On Mon, 25 Aug 2008, Thomas Jarosch wrote:

On Friday, 22. August 2008 23:18:44 Ilpo Järvinen wrote:

quoted

In the meantime, can you check the attached patches. Besides the kernel
patch, you need to build your own patched iproute2 as well to configure
the features (ip tool among them is enough in case the build of some other
part of the toolset fails like it did for me). I somewhat tested them, and
the result seemed to be what I'd expect (I just forced RTOs with some
netem heavy dropping and quickly glanced over the resulting packet
patterns near RTO).

Your patches work fine.

Thanks for testing.

I've noticed two small things:
1. Maybe it's a good idea to add a note above the tcp_use_frto() change
to explain that the value is negated. Took me a while to figure out
why there is no "!" in there :-)
2. Maybe rename the "features" option in iproute2 to "disable_features".
Then it would be more intuitive what it does.

First of all I hate doing anything which has an user interface stamp in 
it... :-)

Second, didn't I write about this negation in some of the log messages... 
/me looks for that... hmm... I think I did ...and that's besides the very 
clear help text :-). 

Yeah, it was just that the earlier ip already prints the field as 
"features" though I guess changing also that is a non-problem to existing 
userspace stuff because of the current usage of the field. There's this 
RTAX_FEATURE_ALLFRAG stuff which somebody could be looking for but I 
don't know how likely that will be.

But anyway, point taken. I'll try to change both to disable_features and 
see if that gets accepted.

-- 
 i.