Date:	Tue, 22 Mar 2011 09:17:07 +0900
From:	Simon Horman <horms@...ge.net.au>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	Ingo Molnar <mingo@...e.hu>, David Miller <davem@...emloft.net>,
	torvalds@...ux-foundation.org, akpm@...ux-foundation.org,
	netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	Thomas Gleixner <tglx@...utronix.de>,
	Arnd Bergmann <arnd@...db.de>,
	Pekka Enberg <penberg@...helsinki.fi>,
	Julian Anastasov <ja@....bg>,
	Hans Schillstrom <hans@...illstrom.com>
Subject: Re: [slab poison overwritten] Re: [GIT] Networking

On Tue, Mar 22, 2011 at 09:01:33AM +0900, Simon Horman wrote:
> On Tue, Mar 22, 2011 at 08:29:21AM +0900, Simon Horman wrote:
> > On Tue, Mar 22, 2011 at 07:13:58AM +0900, Simon Horman wrote:
> > > On Mon, Mar 21, 2011 at 09:15:40PM +0100, Eric Dumazet wrote:
> > > > Le lundi 21 mars 2011 à 19:07 +0100, Eric Dumazet a écrit :
> > > > > Le lundi 21 mars 2011 à 18:39 +0100, Ingo Molnar a écrit :
> > > > > > here's the same but with kallsyms enabled.
> > > > > > 
> > > > > > Thanks,
> > > > > > 
> > > > > > 	Ingo
> > > > > > 
> > > > > > [    9.585627] initcall 0xffffffff81d5b806 returned 0 after 0 usecs
> > > > > > [    9.588960] calling  0xffffffff81d5b9da @ 1
> > > > > > [    9.592303] IPVS: Creating netns size=1272 id=0
> > > > > > [    9.595646] IPVS: __ip_vs_control_init(): alloc_percpu.
> > > > > > [    9.602298] IPVS: cannot register namespace.
> > > > > > [    9.605627] IPVS: can't setup control
> > > > > 
> > > > > It seems IPVS is busted in case of memory allocation error in 
> > > > > __ip_vs_control_init()
> > > > > 
> > > > > IPVS deinits its "struct netns_ipvs" space, but something (in IPVS) uses
> > > > > it after free.
> > > > > 
> > > > > __ip_vs_init() seems to be called before ip_vs_init() completes
> > > > > correctly. We then keep in net->ipvs a pointer to some freed memory.
> > > > > 
> > > > > Commit 14e405461e664b7 did some changes in this area
> > > > > 
> > > > > Simon, any idea ?
> > > > > 
> > > > > 
> > > > 
> > > > For the time being, we can avoid the false memory allocation error (and
> > > > leak)
> > > 
> > > Sorry, that typo is my work.
> > 
> > With your patch applied I now see the following
> > 
> > ffff880003bbf1a0 corresponds to &ipvs->app_key in __ip_vs_app_init().
> > I'll continue looking into this.
> > 
> > [   12.610000] IPVS: Creating netns size=2456 id=0
> > [   12.630000] IPVS: Registered protocols (TCP, UDP, SCTP, AH, ESP)
> > [   12.640000] BUG: key ffff880003bbf1a0 not in .data!
> > [   12.640000] ------------[ cut here ]------------
> > [   12.640000] WARNING: at kernel/lockdep.c:2701
> > lockdep_init_map+0x37b/0x570()
> > [   12.640000] Hardware name: Bochs
> > [   12.640000] Pid: 1, comm: swapper Tainted: G        W
> > 2.6.38-kexec-06330-g69b7efe-dirty #122
> > [   12.650000] Call Trace:
> > [   12.650000]  [<ffffffff8102e685>] warn_slowpath_common+0x75/0xb0
> > [   12.650000]  [<ffffffff8102e6d5>] warn_slowpath_null+0x15/0x20
> > [   12.650000]  [<ffffffff8105967b>] lockdep_init_map+0x37b/0x570
> > [   12.650000]  [<ffffffff8105829d>] ? trace_hardirqs_on+0xd/0x10
> > [   12.650000]  [<ffffffff81055ad8>] debug_mutex_init+0x38/0x50
> > [   12.650000]  [<ffffffff8104bc4c>] __mutex_init+0x5c/0x70
> > [   12.650000]  [<ffffffff81685ee7>] __ip_vs_app_init+0x64/0x86
> > [   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
> > [   12.660000]  [<ffffffff811b1c33>] T.620+0x43/0x170
> > [   12.660000]  [<ffffffff811b1e9a>] ? register_pernet_subsys+0x1a/0x40
> > [   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
> > [   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
> > [   12.660000]  [<ffffffff811b1db7>] register_pernet_operations+0x57/0xb0
> > [   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
> > [   12.670000]  [<ffffffff811b1ea9>] register_pernet_subsys+0x29/0x40
> > [   12.670000]  [<ffffffff81685f19>] ip_vs_app_init+0x10/0x12
> > [   12.670000]  [<ffffffff81685a87>] ip_vs_init+0x4c/0xff
> > [   12.670000]  [<ffffffff8166562c>] do_one_initcall+0x7a/0x12e
> > [   12.670000]  [<ffffffff8166583e>] kernel_init+0x13e/0x1c2
> > [   12.670000]  [<ffffffff8128c134>] kernel_thread_helper+0x4/0x10
> > [   12.670000]  [<ffffffff8128ad40>] ? restore_args+0x0/0x30
> > [   12.680000]  [<ffffffff81665700>] ? kernel_init+0x0/0x1c2
> > [   12.680000]  [<ffffffff8128c130>] ? kernel_thread_helper+0x0/0x10
> > [   12.680000] ---[ end trace 4eaa2a86a8e2da23 ]---
> 
> It seems that the problem above was introduced by
> ab8a5e8408c3 ("IPVS: netns awareness to ip_vs_app").
> I assume the hungs are the cause:

