From mboxrd@z Thu Jan 1 00:00:00 1970 From: Denys Fedoryshchenko Subject: Re: 4.19.4 nf_conntrack_count kernel panic Date: Mon, 26 Nov 2018 22:14:48 +0200 Message-ID: <2088ac5f7c2aaba6776db0f1f8528edf@nuclearcat.com> References: <20181126194638.tpwagr7gqzvi3ogf@m.mifar.in> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII; format=flowed Content-Transfer-Encoding: 7bit Cc: netdev-owner@vger.kernel.org To: Sami Farin , Linux Networking Mailing List , netfilter-devel@vger.kernel.org Return-path: Received: from nuclearcat.com ([144.76.183.226]:51072 "EHLO nuclearcat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726985AbeK0H3d (ORCPT ); Tue, 27 Nov 2018 02:29:33 -0500 In-Reply-To: <20181126194638.tpwagr7gqzvi3ogf@m.mifar.in> Sender: netdev-owner@vger.kernel.org List-ID: On 2018-11-26 21:46, Sami Farin wrote: > 4.18.20 works OK, but unfortunately 4.18 series is EOL. > I have Ryzen 1600X, 32 GB RAM, Fedora 28, gcc-8.2.1-5, nosmt=force, > igb module for Intel I211, > using XFS filesystems only. > > To reproduce, I only do this: connect to VPN using a tunnel (e.g. > tun0), > start downloading a file with qbittorrent (allow port for incoming > TCP connections in qbittorrent and iptables) and wait a couple of > minutes. > I am also using ipset and connlimit modules. > I reproduced this bug three times. > With 4.18 I use fq+htb and with 4.19 I use CAKE for traffic control. > > Only this message in kernel log: > [ 363.935074] TCP: request_sock_TCP: Possible SYN flooding on port > 19044. Dropping request. Check SNMP counters. > I get this message with both 4.18.20 and 4.19.4. > > RIP: 0010:rb_insert_color+0x64 > Call Trace: > nf_conntrack_count [nf_conncount] > ip_set_test [ip_set] > connlimit_mt [xt_connlimit] > set_match_v4 [xt_set] > ipt_do_table [ip_tables] > ip_route_input_noref > nf_hook_slow > ip_local_deliver > inet_add_protocol > ip_rcv > ip_rcv_finish_core > __netif_receive_skb_one_core > netif_receive_skb_internal > tun_rx_batched > tun_get_user > __local_bh_enable_ip > tun_get_user > tun_chr_write_iter > __vfs_write > vfs_write > ksys_write > do_syscall_64 > trace_hardirqs_off_thunk > entry_SYSCALL_64_after_hwframe > > ... > > Kernel panic - not syncing: Fatal exception in interrupt Check this patches: https://patchwork.ozlabs.org/project/netfilter-devel/list/?series=73972&state=* Relevant discussion: https://marc.info/?l=linux-netdev&m=154211826106430&w=2