Re: [Linuxptp-users] Issues with linuxptp and Intel 82574

[Richard C. <ric...@gm...>]

On Sat, Oct 19, 2013 at 12:28:16AM +0000, Fabrizio Giordano wrote:
> Hello everyone,
> 
> I've been trying to use linux with my Intel 82574 with no much luck.
> First I've tried to use it with software timestamping since the version of e1000e I was using didn't support HW timestamping.
> This is what I get:
> 
> [root@199-dorothy13 ~]# ptp4l -i eth0 -v -S
> ptp4l[25333.043]: port 1: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[25333.043]: port 0: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[25334.561]: port 1: new foreign master 003048.fffe.9ea9d6-1
> ptp4l[25338.561]: selected best master clock 003048.fffe.9ea9d6
> ptp4l[25338.561]: port 1: LISTENING to GRAND_MASTER on RS_GRAND_MASTER
> ptp4l[25339.567]: recvmsg tx timestamp failed: Resource temporarily unavailable
> ptp4l[25339.567]: port 1: send sync failed
> ptp4l[25339.567]: port 1: GRAND_MASTER to FAULTY on FAULT_DETECTED
> ptp4l[25354.574]: port 1: FAULTY to LISTENING on FAULT_CLEARED
> 
> What is that supposed to mean? Resource temporarily unavailable?

It means that no transmit time stamp was returned by the networking
stack before the timeout expired.

Looking at the e1000e driver, both the HW and the SW time stamping
first appeared with with kernel 3.9, and so you will not see any
transmit SW time stamps with your 2.6.32 kernel.

However, if the SF driver includes a call to skb_tx_timestamp (and
that code is not conditionally #ifdef'd out), then SW time stamping
should work using that driver on your 2.6.32 kernel.

HTH,
Richard