I think I found the breakage.
Have a look at cbq_set_lss()
Instead of checking the flags with && and working based on that it
actually ANDs the flags every time (which if the flag is not already set
results in an eternal false).
I am rebuilding it at the moment after replacing the offending & w
I have finally gotten around to look at it properly (it has been
annoying me all morning so I did not have choice, but to get to it).
There is no way I can see the current kernel code to work.
It sets borrow to be _ALWAYS_ equal to the parent on line 2077 of
cbq_shed.c. For a bounded class it sho
Bounded classes are allowed to borrow at least under some circumstances.
In my config there is a bounded class parented to root on my DSL uplink
and a hierarchy sitting under it where most classes are allowed to
borrow. If the root class is bounded it all works like a breeze. I have
used to use a
On Wed, Jun 24, 2009 at 10:21:05AM +0100, Anton Ivanov wrote:
> Package: linux-image-2.6.26
> Version: nfsfix.1
> Severity: normal
>
>
> CBQ is completely broken. The borrowed counters never increase
> and from there on the bandwidth computation is totally fubar
Please explain the problem more
Package: linux-image-2.6.26
Version: nfsfix.1
Severity: normal
CBQ is completely broken. The borrowed counters never increase and from there
on the bandwidth computation is totally fubar.
-- System Information:
Debian Release: 5.0.1
APT prefers stable
APT policy: (500, 'stable')
Architectu
5 matches
Mail list logo
