netdev - Re: Performance regression on kernels 3.10 and newer

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:	Sat, 06 Sep 2014 11:21:52 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	David Miller <davem@...emloft.net>,
	Nicolas Dichtel <nicolas.dichtel@...nd.com>
Cc:	therbert@...gle.com, alexander.h.duyck@...el.com,
	netdev@...r.kernel.org
Subject: Re: Performance regression on kernels 3.10 and newer

On Sat, 2014-09-06 at 09:38 -0700, Eric Dumazet wrote:
> On Sat, 2014-09-06 at 08:46 -0700, Eric Dumazet wrote:
> 
> > Works fine on IPv4, not on IPv6, I will submit a v2 (with a proper title
> > btw)
> 
> Oh well, we first need to fix IPv6 early demux.
> 
> Current net-next kernel profile
> 
> +  20.95%      netserver  [kernel.kallsyms]    [k] dst_release     
> +  19.33%      netserver  [kernel.kallsyms]    [k] ip6_pol_route.isra.46
> +  11.75%      netserver  [kernel.kallsyms]    [k] inet6_sk_rx_dst_set
> +   3.72%      netserver  [kernel.kallsyms]    [k] ip6_input_finish
> 
> So we repeat over and over calls to inet6_sk_rx_dst_set(),
> something is surely wrong.
> 

Nicolas, it seems that after your commit
(6f3118b571b8a4c06c7985dc3172c3526cb86253 "ipv6: use net->rt_genid to
check dst validity")  IP early demux is broken.


Basically, we now hit the following condition in ip6_dst_check()

if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev))) 

In my traces, rt->rt6i_genid stays at 2, and the rt_genid_ipv6()
increments every time a route is added/deleted.

So we keep calling dst_release(), and inet6_sk_rx_dst_set(),
but the route we keep installing is already wrong, as its genid is too
old.

Does following fix make any sense ?

diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index f74b0417bd60..a139a16298e3 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -947,9 +947,10 @@ restart:
 		nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 	else if (!(rt->dst.flags & DST_HOST))
 		nrt = rt6_alloc_clone(rt, &fl6->daddr);
-	else
+	else {
+		rt->rt6i_genid = rt_genid_ipv6(net);
 		goto out2;
-
+	}
 	ip6_rt_put(rt);
 	rt = nrt ? : net->ipv6.ip6_null_entry;
 


--
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