Question about failure in PCI power-state change

Question about failure in PCI power-state change

[Larry Finger]

Hi,

On occasion when I unload and reload driver rtl8192ce, I get the message

"Refused to change power state, currently in D3"

I added additional info to the printk and discovered that it was trying to 
change to state D0. The problem seems to occur when I have made a connection, 
unloaded the driver, and reloaded it. If I do not make a connection before 
unloading, then the problem is a lot less likely.

This is with kernel 3.6-rc4 from wireless-testing. A Google search shows that 
the problem is usually due to suspend/resume difficulties, but this machine has 
not been suspended.

I would appreciate any help that you might give in solving this issue.

Thanks,

Larry

Re: Question about failure in PCI power-state change

[Bjorn Helgaas]

+cc Rafael, Huang, linux-pm

On Fri, Sep 7, 2012 at 3:44 PM, Larry Finger <Larry.Finger@lwfinger.net> wrote:
> Hi,
>
> On occasion when I unload and reload driver rtl8192ce, I get the message
>
> "Refused to change power state, currently in D3"
>
> I added additional info to the printk and discovered that it was trying to
> change to state D0. The problem seems to occur when I have made a
> connection, unloaded the driver, and reloaded it. If I do not make a
> connection before unloading, then the problem is a lot less likely.
>
> This is with kernel 3.6-rc4 from wireless-testing. A Google search shows
> that the problem is usually due to suspend/resume difficulties, but this
> machine has not been suspended.
>
> I would appreciate any help that you might give in solving this issue.
>
> Thanks,
>
> Larry

Re: Question about failure in PCI power-state change

[Rafael J. Wysocki]

On Saturday, September 08, 2012, Bjorn Helgaas wrote:
> +cc Rafael, Huang, linux-pm
> 
> On Fri, Sep 7, 2012 at 3:44 PM, Larry Finger <Larry.Finger@lwfinger.net> wrote:
> > Hi,
> >
> > On occasion when I unload and reload driver rtl8192ce, I get the message
> >
> > "Refused to change power state, currently in D3"
> >
> > I added additional info to the printk and discovered that it was trying to
> > change to state D0. The problem seems to occur when I have made a
> > connection, unloaded the driver, and reloaded it. If I do not make a
> > connection before unloading, then the problem is a lot less likely.
> >
> > This is with kernel 3.6-rc4 from wireless-testing. A Google search shows
> > that the problem is usually due to suspend/resume difficulties, but this
> > machine has not been suspended.
> >
> > I would appreciate any help that you might give in solving this issue.

I think writing "on" to the /sys/devices/.../power/control attribute of the device
in question would help.

Thanks,
Rafael

Re: Question about failure in PCI power-state change

[Alan Stern]

On Sat, 8 Sep 2012, Rafael J. Wysocki wrote:

> On Saturday, September 08, 2012, Bjorn Helgaas wrote:
> > +cc Rafael, Huang, linux-pm
> > 
> > On Fri, Sep 7, 2012 at 3:44 PM, Larry Finger <Larry.Finger@lwfinger.net> wrote:
> > > Hi,
> > >
> > > On occasion when I unload and reload driver rtl8192ce, I get the message
> > >
> > > "Refused to change power state, currently in D3"
> > >
> > > I added additional info to the printk and discovered that it was trying to
> > > change to state D0. The problem seems to occur when I have made a
> > > connection, unloaded the driver, and reloaded it. If I do not make a
> > > connection before unloading, then the problem is a lot less likely.
> > >
> > > This is with kernel 3.6-rc4 from wireless-testing. A Google search shows
> > > that the problem is usually due to suspend/resume difficulties, but this
> > > machine has not been suspended.
> > >
> > > I would appreciate any help that you might give in solving this issue.
> 
> I think writing "on" to the /sys/devices/.../power/control attribute of the device
> in question would help.

Are you suggesting that this particular wireless adapter is unable to 
return to D0 from D3?

Larry, what happens if you try the suspend & resume your system while 
there's a connection?  Does a similar error occur during the resume?

Alan Stern

Re: Question about failure in PCI power-state change

[Larry Finger]

On 09/08/2012 04:50 PM, Alan Stern wrote:
> On Sat, 8 Sep 2012, Rafael J. Wysocki wrote:
>
>> On Saturday, September 08, 2012, Bjorn Helgaas wrote:
>>> +cc Rafael, Huang, linux-pm
>>>
>>> On Fri, Sep 7, 2012 at 3:44 PM, Larry Finger <Larry.Finger@lwfinger.net> wrote:
>>>> Hi,
>>>>
>>>> On occasion when I unload and reload driver rtl8192ce, I get the message
>>>>
>>>> "Refused to change power state, currently in D3"
>>>>
>>>> I added additional info to the printk and discovered that it was trying to
>>>> change to state D0. The problem seems to occur when I have made a
>>>> connection, unloaded the driver, and reloaded it. If I do not make a
>>>> connection before unloading, then the problem is a lot less likely.
>>>>
>>>> This is with kernel 3.6-rc4 from wireless-testing. A Google search shows
>>>> that the problem is usually due to suspend/resume difficulties, but this
>>>> machine has not been suspended.
>>>>
>>>> I would appreciate any help that you might give in solving this issue.
>>
>> I think writing "on" to the /sys/devices/.../power/control attribute of the device
>> in question would help.
>
> Are you suggesting that this particular wireless adapter is unable to
> return to D0 from D3?
>
> Larry, what happens if you try the suspend & resume your system while
> there's a connection?  Does a similar error occur during the resume?

After I sent this request, I found an error in the removal process for this 
particular adapter, which I suspect is the real source of the problem. That 
difficulty is only found on the new variant of the card - the older one is OK. I 
think the problem is that the adapter is left in some ill-defined state that can 
only be cleared by doing a power reset. In that state, it is unable to return to D0.

This box has never been able to suspend/resume, thus I cannot answer the other 
question.

Thanks to all that responded. Once I clean up the other error, I will recontact 
the list if the D3 => D0 problem persists.

Larry

Re: Question about failure in PCI power-state change

[Huang Ying]

On Fri, 2012-09-07 at 15:48 -0700, Bjorn Helgaas wrote:
> +cc Rafael, Huang, linux-pm
> 
> On Fri, Sep 7, 2012 at 3:44 PM, Larry Finger <Larry.Finger@lwfinger.net> wrote:
> > Hi,
> >
> > On occasion when I unload and reload driver rtl8192ce, I get the message
> >
> > "Refused to change power state, currently in D3"

Hi, Larry,

We fixed a similar issue recently.  And the patch has been merged by
v3.6-rc5.  Can you try v3.6-rc5?

Best Regards,
Huang Ying

Re: Question about failure in PCI power-state change

[Larry Finger]

On 09/09/2012 07:24 PM, Huang Ying wrote:
> On Fri, 2012-09-07 at 15:48 -0700, Bjorn Helgaas wrote:
>> +cc Rafael, Huang, linux-pm
>>
>> On Fri, Sep 7, 2012 at 3:44 PM, Larry Finger <Larry.Finger@lwfinger.net> wrote:
>>> Hi,
>>>
>>> On occasion when I unload and reload driver rtl8192ce, I get the message
>>>
>>> "Refused to change power state, currently in D3"
>
> Hi, Larry,
>
> We fixed a similar issue recently.  And the patch has been merged by
> v3.6-rc5.  Can you try v3.6-rc5?

V3.6-rc5 was merged with wireless-testing today. It did not change my symptoms, 
but I think I have identified another problem with my driver.

Thanks for the heads-up.

Larry