Re: [PATCH 04/19] v4l: omap3isp: fix handling platform_get_irq result

From: Laurent Pinchart
Date: Tue Nov 10 2015 - 03:52:57 EST


Hi Andrzej,

On Tuesday 10 November 2015 07:48:54 Andrzej Hajda wrote:
> On 11/09/2015 09:16 PM, Laurent Pinchart wrote:
> > On Thursday 24 September 2015 16:00:12 Andrzej Hajda wrote:
> >> The function can return negative value.
> >>
> >> The problem has been detected using proposed semantic patch
> >> scripts/coccinelle/tests/assign_signed_to_unsigned.cocci [1].
> >>
> >> [1]: http://permalink.gmane.org/gmane.linux.kernel/2046107
> >>
> >> Signed-off-by: Andrzej Hajda <a.hajda@xxxxxxxxxxx>
> >> ---
> >> Hi,
> >>
> >> To avoid problems with too many mail recipients I have sent whole
> >> patchset only to LKML. Anyway patches have no dependencies.
> >>
> >> drivers/media/platform/omap3isp/isp.c | 5 +++--
> >> 1 file changed, 3 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/media/platform/omap3isp/isp.c
> >> b/drivers/media/platform/omap3isp/isp.c index 56e683b..df9d2c2 100644
> >> --- a/drivers/media/platform/omap3isp/isp.c
> >> +++ b/drivers/media/platform/omap3isp/isp.c
> >> @@ -2442,12 +2442,13 @@ static int isp_probe(struct platform_device
> >> *pdev)
> >> }
> >>
> >> /* Interrupt */
> >> - isp->irq_num = platform_get_irq(pdev, 0);
> >> - if (isp->irq_num <= 0) {
> >> + ret = platform_get_irq(pdev, 0);
> >> + if (ret <= 0) {
> >
> > Looking at platform_get_irq() all error values are negative. You could
> > just test for ret < 0 here, and remove the ret = -ENODEV assignment below
> > to keep the error code returned by platform_get_irq().
> >
> > If you're fine with that modification there's no need to resubmit, just
> > let me know and I'll fix it when applying it to my tree.
>
> I left it as before, as it was not related to the patch. Additionally I have
> lurked little bit inside platform_get_irq and it looks little bit scary to
> me: platform_get_irq returns value of of_irq_get if ret >= 0,
> of_irq_get calls of_irq_parse_one which can return 0,
> in such case irq_create_of_mapping value is returned which is unsigned int
> and evaluates to 0 in case of failures.
> I am not sure if above scenario can ever occur, but the code looks so messy
> to me, that I gave up :)
>
> Anyway if you are sure about your change I am OK with it also :)

You're right, that's indeed an issue. It looks like a 0 irq is valid or
invalid depending on who you ask. NO_IRQ is defined differently depending on
the architecture :-/ I'll thus keep your version of the patch.

Nonetheless, the core issue should be fixed. Do you feel adventurous ? :-)

> >> dev_err(isp->dev, "No IRQ resource\n");
> >> ret = -ENODEV;
> >> goto error_iommu;
> >>
> >> }
> >>
> >> + isp->irq_num = ret;
> >>
> >> if (devm_request_irq(isp->dev, isp->irq_num, isp_isr, IRQF_SHARED,
> >>
> >> "OMAP3 ISP", isp)) {

--
Regards,

Laurent Pinchart

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/