??: [RESEND 1/3] drm: fsl-dcu: Fix no fb check bug
diff mbox

Message ID HE1PR04MB1051408944D97C4E306D7A17ECE70@HE1PR04MB1051.eurprd04.prod.outlook.com
State New
Headers show

Commit Message

Meng Yi Dec. 24, 2015, 5:28 a.m. UTC
Tested-by: Meng Yi <meng.yi@nxp.com>


-----????-----
???: Dongsheng Wang [mailto:Dongsheng.Wang@freescale.com] 
????: Tuesday, December 01, 2015 4:16 PM
???: airlied@linux.ie
??: stefan@agner.ch; dri-devel@lists.freedesktop.org; Jianwei Wang <jianwei.wang.chn@gmail.com>; Yi Meng-B56799 <B56799@freescale.com>; Wang Dongsheng-B40534 <Dongsheng.Wang@freescale.com>
??: [RESEND 1/3] drm: fsl-dcu: Fix no fb check bug

From: Jianwei Wang <jianwei.wang.chn@gmail.com>


For state->fb may be NULL in fsl_dcu_drm_plane_atomic_check function, if so, return -EINVAL. No need check in fsl_dcu_drm_plane_atomic_update anymore.

Signed-off-by: Jianwei Wang <jianwei.wang.chn@gmail.com>

Signed-off-by: Yi Meng <b56799@freescale.com>

Signed-off-by: Wang Dongsheng <dongsheng.wang@freescale.com>

Tested-by: Stefan Agner <stefan@agner.ch>


 	unsigned int alpha, bpp;
 	int index, ret;
 
-	if (!fb)
-		return;
-
 	index = fsl_dcu_drm_plane_index(plane);
 	if (index < 0)
 		return;
--
2.1.0.27.g96db324

Comments

Stefan Agner Dec. 27, 2015, 3:54 a.m. UTC | #1
Hi Meng,

The only situation I could observe that is when fsl,panel was not
valid...

However, I think with my patch "drm/fsl-dcu: Fix no fb check bug", this
situation can be avoided completely:
https://lkml.org/lkml/2015/11/18/950

With that patch applied, and a non-existing panel assigned, the DRM
driver already fails at probe time gracefully:
[    0.488291] [drm] Initialized drm 1.1.0 20060810
[    0.501576] fsl-dcu 40058000.dcu: failed to initialize mode setting

So I think that a state->fb check for NULL is not necessary at all...

Can you test my patch to see if it fixes the issue for you too?

--
Stefan

On 2015-12-23 21:28, Meng Yi wrote:
> Tested-by: Meng Yi <meng.yi@nxp.com>
> 
> -----????-----
> ???: Dongsheng Wang [mailto:Dongsheng.Wang@freescale.com] 
> ????: Tuesday, December 01, 2015 4:16 PM
> ???: airlied@linux.ie
> ??: stefan@agner.ch; dri-devel@lists.freedesktop.org; Jianwei Wang
> <jianwei.wang.chn@gmail.com>; Yi Meng-B56799 <B56799@freescale.com>;
> Wang Dongsheng-B40534 <Dongsheng.Wang@freescale.com>
> ??: [RESEND 1/3] drm: fsl-dcu: Fix no fb check bug
> 
> From: Jianwei Wang <jianwei.wang.chn@gmail.com>
> 
> For state->fb may be NULL in fsl_dcu_drm_plane_atomic_check function,
> if so, return -EINVAL. No need check in
> fsl_dcu_drm_plane_atomic_update anymore.
> 
> Signed-off-by: Jianwei Wang <jianwei.wang.chn@gmail.com>
> Signed-off-by: Yi Meng <b56799@freescale.com>
> Signed-off-by: Wang Dongsheng <dongsheng.wang@freescale.com>
> Tested-by: Stefan Agner <stefan@agner.ch>
> 
> diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c
> b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c
> index 51daaea..a8932a8 100644
> --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c
> +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c
> @@ -41,6 +41,9 @@ static int fsl_dcu_drm_plane_atomic_check(struct
> drm_plane *plane,  {
>  	struct drm_framebuffer *fb = state->fb;
>  
> +	if (!fb)
> +		return -EINVAL;
> +
>  	switch (fb->pixel_format) {
>  	case DRM_FORMAT_RGB565:
>  	case DRM_FORMAT_RGB888:
> @@ -85,9 +88,6 @@ static void fsl_dcu_drm_plane_atomic_update(struct
> drm_plane *plane,
>  	unsigned int alpha, bpp;
>  	int index, ret;
>  
> -	if (!fb)
> -		return;
> -
>  	index = fsl_dcu_drm_plane_index(plane);
>  	if (index < 0)
>  		return;
> --
> 2.1.0.27.g96db324
Daniel Stone Dec. 28, 2015, 3:14 p.m. UTC | #2
Hi Stefan,

