On Mon, Jan 6, 2020 at 12:37 PM Tony Lindgren <tony@xxxxxxxxxxx> wrote:
>
> Clean-up commit 8c82723414d5 ("clocksource/drivers/timer-ti-dm: Switch to
> platform_get_irq") caused a regression where we now try to access
> uninitialized data for timer:
>
> drivers/clocksource/timer-ti-dm.c: In function 'omap_dm_timer_probe':
> drivers/clocksource/timer-ti-dm.c:798:13: warning: 'timer' may be used
> uninitialized in this function [-Wmaybe-uninitialized]
>
> On boot we now get:
>
> Unable to handle kernel NULL pointer dereference at virtual address
> 00000004
> ...
> (omap_dm_timer_probe) from [<c061ac7c>] (platform_drv_probe+0x48/0x98)
> (platform_drv_probe) from [<c0618c04>] (really_probe+0x1dc/0x348)
> (really_probe) from [<c0618ef4>] (driver_probe_device+0x5c/0x160)
>
> Let's fix the issue by moving platform_get_irq to happen after timer has
> been allocated.
>
> Fixes: 8c82723414d5 ("clocksource/drivers/timer-ti-dm: Switch to platform_get_irq")
> Cc: Yangtao Li <tiny.windzz@xxxxxxxxx>
> Signed-off-by: Tony Lindgren <tony@xxxxxxxxxxx>
Acked-by: Olof Johansson <olof@xxxxxxxxx>
> ---
>
> I did not notice simlar issue with other patches in the series, but
> please do double check Yangtao.
Yeah, this even seems to be caught at build (but our builds have been
so noisy with warnings lately that they're hard to spot):
/build/drivers/clocksource/timer-ti-dm.c: In function 'omap_dm_timer_probe':
/build/drivers/clocksource/timer-ti-dm.c:798:13: warning: 'timer' may
be used uninitialized in this function [-Wmaybe-uninitialized]
798 | timer->irq = platform_get_irq(pdev, 0);
| ~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~
-Olof
