From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: [IPV6]: Audit all ip6_dst_lookup/ip6_dst_store calls Date: Tue, 1 Aug 2006 19:24:10 +0200 Message-ID: <200608011924.10534.ak@suse.de> References: <20060728194531.GA17744@lists.us.dell.com> <20060731090433.GA25192@gondor.apana.org.au> <20060801074133.b5b96f11.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: Herbert Xu , Jan Beulich , mingo@elte.hu, arjan@infradead.org, Matt_Domsch@dell.com, davem@davemloft.net, yoshfuji@linux-ipv6.org, linux-kernel@vger.kernel.org, netdev@vger.kernel.org Return-path: To: Andrew Morton In-Reply-To: <20060801074133.b5b96f11.akpm@osdl.org> Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Tuesday 01 August 2006 16:41, Andrew Morton wrote: > On Mon, 31 Jul 2006 19:04:33 +1000 > Herbert Xu wrote: > > > 2) There is something broken in the x86_64 unwind code which is causing > > it to panic just about everytime somebody calls dump_stack(). > > > > Andi, this is the second time I've seen a report where an otherwise > > harmless dump_stack call (the other one was caused by a WARN_ON) gets > > turned into a panic by the stack unwind code on x86_64. This particular > > report is with 2.6.18-rc3 so it looks like whatever bug is causing it > > hasn't been fixed yet. > > > > Could you please have a look at it? Thanks. > > Jan thinks this might have been fixed by a patch which he sent Andi a > couple of days ago. Andi has sent that patch to Linus I didn't send that particular patch before, just queued it, because I didn't realize that particular crash, but I have now send it yesterday. So far L. hasn't merged it unfortunately, but I will resend. > but I'm not sure > which patch it was "entry-more-unwind" was my version, there was another one from Jan > and I'm not sure whether it has been merged into > mainline. > > But yes, -rc3 unwind has problems. "unwinder stuck" messages are expected and not really fatal because they don't lose any information. I expect it will need some releases to work them all out fully, but then we'll hopefully have a much better unwinder that doesn't generate any false positives anymore. New crashes during unwinding are fatal though and I plan to fix them. So far this one was the only known one. I already got a lot of patches queued for .19 that fix more unwind information in a lot of assembly files. Still not fully complete though. Fixing it all properly unfortunately requires undoing some stuff, e.g. the unwinder cannot deal with separate lock sections, so I was slowly removing them. -Andi