On Sunday, 27 December 2015, Stefan Agner <stefan@agner.ch> wrote:
>
> However, I think with my patch "drm/fsl-dcu: Fix no fb check bug", this
> situation can be avoided completely:
> https://lkml.org/lkml/2015/11/18/950
>
> With that patch applied, and a non-existing panel assigned, the DRM
> driver already fails at probe time gracefully:
> [    0.488291] [drm] Initialized drm 1.1.0 20060810
> [    0.501576] fsl-dcu 40058000.dcu: failed to initialize mode setting
>
> So I think that a state->fb check for NULL is not necessary at all...
>
> Can you test my patch to see if it fixes the issue for you too?
>

This is still required; state->fb being NULL means the plane is being
disabled.

Cheers,
Daniel
Meng Yi Dec. 30, 2015, 7:37 a.m. UTC | #3
Hi, Stefan
I have tested your patch, It seems good to me. 
But I think state->fb check is still necessary, because fb is related to crtc , and panel is related to connector,. When fsl,panel is not valid, it indicate that connector is not available, but fb check is still needed. But I am not so sure, and what do you think?

--
Meng Yi

-----Original Message-----
From: Stefan Agner [mailto:stefan@agner.ch] 

Sent: Sunday, December 27, 2015 11:54 AM
To: Meng Yi <meng.yi@nxp.com>
Cc: Dongsheng Wang <Dongsheng.Wang@freescale.com>; airlied@linux.ie; dri-devel@lists.freedesktop.org; Jianwei Wang <jianwei.wang.chn@gmail.com>; Yi Meng-B56799 <B56799@freescale.com>
Subject: Re: ??: [RESEND 1/3] drm: fsl-dcu: Fix no fb check bug

Hi Meng,

The only situation I could observe that is when fsl,panel was not valid...

However, I think with my patch "drm/fsl-dcu: Fix no fb check bug", this situation can be avoided completely:
https://lkml.org/lkml/2015/11/18/950

With that patch applied, and a non-existing panel assigned, the DRM driver already fails at probe time gracefully:
[    0.488291] [drm] Initialized drm 1.1.0 20060810
[    0.501576] fsl-dcu 40058000.dcu: failed to initialize mode setting

So I think that a state->fb check for NULL is not necessary at all...

Can you test my patch to see if it fixes the issue for you too?

--
Stefan

On 2015-12-23 21:28, Meng Yi wrote:
> Tested-by: Meng Yi <meng.yi@nxp.com>

> 

> -----????-----

> ???: Dongsheng Wang [mailto:Dongsheng.Wang@freescale.com]

> ????: Tuesday, December 01, 2015 4:16 PM

> ???: airlied@linux.ie

> ??: stefan@agner.ch; dri-devel@lists.freedesktop.org; Jianwei Wang 

> <jianwei.wang.chn@gmail.com>; Yi Meng-B56799 <B56799@freescale.com>; 

> Wang Dongsheng-B40534 <Dongsheng.Wang@freescale.com>

> ??: [RESEND 1/3] drm: fsl-dcu: Fix no fb check bug

> 

> From: Jianwei Wang <jianwei.wang.chn@gmail.com>

> 

> For state->fb may be NULL in fsl_dcu_drm_plane_atomic_check function, 

> if so, return -EINVAL. No need check in 

> fsl_dcu_drm_plane_atomic_update anymore.

> 

> Signed-off-by: Jianwei Wang <jianwei.wang.chn@gmail.com>

> Signed-off-by: Yi Meng <b56799@freescale.com>

> Signed-off-by: Wang Dongsheng <dongsheng.wang@freescale.com>

> Tested-by: Stefan Agner <stefan@agner.ch>

> 

> diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c

> b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c

> index 51daaea..a8932a8 100644

> --- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c

> +++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c

> @@ -41,6 +41,9 @@ static int fsl_dcu_drm_plane_atomic_check(struct

> drm_plane *plane,  {

>  	struct drm_framebuffer *fb = state->fb;

>  

> +	if (!fb)

> +		return -EINVAL;

> +

>  	switch (fb->pixel_format) {

>  	case DRM_FORMAT_RGB565:

