Comment # 17 on bug 1075876 from Benjamin Poirier

https://www.mail-archive.com/netdev@vger.kernel.org/msg231452.html
No update from Intel... probably time to send a patch.

The simplest approach would be to remove the problematic TSYNCRXCTL read in
e1000e_get_base_timinca() and set the clock parameters unconditionally, like
for the 82574 case just below. I've prepared a patch which does that and also
adds some logging (for debugging only, not submission). It would also be
important to know if the ptp hardware clock (PHC) still works at that point,
for that I think we can use phc_ctl which leads to a call of
e1000e_phc_gettime().

Please install the updated kernel package from the
home:benjamin_poirier:bsc1075876 project once it has finished building (link
in comment 15, version should contain "g5227f0d") and install the linuxptp
package. Upon a reboot when you see the "No SYSCFI in TSYNCRXCTL [...]" in
dmesg, run
phc_ctl /dev/ptp0 cmp && sleep 60 && phc_ctl /dev/ptp0 cmp

Please provide the dmesg line with the values after "No SYSCFI in TSYNCRXCTL"
and the output of the phc_ctl command.