Re: BUG-lockdep and freeze (was: Arrr! Linux 2.6.18)



On Sat, 30 Sep 2006, Andi Kleen wrote:

We didn't so far find any bug in the unwinder code itself (ok if you don't
count the performance issue Ingo found) just lots in the annotations and one
bug in the dwarf2 standard.

I don't think it matters if it's a bug in the unwinding code or in the
data generated for it. It's still a bug in the unwinder.

Those bugs have been compiler bugs, manual annotation bugs, and it
doesn't _matter_ what kind of bugs. The end result is the same: the
unwinder is buggy.

If you kick the people who add more than three levels of callback
to core driver code to get their acts together too that's fine
to me. Unfortunately I don't think that's realistic. So we clearly
need better unwinding.

I dispute the "clearly". We didn't have _that_ many problems with just
manually filtering out obvious left-overs from some previous callchain.

I mean, really: Andi, point me to anything that was a real problem when we
had no unwinder at all?

Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/