BCM4331 fails to associate after suspend

BCM4331 fails to associate after suspend

[Seth Forshee]

I've been looking into some association problems seen with a MacBook Pro
8,2 that has BCM4331 wireless, which I've confirmed still exist in
wireless-testing. The association failures reliably start appearing
after the machine is resumed from s3 without external power applied
(whether or not power is applied when the machine is suspended doesn't
matter).

When the association failures are happening the BCM4331 seems to be able
to receive packets okay, as it can scan successfully. Transmissions
don't work though; no frames from the BCM4331 are ever seen on the air,
as though the tx path is off. Reloading both the b43 and bcma drivers
gets things working again, but reloading just b43 does not.

I've been looking around a bit in the b43 and bcma drivers, but so far I
haven't found the reason why this happens. Any help would be
appreciated. Let me know if there's additional information I can
provide.

Thanks,
Seth

Re: BCM4331 fails to associate after suspend

[John Schoenick]

On 02/03/2012 02:28 PM, Seth Forshee wrote:
> I've been looking into some association problems seen with a MacBook Pro
> 8,2 that has BCM4331 wireless, which I've confirmed still exist in
> wireless-testing. The association failures reliably start appearing
> after the machine is resumed from s3 without external power applied
> (whether or not power is applied when the machine is suspended doesn't
> matter).
>
> When the association failures are happening the BCM4331 seems to be able
> to receive packets okay, as it can scan successfully. Transmissions
> don't work though; no frames from the BCM4331 are ever seen on the air,
> as though the tx path is off. Reloading both the b43 and bcma drivers
> gets things working again, but reloading just b43 does not.
>
> I've been looking around a bit in the b43 and bcma drivers, but so far I
> haven't found the reason why this happens. Any help would be
> appreciated. Let me know if there's additional information I can
> provide.
>
> Thanks,
> Seth
>
> _______________________________________________
> b43-dev mailing list
> b43-dev@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/b43-dev

I can add that reloading the modules after a suspend/resumes restores 
functionality, though oddly adding them to SUSPEND_MODULES in pm-utils 
config does not.

Re: BCM4331 fails to associate after suspend

[Arend van Spriel]

On 02/03/2012 11:28 PM, Seth Forshee wrote:
> 
> I've been looking around a bit in the b43 and bcma drivers, but so far I
> haven't found the reason why this happens. Any help would be
> appreciated. Let me know if there's additional information I can
> provide.
> 

With which kernel version do you see these issues? BCMA suspend/resume
code has a fix in 3.3-rc1. So if you are seeing it on a 3.2 kernel,
please give it a try.

Gr. AvS

Re: BCM4331 fails to associate after suspend

[Seth Forshee]

On Sat, Feb 04, 2012 at 11:44:23AM +0100, Arend van Spriel wrote:
> On 02/03/2012 11:28 PM, Seth Forshee wrote:
> > 
> > I've been looking around a bit in the b43 and bcma drivers, but so far I
> > haven't found the reason why this happens. Any help would be
> > appreciated. Let me know if there's additional information I can
> > provide.
> > 
> 
> With which kernel version do you see these issues? BCMA suspend/resume
> code has a fix in 3.3-rc1. So if you are seeing it on a 3.2 kernel,
> please give it a try.

I've tried 3.3-rc2 and the wirless-testing tree, and the problem still
exists in both.

Seth

Re: BCM4331 fails to associate after suspend

[Rafał Miłecki]

2012/2/4 Seth Forshee <seth.forshee@canonical.com>:
> On Sat, Feb 04, 2012 at 11:44:23AM +0100, Arend van Spriel wrote:
>> On 02/03/2012 11:28 PM, Seth Forshee wrote:
>> >
>> > I've been looking around a bit in the b43 and bcma drivers, but so far I
>> > haven't found the reason why this happens. Any help would be
>> > appreciated. Let me know if there's additional information I can
>> > provide.
>> >
>>
>> With which kernel version do you see these issues? BCMA suspend/resume
>> code has a fix in 3.3-rc1. So if you are seeing it on a 3.2 kernel,
>> please give it a try.
>
> I've tried 3.3-rc2 and the wirless-testing tree, and the problem still
> exists in both.

