When falling back to SO_TIMESTAMP reply timestamp is probably not working

[payload]

I noticed in this commit
e866063#commitcomment-151310711
that when fping can't setsockopt SO_TIMESTAMPNS it falls back to SO_TIMESTAMP. But the reading of the cmsg data is not considering this fallback. In that case no timestamp would be read. I think the fallback is not completely implemented. In that case the reply time is still coming from this line

fping/src/fping.c

Line 2454 in cb83286

recv_time = current_time_ns;

Relevant documentation: https://docs.kernel.org/networking/timestamping.html

[auerswal]

Thanks for the report!

There seems to be code missing to handle non-nanosecond timestamps. Only nanosecond resolution timestamp messages are handled:

fping/src/fping.c

Lines 2104 to 2117 in cb83286

	#if HAVE_SO_TIMESTAMPNS
	/* ancilliary data */
	{
	struct timespec reply_timestamp_ts;
	for (cmsg = CMSG_FIRSTHDR(&recv_msghdr);
	cmsg != NULL;
	cmsg = CMSG_NXTHDR(&recv_msghdr, cmsg)) {
	if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SCM_TIMESTAMPNS) {
	memcpy(&reply_timestamp_ts, CMSG_DATA(cmsg), sizeof(reply_timestamp_ts));
	*reply_timestamp = timespec_ns(&reply_timestamp_ts);
	}
	}
	}
	#endif

According to the socket(7) man page, the fall back to SO_TIMESTAMP should result in SCM_TIMESTAMP control messages where the timestamp is in struct timeval format. This differs from SO_TIMESTAMPNS which uses SCM_TIMESTAMPNS control messages and struct timespec.