[bug?] bug in the htb netlink messages ?

[Fabrice HOBAYA]

Le 22/02/2012 00:00, Thomas Graf a écrit :
> On Tue, Feb 21, 2012 at 06:01:16PM +0100, Fabrice HOBAYA wrote:
>> Actually the nl_object_dump() I've reported was before I send it to
>> the kernel (via rtnl_class_add()).
>>
>> So before the sending (due to previous messages):
>>
>> class htb dev (null) id 2:700 parent 2:50 rate 1.83KiB/s (15Kbit)
>> log 2147483648
>>     ceil 1.83KiB/s (15Kbit) log 2147483648 prio 0
>>
>> and after the sending:
>>
>> class htb dev (null) id 2:700 parent 2:50 rate 1.83KiB/s (15Kbit) log 8
>>     ceil 1.83KiB/s (15Kbit) log 8 prio 0
>>
>> Seems like the kernel expects something different ? Right ? What can I do ?
> Agreed, what we have to do here is compare the htb rate calculation found
> in iproute2 with the rate calculation of libnl. This used to work so it
> must have been broken by something.
OK, I see the problem but don't know where to look for. Do you think the 
problem could be in the SCH_HTB_HAS_RATE and SCH_HTB_HAS_CEIL mask 
values used by the rtnl_set_htb_rate/ceil? or in the code which 
generates netlink messages in the rtnl_class_add() function ?

-- 
Fabrice HOBAYA
PhD - Ingénieur d'études Viveris Technologies
1 Avenue de l'Europe - Parc technologique du Canal -
Bat C - 1er étage - 31400 Toulouse - France
fabrice.hobaya at toulouse.viveris.com