I don't have access to suspendable machine with Mini PCIe card, not to
mention BCM4331 which is available for Mac-specific slot only (or
routers as embedded wifi). It may be b43 needs to fix but I'm not
really able to work on that. Plus the whole BCM4331 code is really
tricky, as it comes from RE based on dumps only :(

-- 
Rafał

Re: BCM4331 fails to associate after suspend

[Seth Forshee]

On Mon, Feb 06, 2012 at 05:56:20PM +0100, Rafał Miłecki wrote:
> 2012/2/4 Seth Forshee <seth.forshee@canonical.com>:
> > On Sat, Feb 04, 2012 at 11:44:23AM +0100, Arend van Spriel wrote:
> >> On 02/03/2012 11:28 PM, Seth Forshee wrote:
> >> >
> >> > I've been looking around a bit in the b43 and bcma drivers, but so far I
> >> > haven't found the reason why this happens. Any help would be
> >> > appreciated. Let me know if there's additional information I can
> >> > provide.
> >> >
> >>
> >> With which kernel version do you see these issues? BCMA suspend/resume
> >> code has a fix in 3.3-rc1. So if you are seeing it on a 3.2 kernel,
> >> please give it a try.
> >
> > I've tried 3.3-rc2 and the wirless-testing tree, and the problem still
> > exists in both.
> 
> I don't have access to suspendable machine with Mini PCIe card, not to
> mention BCM4331 which is available for Mac-specific slot only (or
> routers as embedded wifi). It may be b43 needs to fix but I'm not
> really able to work on that. Plus the whole BCM4331 code is really
> tricky, as it comes from RE based on dumps only :(

Okay, I'll keep poking at it to see if I can come up with anything. If
you've got any suggestions on things I should be looking at or trying,
I'm all ears.

Seth

Re: BCM4331 fails to associate after suspend

[Hauke Mehrtens]

On 02/06/2012 06:37 PM, Seth Forshee wrote:
> On Mon, Feb 06, 2012 at 05:56:20PM +0100, Rafał Miłecki wrote:
>> 2012/2/4 Seth Forshee <seth.forshee@canonical.com>:
>>> On Sat, Feb 04, 2012 at 11:44:23AM +0100, Arend van Spriel wrote:
>>>> On 02/03/2012 11:28 PM, Seth Forshee wrote:
>>>>>
>>>>> I've been looking around a bit in the b43 and bcma drivers, but so far I
>>>>> haven't found the reason why this happens. Any help would be
>>>>> appreciated. Let me know if there's additional information I can
>>>>> provide.
>>>>>
>>>>
>>>> With which kernel version do you see these issues? BCMA suspend/resume
>>>> code has a fix in 3.3-rc1. So if you are seeing it on a 3.2 kernel,
>>>> please give it a try.
>>>
>>> I've tried 3.3-rc2 and the wirless-testing tree, and the problem still
>>> exists in both.
>>
>> I don't have access to suspendable machine with Mini PCIe card, not to
>> mention BCM4331 which is available for Mac-specific slot only (or
>> routers as embedded wifi). It may be b43 needs to fix but I'm not
>> really able to work on that. Plus the whole BCM4331 code is really
>> tricky, as it comes from RE based on dumps only :(
> 
> Okay, I'll keep poking at it to see if I can come up with anything. If
> you've got any suggestions on things I should be looking at or trying,
> I'm all ears.
> 
> Seth

According to the Broadcom SDK for SoCs bcma_pcicore_serdes_workaround()
should be called when coming out of standby/hibernate.
The code wl uses to do the same thing as bcma does is open source. You
find it in the GPL package of many Broadcom based Wifi Routers.

Hauke

Re: BCM4331 fails to associate after suspend

[Seth Forshee]

On Mon, Feb 06, 2012 at 08:13:11PM +0100, Hauke Mehrtens wrote:
> On 02/06/2012 06:37 PM, Seth Forshee wrote:
> > On Mon, Feb 06, 2012 at 05:56:20PM +0100, Rafał Miłecki wrote:
> >> 2012/2/4 Seth Forshee <seth.forshee@canonical.com>:
> >>> On Sat, Feb 04, 2012 at 11:44:23AM +0100, Arend van Spriel wrote:
> >>>> On 02/03/2012 11:28 PM, Seth Forshee wrote:
> >>>>>
> >>>>> I've been looking around a bit in the b43 and bcma drivers, but so far I
> >>>>> haven't found the reason why this happens. Any help would be
> >>>>> appreciated. Let me know if there's additional information I can
> >>>>> provide.
> >>>>>
> >>>>
> >>>> With which kernel version do you see these issues? BCMA suspend/resume
> >>>> code has a fix in 3.3-rc1. So if you are seeing it on a 3.2 kernel,
> >>>> please give it a try.
> >>>
> >>> I've tried 3.3-rc2 and the wirless-testing tree, and the problem still
> >>> exists in both.
> >>
> >> I don't have access to suspendable machine with Mini PCIe card, not to
> >> mention BCM4331 which is available for Mac-specific slot only (or
> >> routers as embedded wifi). It may be b43 needs to fix but I'm not
> >> really able to work on that. Plus the whole BCM4331 code is really
> >> tricky, as it comes from RE based on dumps only :(
> > 
> > Okay, I'll keep poking at it to see if I can come up with anything. If
> > you've got any suggestions on things I should be looking at or trying,
> > I'm all ears.
> > 
> > Seth
> 
> According to the Broadcom SDK for SoCs bcma_pcicore_serdes_workaround()
> should be called when coming out of standby/hibernate.
> The code wl uses to do the same thing as bcma does is open source. You
> find it in the GPL package of many Broadcom based Wifi Routers.

