From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 72EA9C4167B for ; Fri, 23 Dec 2022 08:04:55 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230060AbiLWIEy (ORCPT ); Fri, 23 Dec 2022 03:04:54 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:47088 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229630AbiLWIEx (ORCPT ); Fri, 23 Dec 2022 03:04:53 -0500 Received: from szxga08-in.huawei.com (szxga08-in.huawei.com [45.249.212.255]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 14F0832BA7; Fri, 23 Dec 2022 00:04:50 -0800 (PST) Received: from dggpemm500024.china.huawei.com (unknown [172.30.72.57]) by szxga08-in.huawei.com (SkyGuard) with ESMTP id 4Ndfnl1fNLz16LkF; Fri, 23 Dec 2022 16:03:39 +0800 (CST) Received: from [10.67.110.173] (10.67.110.173) by dggpemm500024.china.huawei.com (7.185.36.203) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.34; Fri, 23 Dec 2022 16:04:48 +0800 Message-ID: Date: Fri, 23 Dec 2022 16:04:48 +0800 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101 Thunderbird/102.3.2 Subject: Re: [RFC] IMA LSM based rule race condition issue on 4.19 LTS Content-Language: en-US From: "Guozihua (Scott)" To: Mimi Zohar , Paul Moore CC: , , , Greg KH , , , "linux-integrity@vger.kernel.org" , , luhuaxin References: <389334fe-6e12-96b2-6ce9-9f0e8fcb85bf@huawei.com> <6a5bc829-b788-5742-cbfc-dba348065dbe@huawei.com> <566721e9e8d639c82d841edef4d11d30a4d29694.camel@linux.ibm.com> <40cf70a96d2adbff1c0646d3372f131413989854.camel@linux.ibm.com> <757bc525f7d3fe6db5f3ee1f86de2f4d02d8286b.camel@linux.ibm.com> <381efcb7-604f-7f89-e950-efc142350417@huawei.com> <6348a26f165c27c562db48eb39b04417cbe1380c.camel@linux.ibm.com> <944ea86a-2e6b-ce95-a6cb-fcf6b30ad78b@huawei.com> <578081a5-9ddd-b9bd-002d-f4f14bee79a3@huawei.com> In-Reply-To: <578081a5-9ddd-b9bd-002d-f4f14bee79a3@huawei.com> Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit X-Originating-IP: [10.67.110.173] X-ClientProxiedBy: dggems702-chm.china.huawei.com (10.3.19.179) To dggpemm500024.china.huawei.com (7.185.36.203) X-CFilter-Loop: Reflected Precedence: bulk List-ID: X-Mailing-List: stable@vger.kernel.org On 2022/12/21 18:51, Guozihua (Scott) wrote: > On 2022/12/20 9:11, Guozihua (Scott) wrote: >> On 2022/12/19 21:11, Mimi Zohar wrote: >>> On Mon, 2022-12-19 at 15:10 +0800, Guozihua (Scott) wrote: >>>> On 2022/12/16 11:04, Paul Moore wrote: >>>>> On Thu, Dec 15, 2022 at 9:36 PM Guozihua (Scott) wrote: >>>>>> On 2022/12/16 5:04, Paul Moore wrote: >>>>> >>>>> ... >>>>> >>>>>>> How bad is the backport really? Perhaps it is worth doing it to see >>>>>>> what it looks like? >>>>>>> >>>>>> It might not be that bad, I'll try to post a version next Monday. >>>>> >>>>> Thanks for giving it a shot. >>>>> >>>> When I am trying a partial backport of b16942455193 ("ima: use the lsm >>>> policy update notifier"), I took a closer look into it and if we rip off >>>> the RCU and the notifier part, there would be a potential UAF issue when >>>> multiple processes are calling ima_lsm_update_rule() and >>>> ima_match_rules() at the same time. ima_lsm_update_rule() would free the >>>> old rule if the new rule is successfully copied and initialized, leading >>>> to ima_match_rules() accessing a freed rule. >>>> >>>> To reserve the mainline solution, we would have to either introduce RCU >>>> for rule access, which would work better with notifier mechanism or the >>>> same rule would be updated multiple times, or we would have to introduce >>>> a lock for LSM based rule update. >>> >>> Even with the RCU changes, the rules will be updated multiple times. >>> With your "ima: Handle -ESTALE returned by ima_filter_rule_match()" >>> patch, upstream makes a single local copy of the rule to avoid updating >>> it multiple times. Without the notifier, it's updating all the rules. >> That's true. However, in the mainline solution, we are only making a >> local copy of the rule. In 4.19, because of the lazy update mechanism, >> we are replacing the rule on the rule list multiple times and is trying >> to free the original rule. >>> >>> Perhaps an atomic variable to detect if the rules are already being >>> updated would suffice. If the atomic variable is set, make a single >>> local copy of the rule. >> That should do it. I'll send a patch set soon. >> > Including Huaxin Lu in the loop. Sorry for forgotten about it for quite > some time. > > I tried the backported solution, it seems that it's causing RCU stall. > It seems on 4.19.y IMA is already accessing rules through RCU. Still > debugging it. It seems that after the backport, a NULL pointer deference pops out. I'll have to look into it. -- Best GUO Zihua