From mboxrd@z Thu Jan 1 00:00:00 1970 From: Colin Ian King Subject: rhashtable concurrent writers issue, 4.2+ Date: Fri, 11 Dec 2015 16:09:04 +0000 Message-ID: <566AF520.4000904@canonical.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Cc: Thomas Graf , "linux-kernel@vger.kernel.org" To: netdev@vger.kernel.org Return-path: Received: from youngberry.canonical.com ([91.189.89.112]:45836 "EHLO youngberry.canonical.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751134AbbLKQJH (ORCPT ); Fri, 11 Dec 2015 11:09:07 -0500 Sender: netdev-owner@vger.kernel.org List-ID: Hi, After testing with commit 7def0f952eccdd0edb3c504f4dab35ee0d3aba1f ("lib: fix data race in rhashtable_rehash_one") on the current 4.4-rc4 kernel, I believe I am still seeing race conditions that seem to bite concurrent readers. The reproducer is found in my stress-ng system stress tool: git://kernel.ubuntu.com/cking/stress-ng.git one needs libattr1-dev, libkeyutils-dev to build this. run the concurrent procfs stressor on 2 CPUs: ./stress-ng --procfs 2 ..sometimes this hangs w/o a kernel stack dump, but sometimes it oopses and one can see that we get a concurrent reader failure. Colin