| Message ID | 20170703130926.32315-1-quentin.schulz@free-electrons.com |
|---|---|
| State | Accepted |
| Commit | be2ea53330f1865ac5b00377f2074f8e255bf74c |
| Headers | show |
On Mon, Jul 03, 2017 at 03:09:26PM +0200, Quentin Schulz wrote: > When initializing interrupts, the devm_request_any_context_irq will > enable them right away. An atomic flag was set in sun4i_irq_init and read > in the interrupt handler to make sure no unwanted interrupts were > handled. If an unwanted interrupt occurred, the handler would disable > the irq and return IRQ_HANDLED. However, at the end of sun4i_irq_init, > the irq would be disabled as well, resulting in an unbalanced enable > (since there are more disables than enables, the code enabling the > interrupt would never be called). > > When reading the ADC or the temperature, the respective irq would be > enabled in the read function and disabled in the irq handler. In the > read function, we would wait for a completion (with a timeout) that will > be set in the irq handler. However, if the completion is never set or if > the wait for completion times out, the irq would not be disabled in the > read function resulting in an unbalanced enable once the read function > is called again (since there are 2+ enables for no disable). > > Moving disable_irq from the irq handler to the read function get rid of > these two cases of unbalanced enable. > > Fixes: d1caa9905538 ("iio: adc: add support for Allwinner SoCs ADC") > > Reported-by: Andreas Färber <afaerber@suse.de> > Signed-off-by: Quentin Schulz <quentin.schulz@free-electrons.com> Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com> Thanks! Maxime -- Maxime Ripard, Free Electrons Embedded Linux and Kernel engineering http://free-electrons.com
On Mon, 3 Jul 2017 23:19:45 +0200 Maxime Ripard <maxime.ripard@free-electrons.com> wrote: > On Mon, Jul 03, 2017 at 03:09:26PM +0200, Quentin Schulz wrote: > > When initializing interrupts, the devm_request_any_context_irq will > > enable them right away. An atomic flag was set in sun4i_irq_init and read > > in the interrupt handler to make sure no unwanted interrupts were > > handled. If an unwanted interrupt occurred, the handler would disable > > the irq and return IRQ_HANDLED. However, at the end of sun4i_irq_init, > > the irq would be disabled as well, resulting in an unbalanced enable > > (since there are more disables than enables, the code enabling the > > interrupt would never be called). > > > > When reading the ADC or the temperature, the respective irq would be > > enabled in the read function and disabled in the irq handler. In the > > read function, we would wait for a completion (with a timeout) that will > > be set in the irq handler. However, if the completion is never set or if > > the wait for completion times out, the irq would not be disabled in the > > read function resulting in an unbalanced enable once the read function > > is called again (since there are 2+ enables for no disable). > > > > Moving disable_irq from the irq handler to the read function get rid of > > these two cases of unbalanced enable. > > > > Fixes: d1caa9905538 ("iio: adc: add support for Allwinner SoCs ADC") > > > > Reported-by: Andreas Färber <afaerber@suse.de> > > Signed-off-by: Quentin Schulz <quentin.schulz@free-electrons.com> > > Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com> Nice detailed explanation - thanks. Applied to the fixes-togreg branch of iio.git. Jonathan > > Thanks! > Maxime >
diff --git a/drivers/iio/adc/sun4i-gpadc-iio.c b/drivers/iio/adc/sun4i-gpadc-iio.c index 81d4c39e414a..137f577d9432 100644 --- a/drivers/iio/adc/sun4i-gpadc-iio.c +++ b/drivers/iio/adc/sun4i-gpadc-iio.c @@ -256,6 +256,7 @@ static int sun4i_gpadc_read(struct iio_dev *indio_dev, int channel, int *val, err: pm_runtime_put_autosuspend(indio_dev->dev.parent); + disable_irq(irq); mutex_unlock(&info->mutex); return ret; @@ -365,7 +366,6 @@ static irqreturn_t sun4i_gpadc_temp_data_irq_handler(int irq, void *dev_id) complete(&info->completion); out: - disable_irq_nosync(info->temp_data_irq); return IRQ_HANDLED; } @@ -380,7 +380,6 @@ static irqreturn_t sun4i_gpadc_fifo_data_irq_handler(int irq, void *dev_id) complete(&info->completion); out: - disable_irq_nosync(info->fifo_data_irq); return IRQ_HANDLED; }
When initializing interrupts, the devm_request_any_context_irq will enable them right away. An atomic flag was set in sun4i_irq_init and read in the interrupt handler to make sure no unwanted interrupts were handled. If an unwanted interrupt occurred, the handler would disable the irq and return IRQ_HANDLED. However, at the end of sun4i_irq_init, the irq would be disabled as well, resulting in an unbalanced enable (since there are more disables than enables, the code enabling the interrupt would never be called). When reading the ADC or the temperature, the respective irq would be enabled in the read function and disabled in the irq handler. In the read function, we would wait for a completion (with a timeout) that will be set in the irq handler. However, if the completion is never set or if the wait for completion times out, the irq would not be disabled in the read function resulting in an unbalanced enable once the read function is called again (since there are 2+ enables for no disable). Moving disable_irq from the irq handler to the read function get rid of these two cases of unbalanced enable. Fixes: d1caa9905538 ("iio: adc: add support for Allwinner SoCs ADC") Reported-by: Andreas Färber <afaerber@suse.de> Signed-off-by: Quentin Schulz <quentin.schulz@free-electrons.com> --- drivers/iio/adc/sun4i-gpadc-iio.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) -- 2.11.0