Hi David,
On Tue, Jun 03, 2014 at 05:01:14PM +0200, David Sommerseth wrote:
Sorry for the long delay. It's just been busy to do a full
review of
these patches.
no worries at all, I just pinged after a couple of weeks to make sure
they were not forgotten.
Generally speaking, patches 3-5 looks fine at first glance. In
regards
to the statistics patches, I like such a feature! However, I'm trying
to move away from the old API ( ethtool.{get,set}_*($DEVICE) ) and
rather put it into the object oriented API. So moving in that direction
is highly appreciated.
Ah ok, I'll look things up and resubmit patch 1-2 to fit them in the
object oriented API. Feel free to apply 3-5 if they are okay for you.
(I'll be away in the next couple of weeks, so it might take a bit longer
to redo 1-2)
There are plans on my TODO list to move the old API completely over
to
the new one as well. But time ..... In addition, I've been holding off
a little bit to see what would happen with libnl3's Python support - and
it seems the libnl3 team is not putting much efforts into that area now,
so python-ethtool will be needed too.
Just one question. I see that your stats patches uses ioctl() to
extract this information. Do you know if these stats are comparable to
the information which can be extracted from libnl3?
<
http://www.carisma.slowglass.com/~tgr/libnl/doc/route.html#tc_stats>
Nope, they are an entirely different beast. Basically, ethtool statistics
are statistics returned from the firmware of the NIC itself. They are
very NIC/driver dependent and some drivers do not support statistics at
all (most used NICs do though). TC stats or any other stat coming from
'tc' or 'netstat' measures metrics from the Linux stack, not from the
NIC itself.
Thanks a lot for the review & comments :)
regards,
Michele
--
Michele Baldessari <michele(a)acksyn.org>
C2A5 9DA3 9961 4FFB E01B D0BC DDD4 DCCB 7515 5C6D