Re: [PATCH] soc: renesas: rmobile-sysc: Fix some leaks in rmobile_init_pm_domains()

From: Geert Uytterhoeven
Date: Wed Sep 23 2020 - 05:10:34 EST

Hi Dan,

On Wed, Sep 23, 2020 at 10:47 AM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
> This code needs to call iounmap() on the error paths.

Thanks for your patch!

> Fixes: 2ed29e15e4b2 ("ARM: shmobile: R-Mobile: Move pm-rmobile to drivers/soc/renesas/")

This is not the commit that introduced the issue.

Fixes: 2173fc7cb681c38b ("ARM: shmobile: R-Mobile: Add DT support for
PM domains")

> Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>

> --- a/drivers/soc/renesas/rmobile-sysc.c
> +++ b/drivers/soc/renesas/rmobile-sysc.c
> @@ -328,6 +328,7 @@ static int __init rmobile_init_pm_domains(void)
> pmd = of_get_child_by_name(np, "pm-domains");
> if (!pmd) {
> pr_warn("%pOF lacks pm-domains node\n", np);
> + iounmap(base);

This one I can agree with, although that case is a bug in the DTS.

> continue;
> }
> @@ -341,6 +342,7 @@ static int __init rmobile_init_pm_domains(void)
> of_node_put(pmd);
> if (ret) {
> of_node_put(np);
> + iounmap(base);

This one I cannot: in the (unlikely, only if OOM) case
rmobile_add_pm_domains() returns an error, one or more PM subdomains may
have been registered already. Hence if you call iounmap() here, the
code will try to access unmapped registers later, leading to a crash.



Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds