Re: [PATCH] media: intel/ipu6: Fix an error handling path in isys_probe()

[Christophe JAILLET]

Le 20/05/2024 à 12:47, Sakari Ailus a écrit :
> Hi Christophe,
>
> Thanks for the patch.
>
> On Mon, May 20, 2024 at 12:32:30PM +0200, Christophe JAILLET wrote:
> > If an error occurs after a successful alloc_fw_msg_bufs() call, some
> > resources should be released as already done in the remove function.
> >
> > Fixes: f50c4ca0a820 ("media: intel/ipu6: add the main input system driver")
> > Signed-off-by: Christophe JAILLET <christophe.jaillet@xxxxxxxxxx>
> > ---
> > Compile tested only
> > ---
> >   drivers/media/pci/intel/ipu6/ipu6-isys.c | 8 +++++++-
> >   1 file changed, 7 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/media/pci/intel/ipu6/ipu6-isys.c b/drivers/media/pci/intel/ipu6/ipu6-isys.c
> > index 5992138c7290..d9e1e1a135b9 100644
> > --- a/drivers/media/pci/intel/ipu6/ipu6-isys.c
> > +++ b/drivers/media/pci/intel/ipu6/ipu6-isys.c
> > @@ -1062,6 +1062,7 @@ static int isys_probe(struct auxiliary_device *auxdev,
> >   	const struct ipu6_isys_internal_csi2_pdata *csi2_pdata;
> >   	struct ipu6_bus_device *adev = auxdev_to_adev(auxdev);
> >   	struct ipu6_device *isp = adev->isp;
> > +	struct isys_fw_msgs *fwmsg, *safe;
> >   	const struct firmware *fw;
> >   	struct ipu6_isys *isys;
> >   	unsigned int i;
> > @@ -1140,12 +1141,17 @@ static int isys_probe(struct auxiliary_device *auxdev,
> >   
> >   	ret = isys_register_devices(isys);
> >   	if (ret)
> > -		goto out_remove_pkg_dir_shared_buffer;
> > +		goto free_fw_msg_bufs;
> >   
> >   	ipu6_mmu_hw_cleanup(adev->mmu);
> >   
> >   	return 0;
> >   
> > +free_fw_msg_bufs:
> > +	list_for_each_entry_safe(fwmsg, safe, &isys->framebuflist, head)
> > +		dma_free_attrs(&auxdev->dev, sizeof(struct isys_fw_msgs),
> > +			       fwmsg, fwmsg->dma_addr, 0);
>
> This is also done in isys_remove(). How about moving these into a new
> function that releases memory from both lists?

I agree. I thought about that too but decided to propose a simple 
solution first.

> There's no harm in traversing framebuflist_fw as well, it's empty at this
> point.

Yes, that's my understanding.
That's why I only copied one list_for_each_entry_safe().


I'll send a v2 with the list_for_each_entry_safe() in a new function.
It will be much cleaner.

CJ

> > +
> >   out_remove_pkg_dir_shared_buffer:
> >   	if (!isp->secure_mode)
> >   		ipu6_cpd_free_pkg_dir(adev);