netdev - Re: [PATCH net] ipv6: prevent fib6_run

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 1 Aug 2013 00:19:31 +0200
From:	Michal Kubecek <mkubecek@...e.cz>
To:	David Miller <davem@...emloft.net>
Cc:	netdev@...r.kernel.org, kuznet@....inr.ac.ru, jmorris@...ei.org,
	yoshfuji@...ux-ipv6.org, kaber@...sh.net,
	Eric Dumazet <eric.dumazet@...il.com>
Subject: Re: [PATCH net] ipv6: prevent fib6_run_gc() contention

On Tue, Jun 11, 2013 at 02:22:29PM -0700, David Miller wrote:
> From: Michal Kubecek <mkubecek@...e.cz>
> Date: Tue, 11 Jun 2013 22:49:12 +0200
> 
> > On Tue, Jun 11, 2013 at 01:01:14PM -0700, David Miller wrote:
> >> From: Michal Kubecek <mkubecek@...e.cz>
> >> Date: Tue, 11 Jun 2013 12:07:18 +0200
> >> 
> >> > On Mon, Jun 10, 2013 at 02:26:42PM -0700, David Miller wrote:
> >> >> 
> >> >> It seems to me that it would be much simpler to simply update
> >> >> ip6_rt_last_gc first, that way the other threads would elide the GC
> >> >> call.
> >> > 
> >> > That was my original idea but I was afraid that while the remaining
> >> > window in ip6_dst_gc() would be very short and probably safe, we could
> >> > still run into problem if fib6_gc_lock was locked by some other caller
> >> > of fib6_run_gc() which doesn't update net->ipv6.ip6_rt_last_gc,
> >> > especially via a timer.
> >> 
> >> Why don't you simply try it and find out?
> > 
> > I don't have the system where the issue was observed at hand. I'll have
> > to ask the customer who reported it to run the test so that it may take
> > some time.
> 
> Please also take care of the issue Eric mentioned, in that some GC
> calls erroneously do not update the last GC timestamp.

I received the feedback today. The customer reported the patch moving
update of ip6_rt_last_gc into fib6_run_gc() before the call of
fib6_clean_all() was not sufficient to prevent the contention with their
workload.

So I suggest to return back to the approach I submitted originally but
to move the update into fib6_run_gc() anyway so that every run of the
garbage collector updates it. I will submit the patches tomorrow.

                                                        Michal Kubecek

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html