From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751173AbdEaJA1 (ORCPT ); Wed, 31 May 2017 05:00:27 -0400 Received: from mail-wr0-f195.google.com ([209.85.128.195]:32821 "EHLO mail-wr0-f195.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751000AbdEaJA0 (ORCPT ); Wed, 31 May 2017 05:00:26 -0400 Date: Wed, 31 May 2017 11:00:21 +0200 From: Richard Cochran To: Thomas Gleixner Cc: LKML , John Stultz , Peter Zijlstra , Ingo Molnar Subject: Re: [patch 05/26] posix-clocks: Remove interval timer facility and mmap/fasync callbacks Message-ID: <20170531090021.GA1910@localhost.localdomain> References: <20170530211533.216608851@linutronix.de> <20170530211656.145036286@linutronix.de> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170530211656.145036286@linutronix.de> User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, May 30, 2017 at 11:15:38PM +0200, Thomas Gleixner wrote: > The only user of this facility is ptp_clock, which does not implement any of > those functions. > > Remove them to prevent accidental users. Especially the interval timer > interfaces are now more or less impossible to implement because the > necessary infrastructure has been confined to the core code. Aside of that > it's really complex to make these callbacks implemented according to spec > as the alarm timer implementation demonstrates. If at all then a nanosleep > callback might be a reasonable extension. For now keep just what ptp_clock > needs. > > Signed-off-by: Thomas Gleixner Acked-by: Richard Cochran For the record, recently someone did try to implement this interface for the i210 card. Although there were issues in the implementation, still I was curious enough to test the performance. https://www.mail-archive.com/linuxptp-devel@lists.sourceforge.net/msg01738.html The result confirmed my expectation that using the normal system clock sychronized to the PHC using the phc2sys utility yields better performance than the direct timer_settime() implementation. The nanosleep() idea might be worthwhile, but only for devices with directly mapped registers (like in some SoCs) and not for the common PCIe devices. In any case, I am happy to see the timer interface removed, as it is basically useless and gives people false impressions. Could you please include this documentation fix, too? Thanks, Richard ---8<--- diff --git a/Documentation/ptp/ptp.txt b/Documentation/ptp/ptp.txt index ae8fef8..11e904e 100644 --- a/Documentation/ptp/ptp.txt +++ b/Documentation/ptp/ptp.txt @@ -18,7 +18,6 @@ - Adjust clock frequency + Ancillary clock features - - One short or periodic alarms, with signal delivery to user program - Time stamp external events - Period output signals configurable from user space - Synchronization of the Linux system time via the PPS subsystem @@ -48,9 +47,7 @@ User space programs may control the clock using standardized ioctls. A program may query, enable, configure, and disable the ancillary clock features. User space can receive time stamped - events via blocking read() and poll(). One shot and periodic - signals may be configured via the POSIX timer_settime() system - call. + events via blocking read() and poll(). ** Writing clock drivers