Issue with ifconfig down and 4965 driver

Issue with ifconfig down and 4965 driver

[Marcel Holtmann]

Hi guys,

so I am running a 2.6.28-rc4 kernel and every time I stop  
wpa_supplicant or call ifconfig down on the 4965 card, I get a user  
noticeable delay (mouse freezes). Seems that some part inside the  
driver is actively sleeping with a busy-loop. This is really not an  
acceptable behavior since it is a day-to-day task to bring the device  
down. And for power concerns, we really wanna allow to bring down the  
interface.

Regards

Marcel

Re: Issue with ifconfig down and 4965 driver

[reinette chatre]

On Wed, 2008-11-12 at 20:14 -0800, Marcel Holtmann wrote:
> Hi guys,
> 
> so I am running a 2.6.28-rc4 kernel and every time I stop  
> wpa_supplicant or call ifconfig down on the 4965 card, I get a user  
> noticeable delay (mouse freezes). Seems that some part inside the  
> driver is actively sleeping with a busy-loop. This is really not an  
> acceptable behavior since it is a day-to-day task to bring the device  
> down. And for power concerns, we really wanna allow to bring down the  
> interface.
> 

Please add your information to
http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790

Reinette

Re: Issue with ifconfig down and 4965 driver

[Marcel Holtmann]

Hi Reinette,

>> so I am running a 2.6.28-rc4 kernel and every time I stop  
>> wpa_supplicant or call ifconfig down on the 4965 card, I get a user  
>> noticeable delay (mouse freezes). Seems that some part inside the  
>> driver is actively sleeping with a busy-loop. This is really not an  
>> acceptable behavior since it is a day-to-day task to bring the device  
>> down. And for power concerns, we really wanna allow to bring down the  
>> interface.
>>
> 
> Please add your information to
> http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790

all information are already in the Bugzilla. It is nothing fancy and 
easy to reproduce. Do we have a patch for it?

Regards

Marcel

Re: Issue with ifconfig down and 4965 driver

[reinette chatre]

On Thu, 2008-11-13 at 01:45 -0800, Marcel Holtmann wrote:
> Hi Reinette,
> 
> >> so I am running a 2.6.28-rc4 kernel and every time I stop
> >> wpa_supplicant or call ifconfig down on the 4965 card, I get a user
> >> noticeable delay (mouse freezes). Seems that some part inside the
> >> driver is actively sleeping with a busy-loop. This is really not an
> >> acceptable behavior since it is a day-to-day task to bring the device
> >> down. And for power concerns, we really wanna allow to bring down the
> >> interface.
> >>
> >
> > Please add your information to
> > http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790
> 
> all information are already in the Bugzilla. It is nothing fancy and
> easy to reproduce. Do we have a patch for it?
> 

http://marc.info/?l=linux-wireless&m=122824886623698&w=2


Reinette

Re: Issue with ifconfig down and 4965 driver

[Marcel Holtmann]

Hi Reinette,

> > >> so I am running a 2.6.28-rc4 kernel and every time I stop
> > >> wpa_supplicant or call ifconfig down on the 4965 card, I get a user
> > >> noticeable delay (mouse freezes). Seems that some part inside the
> > >> driver is actively sleeping with a busy-loop. This is really not an
> > >> acceptable behavior since it is a day-to-day task to bring the device
> > >> down. And for power concerns, we really wanna allow to bring down the
> > >> interface.
> > >>
> > >
> > > Please add your information to
> > > http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790
> > 
> > all information are already in the Bugzilla. It is nothing fancy and
> > easy to reproduce. Do we have a patch for it?
> > 
> 
> http://marc.info/?l=linux-wireless&m=122824886623698&w=2

so I finally got around testing the patch. It is great that it reduces
the time drastically, but I think that it is still not good enough. So I
don't care how long it takes to bring the device down. My concern is
that we are busy-looping during this time. Meaning the kernel is not
responding to anything. Simple symptom is that the mouse is stuck.

