From mboxrd@z Thu Jan 1 00:00:00 1970 From: John Fastabend Subject: Re: [Patch net-next] net_sched: move the empty tp check from ->destroy() to ->delete() Date: Sun, 27 Nov 2016 18:57:38 -0800 Message-ID: <583B9D22.8090906@gmail.com> References: <1479952708-26763-1-git-send-email-xiyou.wangcong@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Cc: roid@mellanox.com, jiri@mellanox.com, Daniel Borkmann To: Cong Wang , netdev@vger.kernel.org Return-path: Received: from mail-pg0-f42.google.com ([74.125.83.42]:36040 "EHLO mail-pg0-f42.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753944AbcK1C65 (ORCPT ); Sun, 27 Nov 2016 21:58:57 -0500 Received: by mail-pg0-f42.google.com with SMTP id f188so51405183pgc.3 for ; Sun, 27 Nov 2016 18:58:19 -0800 (PST) In-Reply-To: <1479952708-26763-1-git-send-email-xiyou.wangcong@gmail.com> Sender: netdev-owner@vger.kernel.org List-ID: On 16-11-23 05:58 PM, Cong Wang wrote: > Roi reported we could have a race condition where in ->classify() path > we dereference tp->root and meanwhile a parallel ->destroy() makes it > a NULL. > > This is possible because ->destroy() could be called when deleting > a filter to check if we are the last one in tp, this tp is still > linked and visible at that time. > > The root cause of this problem is the semantic of ->destroy(), it > does two things (for non-force case): > > 1) check if tp is empty > 2) if tp is empty we could really destroy it > > and its caller, if cares, needs to check its return value to see if > it is really destroyed. Therefore we can't unlink tp unless we know > it is empty. > > As suggested by Daniel, we could actually move the test logic to ->delete() > so that we can safely unlink tp after ->delete() tells us the last one is > just deleted and before ->destroy(). > > What's more, even we unlink it before ->destroy(), it could still have > readers since we don't wait for a grace period here, we should not modify > tp->root in ->destroy() either. > > Fixes: 1e052be69d04 ("net_sched: destroy proto tp when all filters are gone") > Reported-by: Roi Dayan > Cc: Daniel Borkmann > Cc: John Fastabend > Signed-off-by: Cong Wang > --- Hi Cong, Thanks a lot for doing this. Can you rebase it on top of Daniel's patch though, [PATCH net] net, sched: respect rcu grace period on cls destruction And then push the NULL pointer work for the cls_fw and cls_route classifiers into another patch. Then I believe the last thing to make this correct is to convert the call_rcu() paths to call_rcu_bh(). .John