Skip to contentSkip to conversation

spi: gxp: removed unneeded call to platform_set_drvdata()

PATCH In review Andrei Coardos 3 participants 5 messages Aug 7, 2023

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 spi: gxp: removed unneeded call to platform_set_drvdata()

This function call was found to be unnecessary as there is no equivalent platform_get_drvdata() call to access the private data of the driver. Also, the private data is defined in this driver, so there is no risk of it being accessed outside of this driver file.

---
 drivers/spi/spi-gxp.c | 1 -
 1 file changed, 1 deletion(-)
drivers/spi/spi-gxp.c +0 -1
index 684d63f402f3..6261d9f036bc 100644
@@ -264,7 +264,6 @@ static int gxp_spifi_probe(struct platform_device *pdev)
 
 	spifi = spi_controller_get_devdata(ctlr);
 
-	platform_set_drvdata(pdev, spifi);
 	spifi->data = data;
 	spifi->dev = dev;
 
Signature · 1 line
2.34.1
maintainer
#1 lore raw
Quoted 4 lines · Andrei Coardos

This function call was found to be unnecessary as there is no equivalent platform_get_drvdata() call to access the private data of the driver. Also, the private data is defined in this driver, so there is no risk of it being accessed outside of this driver file.

That isn't enough of a check here - people can still reference the driver data without going through the accessor function.

#2 lore raw
Quoted 10 lines · Mark Brown
Quoted 4 lines · Andrei Coardos

This function call was found to be unnecessary as there is no equivalent platform_get_drvdata() call to access the private data of the driver. Also, the private data is defined in this driver, so there is no risk of it being accessed outside of this driver file.

That isn't enough of a check here - people can still reference the driver data without going through the accessor function.

So, is that like calling platform_get_drvdata() in a parent/chid device, to check if the driver-data is set?
Would it make sense for another driver to do that (i.e. check the driver-data is non-null, but not access the data)?
I can imagine that being possible, but it's a bit quirky.

Or, is the issue with the wording of the comment?

maintainer
#3 lore raw
Quoted 9 lines · Alexandru Ardelean
Quoted 6 lines · Mark Brown
Quoted 4 lines · Andrei Coardos

This function call was found to be unnecessary as there is no equivalent platform_get_drvdata() call to access the private data of the driver. Also, the private data is defined in this driver, so there is no risk of it being accessed outside of this driver file.

Quoted 5 lines
Quoted 2 lines

That isn't enough of a check here - people can still reference the driver data without going through the accessor function.

So, is that like calling platform_get_drvdata() in a parent/chid device, to check if the driver-data is set?

That wasn't what I was thinking of, waht I was thinking of was just open coding platform_get_drvdata() and looking directly at struct device. Another common case is where drivers that support multiple bus types will pass around the struct device and use dev_get_drvdata() to read the data rather than using platform_get_drvdata(). The driver data can be allocated and initialised with bus specific bits before being passed off to the generic code.

That said the looking at the parent's driver data is definitely a thing that happens with MFDs.

#4 lore raw
Quoted 11 lines · Mark Brown
Quoted 8 lines · Alexandru Ardelean
Quoted 6 lines · Mark Brown
Quoted 4 lines · Andrei Coardos

This function call was found to be unnecessary as there is no equivalent platform_get_drvdata() call to access the private data of the driver. Also, the private data is defined in this driver, so there is no risk of it being accessed outside of this driver file.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

That isn't enough of a check here - people can still reference the driver data without going through the accessor function.

So, is that like calling platform_get_drvdata() in a parent/chid device, to check if the driver-data is set?

That wasn't what I was thinking of, waht I was thinking of was just open coding platform_get_drvdata() and looking directly at struct device.

Ah. Right.
I hadn't thought of checking "dev->driver_data" access.

Quoted 2 lines

Another common case is where drivers that support multiple bus types will pass around the struct device and use dev_get_drvdata() to read the

Agree. I see that happening with PM routines. It doesn't look like it's the case in this driver.

Quoted 3 lines

data rather than using platform_get_drvdata(). The driver data can be allocated and initialised with bus specific bits before being passed off to the generic code.

If I'm looking more closely, I am seeing that the "platform_set_drvdata(pdev, spifi);" has no equivalent access to "pdev->dev.driver_data"
Nor by open-coding, nor by "dev_get_drvdata(&pdev->dev)" But I do see that "spi_controller_get_devdata()" is calling "dev_get_drvdata()" on a device object allocated here via "devm_spi_alloc_master()"

So, I agree. That a more thorough check is needed here.

Quoted 3 lines

That said the looking at the parent's driver data is definitely a thing that happens with MFDs.

Yep. MFDs is one case I was thinking of too (with respect to parent/child lookup of driver data).