[PATCH RFT v2] media: atomisp: Add error check in create_host_regular_capture_pipeline()

[PATCH RFT v2] media: atomisp: Add error check in create_host_regular_capture_pipeline()

[Robertus Diawan Chris]

When calling ia_css_pipeline_create_and_add_stage() function, there's no
condition to check whether ia_css_pipeline_create_and_add_stage() throw
an error or not. So add conditional check for the return value of
ia_css_pipeline_create_and_add_stage() function.

This is reported by Coverity Scan with CID 1408952 as UNUSED_VALUE.

Fixes: a49d25364dfb ("staging/atomisp: Add support for the Intel IPU v2")
Signed-off-by: Robertus Diawan Chris <robertusdchris@gmail.com>
---
v1 -> v2: Add Fixes tag (suggested by Dan Carpenter).

v1:
https://lore.kernel.org/all/20260423060246.296986-1-robertusdchris@gmail.com/

I am not sure if omitting the error check is intentional or not, but from
what I see from another ia_css_pipeline_create_and_add_stage() usage
in the same function, we check the return value after the call and return
the error code if it's non-zero.

I also don't have the device to test this change.

 drivers/staging/media/atomisp/pci/sh_css.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/staging/media/atomisp/pci/sh_css.c b/drivers/staging/media/atomisp/pci/sh_css.c
index 6cda5925fa45..6aa8a883cc76 100644
--- a/drivers/staging/media/atomisp/pci/sh_css.c
+++ b/drivers/staging/media/atomisp/pci/sh_css.c
@@ -7041,6 +7041,10 @@ create_host_regular_capture_pipeline(struct ia_css_pipe *pipe)
 			err = ia_css_pipeline_create_and_add_stage(me,
 								   &stage_desc,
 								   &current_stage);
+			if (err) {
+				IA_CSS_LEAVE_ERR_PRIVATE(err);
+				return err;
+			}
 			local_in_frame = current_stage->args.out_frame[0];
 		}
 		err = add_capture_pp_stage(pipe, me, local_in_frame,

base-commit: 028ef9c96e96197026887c0f092424679298aae8
-- 
2.53.0

Re: [PATCH RFT v2] media: atomisp: Add error check in create_host_regular_capture_pipeline()

[Andy Shevchenko]

On Fri, Apr 24, 2026 at 03:15:38PM +0700, Robertus Diawan Chris wrote:
> When calling ia_css_pipeline_create_and_add_stage() function, there's no
> condition to check whether ia_css_pipeline_create_and_add_stage() throw
> an error or not. So add conditional check for the return value of
> ia_css_pipeline_create_and_add_stage() function.

> This is reported by Coverity Scan with CID 1408952 as UNUSED_VALUE.

Unneeded detail in the commit message. Use the comment block for that (as there
changelog right now and other stuff you explained).

> Fixes: a49d25364dfb ("staging/atomisp: Add support for the Intel IPU v2")
> Signed-off-by: Robertus Diawan Chris <robertusdchris@gmail.com>
> ---
> v1 -> v2: Add Fixes tag (suggested by Dan Carpenter).
> 
> v1:
> https://lore.kernel.org/all/20260423060246.296986-1-robertusdchris@gmail.com/
> 
> I am not sure if omitting the error check is intentional or not, but from
> what I see from another ia_css_pipeline_create_and_add_stage() usage
> in the same function, we check the return value after the call and return
> the error code if it's non-zero.

Okay, while this observation sounds like a +1 to the suggested change,
you still need to dive into the code and try to get if
ia_css_pipeline_create_and_add_stage() failure is fatal or not for
create_host_regular_capture_pipeline().

> I also don't have the device to test this change.

It's a pity as this driver is quite complicated thing...

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH RFT v2] media: atomisp: Add error check in create_host_regular_capture_pipeline()

[Robertus Diawan Chris]

Hello Andy,

On Fri, Apr 24, 2026 at 11:48:52AM +0300, Andy Shevchenko wrote:
> On Fri, Apr 24, 2026 at 03:15:38PM +0700, Robertus Diawan Chris wrote:
> > When calling ia_css_pipeline_create_and_add_stage() function, there's no
> > condition to check whether ia_css_pipeline_create_and_add_stage() throw
> > an error or not. So add conditional check for the return value of
> > ia_css_pipeline_create_and_add_stage() function.
>
> > This is reported by Coverity Scan with CID 1408952 as UNUSED_VALUE.
>
> Unneeded detail in the commit message. Use the comment block for that (as there
> changelog right now and other stuff you explained).

Alright, I will keep that in mind for future patches.

> > Fixes: a49d25364dfb ("staging/atomisp: Add support for the Intel IPU v2")
> > Signed-off-by: Robertus Diawan Chris <robertusdchris@gmail.com>
> > ---
> > v1 -> v2: Add Fixes tag (suggested by Dan Carpenter).
> >
> > v1:
> > https://lore.kernel.org/all/20260423060246.296986-1-robertusdchris@gmail.com/
> >
> > I am not sure if omitting the error check is intentional or not, but from
> > what I see from another ia_css_pipeline_create_and_add_stage() usage
> > in the same function, we check the return value after the call and return
> > the error code if it's non-zero.
>
> Okay, while this observation sounds like a +1 to the suggested change,
> you still need to dive into the code and try to get if
> ia_css_pipeline_create_and_add_stage() failure is fatal or not for
> create_host_regular_capture_pipeline().

After diving further into the code, I am not confident enough to
continue this patch. I lack the context and not sure if this _seemingly_
small change can have a big side effect for the overall driver.

With that in mind, I will not continue with this patch. I will try to take a
look at another driver.

> > I also don't have the device to test this change.
>
> It's a pity as this driver is quite complicated thing...

Yeah... Let's just say it is _a sight to behold_ for newcomer like
myself :)

Thank you for taking the time to check this patch.

Best regards,
Robertus Diawan Chris