Skip to contentSkip to conversation

iio: adc: ad7779: Initialize completion before requesting IRQ

PATCH In review Moksh Panicker 3 participants 7 messages Jun 27 – Jun 28

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iio: adc: ad7779: Initialize completion before requesting IRQ

init_completion() is called after devm_request_irq() in ad7779_setup_trigger(). If the IRQ fires before init_completion() runs, the completion is in an undefined state.

Move init_completion() before devm_request_irq() to ensure the completion is ready before the IRQ handler can signal it.

Fixes c9a3f8c7bfcb ("drivers: iio: adc: add support for ad777x family")
---
 drivers/iio/adc/ad7779.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/iio/adc/ad7779.c +2 -2
index 695cc79e78da..db8f5f4c6d6a 100644
@@ -838,6 +838,8 @@ static int ad7779_setup_without_backend(struct ad7779_state *st, struct iio_dev
 	st->trig->ops = &ad7779_trigger_ops;
 
 	iio_trigger_set_drvdata(st->trig, st);
+	init_completion(&st->completion);
+
 
 	ret = devm_request_irq(dev, st->spi->irq, iio_trigger_generic_data_rdy_poll,
 			       IRQF_NO_THREAD | IRQF_NO_AUTOEN, indio_dev->name,
@@ -852,8 +854,6 @@ static int ad7779_setup_without_backend(struct ad7779_state *st, struct iio_dev
 
 	indio_dev->trig = iio_trigger_get(st->trig);
 
-	init_completion(&st->completion);
-
 	ret = devm_iio_triggered_buffer_setup(dev, indio_dev,
 					      &iio_pollfunc_store_time,
 					      &ad7779_trigger_handler,
Signature · 1 line
2.34.1
#1 lore raw
Quoted 3 lines · On Sat, 27 Jun 2026 11:22:05 +0000 Moksh Panicker <[email protected]>…

init_completion() is called after devm_request_irq() in ad7779_setup_trigger(). If the IRQ fires before init_completion() runs, the completion is in an undefined state.

This is probably impossible, as devm_request_irq() is called with IRQF_NO_AUTOEN as a parameter, therefore IRQs are enabled manually in the code.

Quoted 15 lines

Move init_completion() before devm_request_irq() to ensure the completion is ready before the IRQ handler can signal it.

Fixes c9a3f8c7bfcb ("drivers: iio: adc: add support for ad777x family")
---
 drivers/iio/adc/ad7779.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/iio/adc/ad7779.c +0 -0
index 695cc79e78da..db8f5f4c6d6a 100644

drivers/iio/adc/ad7779.c static int ad7779_setup_without_backend(struct ad7779_state *st, struct iio_dev · line 841

Quoted 4 lines
drivers/iio/adc/ad7779.c +1 -0
@@ -838,6 +838,8 @@ static int ad7779_setup_without_backend(struct ad7779_state *st, struct iio_dev
 	st->trig->ops = &ad7779_trigger_ops;
 
 	iio_trigger_set_drvdata(st->trig, st);
+	init_completion(&st->completion);

Is there really any point in having this when st->completion isn't used anywhere in the driver? I'd rather be for removing this function call and the struct completion from struct ad7779_state.

drivers/iio/adc/ad7779.c static int ad7779_setup_without_backend(struct ad7779_state *st, struct iio_dev · line 845

Quoted 4 lines
drivers/iio/adc/ad7779.c +1 -0
+
 
 	ret = devm_request_irq(dev, st->spi->irq, iio_trigger_generic_data_rdy_poll,
 			       IRQF_NO_THREAD | IRQF_NO_AUTOEN, indio_dev->name,

Unrelated, however the dev_err_probe() call below this is unnecessary, as devm_request_irq already returns a dev_err_probe on failure. Consider sending a patch for this as well.

Signature · 3 lines
Kind regards

CJD
#2 lore raw

PATCH 1/2 iio: adc: ad7779: Remove unused completion field

struct ad7779_state contains a completion field that is initialized in ad7779_setup_without_backend() but never waited on or signaled anywhere in the driver. Remove the dead code.

