[Devel] [PATCH] ve/net/ip6tables: fix NULL ptr dereference at ip6t_unregister_table

[Kirill Tkhai]

On 28.10.2016 22:44, Dmitry Safonov wrote:
> We do not init raw table when IPTABLES6 is forbidden in VE.
> So we shouldn't unregister/free it.
> 
> The following actions:
> [root at s186 ~]# vzctl create 100
> [root at s186 ~]# prlctl set 100 --netfilter disabled
> [root at s186 ~]# prlctl start 100
> Starting the CT...
> The CT has been successfully started.
> [root at s186 ~]# prlctl stop 100
> Stopping the CT...
> 
> May lead to memory corruption/NULL pointer dereference:
> [13990.985979] BUG: unable to handle kernel NULL pointer dereference at 0000000000000020
> [13990.995425] Workqueue: netns cleanup_net
> [13990.996030] task: ffff880423302610 ti: ffff880422dd0000 task.ti: ffff880422dd0000
> [13990.996637] RIP: 0010:[<ffffffffa051f178>]  [<ffffffffa051f178>] ip6t_unregister_table+0x18/0x80 [ip6_tables]
> [13991.006983] Call Trace:
> [13991.007581]  [<ffffffffa05d6055>] ip6table_raw_net_exit+0x15/0x20 [ip6table_raw]
> [13991.008187]  [<ffffffff81527dc9>] ops_exit_list.isra.5+0x39/0x60
> [13991.008788]  [<ffffffff81528f50>] cleanup_net+0x260/0x480
> [13991.009380]  [<ffffffff8109ef4b>] process_one_work+0x17b/0x470
> [13991.009980]  [<ffffffff8109fd1b>] worker_thread+0x11b/0x400
> [13991.010576]  [<ffffffff8109fc00>] ? rescuer_thread+0x400/0x400
> [13991.011150]  [<ffffffff810a779f>] kthread+0xcf/0xe0
> [13991.011717]  [<ffffffff810a76d0>] ? create_kthread+0x60/0x60
> [13991.012619]  [<ffffffff81643cd8>] ret_from_fork+0x58/0x90
> [13991.013153]  [<ffffffff810a76d0>] ? create_kthread+0x60/0x60
> [13991.013668] Code: c4 28 5b 41 5c 41 5d 5d c3 e8 d5 c2 b5 e0 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 57 41 56 49 89 fe 48 89 f7 41 55 41 54 53 <4c> 8b 7e 20 e8 7f 5f 04 e1 4c 8d 68 58 49 89 c4 8b 00 4c 89 eb
> [13991.014765] RIP  [<ffffffffa051f178>] ip6t_unregister_table+0x18/0x80 [ip6_tables]
> 
> Fixes: commit bff5233618d8 ("ve/net/ip6tables: fix autoloading of the
> ip6table_raw module from CT")
> 
> https://jira.sw.ru/browse/PSBM-54244
> 
> Cc: Andrey Ryabinin <aryabinin at virtuozzo.com>
> Cc: Kirill Tkhai <ktkhai at virtuozzo.com>
> Signed-off-by: Dmitry Safonov <dsafonov at virtuozzo.com>
> ---
>  net/ipv6/netfilter/ip6table_raw.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/net/ipv6/netfilter/ip6table_raw.c b/net/ipv6/netfilter/ip6table_raw.c
> index 271835df64ce..f7ad588586e8 100644
> --- a/net/ipv6/netfilter/ip6table_raw.c
> +++ b/net/ipv6/netfilter/ip6table_raw.c
> @@ -48,7 +48,8 @@ static int __net_init ip6table_raw_net_init(struct net *net)
>  
>  static void __net_exit ip6table_raw_net_exit(struct net *net)
>  {
> -	ip6t_unregister_table(net, net->ipv6.ip6table_raw);
> +	if (net_ipt_permitted(net, VE_IP_IPTABLES6))
> +		ip6t_unregister_table(net, net->ipv6.ip6table_raw);

As I remember, we was going to make ipt_mask be online configurable one day.
If so, this won't be safe...

Could you change you patch in the way we did for ipv4?

>  }
>  
>  static struct pernet_operations ip6table_raw_net_ops = {
>