Skip to contentSkip to conversation

staging: iio: addac: adt7316: check SPI write errors in probe

PATCH In review Hungyu Lin 3 participants 6 messages May 9 – May 10

Subsystems (4)
lore
Share & tools
lore

Hungyu Lin

author
#0 lore raw

PATCH v2 1/1 staging: iio: addac: adt7316: check SPI write errors in probe

The probe function issues three SPI writes to switch the device from the default I2C protocol to SPI protocol, but ignores their return values.

Switch the repeated SPI writes to a loop and return an error if any of them fail, as failing to switch the protocol should abort the probe.

---
Changes in v2:
- Return error instead of just warning
- Clarify that repeated SPI writes are converted to a loop

---
 drivers/staging/iio/addac/adt7316-spi.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/staging/iio/addac/adt7316-spi.c +6 -3
index f91325d11394..056a4505bf30 100644
@@ -98,6 +98,7 @@ static int adt7316_spi_probe(struct spi_device *spi_dev)
 		.multi_read = adt7316_spi_multi_read,
 		.multi_write = adt7316_spi_multi_write,
 	};
+	int i, ret;
 
 	/* don't exceed max specified SPI CLK frequency */
 	if (spi_dev->max_speed_hz > ADT7316_SPI_MAX_FREQ_HZ) {
@@ -107,9 +108,11 @@ static int adt7316_spi_probe(struct spi_device *spi_dev)
 	}
 
 	/* switch from default I2C protocol to SPI protocol */
-	adt7316_spi_write(spi_dev, 0, 0);
-	adt7316_spi_write(spi_dev, 0, 0);
-	adt7316_spi_write(spi_dev, 0, 0);
+	for (i = 0; i < 3; i++) {
+		ret = adt7316_spi_write(spi_dev, 0, 0);
+		if (ret)
+			return ret;
+	}
 
 	return adt7316_probe(&spi_dev->dev, &bus, spi_dev->modalias);
 }
Signature · 1 line
2.34.1
#1 lore raw
Quoted 9 lines · Hungyu Lin

The probe function issues three SPI writes to switch the device from the default I2C protocol to SPI protocol, but ignores their return values.

Switch the repeated SPI writes to a loop and return an error if any of them fail, as failing to switch the protocol should abort the probe.

Quoted 9 lines

---
Changes in v2:
- Return error instead of just warning
- Clarify that repeated SPI writes are converted to a loop

---
 drivers/staging/iio/addac/adt7316-spi.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

[snip]

Hmmm...has this been tested? Normally patches like these don't get merged unless tested since we need to know how the hardware would respond when we hit an error case and quit sending data, and also some of the timing changes introduced by writing, checking if its an error, then jumping back. I've personally tried this patch before and it got rejected so I'll link it below since the questions that you might get asked will likely be similar.

best regards,
max

author
#2 lore raw

Hi Nuno, Maxwell,

Thanks for the feedback.

I agree that just warning is not very useful, and we should bail out on errors if we check them. However, I currently cannot test the failure case and I'm not sure how the device behaves if the switching sequence is interrupted.

I've reached out to Analog Devices to clarify this. For now, I think it's safer not to change the behavior until we have confirmation.

I'll follow up once I have more information.

Thanks,
Hungyu

Quoted 34 lines · Maxwell Doose
Quoted 9 lines · Hungyu Lin

The probe function issues three SPI writes to switch the device from the default I2C protocol to SPI protocol, but ignores their return values.

Switch the repeated SPI writes to a loop and return an error if any of them fail, as failing to switch the protocol should abort the probe.

Quoted 9 lines

---
Changes in v2:
- Return error instead of just warning
- Clarify that repeated SPI writes are converted to a loop

---
 drivers/staging/iio/addac/adt7316-spi.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

[snip]

Hmmm...has this been tested? Normally patches like these don't get merged unless tested since we need to know how the hardware would respond when we hit an error case and quit sending data, and also some of the timing changes introduced by writing, checking if its an error, then jumping back. I've personally tried this patch before and it got rejected so I'll link it below since the questions that you might get asked will likely be similar.

best regards,
max

#3 lore raw
Quoted 10 lines · Denny Lin

Hi Nuno, Maxwell,

Thanks for the feedback.

I agree that just warning is not very useful, and we should bail out on errors if we check them. However, I currently cannot test the failure case and I'm not sure how the device behaves if the switching sequence is interrupted.

Quoted 8 lines

I've reached out to Analog Devices to clarify this. For now, I think it's safer not to change the behavior until we have confirmation.

I'll follow up once I have more information.

Thanks,
Hungyu

Two things. One, please don't top-post, and two, from what I've heard the switching sequence tends to be very error-prone since (I think) we're essentially spamming the bus with spi data while we're still in i2c mode, so we may want to ignore error codes anyways. Just a heads-up.

best regards,
max

#4 lore raw
In reply to Hungyu Lin
Quoted 7 lines · Hungyu Lin

The probe function issues three SPI writes to switch the device from the default I2C protocol to SPI protocol, but ignores their return values.

Switch the repeated SPI writes to a loop and return an error if any of them fail, as failing to switch the protocol should abort the probe.

...

drivers/staging/iio/addac/adt7316-spi.c static int adt7316_spi_probe(struct spi_device *spi_dev) · line 115

Quoted 9 lines
drivers/staging/iio/addac/adt7316-spi.c +5 -3
 	/* switch from default I2C protocol to SPI protocol */
-	adt7316_spi_write(spi_dev, 0, 0);
-	adt7316_spi_write(spi_dev, 0, 0);
-	adt7316_spi_write(spi_dev, 0, 0);
+	for (i = 0; i < 3; i++) {
+		ret = adt7316_spi_write(spi_dev, 0, 0);
+		if (ret)
+			return ret;
+	}

There was an attempt in the past to do the same and it was NAKed, so is this now. NAK.

Instead, add a better comment, if you wish, with the citing from the datasheet to explain the magic here and why we don't check for the error codes.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#5 lore raw
Quoted 1 line · Andy Shevchenko

On Sat, May 09, 2026 at 09:03:20AM +0000, Hungyu Lin wrote:

[snip]

Quoted 7 lines

There was an attempt in the past to do the same and it was NAKed, so is this now. NAK.

Instead, add a better comment, if you wish, with the citing from the datasheet to explain the magic here and why we don't check for the error codes.

Agree with Andy here on the comment, I'm personally not a fan of how undocumented this "spam data on the bus and see if it works" stuff is. Make sure to mention:

1. We're in the transitional state between I2C and SPI
2. Facts from the data sheet about why this is what we're doing
3. Why we shouldn't check the return values from the spi_write() calls

best regards,
max

Quoted 6 lines

--
With Best Regards,
Andy Shevchenko