[PATCH] ARM: imx: introduce function imx_free_mx3_camera

[Emil Goode]

Hello Uwe,

On Thu, May 22, 2014 at 08:10:24PM +0200, Uwe Kleine-König wrote:
> Hello Emil,
> 
> On Thu, May 22, 2014 at 07:51:19PM +0200, Emil Goode wrote:
> > We forgot to free pdev->dev.dma_mask on error after
> > having called the imx_alloc_mx3_camera function.
> > This patch introduces the imx_free_mx3_camera function
> > that adds the missing kfree call and is practical for
> > future usage with imx_alloc_mx3_camera().
> > 
> > Signed-off-by: Emil Goode <emilgoode at gmail.com>
> > ---
> >  arch/arm/mach-imx/devices-imx31.h             |    2 ++
> >  arch/arm/mach-imx/devices-imx35.h             |    2 ++
> >  arch/arm/mach-imx/devices/devices-common.h    |    1 +
> >  arch/arm/mach-imx/devices/platform-ipu-core.c |   11 +++++++++--
> >  arch/arm/mach-imx/mach-mx31_3ds.c             |    2 +-
> >  arch/arm/mach-imx/mach-mx31moboard.c          |    3 +--
> >  arch/arm/mach-imx/mach-mx35_3ds.c             |    2 +-
> >  arch/arm/mach-imx/mach-pcm037.c               |    2 +-
> >  8 files changed, 18 insertions(+), 7 deletions(-)
> > 
> > diff --git a/arch/arm/mach-imx/devices-imx31.h b/arch/arm/mach-imx/devices-imx31.h
> > index e8d1611..900d3b0 100644
> > --- a/arch/arm/mach-imx/devices-imx31.h
> > +++ b/arch/arm/mach-imx/devices-imx31.h
> > @@ -45,6 +45,8 @@ extern const struct imx_ipu_core_data imx31_ipu_core_data;
> >  	imx_add_ipu_core(&imx31_ipu_core_data)
> >  #define imx31_alloc_mx3_camera(pdata)	\
> >  	imx_alloc_mx3_camera(&imx31_ipu_core_data, pdata)
> > +#define imx31_free_mx3_camera(pdev)	\
> > +	imx_free_mx3_camera(pdev)
> I wouldn't make this a globally visible function. Today all imx machines
> should get their devices from an oftree, so the various functions to add
> devices started to bitrot. Moreover there is no reason to remove a
> device once it was successfully added.

Ok I see. In mx31_3ds_init_camera() there are two checks that could fail
before the device is added though.

> Note that platform_device_register_full has the same problem (i.e.
> pdev->dev.dma_mask isn't freed when the last reference to a device is
> dropped.) You'd do a better deed if you picked up
> http://thread.gmane.org/gmane.linux.kernel/1613364/focus=1635995
> instead of fixing dead code. But feel free to choose yourself where you
> want to patch.

Thank you for the hint, I was about to leave this code alone and move on
but couldn't resist one more patch :) Yes I realized that leaking dma_mask
is a general problem, I will look into that thread, thanks.

Best regards,

Emil Goode