>  	case DRM_FORMAT_RGB888:

> @@ -85,9 +88,6 @@ static void fsl_dcu_drm_plane_atomic_update(struct

> drm_plane *plane,

>  	unsigned int alpha, bpp;

>  	int index, ret;

>  

> -	if (!fb)

> -		return;

> -

>  	index = fsl_dcu_drm_plane_index(plane);

>  	if (index < 0)

>  		return;

> --

> 2.1.0.27.g96db324
Daniel Stone Jan. 1, 2016, 3:10 p.m. UTC | #4
Hi,

On 30 December 2015 at 07:37, Meng Yi <meng.yi@nxp.com> wrote:
> I have tested your patch, It seems good to me.
> But I think state->fb check is still necessary, because fb is related to crtc , and panel is related to connector,. When fsl,panel is not valid, it indicate that connector is not available, but fb check is still needed. But I am not so sure, and what do you think?

fb is tied to planes. Planes are tied to CRTCs. CRTCs are tied to
connectors. Connectors are tied to the panel. If the panel is not
found, no connectors will be activated, so no CRTCs will be enabled,
so no planes will be enabled.

Please let me be very clear though: setting fb == NULL is legitimate.
Userspace may do this at any time. Crashing when plane->state->fb ==
NULL is an error in your driver and must be fixed.

Cheers,
Daniel
Stefan Agner Jan. 4, 2016, 6:07 a.m. UTC | #5
On 2016-01-01 07:10, Daniel Stone wrote:
> Hi,
> 
> On 30 December 2015 at 07:37, Meng Yi <meng.yi@nxp.com> wrote:
>> I have tested your patch, It seems good to me.
>> But I think state->fb check is still necessary, because fb is related to crtc , and panel is related to connector,. When fsl,panel is not valid, it indicate that connector is not available, but fb check is still needed. But I am not so sure, and what do you think?
> 
> fb is tied to planes. Planes are tied to CRTCs. CRTCs are tied to
> connectors. Connectors are tied to the panel. If the panel is not
> found, no connectors will be activated, so no CRTCs will be enabled,
> so no planes will be enabled.
> 
> Please let me be very clear though: setting fb == NULL is legitimate.
> Userspace may do this at any time. Crashing when plane->state->fb ==
> NULL is an error in your driver and must be fixed.

Thanks for this clarification. Ok, I agree the patch is needed despite
my fix then...

Acked-by: Stefan Agner <stefan@agner.ch>

--
Stefan
Meng Yi Jan. 4, 2016, 6:48 a.m. UTC | #6
Hi, Daniel
	Thanks for your clarification, and I will fix this error.
--
Meng

-----Original Message-----
From: Daniel Stone [mailto:daniel@fooishbar.org] 

Sent: Friday, January 01, 2016 11:10 PM
To: Meng Yi <meng.yi@nxp.com>
Cc: Stefan Agner <stefan@agner.ch>; dri-devel@lists.freedesktop.org; Yi Meng-B56799 <B56799@freescale.com>
Subject: Re: ??: [RESEND 1/3] drm: fsl-dcu: Fix no fb check bug

Hi,

On 30 December 2015 at 07:37, Meng Yi <meng.yi@nxp.com> wrote:
> I have tested your patch, It seems good to me.

> But I think state->fb check is still necessary, because fb is related to crtc , and panel is related to connector,. When fsl,panel is not valid, it indicate that connector is not available, but fb check is still needed. But I am not so sure, and what do you think?


fb is tied to planes. Planes are tied to CRTCs. CRTCs are tied to connectors. Connectors are tied to the panel. If the panel is not found, no connectors will be activated, so no CRTCs will be enabled, so no planes will be enabled.

Please let me be very clear though: setting fb == NULL is legitimate.
Userspace may do this at any time. Crashing when plane->state->fb == NULL is an error in your driver and must be fixed.

Cheers,
Daniel

Patch
diff mbox

diff --git a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c
index 51daaea..a8932a8 100644
--- a/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c
+++ b/drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.c
@@ -41,6 +41,9 @@  static int fsl_dcu_drm_plane_atomic_check(struct drm_plane *plane,  {
 	struct drm_framebuffer *fb = state->fb;
 
+	if (!fb)
+		return -EINVAL;
+
 	switch (fb->pixel_format) {
 	case DRM_FORMAT_RGB565:
 	case DRM_FORMAT_RGB888:
@@ -85,9 +88,6 @@  static void fsl_dcu_drm_plane_atomic_update(struct drm_plane *plane,