Re: [PATCH v2] serdev: Replace serdev_device_write_buf with serdev_device_write

[Andrey Smirnov]

On Tue, Mar 28, 2017 at 10:07 AM, Andy Shevchenko
<andy.shevchenko@xxxxxxxxx> wrote:
> On Tue, Mar 28, 2017 at 7:01 PM, Andrey Smirnov
> <andrew.smirnov@xxxxxxxxx> wrote:
>> Convert serdev_device_write_buf's code to be able to transfer amount of
>> data potentially exceeding "write room" at the moment of invocation.
>>
>> To support that, also add serdev_device_write_wakeup.
>>
>> Drivers wanting to use full extent of serdev_device_write
>> functionality are expected to provide serdev_device_write_wakeup as a
>> sole handler of .write_wakeup event or call it as a part of driver's
>> custom .write_wakeup code.
>>
>> Drivers wanting to retain old serdev_device_write_buf behaviour can
>
>> replace those call to calls to serdev_device_write with timeout of
>> 0. Providing .write_wakeup handler in such case is optional.
>
> Some indentation would be better if, for example, 0 will be kept on
> previous line.
>

OK, sure.

> So, what I would see if no one objects is patch series of two:
> 1) introduction of new API
> 2) removing old one.
>
> It will benefit for easier review and any potential code anthropologist.
>

Second version of the patch preserves the old API an just
re-implements it in terms of a new one. I am not sure I see the
benefit in splitting it into two patches, but I'll leave it up to Rob
to decide.

>> --- a/drivers/tty/serdev/core.c
>> +++ b/drivers/tty/serdev/core.c
>> @@ -116,17 +116,41 @@ void serdev_device_close(struct serdev_device *serdev)
>>  }
>>  EXPORT_SYMBOL_GPL(serdev_device_close);
>>
>> -int serdev_device_write_buf(struct serdev_device *serdev,
>> -                           const unsigned char *buf, size_t count)
>> +void serdev_device_write_wakeup(struct serdev_device *serdev)
>> +{
>> +       complete(&serdev->write_comp);
>> +}
>> +EXPORT_SYMBOL_GPL(serdev_device_write_wakeup);
>> +
>> +int serdev_device_write(struct serdev_device *serdev,
>> +                       const unsigned char *buf, size_t count,
>> +                       unsigned long timeout)
>>  {
>>         struct serdev_controller *ctrl = serdev->ctrl;
>> +       int ret;
>>
>> -       if (!ctrl || !ctrl->ops->write_buf)
>> +       if (!ctrl || !ctrl->ops->write_buf ||
>> +           (timeout && !serdev->ops->write_wakeup))
>>                 return -EINVAL;
>>
>> -       return ctrl->ops->write_buf(ctrl, buf, count);
>> +       mutex_lock(&serdev->write_lock);
>> +       do {
>> +               reinit_completion(&serdev->write_comp);
>> +
>> +               ret = ctrl->ops->write_buf(ctrl, buf, count);
>> +               if (ret < 0)
>> +                       break;
>> +
>
>> +               buf   += ret;
>
> Extra white spaces.

Which is there on purpose to re-align "+=" with "-=" on the next line.
I'll remove it.

>
>> +               count -= ret;
>> +
>
>> +       } while (count &&
>> +                (timeout = wait_for_completion_timeout(&serdev->write_comp,
>> +                                                       timeout)));
>
> So, would it be better to support interrupts here and return a
> corresponding error code to the user?
>

I don't have a use-case for that and as far as I can tell, neither SPI
nor I2C slave device API offer such functionality universally, so I am
inclined to say no. Since the change from wait_for_completion to
wait_for_completion_timeout was made per Rob's request, I'd leave it
up to him to decided about this change as well.

> Besides that question, readability might be better if you use
> temporary variable and pack above on one line:
>
> unsigned long to = timeout;
>
> } while (count && (to = ...(to)));
>

Even if you shorten 'timeout' to 'to', formatted as a single line,
it'd still exceed line length limitations.

>
>
>> +       mutex_unlock(&serdev->write_lock);
>> +       return ret < 0 ? ret : (count ? -ETIMEDOUT : 0);
>>  }
>> -EXPORT_SYMBOL_GPL(serdev_device_write_buf);
>> +EXPORT_SYMBOL_GPL(serdev_device_write);
>
>> + * @write_comp Completion used by serdev_device_write internally
>
> Links to the functions are func()-like. Please check kernel doc howto:s.

OK, will do.

>
>> + * @write_lock Mutext used to esure exclusive access to the bus when
>> + *             writing data with serdev_device_write()
>
> checkpatch.pl has integrated spellchecker AFAIU.

My bad, forgot to enable it as a git hook, will fix.

> Moreover, can you try harder to make that description shorter?
>

I am all ears for suggestions alternative phrasing, otherwise, no,
that's about as hard as I try.

>>  void serdev_device_write_flush(struct serdev_device *);
>>  int serdev_device_write_room(struct serdev_device *);
>>
>> +
>>  /*
>>   * serdev device driver functions
>>   */
>
> This doesn't belong to the change.

Oops, didn't notice this. Will remove.

Thanks,
Andrey Smirnov