performance. >> >> My feeling as well. At least, it can't be important enough to warrant >> a sustained exposure to the kind of bug we're discussing here. > > Is it a regression? Changing this in stage 4, and this late in stage 4, > is super invasive. Wrt performance, well, I'd like to see numbers :-/
Sorry, my comment was ambiguous: "My feeling as well" etc was in reaction to Richard's second sentence. I didn't mean to comment on what we want to do or not for stage4. Regarding perfs, I agree that having numbers would be good. Nevertheless, I'd rather see some perf drop than just hoping for this not to show up, in any case, and I remain convinced that whatever we gain seems unlikely to be worth the risk of hitting this bug. Olivier