From mboxrd@z Thu Jan 1 00:00:00 1970 From: Julia Cartwright Subject: Re: Kernel 4.6.7-rt13: Intel Ethernet driver igb causes huge latencies in cyclictest Date: Mon, 17 Oct 2016 13:36:45 -0500 Message-ID: <20161017183645.GA9435@jcartwri.amer.corp.natinst.com> References: <20161005155959.GH10625@jcartwri.amer.corp.natinst.com> <13c3cd3ffee4490fb22b8de383e51361@FE-MBX1012.de.bosch.com> <29250f87b1d84aacb8aa312935582291@FE-MBX1012.de.bosch.com> <20161010193958.GE22235@jcartwri.amer.corp.natinst.com> <20161013161839.GV10625@jcartwri.amer.corp.natinst.com> <20161014220633.GA1811@netboy> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Cc: "Koehrer Mathias (ETAS/ESW5)" , "Williams, Mitch A" , "Kirsher, Jeffrey T" , "linux-rt-users@vger.kernel.org" , Sebastian Andrzej Siewior , "netdev@vger.kernel.org" , "intel-wired-lan@lists.osuosl.org" , Greg To: Richard Cochran Return-path: Received: from skprod2.natinst.com ([130.164.80.23]:45278 "EHLO ni.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S932863AbcJQShB (ORCPT ); Mon, 17 Oct 2016 14:37:01 -0400 In-Reply-To: <20161014220633.GA1811@netboy> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: On Sat, Oct 15, 2016 at 12:06:33AM +0200, Richard Cochran wrote: > On Fri, Oct 14, 2016 at 08:58:22AM +0000, Koehrer Mathias (ETAS/ESW5) wrote: > > @@ -753,7 +756,9 @@ u32 igb_rd32(struct e1000_hw *hw, u32 re > > if (E1000_REMOVED(hw_addr)) > > return ~value; > > > > + trace_igb(801); > > value = readl(&hw_addr[reg]); > > + trace_igb(802); > > Nothing prevents this code from being preempted between the two trace > points, and so you can't be sure whether the time delta in the trace > is caused by the PCIe read stalling or not. While that is certainly the case, and would explain the most egregious of measured latency spikes, it doesn't invalidate the test if you consider the valuable data point(s) to be the minimum and/or median latencies. Julia From mboxrd@z Thu Jan 1 00:00:00 1970 From: Julia Cartwright Date: Mon, 17 Oct 2016 13:36:45 -0500 Subject: [Intel-wired-lan] Kernel 4.6.7-rt13: Intel Ethernet driver igb causes huge latencies in cyclictest In-Reply-To: <20161014220633.GA1811@netboy> References: <20161005155959.GH10625@jcartwri.amer.corp.natinst.com> <13c3cd3ffee4490fb22b8de383e51361@FE-MBX1012.de.bosch.com> <29250f87b1d84aacb8aa312935582291@FE-MBX1012.de.bosch.com> <20161010193958.GE22235@jcartwri.amer.corp.natinst.com> <20161013161839.GV10625@jcartwri.amer.corp.natinst.com> <20161014220633.GA1811@netboy> Message-ID: <20161017183645.GA9435@jcartwri.amer.corp.natinst.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: intel-wired-lan@osuosl.org List-ID: On Sat, Oct 15, 2016 at 12:06:33AM +0200, Richard Cochran wrote: > On Fri, Oct 14, 2016 at 08:58:22AM +0000, Koehrer Mathias (ETAS/ESW5) wrote: > > @@ -753,7 +756,9 @@ u32 igb_rd32(struct e1000_hw *hw, u32 re > > if (E1000_REMOVED(hw_addr)) > > return ~value; > > > > + trace_igb(801); > > value = readl(&hw_addr[reg]); > > + trace_igb(802); > > Nothing prevents this code from being preempted between the two trace > points, and so you can't be sure whether the time delta in the trace > is caused by the PCIe read stalling or not. While that is certainly the case, and would explain the most egregious of measured latency spikes, it doesn't invalidate the test if you consider the valuable data point(s) to be the minimum and/or median latencies. Julia