From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mathieu Desnoyers Subject: Re: linux-next: manual merge of the tip tree with Linus' tree Date: Tue, 6 Feb 2018 17:05:52 +0000 (UTC) Message-ID: <1409139836.17054.1517936752110.JavaMail.zimbra@efficios.com> References: <20180206114048.6c8ced1c@canb.auug.org.au> <750091485.16756.1517921554396.JavaMail.zimbra@efficios.com> <20180206135512.GB22740@arm.com> <1815310787.16844.1517926010063.JavaMail.zimbra@efficios.com> <20180206141118.GC22740@arm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Return-path: Received: from mail.efficios.com ([167.114.142.141]:41240 "EHLO mail.efficios.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753360AbeBFRFH (ORCPT ); Tue, 6 Feb 2018 12:05:07 -0500 In-Reply-To: <20180206141118.GC22740@arm.com> Sender: linux-next-owner@vger.kernel.org List-ID: To: Will Deacon Cc: Stephen Rothwell , Thomas Gleixner , Ingo Molnar , "H. Peter Anvin" , Peter Zijlstra , Linux-Next Mailing List , linux-kernel ----- On Feb 6, 2018, at 9:11 AM, Will Deacon will.deacon@arm.com wrote: > On Tue, Feb 06, 2018 at 02:06:50PM +0000, Mathieu Desnoyers wrote: >> ----- On Feb 6, 2018, at 8:55 AM, Will Deacon will.deacon@arm.com wrote: >> > On Tue, Feb 06, 2018 at 12:52:34PM +0000, Mathieu Desnoyers wrote: >> >> One approach I would consider for this is to duplicate this >> >> comment and add it just above the "eret" instruction within the >> >> macro: >> >> >> >> /* >> >> * ARCH_HAS_MEMBARRIER_SYNC_CORE rely on eret context synchronization >> >> * when returning from IPI handler, and when returning to user-space. >> >> */ >> >> >> >> Or perhaps Will has something else in mind ? >> > >> > To be honest with you, I'd just drop the comment entirely. entry.S is >> > terrifying these days and nobody should have to go in there to understand >> > why we select ARCH_HAS_MEMBARRIER_SYNC_CORE. If you really feel a justification >> > is needed, I'd be happy with a line in the Kconfig file. >> >> My concern is that someone wanting to optimize away a few cycles by changing >> eret to something else in the future will not be looking at Kconfig: that >> person will be staring at entry.S. > > That person will probably also be me, or somebody who sits within punching > distance. I really wouldn't worry about it. There a bunch of other > things that will break if we don't use ERET here and, if it's a real > concern, we're making the *huge* assumption that developers actually > read and pay attention to comments. > >> One alternative presented by PeterZ on irc is to do like ppc: define a >> macro for eret, and stick all appropriate comments near the macro. This >> way, it won't hurt when reading the code, but at least it keeps the >> comments near the instruction being discussed. > > For the sake of avoiding the conflict, can we just drop it for now, please? > Having an "eret" macro isn't obvious, because people won't realise that it's > a macro. Having "exception_return" is cryptic as hell to people familiar > with the ISA. I'd be OK not adding comments in the assembly provided that we document this within the new documentation file as I just posted as RFC: http://lkml.kernel.org/r/1517936413-19675-1-git-send-email-mathieu.desnoyers@efficios.com Thoughts ? Thanks, Mathieu > > Will -- Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com