> -----Original Message----- > From: Miroslav Lichvar [mailto:[email protected]] > Sent: Wednesday, May 09, 2018 11:17 AM > To: Keller, Jacob E <[email protected]> > Cc: [email protected] > Subject: Re: [Linuxptp-devel] Delay requests causing ptp4l to miss sync TX > timestamp? > > On Wed, May 09, 2018 at 04:05:48PM +0000, Keller, Jacob E wrote: > > > Delay responses don't generate TX timestamps (unlike NSM responses). > > > How can a high rate of RX timestamps cause a TX timestamp to be > > > missed? Is this a common HW limitation? I see this happening with > > > three different drivers (tg3, igb, ixgbe). > > > > > > > I'm not aware of such a limitation... Is it possible that it's fast enough > > that the > general latency in the system exceeds the polling timeout? What about if you > increase the timeout window when waiting for the timestamp response? > > I tried increasing the timeout to 1000 milliseconds and it made no > difference. > > > For the Intel parts, they should share a common tx_hwtstamp_skipped or > tx_hwtstamp_missed counters in ethtool -S which should indicate if the driver > thinks timestamps are being missed. > > Thanks for the hints. > > # ethtool -S enp2s0f0 | grep skip > tx_hwtstamp_skipped: 0 > > I tried it again with ptp4l using SW timestamping and it's the same > problem appearing at the same rate of requests. Does that mean that > the error queue cannot accept a single packet because the buffers are > full due to the non-error queue? > > -- > Miroslav Lichvar
Oh. That could explain this. I'm not sure how the error queue works here... there might be a sysctl value we could set to increase the buffer size and figure that out? Thanks, Jake ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Linuxptp-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
