From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alexander Fyodorov Subject: Re: [PATCH rt] Fix races in ptrace Date: Tue, 13 Aug 2013 01:13:56 +0400 Message-ID: <393831376342036@web15d.yandex.ru> References: <102271374683007@web8f.yandex.ru> <20130812164145.GJ23040@linutronix.de> Mime-Version: 1.0 Content-Type: text/plain; charset=koi8-r Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: "linux-rt-users@vger.kernel.org" To: Sebastian Andrzej Siewior Return-path: Received: from forward4.mail.yandex.net ([77.88.46.9]:35368 "EHLO forward4.mail.yandex.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753010Ab3HLVTT (ORCPT ); Mon, 12 Aug 2013 17:19:19 -0400 In-Reply-To: <20130812164145.GJ23040@linutronix.de> Sender: linux-rt-users-owner@vger.kernel.org List-ID: 12.08.2013, 20:41, "Sebastian Andrzej Siewior" : > * Alexander Fyodorov | 2013-07-24 20:23:27 [+0400]: > Hi Alexander, Hi Sebastian, >> I ran into a race on 2.6.33-rt kernel which I think still exists in = the latest RT patches. The following patch fixes it but it may be incom= plete: probably spinlock should be taken in all other places that check= saved_state too. =9AI only compile-tested this patch for 3.6-rt. > > It looks interresting, thanks for the report. Do you have maybe a > test-case which can provoke the bug? I can't publish the test but essentially it was doing the following: on= e process (application) has 20 threads sending signals to each other (c= reating contention for tasklist_lock and profiling events), and another= one (profiler) attaches to the application and then does: for (;;) { pid =3D wait(); ptrace(PTRACE_CONT, pid); } Is there a need to write a test? -- To unsubscribe from this list: send the line "unsubscribe linux-rt-user= s" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html