Re: [PATCH] clocksource/drivers/davinci: fix memory leak on clockevent on error return

[Thomas Gleixner <tglx@linutronix.de>]

[-- Attachment #1: Type: text/plain, Size: 1912 bytes --]

Bartosz,

On Sun, 10 Nov 2019, Bartosz Golaszewski wrote:
> sob., 9 lis 2019 o 16:58 Colin King <colin.king@canonical.com> napisał(a):
> >
> > From: Colin Ian King <colin.king@canonical.com>
> >
> > In the case where request_irq fails, the return path does not kfree
> > clockevent and hence we have a memory leak.  Fix this by kfree'ing

s/we have/creates/  or whatever verb you prefer.

> > clockevent before returning.
> >
> > Addresses-Coverity: ("Resource leak")
> > Fixes: 721154f972aa ("clocksource/drivers/davinci: Add support for clockevents")
> > Signed-off-by: Colin Ian King <colin.king@canonical.com>
> > ---
> >  drivers/clocksource/timer-davinci.c | 1 +
> >  1 file changed, 1 insertion(+)
> >
> > diff --git a/drivers/clocksource/timer-davinci.c b/drivers/clocksource/timer-davinci.c
> > index 62745c962049..910d4d2f0d64 100644
> > --- a/drivers/clocksource/timer-davinci.c
> > +++ b/drivers/clocksource/timer-davinci.c
> > @@ -299,6 +299,7 @@ int __init davinci_timer_register(struct clk *clk,
> >                          "clockevent/tim12", clockevent);
> >         if (rv) {
> >                 pr_err("Unable to request the clockevent interrupt");
> > +               kfree(clockevent);
> >                 return rv;
> >         }
> >
> > --
> > 2.20.1
> >
> 
> Hi Daniel,
> 
> this is what I think the third time someone tries to "fix" this
> driver's "memory leaks". I'm not sure what the general approach in
> clocksource is but it doesn't make sense to free resources on
> non-recoverable errors, does it? Should I add a comment about it or
> you'll just take those "fixes" to stop further such submissions?

There are two ways to deal with that:

  1) If the error is really unrecoverable, panic right there. No point
     to continue.

  2) If there is even a minimal chance to survive, free the memory and
     return.

Adding a comment is just a useless non-option.

Thanks,

	tglx