Re: [Madwifi-devel] Big loss of small packets

[Benoit P. <ben...@fr...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ignacy Gawedzki a écrit :
> Okay, I agree with that. But first, in the case of my measurements,
>  we can say that the kernel should consider the "cable" as plugged,
> since the destination MAC address *is* reachable.  And second, how
> come changing the debug flags (athdebug +xmit) changes anything at
> all in the way the kernel would consider the "cable" as plugged or
> not?
>
> I would like to understand what kind of side-effects are caused by
> the activation of the debug flag.  Note that it seems only the xmit
> flag causes this change, though I have tried only a few others
> (namely tsf and xmit_proc).
>

That is the good questions :-). In fact, I think it has to do with
Linux queue management which is something that is partially handled by
the kernel (where the packets are stored in a queue) and the drivers
(which reports if its hardware queue are full or not). Depending on
your platform, when you set the xmit flags, it might consumes a lot of
CPU and as such, break the CPU scheduling and as a consequence the
queue scheduling.

What CPU are you using? Is it an embedded platform?

Regards,
Benoît

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkpPWJcACgkQOR6EySwP7oJiAgCeOP2Z53b4LETB5w/8DgbktnY2
WzUAn0Vmb8ZBTGKx9JO4U8xxv/fvLJM3
=DCBV
-----END PGP SIGNATURE-----