From mboxrd@z Thu Jan 1 00:00:00 1970 From: Miroslav Lichvar Subject: Re: [patch 4/5] PTP: add PTP_SYS_OFFSET emulation via cross timestamps infrastructure Date: Tue, 24 Jan 2017 09:15:56 +0100 Message-ID: <20170124081556.GY14227@localhost> References: <20170120122025.665985919@redhat.com> <20170120122503.746158230@redhat.com> <20170120202502.GA10368@localhost.localdomain> <20170123131913.GA28104@amt.cnet> <20170123184415.GA2158@localhost.localdomain> <20170123230617.GA14752@amt.cnet> <20170124053243.GA1906@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Marcelo Tosatti , kvm@vger.kernel.org, linux-kernel@vger.kernel.org, Paolo Bonzini , Radim Krcmar To: Richard Cochran Return-path: Content-Disposition: inline In-Reply-To: <20170124053243.GA1906@localhost.localdomain> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On Tue, Jan 24, 2017 at 06:32:43AM +0100, Richard Cochran wrote: > On Mon, Jan 23, 2017 at 09:06:20PM -0200, Marcelo Tosatti wrote: > > Can you please describe what problem exists with this scheme? > > This new kernel code exists just because chrony doesn't implement the > PRECISE ioctl. Instead of adding new "fake" modes, just teach chrony > about the better method. The latest development code of chrony now supports the PRECISE ioctl. I did some tests with an e1000e NIC (i219) and it seemed the stability was slightly worse than with the non-PRECISE ioctl, but there was a 400-500ns offset between the two, so it should be much more accurate (we finally have something that avoids the asymmetry on the PCI-e bus?). Configuring the refclock with a shorter dpoll should compensate for the decrease in stability. In any case, there is a "nocrossts" option to not use the PRECISE ioctl even if it's supported. -- Miroslav Lichvar