Miroslav Lichvar wrote:
> One thing that is not clear to me, how could this bug impact transmit
> timestamps? The SCM_TIMESTAMPING_PKTINFO option should apply only to
> received packets, at least that's what it originally was support to do.

You're right, it doesn't. The fix is RX-only. TX resolves through
the error queue IP_PKTINFO, which carries the egress interface, and
never saw the zero if_index.

The before/after in my note is misleading, and I should have caught
it: it spans bringing up the hardware timestamps at all, not just
this fix. This board's CPTS stamps only PTP frames, so NTP is
stamped by a PRU core on the GPS PPS timebase, fed into the kernel
paths by a small module that went live the same night. TX read
hardware as soon as that landed; only RX needed the zero check.

-- 
To unsubscribe email [email protected] with "unsubscribe" 
in the subject.
For help email [email protected] with "help" in the 
subject.
Trouble?  Email [email protected].

Reply via email to