Re: [RFC/RFT 2/5] b43: Load firmware from a work queue and not from the probe routine

[Larry Finger <Larry.Finger@lwfinger.net>]

On 02/13/2012 05:41 PM, Julian Calaby wrote:
> Hi Larry,
>
> On Tue, Feb 14, 2012 at 10:28, Larry Finger<Larry.Finger@lwfinger.net>  wrote:
>> On 02/13/2012 05:06 PM, Julian Calaby wrote:
>>>
>>> Hi Larry,
>>>
>>> On Tue, Feb 14, 2012 at 06:37, Larry Finger<Larry.Finger@lwfinger.net>
>>>   wrote:
>>>>
>>>> Recent changes in udev are causing problems for drivers that load
>>>> firmware
>>>> from the probe routine. As b43 has such a structure, it must be changed.
>>>> As this driver loads more than 1 firmware file, changing to the
>>>> asynchronous routine
>>>> request_firmware_nowait() would be complicated. In this implementation,
>>>> the probe
>>>> routine starts a delayed_work queue that calls the firmware loading
>>>> routines when
>>>> the delay (1 sec) expires..
>>>>
>>>> Signed-off-by: Larry Finger<Larry.Finger@lwfinger.net>
>>>> ---
>>>> diff --git a/drivers/net/wireless/b43/b43.h
>>>> b/drivers/net/wireless/b43/b43.h
>>>> index 835462d..532ba79 100644
>>>> --- a/drivers/net/wireless/b43/b43.h
>>>> +++ b/drivers/net/wireless/b43/b43.h
>>>> @@ -932,6 +932,9 @@ struct b43_wl {
>>>>         /* Flag that implement the queues stopping. */
>>>>         bool tx_queue_stopped[B43_QOS_QUEUE_NUM];
>>>>
>>>> +       /* delayed firmware loading */
>>>> +       struct delayed_work firmware_load;
>>>> +
>>>>         /* The device LEDs. */
>>>>         struct b43_leds leds;
>>>>
>>>> diff --git a/drivers/net/wireless/b43/main.c
>>>> b/drivers/net/wireless/b43/main.c
>>>> index 1b540d2..903e1ea 100644
>>>> --- a/drivers/net/wireless/b43/main.c
>>>> +++ b/drivers/net/wireless/b43/main.c
>>>> @@ -2427,9 +2433,17 @@ static int b43_request_firmware(struct b43_wldev
>>>> *dev)
>>>>         b43_print_fw_helptext(dev->wl, 1);
>>>>         err = -ENOENT;
>>>
>>>
>>> Should something be done here rather than immediately going to
>>> register the card with ieee80211?
>>
>>
>> I don't think so. When firmware is loaded from the probe routine, it
>> registers the card as the next step. I did miss the step that registers the
>> leds - that has now been added.
>
> My point here is that in the original flow, the -ENOENT would have
> been returned to b43_chip_init() and the hardware would never have
> been registered with mac80211. After this patch, it appears that if
> the firmware cannot be found, it still registers the hardware with
> mac80211.

I see what you mean. The statement before the start_ieee80211 label should be 
'goto out' rather than falling through.

Larry