From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 3DB81171A9 for ; Mon, 22 May 2023 19:37:18 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id B1E47C433EF; Mon, 22 May 2023 19:37:17 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1684784238; bh=l1ub/ZX4cJgsjKIFpwKGykVmCi/zKer53Sy9LcDGVwI=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=QoU85IQVeXzsTcm3SfmKrE+A+UVCAFtZt0cCSByJgy7SHwEZHxmlOP7Ja6ZUwl35B dzOXQPmH25cMXR2p5HsaXu7POjWVmyTDSmZCuENFKzksqJY1yjyJsFnmEdt/ob/UPL F9QUlUT13q8xV/q1S+s4Q7qFEEVn+OSdahweZIGE= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Florian Westphal , Pablo Neira Ayuso , Sasha Levin Subject: [PATCH 6.3 016/364] netfilter: nf_tables: always release netdev hooks from notifier Date: Mon, 22 May 2023 20:05:21 +0100 Message-Id: <20230522190413.232216152@linuxfoundation.org> X-Mailer: git-send-email 2.40.1 In-Reply-To: <20230522190412.801391872@linuxfoundation.org> References: <20230522190412.801391872@linuxfoundation.org> User-Agent: quilt/0.67 Precedence: bulk X-Mailing-List: patches@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From: Florian Westphal [ Upstream commit dc1c9fd4a8bbe1e06add9053010b652449bfe411 ] This reverts "netfilter: nf_tables: skip netdev events generated on netns removal". The problem is that when a veth device is released, the veth release callback will also queue the peer netns device for removal. Its possible that the peer netns is also slated for removal. In this case, the device memory is already released before the pre_exit hook of the peer netns runs: BUG: KASAN: slab-use-after-free in nf_hook_entry_head+0x1b8/0x1d0 Read of size 8 at addr ffff88812c0124f0 by task kworker/u8:1/45 Workqueue: netns cleanup_net Call Trace: nf_hook_entry_head+0x1b8/0x1d0 __nf_unregister_net_hook+0x76/0x510 nft_netdev_unregister_hooks+0xa0/0x220 __nft_release_hook+0x184/0x490 nf_tables_pre_exit_net+0x12f/0x1b0 .. Order is: 1. First netns is released, veth_dellink() queues peer netns device for removal 2. peer netns is queued for removal 3. peer netns device is released, unreg event is triggered 4. unreg event is ignored because netns is going down 5. pre_exit hook calls nft_netdev_unregister_hooks but device memory might be free'd already. Fixes: 68a3765c659f ("netfilter: nf_tables: skip netdev events generated on netns removal") Signed-off-by: Florian Westphal Signed-off-by: Pablo Neira Ayuso Signed-off-by: Sasha Levin --- net/netfilter/nft_chain_filter.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/net/netfilter/nft_chain_filter.c b/net/netfilter/nft_chain_filter.c index c3563f0be2692..680fe557686e4 100644 --- a/net/netfilter/nft_chain_filter.c +++ b/net/netfilter/nft_chain_filter.c @@ -344,6 +344,12 @@ static void nft_netdev_event(unsigned long event, struct net_device *dev, return; } + /* UNREGISTER events are also happening on netns exit. + * + * Although nf_tables core releases all tables/chains, only this event + * handler provides guarantee that hook->ops.dev is still accessible, + * so we cannot skip exiting net namespaces. + */ __nft_release_basechain(ctx); } @@ -362,9 +368,6 @@ static int nf_tables_netdev_event(struct notifier_block *this, event != NETDEV_CHANGENAME) return NOTIFY_DONE; - if (!check_net(ctx.net)) - return NOTIFY_DONE; - nft_net = nft_pernet(ctx.net); mutex_lock(&nft_net->commit_mutex); list_for_each_entry(table, &nft_net->tables, list) { -- 2.39.2