Messages in this thread | | | Date | Wed, 16 Jul 2008 00:01:42 +0300 | From | "Pekka Enberg" <> | Subject | Re: [git pull] RCU updates for v2.6.27 |
| |
Hi Linus,
On Tue, 15 Jul 2008 10:38:27 -0700 (PDT) Linus Torvalds <torvalds@linux-foundation.org> wrote: >>> Hmm. I really wanted to wait with kmemcheck, which is huge and I'm not >>> entirely convinced yet.
On Tue, 15 Jul 2008, Andrew Morton wrote: >> Any particular concerns?
On Tue, Jul 15, 2008 at 11:29 PM, Linus Torvalds <torvalds@linux-foundation.org> wrote: > It's big, complex and looks fragile. And it reports lots of false > positives, with big warnings to people to generally not use or make > reports about itit unless they are already experts. So it looks like > something pretty specific. > > I also want to feel like it actually has been helpful.
Besides catching reads to uninitialized memory, kmemcheck detects use-after-free now and can be extended to catch red-zone overwrites *as they happen* which will make debugging kernel memory corruptions much easier for everybody.
There are now three commits in your tree that fix a real bug reported by kmemcheck: 8410565f540db87ca938f56f92780d251e4f157d ("ACPICA: Fix for access to deleted object <regression>"), adeed48090fc370afa0db8d007748ee72a40b578 ("rc80211_pid: Fix fast_start parameter handling"), and 62f75532b583c03840f31e40386ce2df73be9ca0 ("slub: Initialize per-cpu stats").
So I think it has been helpful and that the infrastructure it adds is valuable for future debugging aids.
Pekka
| |