[Linuxptp-users] Overriding hardware timestamping with /etc/timemaster.conf

[Matthew Huff]

We are running Redhat 6.8 on a HPE BL460c Gen 9 blade with a HP FlexFabric 20Gb 2-port 630FLB Adapter. The NIC uses the qlogic netxtreme2 driver. There is a known bug with older versions of the bnx2x driver that prevents PTP from working. Newever releases of the kmod-netxtreme2 module resolves this. We are running Redhat's MRG realtime kernel that doesn't allow kmod-* kernels to load, so it's stuck using version 1.710.51-0 from 2014.

While I'm working with Redhat to resolve this, I'm trying to find a way to force timemaster to use software timestamping instead of the hardware timestamping it discovers that the card supports. Any suggestions?

I've tried:

[ptp_domain 0]
time_stamping software

and

[ptp_domain 0]
ptp4l_option time_stamping software

And

[ptp4l.conf]
time_stamping software

and

[ptp4l]
options -S

either the clause has no effect or it prevents the processes from starting.

Any suggestions?


----
Matthew Huff             | 1 Manhattanville Rd
Director of Operations   | Purchase, NY 10577
OTA Management LLC       | Phone: [phone]
aim: matthewbhuff        | Fax:   [phone]

[Miroslav Lichvar]

Post by Matthew Huff
We are running Redhat 6.8 on a HPE BL460c Gen 9 blade with a HP FlexFabric 20Gb 2-port 630FLB Adapter. The NIC uses the qlogic netxtreme2 driver. There is a known bug with older versions of the bnx2x driver that prevents PTP from working. Newever releases of the kmod-netxtreme2 module resolves this. We are running Redhat's MRG realtime kernel that doesn't allow kmod-* kernels to load, so it's stuck using version 1.710.51-0 from 2014.
While I'm working with Redhat to resolve this, I'm trying to find a way to force timemaster to use software timestamping instead of the hardware timestamping it discovers that the card supports. Any suggestions?

There is currently no option to tell timemaster it shouldn't use HW
timestamping when it's available. I think the only way is to define a
dummy PTP domain which would take HW timestamping on the interface
first and the real domain would be left with SW timestamping as only
one domain per interface can use HW timestamping. E.g.:

[ptp_domain 1]
interfaces eth0

[ptp_domain 0]
interfaces eth0

Domain 1 would use HW timestamping on eth0, but if there is no master
in domain 1 on your network, it wouldn't do anything.