Re: [PATCH v2] iommu/qcom: Fix local_base status check

Re: [PATCH v2] iommu/qcom: Fix local_base status check

[Markus Elfring]

> Release resources when exiting on error.

I have got doubts that such a change description fits to
the proposed source code adjustment.


…
> +++ b/drivers/iommu/qcom_iommu.c
> @@ -813,8 +813,11 @@ static int qcom_iommu_device_probe(struct platform_device *pdev)
>  	qcom_iommu->dev = dev;
>
>  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> -	if (res)
> +	if (res) {
>  		qcom_iommu->local_base = devm_ioremap_resource(dev, res);
> +		if (IS_ERR(qcom_iommu->local_base))
> +			return PTR_ERR(qcom_iommu->local_base);
> +	}
>
>  	qcom_iommu->iface_clk = devm_clk_get(dev, "iface");
…

Will the commit message be improved any further?

Would you like to add the tag “Fixes”?

Regards,
Markus

Re: [PATCH v2] iommu/qcom: Fix local_base status check

[Markus Elfring]

> > ------v2-----------------
> > As requested, add some {} around this chunk.
> >
> > ------v1-----------------
>
> The changelog typically goes after the ---, as it doesn't add value to
> the resulting git log. So please drop the above 4 lines from the commit
> message. And please use ./scripts/get_maintainer.pl to find your
> recipient list.
>
> Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>

How does this tag fit to the requested changes?

Regards,
Markus

Re: [PATCH v2] iommu/qcom: Fix local_base status check

[Markus Elfring]

> > ------v2-----------------
> > As requested, add some {} around this chunk.
> >
> > ------v1-----------------
>
> The changelog typically goes after the ---, as it doesn't add value to
> the resulting git log. So please drop the above 4 lines from the commit
> message. And please use ./scripts/get_maintainer.pl to find your
> recipient list.
>
> Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>

How does this tag fit to the requested changes?

Regards,
Markus

Re: [PATCH v2] iommu/qcom: Fix local_base status check

[Markus Elfring]

> The function qcom_iommu_device_probe() does not perform sufficient
> error checking after executing devm_ioremap_resource(), which can
> result in crashes if a critical error path is encountered.

Your update suggestion will be rechecked once more.

* Can it be that the patch would need a higher version number
  according to previous review comments?

* Would you like to adjust the patch subject?


…
> +++ b/drivers/iommu/qcom_iommu.c
> @@ -813,8 +813,11 @@ static int qcom_iommu_device_probe(struct platform_device *pdev)
>  	qcom_iommu->dev = dev;
>
>  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> -	if (res)

I find the deletion of this check appropriate.


> +	if (res) {
>  		qcom_iommu->local_base = devm_ioremap_resource(dev, res);

But I do not see a need to preserve such a check because this function
performs input parameter validation.
https://elixir.bootlin.com/linux/v5.7-rc1/source/lib/devres.c#L116
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/lib/devres.c?id=50cc09c18985eacbbd666acfd7be2391394733f5#n116

Please take another look at a corresponding usage example.

Regards,
Markus

Re: [PATCH v2] iommu/qcom: Fix local_base status check

[Markus Elfring]

> …
>> +++ b/drivers/iommu/qcom_iommu.c
>> @@ -813,8 +813,11 @@ static int qcom_iommu_device_probe(struct platform_device *pdev)
>>  	qcom_iommu->dev = dev;
>>
>>  	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
…
>>  		qcom_iommu->local_base = devm_ioremap_resource(dev, res);
…
> Please take another look at a corresponding usage example.

I would like to point another possibility out for desirable software evolution.
How do you think about to call a “known” wrapper function instead?

devm_platform_get_and_ioremap_resource
https://elixir.bootlin.com/linux/v5.7-rc1/source/drivers/base/platform.c#L66
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/base/platform.c

Regards,
Markus