---
 drivers/iio/adc/ad7779.c | 2 --
 1 file changed, 2 deletions(-)
drivers/iio/adc/ad7779.c +0 -2
index 695cc79e78da..c1a99b4c1256 100644
@@ -143,7 +143,6 @@ struct ad7779_state {
 	const struct ad7779_chip_info *chip_info;
 	struct clk *mclk;
 	struct iio_trigger *trig;
-	struct completion completion;
 	unsigned int sampling_freq;
 	enum ad7779_filter filter_enabled;
 	struct iio_backend *back;
@@ -852,7 +851,6 @@ static int ad7779_setup_without_backend(struct ad7779_state *st, struct iio_dev
 
 	indio_dev->trig = iio_trigger_get(st->trig);
 
-	init_completion(&st->completion);
 
 	ret = devm_iio_triggered_buffer_setup(dev, indio_dev,
 					      &iio_pollfunc_store_time,
Signature · 1 line
2.34.1
#3 lore raw

PATCH 2/2 iio: adc: ad7779: Remove redundant dev_err_probe() after devm_request_irq()

devm_request_irq() already prints an error message on failure via the IRQ core, so wrapping its return value in dev_err_probe() results in a duplicate error message. Return the error directly instead.

---
 drivers/iio/adc/ad7779.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/iio/adc/ad7779.c +1 -2
index c1a99b4c1256..a5b8bfa99a74 100644
@@ -842,8 +842,7 @@ static int ad7779_setup_without_backend(struct ad7779_state *st, struct iio_dev
 			       IRQF_NO_THREAD | IRQF_NO_AUTOEN, indio_dev->name,
 			       st->trig);
 	if (ret)
-		return dev_err_probe(dev, ret, "request IRQ %d failed\n",
-				     st->spi->irq);
+		return ret;
 
 	ret = devm_iio_trigger_register(dev, st->trig);
 	if (ret)
Signature · 1 line
2.34.1
Reviewed-by on v1 2/2
#4 lore raw

Re: [PATCH 2/2] iio: adc: ad7779: Remove redundant dev_err_probe() after devm_request_irq()

Quoted 7 lines · On Sun, 28 Jun 2026 07:14:05 +0000 Moksh Panicker <[email protected]>…

devm_request_irq() already prints an error message on failure via the IRQ core, so wrapping its return value in dev_err_probe() results in a duplicate error message. Return the error directly instead.

---

Signature · 3 lines
Kind regards

CJD
Reviewed-by on v1 1/2
#5 lore raw

Re: [PATCH 1/2] iio: adc: ad7779: Remove unused completion field

In reply to Moksh Panicker
Quoted 7 lines · On Sun, 28 Jun 2026 07:14:04 +0000 Moksh Panicker <[email protected]>…

struct ad7779_state contains a completion field that is initialized in ad7779_setup_without_backend() but never waited on or signaled anywhere in the driver. Remove the dead code.

---

If you're sending a patch series, please use a cover letter (if you're using git send-email, add a cover letter with the --cover-letter flag) where you can describe your changes in detail, along with additional useful info (testing etc.)

Nevertheless,

Signature · 3 lines
Kind regards

CJD
reviewer
#6 lore raw

Re: [PATCH 1/2] iio: adc: ad7779: Remove unused completion field

Quoted 7 lines · Joshua Crofts
Quoted 4 lines · On Sun, 28 Jun 2026 07:14:04 +0000 Moksh Panicker <[email protected]>…

struct ad7779_state contains a completion field that is initialized in ad7779_setup_without_backend() but never waited on or signaled anywhere in the driver. Remove the dead code.

Quoted 8 lines
Quoted 3 lines

---

If you're sending a patch series, please use a cover letter (if you're using git send-email, add a cover letter with the --cover-letter flag) where you can describe your changes in detail, along with additional useful info (testing etc.)

Also, please don't send new series or revisions of a series in reply to other series or previous revisions. It breaks existing workflows and tooling.

Quoted 5 lines

Nevertheless,