Apologies for this late follow-up. I have spent an entire
day going through the patches again, checking again the
changes made on Friday.
I had missed one or two instances of skb_get_timestamp, these
were better hidden. There are now no more left and the patch
set has been compile tested on 5 different machines.
Furthermore, I found a bug in the computation of CCID3 elapsed
time when sending feedback, it is one of the older bugs and
had not occurred to anyone before.
The problem is that the `elapsed time' reported by the receiver
is merely the time difference between calling the functions
rx_packet_recv() and rx_send_feedback(). This amounts on the average
to 50 .. 100 usec.
With regard to RTT estimation, this is almost entirely useless.
Following is therefore a patch which tackles this problem with an
alternative approach, I believe it is a bit better.
I have left in the clean-up operation of removing the last instance
of skb_get_timestamp, hope that this is ok.
Due to the patch inter-dependencies, this patch is between the others,
uploaded as
10e_CCID3-TIMESTAMP-FIX.diff
Thanks.
-
To unsubscribe from this list: send the line "unsubscribe dccp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
