Re: [syzbot] WARNING in destroy_conntrack

[Pavel Skripkin]

On 8/9/21 6:38 PM, syzbot wrote:
> Hello,
>
> syzbot found the following issue on:
>
> HEAD commit:    f9be84db09d2 net: bonding: bond_alb: Remove the dependency..
> git tree:       net-next
> console output: https://syzkaller.appspot.com/x/log.txt?x=12755626300000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=8075b2614f3db143
> dashboard link: https://syzkaller.appspot.com/bug?extid=a1eb62c681423ee5c0d7
> compiler:       gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.1
> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=1241d9d6300000
> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=168e66e9300000
>
> The issue was bisected to:
>
> commit 65038428b2c6c5be79d3f78a6b79c0cdc3a58a41
> Author: Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx>
> Date:   Tue Mar 17 13:13:46 2020 +0000
>
>      netfilter: nf_tables: allow to specify stateful expression in set definition
>
> bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=1683f1f1300000
> final oops:     https://syzkaller.appspot.com/x/report.txt?x=1583f1f1300000
> console output: https://syzkaller.appspot.com/x/log.txt?x=1183f1f1300000
>
> IMPORTANT: if you fix the issue, please add the following tag to the commit:
> Reported-by: syzbot+a1eb62c681423ee5c0d7@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: 65038428b2c6 ("netfilter: nf_tables: allow to specify stateful expression in set definition")
>
> ------------[ cut here ]------------
> WARNING: CPU: 1 PID: 8989 at net/netfilter/nf_conntrack_core.c:610 destroy_conntrack+0x232/0x2c0 net/netfilter/nf_conntrack_core.c:610
> Modules linked in:
> CPU: 0 PID: 8989 Comm: syz-executor188 Not tainted 5.14.0-rc4-syzkaller #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> RIP: 0010:destroy_conntrack+0x232/0x2c0 net/netfilter/nf_conntrack_core.c:610
> Code: da fc ff eb 90 e8 ae 27 19 fa 48 89 ef e8 c6 53 02 00 48 89 ef e8 ee 1c 5f fa 5b 5d 41 5c 41 5d e9 93 27 19 fa e8 8e 27 19 fa <0f> 0b e9 2f fe ff ff e8 82 27 19 fa 4c 8d a5 e8 00 00 00 48 b8 00
> RSP: 0018:ffffc90002d7f080 EFLAGS: 00010293
> RAX: 0000000000000000 RBX: 00000000ffffffff RCX: 0000000000000000
> RDX: ffff88802eeb8000 RSI: ffffffff875c8632 RDI: 0000000000000003
> RBP: ffff888147d35400 R08: 0000000000000000 R09: ffff888147d35403
> R10: ffffffff875c8460 R11: 0000000000000000 R12: ffff888147d35400
> R13: ffffffff8b31b880 R14: 0000000000000000 R15: 0000000000000001
> FS:  00007fb18d49a700(0000) GS:ffff8880b9c00000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00007f37c403d088 CR3: 00000000182c1000 CR4: 00000000001506f0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> Call Trace:
>   nf_conntrack_destroy+0xab/0x230 net/netfilter/core.c:677
>   nf_conntrack_put include/linux/netfilter/nf_conntrack_common.h:34 [inline]
>   nf_ct_put include/net/netfilter/nf_conntrack.h:176 [inline]
>   nft_ct_tmpl_put_pcpu+0x15e/0x1e0 net/netfilter/nft_ct.c:356


Hm... Calltrace looks similar to 
https://syzkaller.appspot.com/bug?id=34b3d29c783f19d70086206194da85e59c448167 
and my debug logs say that right before this check

	WARN_ON(atomic_read(&nfct->use) != 0);

nfct->use is zero. Reproducer is threaded, so, I think, 2 netns 
concurrently doing inc/dec somewhere. This should be fixed by my 
previous netfilter patch.


Also, with my previous netfilter patch applied I cannot reproduce this 
bug locally, so

#syz fix: netfilter: nft_ct: protect nft_ct_pcpu_template_refcnt with mutex


With regards,
Pavel Skripkin