s/hungs/hunks below/

I am a little unsure of what to do about this.

The problem seems to be that ipvs->app_key is not in static storage.
But I'm not sure how to resolve that given that the struct netns_ipvs is
per-network namespace. So I guess that the locking needs to be re-worked.
Again, I'm a little unsure of what the best way forward is.

> diff --git a/net/netfilter/ipvs/ip_vs_app.c b/net/netfilter/ipvs/ip_vs_app.c
> index 40b09cc..286f465 100644
> --- a/net/netfilter/ipvs/ip_vs_app.c
> +++ b/net/netfilter/ipvs/ip_vs_app.c
> @@ -43,11 +43,6 @@ EXPORT_SYMBOL(register_ip_vs_app);
>  EXPORT_SYMBOL(unregister_ip_vs_app);
>  EXPORT_SYMBOL(register_ip_vs_app_inc);
>  
> -/* ipvs application list head */
> -static LIST_HEAD(ip_vs_app_list);
> -static DEFINE_MUTEX(__ip_vs_app_mutex);
> -
> -
>  /*
>   *	Get an ip_vs_app object
>   */
> 
> @@ -571,9 +580,13 @@ static const struct file_operations ip_vs_app_fops = {
>  
>  static int __net_init __ip_vs_app_init(struct net *net)
>  {
> +	struct netns_ipvs *ipvs = net_ipvs(net);
> +
>  	if (!net_eq(net, &init_net))	/* netns not enabled yet */
>  		return -EPERM;
>  
> +	INIT_LIST_HEAD(&ipvs->app_list);
> +	__mutex_init(&ipvs->app_mutex, "ipvs->app_mutex", &ipvs->app_key);
>  	proc_net_fops_create(net, "ip_vs_app", 0, &ip_vs_app_fops);
>  	return 0;
>  }
--
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

Powered by blists - more mailing lists