Re: [PATCH] gpio: pcf857x: restore the initial line state of all pcf lines

[Kishon Vijay Abraham I <kishon@ti.com>]

Hi,

On Wednesday 14 January 2015 05:28 PM, Linus Walleij wrote:
> On Mon, Jan 5, 2015 at 7:26 AM, Kishon Vijay Abraham I <kishon@ti.com> wrote:
>> On Thursday 18 December 2014 07:41 PM, Nishanth Menon wrote:
>>> On 12/18/2014 12:18 AM, Kishon Vijay Abraham I wrote:
>>>>
>>>>
>>>> On Tuesday 16 December 2014 02:20 AM, Nishanth Menon wrote:
>>>>> On 12/12/2014 02:06 AM, Kishon Vijay Abraham I wrote:
>>>>>> The reset values for all the PCF lines are high and hence on shutdown
>>>>>> we should drive all the lines high in order to bring it to the reset state.
>>>>>>
>>>>>> This is actually required since pcf doesn't have a reset line and even after
>>>>>> warm reset (by invoking "reboot" in prompt) the pcf lines maintains it's
>>>>>> previous programmed state. This becomes a problem if the boards are designed
>>>>>> to work with the default initial state.
>>>>>>
>>>>>> DRA7XX_evm uses PCF8575 and one of the PCF output lines feeds to MMC/SD and
>>>>>> this line should be driven high in order for the MMC/SD to be detected.
>>>>>> This line is modelled as regulator and the hsmmc driver takes care of enabling
>>>>>> and disabling it. In the case of 'reboot', during shutdown path as part of it's
>>>>>> cleanup process the hsmmc driver disables this regulator. This makes MMC boot
>>>>>> not functional.
>>>>>>
>>>>>> Fixed it by driving high all the pcf lines.
>>>>>>
>>>>>> Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
>>>>>> ---
>>>>>>   drivers/gpio/gpio-pcf857x.c |    9 +++++++++
>>>>>>   1 file changed, 9 insertions(+)
>>>>>>
>>>>>> diff --git a/drivers/gpio/gpio-pcf857x.c b/drivers/gpio/gpio-pcf857x.c
>>>>>> index 236708a..00b15b2 100644
>>>>>> --- a/drivers/gpio/gpio-pcf857x.c
>>>>>> +++ b/drivers/gpio/gpio-pcf857x.c
>>>>>> @@ -448,6 +448,14 @@ static int pcf857x_remove(struct i2c_client *client)
>>>>>>     return status;
>>>>>>   }
>>>>>>
>>>>>> +static void pcf857x_shutdown(struct i2c_client *client)
>>>>>> +{
>>>>>> +  struct pcf857x *gpio = i2c_get_clientdata(client);
>>>>>> +
>>>>>> +  /* Drive all the I/O lines high */
>>>>>> +  gpio->write(gpio->client, BIT(gpio->chip.ngpio) - 1);
>>>>>
>>>>> you might force a contention here - depending on System configuration.
>>>>> example:
>>>>> +-------+
>>>>> |       |
>>>>> |  U1   |         +------+      +-----------+
>>>>> |       +--------->      |      |           |
>>>>> +-------+         |      |      |           |
>>>>>                    | Switch<-----+    SoC    |
>>>>> +-------+         |      |      |           |
>>>>> |       |         |      |      |           |
>>>>> | U2    <---------+--^---+      +-----------+
>>>>> |       |            |
>>>>> |       |            |
>>>>> +-------+            |
>>>>>                    +--+--+
>>>>>                    |     |
>>>>>                    | PCF |
>>>>>                    |     |
>>>>>                    +-----+
>>>>>
>>>>> At low, SoC pin is connected to U2 as drive. when reset to high, you
>>>>> now have U1 driving to the same pin that SoC has, potentially
>>>>> resulting in contention.
>>>>>
>>>>>
>>>>> Unfortunately, at this level, you do not know what the state of the
>>>>> system is, blindly forcing a pin level will potentially cause
>>>>> contention risk depending on pin configuration.
>>>>
>>>> Assume we are doing a reset when the system is powered on, irrespective of the
>>>> state of the system, we'll be forcing the pin level to the default state.
>>>
>>> Yes, I dont deny that system will be fine *after* reset sequence is
>>> started or completed. However there is a duration between the pcf
>>> shutdown handler is called and the final reset handler is invoked -
>>> that is the duration when  the contention might cause device behavior.
>>> Essentially ignoring the state various drivers have asked PCF to setup
>>> the pins and doing a hands down configuration may have side effects we
>>> cant properly expect.
>>
>> The solution might be to invoke the shutdown handler of the various drivers
>> using the PCF before the shutdown handler of the PCF driver itself gets
>> invoked? But I'm not sure how can that be achieved in linux kernel :-(
>
> #include <linux/reboot.h>
>
> static int foo_reboot_handler(struct notifier_block *this,
>                                  unsigned long code,
>                                  void *unused)
> {
>          pr_crit("do some last minute stuff\n");
>          return NOTIFY_OK;
> }
>
> static struct notifier_block foo_reboot_notifier = {
>          .notifier_call = foo_reboot_handler,
> };
>
> register_reboot_notifier(&foo_reboot_notifier);

Added debug prints and found the reboot notifier gets invoked before the
shutdown handler which means some stuff can be done after this reboot
notifier:-(

Thanks
Kishon