Thanks for the suggestion. Unfortunately calling
bcma_pcicore_serdes_workaround() during resume doesn't help. I'll look
at the wl code to see if I can learn anything there.

Seth

Re: BCM4331 fails to associate after suspend

[Rafał Miłecki]

2012/2/6 Seth Forshee <seth.forshee@canonical.com>:
> On Mon, Feb 06, 2012 at 08:13:11PM +0100, Hauke Mehrtens wrote:
>> According to the Broadcom SDK for SoCs bcma_pcicore_serdes_workaround()
>> should be called when coming out of standby/hibernate.
>> The code wl uses to do the same thing as bcma does is open source. You
>> find it in the GPL package of many Broadcom based Wifi Routers.
>
> Thanks for the suggestion. Unfortunately calling
> bcma_pcicore_serdes_workaround() during resume doesn't help. I'll look
> at the wl code to see if I can learn anything there.

Can you try


diff --git a/drivers/bcma/main.c b/drivers/bcma/main.c
index febbc0a..a097a26 100644
--- a/drivers/bcma/main.c
+++ b/drivers/bcma/main.c
@@ -267,6 +267,13 @@ int bcma_bus_resume(struct bcma_bus *bus)
                bcma_core_chipcommon_init(&bus->drv_cc);
        }

+       /* Init PCIE core */
+       core = bcma_find_core(bus, BCMA_CORE_PCIE);
+       if (core) {
+               bus->drv_pci.setup_done = false;
+               bcma_core_pci_init(&bus->drv_pci);
+       }
+
        list_for_each_entry(core, &bus->cores, list) {
                struct device_driver *drv = core->dev.driver;
                if (drv) {

Re: BCM4331 fails to associate after suspend

[Rafał Miłecki]

W dniu 7 lutego 2012 08:04 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
> 2012/2/6 Seth Forshee <seth.forshee@canonical.com>:
>> On Mon, Feb 06, 2012 at 08:13:11PM +0100, Hauke Mehrtens wrote:
>>> According to the Broadcom SDK for SoCs bcma_pcicore_serdes_workaround()
>>> should be called when coming out of standby/hibernate.
>>> The code wl uses to do the same thing as bcma does is open source. You
>>> find it in the GPL package of many Broadcom based Wifi Routers.
>>
>> Thanks for the suggestion. Unfortunately calling
>> bcma_pcicore_serdes_workaround() during resume doesn't help. I'll look
>> at the wl code to see if I can learn anything there.
>
> Can you try
>
>
> diff --git a/drivers/bcma/main.c b/drivers/bcma/main.c
> index febbc0a..a097a26 100644
> --- a/drivers/bcma/main.c
> +++ b/drivers/bcma/main.c
> @@ -267,6 +267,13 @@ int bcma_bus_resume(struct bcma_bus *bus)
>                bcma_core_chipcommon_init(&bus->drv_cc);
>        }
>
> +       /* Init PCIE core */
> +       core = bcma_find_core(bus, BCMA_CORE_PCIE);
> +       if (core) {
> +               bus->drv_pci.setup_done = false;
> +               bcma_core_pci_init(&bus->drv_pci);
> +       }
> +
>        list_for_each_entry(core, &bus->cores, list) {
>                struct device_driver *drv = core->dev.driver;
>                if (drv) {

If this doesn't work out of box, please check if reloading b43 is
enough. AFAIR now you have to reload both: b43 and bcma. I hope this
patch will let you avoid reloading at least bcma.

