iio: adc: sophgo-saradc: Handle errors from optional IRQ lookup
Subsystems (2)
PATCH 1/1 iio: adc: sophgo-saradc: Handle errors from optional IRQ lookup
From: bui duc phuc <[email protected]>
platform_get_irq_optional() returns a positive IRQ number on success or a negative error code on failure. For an optional IRQ, -ENXIO indicates that no optional IRQ is available. Other errors, such as -EPROBE_DEFER and -EINVAL, should be propagated so that the caller can handle them appropriately.
Propagate negative errors other than -ENXIO.
--- drivers/iio/adc/sophgo-cv1800b-adc.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/iio/adc/sophgo-cv1800b-adc.c +2 -0
@@ -182,6 +182,8 @@ static int cv1800b_adc_probe(struct platform_device *pdev)return PTR_ERR (saradc->regs); saradc->irq =platform_get_irq_optional (pdev,0 ); +if (saradc->irq <0 && saradc->irq != -ENXIO ) +return saradc->irq;if (saradc->irq >0 ) {init_completion (&saradc->completion); ret = devm_request_irq(dev, saradc->irq,
Signature · 1 line
2.43.0
Jonathan Cameron
maintainerQuoted 7 lines · On Thu, 13 Aug 2026 14:48:10 +0700 [email protected] wrote:
From: bui duc phuc <[email protected]>
platform_get_irq_optional()returns a positive IRQ number on success or a negative error code on failure. For an optional IRQ,-ENXIOindicates that no optional IRQ is available. Other errors, such as -EPROBE_DEFER and-EINVAL, should be propagated so that the caller can handle them appropriately.
That function is very much undocumented other than not printing a message when it returns an error. However I think you analysis is correct.
I'm not going to rush this is because it is not known to have been a problem in the wild (only odd loading orders should have caused deferal).
Quoted 21 lines
Propagate negative errors other than
-ENXIO.Signed-off-by bui duc phuc <[email protected]>--- drivers/iio/adc/sophgo-cv1800b-adc.c | 2 ++ 1 file changed, 2 insertions(+)drivers/iio/adc/sophgo-cv1800b-adc.c +2
-0@@ -182,6 +182,8 @@ static int cv1800b_adc_probe(struct platform_device *pdev)return PTR_ERR (saradc->regs); saradc->irq =platform_get_irq_optional (pdev,0 ); +if (saradc->irq <0 && saradc->irq != -ENXIO ) +return saradc->irq;if (saradc->irq >0 ) {init_completion (&saradc->completion); ret = devm_request_irq(dev, saradc->irq,
Hi Jonathan,
Thank you for your feedback.
Quoted 11 lines
Quoted 6 lines
platform_get_irq_optional()returns a positive IRQ number on success or a negative error code on failure. For an optional IRQ,-ENXIOindicates that no optional IRQ is available. Other errors, such as -EPROBE_DEFER and-EINVAL, should be propagated so that the caller can handle them appropriately.That function is very much undocumented other than not printing a message when it returns an error. However I think you analysis is correct.
Yes, I agree. The documentation for this function may not be clear enough, which could have led to error handling being implemented incorrectly in some places.
There is also an inconsistency in this driver: if devm_request_irq() fails, the error is returned and the probe fails. But if platform_get_irq_optional() fails, the error is ignored.
I'm not sure whether the author misunderstood and assumed that any negative return value simply means that there is no IRQ.
Quoted 4 lines
I'm not going to rush this is because it is not known to have been a problem in the wild (only odd loading orders should have caused deferal).
I understand your point. However, in this case the error can be completely hidden:
There is no error message or log, the error is not returned, and the driver falls back
to polling:
-----------------------------------
if (saradc->irq < 0) {
u32 reg;
return readl_poll_timeout(saradc->regs + CV1800B_ADC_STATUS_REG,
reg, !(reg & CV1800B_ADC_BUSY),
500, CV1800B_READ_TIMEOUT_US);
}------------------------------------
So there may never be an obvious failure for a user to report.
Best regards,
Phuc
Jonathan Cameron
maintainerQuoted 20 lines · On Sun, 23 Aug 2026 05:27:21 +0700 Bui Duc Phuc <[email protected]> wrote:
Hi Jonathan,
Thank you for your feedback.
Quoted 11 lines
Quoted 6 lines
platform_get_irq_optional()returns a positive IRQ number on success or a negative error code on failure. For an optional IRQ,-ENXIOindicates that no optional IRQ is available. Other errors, such as -EPROBE_DEFER and-EINVAL, should be propagated so that the caller can handle them appropriately.That function is very much undocumented other than not printing a message when it returns an error. However I think you analysis is correct.
Yes, I agree. The documentation for this function may not be clear enough, which could have led to error handling being implemented incorrectly in some places.
Quoted 5 lines
There is also an inconsistency in this driver: if
devm_request_irq()fails, the error is returned and the probe fails. But ifplatform_get_irq_optional()fails, the error is ignored.
I'm not sure whether the author misunderstood and assumed that any negative return value simply means that there is no IRQ.
If the platform irq get fails because there isn't one in firmware, we expect to just carry on (no interrupt support). What we are missing is failing when there is one but we get an error anyway.
For the later devm_request_irq() that is only called if we have an irq from firmware, but something else goes wrong. That one should definitely always fail probe as it indicates a probe (rather than lack of interrupt support)
Quoted 19 lines
Quoted 5 lines
I'm not going to rush this is because it is not known to have been a problem in the wild (only odd loading orders should have caused deferal).
I understand your point. However, in this case the error can be completely hidden:
There is no error message or log, the error is not returned, and the driver falls back
to polling:
-----------------------------------if (saradc->irq < 0) { u32 reg; return readl_poll_timeout(saradc->regs + CV1800B_ADC_STATUS_REG, reg, !(reg & CV1800B_ADC_BUSY), 500, CV1800B_READ_TIMEOUT_US); }
Quoted 3 lines
------------------------------------
So there may never be an obvious failure for a user to report.
Without evidence that there are real setups where interrupt controller loads late enough to result in a deferral + are used with this chip I'm fine with the small risk of just using polling in kernels prior to the fix. Everything still works, just potentially less efficiently.
Jonathan
Quoted 3 lines
Best regards,
Phuc