On Wed, May 20, 2015 at 9:25 AM, Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote: > On Wed, May 20, 2015 at 09:03:37AM -0700, Andy Lutomirski wrote: >> On Wed, May 20, 2015 at 7:48 AM, Ingo Molnar <mingo@xxxxxxxxxx> wrote: >> > Yeah, so many of these seem to be 'leaf only' functions: functions >> > that don't ever call functions themselves. >> > >> > So lets assume we always have CONFIG_FRAME_POINTERS=y. >> > >> > If they don't set up a frame pointer then they in essence won't show >> > up in the call chain - but normally they wouldn't because they call >> > nothing. >> > >> > If they trigger an exception/fault or if they get hit by an interrupt >> > then I think we'll still correctly walk the stack - just those >> > functions might be missing from the deterministic call chain, right? >> > (it will still show up as a '?' entry.) >> >> I've never quite understood what the '?' means. > > It basically means "here's a function address we found on the stack, > which may or may not have been called." It's needed because stack > walking isn't currently 100% reliable. > >> > If they crash then we'll see them because the crashing RIP will be >> > printed. >> > >> > So I'm wondering what the x86 policy here should be: to create frame >> > pointers in them or not. Cc:-ed a few more gents for thoughts. >> > >> >> I think it would be nice to have full DWARF unwind support for >> everything at some point. Unfortunately, I don't see any easy path to >> getting there. It doesn't help that AFAIK no one has ever proposed a >> usable in-kernel DWARF unwinder. >> >> It also doesn't help that writing correct CFI annotations in inline >> asm can be very complicated. >> >> I think that ia64 manages to have complete unwind support. How did >> they manage it? >> >> If we had an unwinder, it would be relatively straightforward to write >> something perf-based that would frequently check that we can unwind >> all the way out of an NMI back to userspace and warn if we couldn't. > > I agree that DWARF unwind support would be nice. I have some plans > about how to achieve that in future patch sets. It includes several > pieces: > > - compile-time DWARF data validation (using some similar approaches to > this patch set) > > - run time DWARF data validation, including: > - a DWARF unwinder which doesn't blindly trust any of the DWARF data Fantastic! > - ensuring DWARF and frame pointer data are consistent with each other > - ensuring it can walk all the way to the bottom of the stack > - a DEBUG option which validates the stack periodically from an NMI > and/or schedule() We think alike :) NMI will be much more interesting than schedule. --Andy -- To unsubscribe from this list: send the line "unsubscribe live-patching" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