Regards

Marcel

Re: Issue with ifconfig down and 4965 driver

[reinette chatre]

On Wed, 2008-12-03 at 15:38 -0800, Marcel Holtmann wrote:
> Hi Reinette,
> 
> > > >> so I am running a 2.6.28-rc4 kernel and every time I stop
> > > >> wpa_supplicant or call ifconfig down on the 4965 card, I get a user
> > > >> noticeable delay (mouse freezes). Seems that some part inside the
> > > >> driver is actively sleeping with a busy-loop. This is really not an
> > > >> acceptable behavior since it is a day-to-day task to bring the device
> > > >> down. And for power concerns, we really wanna allow to bring down the
> > > >> interface.
> > > >>
> > > >
> > > > Please add your information to
> > > > http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790
> > > 
> > > all information are already in the Bugzilla. It is nothing fancy and
> > > easy to reproduce. Do we have a patch for it?
> > > 
> > 
> > http://marc.info/?l=linux-wireless&m=122824886623698&w=2
> 
> so I finally got around testing the patch. It is great that it reduces
> the time drastically, but I think that it is still not good enough. So I
> don't care how long it takes to bring the device down. My concern is
> that we are busy-looping during this time. Meaning the kernel is not
> responding to anything. Simple symptom is that the mouse is stuck.
> 

We have found that this one patch is not the full solution. More work is
currently being done with activity tracked at
http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790

Reinette

Re: Issue with ifconfig down and 4965 driver

[Marcel Holtmann]

Hi Reinette,

> > > > >> so I am running a 2.6.28-rc4 kernel and every time I stop
> > > > >> wpa_supplicant or call ifconfig down on the 4965 card, I get a user
> > > > >> noticeable delay (mouse freezes). Seems that some part inside the
> > > > >> driver is actively sleeping with a busy-loop. This is really not an
> > > > >> acceptable behavior since it is a day-to-day task to bring the device
> > > > >> down. And for power concerns, we really wanna allow to bring down the
> > > > >> interface.
> > > > >>
> > > > >
> > > > > Please add your information to
> > > > > http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790
> > > > 
> > > > all information are already in the Bugzilla. It is nothing fancy and
> > > > easy to reproduce. Do we have a patch for it?
> > > > 
> > > 
> > > http://marc.info/?l=linux-wireless&m=122824886623698&w=2
> > 
> > so I finally got around testing the patch. It is great that it reduces
> > the time drastically, but I think that it is still not good enough. So I
> > don't care how long it takes to bring the device down. My concern is
> > that we are busy-looping during this time. Meaning the kernel is not
> > responding to anything. Simple symptom is that the mouse is stuck.
> > 
> 
> We have found that this one patch is not the full solution. More work is
> currently being done with activity tracked at
> http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790

I quickly checked Mohamed's last patch. It actually makes it worse on my
system.

Regards

Marcel

Re: Issue with ifconfig down and 4965 driver

[reinette chatre]

On Thu, 2008-12-04 at 03:19 -0800, Marcel Holtmann wrote:
> Hi Reinette,
> 
> > > > > >> so I am running a 2.6.28-rc4 kernel and every time I stop
> > > > > >> wpa_supplicant or call ifconfig down on the 4965 card, I get a user
> > > > > >> noticeable delay (mouse freezes). Seems that some part inside the
> > > > > >> driver is actively sleeping with a busy-loop. This is really not an
> > > > > >> acceptable behavior since it is a day-to-day task to bring the device
> > > > > >> down. And for power concerns, we really wanna allow to bring down the
> > > > > >> interface.
> > > > > >>
> > > > > >
> > > > > > Please add your information to
> > > > > > http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790
> > > > > 
> > > > > all information are already in the Bugzilla. It is nothing fancy and
> > > > > easy to reproduce. Do we have a patch for it?
> > > > > 
> > > > 
> > > > http://marc.info/?l=linux-wireless&m=122824886623698&w=2
> > > 
> > > so I finally got around testing the patch. It is great that it reduces
> > > the time drastically, but I think that it is still not good enough. So I
> > > don't care how long it takes to bring the device down. My concern is
> > > that we are busy-looping during this time. Meaning the kernel is not
> > > responding to anything. Simple symptom is that the mouse is stuck.
> > > 
> > 
> > We have found that this one patch is not the full solution. More work is
> > currently being done with activity tracked at
> > http://www.intellinuxwireless.org/bugzilla/show_bug.cgi?id=1790
> 
> I quickly checked Mohamed's last patch. It actually makes it worse on my
> system.

Thank you very much for testing - I will let him know. Would it be
possible to add your feedback to that bug report for visibility to
everybody tracking that problem? Not everybody there are subscribed to
linux-wireless and other people testing the patch (see last comment #12)
reports that the issue is solved for them.

Thank you very much

Reinette