-- 
Rafał

Re: BCM4331 fails to associate after suspend

[Seth Forshee]

On Tue, Feb 07, 2012 at 08:08:11AM +0100, Rafał Miłecki wrote:
> W dniu 7 lutego 2012 08:04 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
> > 2012/2/6 Seth Forshee <seth.forshee@canonical.com>:
> >> On Mon, Feb 06, 2012 at 08:13:11PM +0100, Hauke Mehrtens wrote:
> >>> According to the Broadcom SDK for SoCs bcma_pcicore_serdes_workaround()
> >>> should be called when coming out of standby/hibernate.
> >>> The code wl uses to do the same thing as bcma does is open source. You
> >>> find it in the GPL package of many Broadcom based Wifi Routers.
> >>
> >> Thanks for the suggestion. Unfortunately calling
> >> bcma_pcicore_serdes_workaround() during resume doesn't help. I'll look
> >> at the wl code to see if I can learn anything there.
> >
> > Can you try
> >
> >
> > diff --git a/drivers/bcma/main.c b/drivers/bcma/main.c
> > index febbc0a..a097a26 100644
> > --- a/drivers/bcma/main.c
> > +++ b/drivers/bcma/main.c
> > @@ -267,6 +267,13 @@ int bcma_bus_resume(struct bcma_bus *bus)
> >                bcma_core_chipcommon_init(&bus->drv_cc);
> >        }
> >
> > +       /* Init PCIE core */
> > +       core = bcma_find_core(bus, BCMA_CORE_PCIE);
> > +       if (core) {
> > +               bus->drv_pci.setup_done = false;
> > +               bcma_core_pci_init(&bus->drv_pci);
> > +       }
> > +
> >        list_for_each_entry(core, &bus->cores, list) {
> >                struct device_driver *drv = core->dev.driver;
> >                if (drv) {
> 
> If this doesn't work out of box, please check if reloading b43 is
> enough. AFAIR now you have to reload both: b43 and bcma. I hope this
> patch will let you avoid reloading at least bcma.

I tried exactly that yesterday, but it doesn't fix the issue. I just
checked to see if it allows reloading only b43, and it does not. I still
have to reload bcma to get the wireless working again.

Seth

Re: BCM4331 fails to associate after suspend

[Seth Forshee]

On Tue, Feb 07, 2012 at 09:29:21AM -0600, Seth Forshee wrote:
> On Tue, Feb 07, 2012 at 08:08:11AM +0100, Rafał Miłecki wrote:
> > W dniu 7 lutego 2012 08:04 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
> > > 2012/2/6 Seth Forshee <seth.forshee@canonical.com>:
> > >> On Mon, Feb 06, 2012 at 08:13:11PM +0100, Hauke Mehrtens wrote:
> > >>> According to the Broadcom SDK for SoCs bcma_pcicore_serdes_workaround()
> > >>> should be called when coming out of standby/hibernate.
> > >>> The code wl uses to do the same thing as bcma does is open source. You
> > >>> find it in the GPL package of many Broadcom based Wifi Routers.
> > >>
> > >> Thanks for the suggestion. Unfortunately calling
> > >> bcma_pcicore_serdes_workaround() during resume doesn't help. I'll look
> > >> at the wl code to see if I can learn anything there.
> > >
> > > Can you try
> > >
> > >
> > > diff --git a/drivers/bcma/main.c b/drivers/bcma/main.c
> > > index febbc0a..a097a26 100644
> > > --- a/drivers/bcma/main.c
> > > +++ b/drivers/bcma/main.c
> > > @@ -267,6 +267,13 @@ int bcma_bus_resume(struct bcma_bus *bus)
> > >                bcma_core_chipcommon_init(&bus->drv_cc);
> > >        }
> > >
> > > +       /* Init PCIE core */
> > > +       core = bcma_find_core(bus, BCMA_CORE_PCIE);
> > > +       if (core) {
> > > +               bus->drv_pci.setup_done = false;
> > > +               bcma_core_pci_init(&bus->drv_pci);
> > > +       }
> > > +
> > >        list_for_each_entry(core, &bus->cores, list) {
> > >                struct device_driver *drv = core->dev.driver;
> > >                if (drv) {
> > 
> > If this doesn't work out of box, please check if reloading b43 is
> > enough. AFAIR now you have to reload both: b43 and bcma. I hope this
> > patch will let you avoid reloading at least bcma.
> 
> I tried exactly that yesterday, but it doesn't fix the issue. I just
> checked to see if it allows reloading only b43, and it does not. I still
> have to reload bcma to get the wireless working again.

I've been pounding on this quite a bit in the past week, but I haven't
been able to get anywhere so far. Any relevant workarounds I found in wl
have no effect, and neither does any amount of reinitialization.

I'm still digging into the drivers to see if I can learn anything, but
if anyone has more suggestions I'll happily give them a try.

Seth