On Sun, May 7, 2017 at 5:21 AM, Jonathan Cameron <jic23@xxxxxxxxxx> wrote:
> On 05/05/17 01:32, Matt Ranostay wrote:
>> Using iio_trigger_poll() can oops when multiple interrupts
>> happen before the first is handled.
>>
> Ah. I see. This should cause all sorts of trouble and warnings.
>
> It's calling the irq_chip infrastructure without going through the
> dance necessary to make that happen from an irq context.
>
> This is indeed the simplest fix.
Yeah only saw this because of a lightning storm last week (which never
happens here :)).
- Matt
>
> Jonathan
>> Use iio_trigger_poll_chained() instead and use the timestamp
>> when processed, since it will be in theory be 2 ms max latency.
>>
>> Fixes: 24ddb0e4bba4 ("iio: Add AS3935 lightning sensor support")
>> Cc: stable@xxxxxxxxxxxxxxx
>> Signed-off-by: Matt Ranostay <matt.ranostay@xxxxxxxxxxxx>
>> ---
>> drivers/iio/proximity/as3935.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/iio/proximity/as3935.c b/drivers/iio/proximity/as3935.c
>> index ddf9bee89f77..d98d38227f26 100644
>> --- a/drivers/iio/proximity/as3935.c
>> +++ b/drivers/iio/proximity/as3935.c
>> @@ -215,7 +215,7 @@ static irqreturn_t as3935_trigger_handler(int irq, void *private)
>>
>> st->buffer[0] = val & AS3935_DATA_MASK;
>> iio_push_to_buffers_with_timestamp(indio_dev, &st->buffer,
>> - pf->timestamp);
>> + iio_get_time_ns(indio_dev));
>> err_read:
>> iio_trigger_notify_done(indio_dev->trig);
>>
>> @@ -244,7 +244,7 @@ static void as3935_event_work(struct work_struct *work)
>>
>> switch (val) {
>> case AS3935_EVENT_INT:
>> - iio_trigger_poll(st->trig);
>> + iio_trigger_poll_chained(st->trig);
>> break;
>> case AS3935_NOISE_INT:
>> dev_warn(&st->spi->dev, "noise level is too high\n");
>>
>