Re: [PATCH] ptp: ixp46x: Fix error handling in ptp_ixp_probe()

[Linus Walleij]

On Fri, Nov 26, 2021 at 2:40 AM tangbin <tangbin@xxxxxxxxxxxxxxxxxxxx> wrote:
> On 2021/11/26 9:26, Linus Walleij wrote:
> > On Fri, Nov 26, 2021 at 2:10 AM Tang Bin <tangbin@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> >> In the function ptp_ixp_probe(), when get irq failed
> >> after executing platform_get_irq(), the negative value
> >> returned will not be detected here. So fix error handling
> >> in this place.
> >>
> >> Fixes: 9055a2f591629 ("ixp4xx_eth: make ptp support a platform driver")
> >> Signed-off-by: Tang Bin <tangbin@xxxxxxxxxxxxxxxxxxxx>
> > OK the intention is right but:
> >
> >> -           !ixp_clock.master_irq || !ixp_clock.slave_irq)
> >> +           (ixp_clock.master_irq < 0) || (ixp_clock.slave_irq < 0))
> > Keep disallowing 0. Because that is not a valid IRQ.
> >
> > ... <= 0 ...
>
> Please look at the function platform_get_irq() in the file
> drivers/base/platform.c,
>
> the example is :
>
>      * int irq = platform_get_irq(pdev, 0);
>
>      * if (irq < 0)
>
>      *        return irq;

In this case, reading the code is a bad idea. IRQ 0 is not valid,
and the fact that platform_get_irq() can return 0 does not make
it valid.

See:
https://lwn.net/Articles/470820/

Yours,
Linus Walleij