Bug#534444: Acknowledgement (linux-image-2.6.26-2-486: CBQ broken)
I thought we had it closed :)
I have not tried with anything past 2.6.26.
2.6.26 as discussed before on this bug has two problems:
1. Counterintuitive use of the word "borrow" in the stats.
2. Bad estimator precision compared to 2.6.18 and older kernels.
If you know about both you can workaround. I do not believe that the use
of "Borrow" will be changed any time soon. It will be interesting to see
if the timers have improved, but I have no test rig to test it right now.
Brgds,
Moritz Muehlenhoff wrote:
> tags 534444 moreinfo
> thanks
>
> On Wed, Jun 24, 2009 at 11:08:36PM +0100, Ben Hutchings wrote:
>
>> On Wed, 2009-06-24 at 16:42 +0100, Anton Ivanov wrote:
>>
>>> Additional information.
>>>
>>> It does not do it on all classes. I can observe it on a particular class
>>> parented to the root CBQ qdisc with multiple burstable children.
>>>
>>> "isolated" put on another class parented to the root qdisc is similarly
>>> ignored.
>>>
>>> I will try to dig through the source to see exactly where the bug is,
>>> but it is definitely a bug (the results of tc are confirmed by delay
>>> measurements and bandwidth measurements in the relevant classes).
>>>
>> Please also test whether the configuration you're trying is still broken
>> in kernel version 2.6.30 (from unstable). If so, please report this bug
>> upstream on bugzilla.kernel.org.
>>
>
> Anton,
> does still occur with current kernels? Did you report it in the kernel.org
> bugzilla?
>
> Cheers,
> Moritz
>
>
--
Understanding is a three-edged sword:
your side, their side, and the truth. --Kosh Naranek
A. R. Ivanov
E-mail: aivanov@sigsegv.cx
WWW: http://www.sigsegv.cx/
pub 1024D/DDE5E715 2002-03-03 Anton R. Ivanov <ai1-n@sigsegv.cx>
Fingerprint: C824 CBD7 EE4B D7F8 5331 89D5 FCDA 572E DDE5 E715
Reply to: