Skip to contentSkip to conversation

drm_crtc: check if fb_create return NULL

PATCH In review Su, Xuemin 3 participants 6 messages Jan 24, 2013 – Jan 31, 2013

Subsystems (2)
lore
Share & tools
lore

Su, Xuemin

author
#0 lore raw

PATCH v3 1/1 drm_crtc: check if fb_create return NULL

From: xueminsu <[email protected]>
Date: Tue, 22 Jan 2013 22:39:39 +0800
Subject: [PATCH] drm_crtc: check if fb_create return NULL

Some buggy driver may still return NULL in fb_create, which leads to kernel panic.

---
 drivers/gpu/drm/drm_crtc.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
drivers/gpu/drm/drm_crtc.c +2 -0
index f2d667b..ae613ec 100644
@@ -2172,6 +2172,8 @@ int drm_mode_addfb(struct drm_device *dev,
 		ret = PTR_ERR(fb);
 		goto out;
 	}
+	/* some buggy driver may return NULL here, which may cause panic */
+	BUG_ON(!fb);
 
 	or->fb_id = fb->base.id;
 	list_add(&fb->filp_head, &file_priv->fbs);
Signature · 1 line
1.7.6
#1 lore raw
Quoted 17 lines · Xuemin

From: xueminsu <[email protected]>
Date: Tue, 22 Jan 2013 22:39:39 +0800
Subject: [PATCH] drm_crtc: check if fb_create return NULL

Some buggy driver may still return NULL in fb_create, which leads to kernel panic.

---
 drivers/gpu/drm/drm_crtc.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
drivers/gpu/drm/drm_crtc.c +0 -0
index f2d667b..ae613ec 100644

drivers/gpu/drm/drm_crtc.c int drm_mode_addfb(struct drm_device *dev, · line 2176

Quoted 5 lines
drivers/gpu/drm/drm_crtc.c +2 -0
@@ -2172,6 +2172,8 @@ int drm_mode_addfb(struct drm_device *dev,
 		ret = PTR_ERR(fb);
 		goto out;
 	}
+	/* some buggy driver may return NULL here, which may cause panic */
+	BUG_ON(!fb);

I fail to see the benefit of this compared to just letting it oops...

drivers/gpu/drm/drm_crtc.c int drm_mode_addfb(struct drm_device *dev, · line 2178

Quoted 1 line
drivers/gpu/drm/drm_crtc.c +0 -0
 	or->fb_id = fb->base.id;

...right here.

BR,
Jani.

Quoted 10 lines
drivers/gpu/drm/drm_crtc.c +0 -0
 	list_add(&fb->filp_head, &file_priv->fbs);
Signature · 8 lines
1.7.6



_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel
author
#2 lore raw
Quoted 11 lines · Jani Nikula

drivers/gpu/drm/drm_crtc.c int drm_mode_addfb(struct drm_device *dev, · line 2176

Quoted 3 lines
drivers/gpu/drm/drm_crtc.c +2 -0
 	}
+	/* some buggy driver may return NULL here, which may cause panic */
+	BUG_ON(!fb);

I fail to see the benefit of this compared to just letting it oops...

Quoted 1 line

or->fb_id = fb->base.id;

...right here.

For PATCH V3, I think a BUG_ON may give the user clearer information about the reason of panic. Easier to debug.
I submitted a PATCH V2 yesterday which gives a warning and then return -EVAL if fb==NULL, preventing the panic. Do you think this is acceptable?

Reviewed-by
#3 lore raw
Quoted 7 lines · Xuemin
Quoted 6 lines · Jani Nikula

drivers/gpu/drm/drm_crtc.c int drm_mode_addfb(struct drm_device *dev, · line 2176

Quoted 3 lines
drivers/gpu/drm/drm_crtc.c +2 -0
 	}
+	/* some buggy driver may return NULL here, which may cause panic */
+	BUG_ON(!fb);

I fail to see the benefit of this compared to just letting it oops...

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

or->fb_id = fb->base.id;

...right here.

For PATCH V3, I think a BUG_ON may give the user clearer information about the reason of panic. Easier to debug.

Ah, sorry, never mind, I missed Daniel's comment. The benefit of the BUG_ON() is making it clear what's expected of the drivers.

author
#4 lore raw

-----Original Message-----
From: Jani Nikula [mailto:[email protected]]
Sent: Thursday, January 24, 2013 5:05 PM
To: Su, Xuemin
Cc: [email protected]; [email protected]; [email protected]; [email protected]; He, Bo
Subject: Re: [PATCH V3] drm_crtc: check if fb_create return NULL

Quoted 4 lines

Ah, sorry, never mind, I missed Daniel's comment. The benefit of the BUG_ON() is making it clear what's expected of the drivers.

Do you think this patch is still needed?
Currently I fix a buggy function radeon_user_framebuffer_create() which returns NULL and patch is added to 3.7-stable tree. Do you think it's also needed to do something in drm_mode_addfb()?
Thanks.

#5 lore raw
Quoted 7 lines · Xuemin

-----Original Message-----
From: Jani Nikula [mailto:[email protected]]
Sent: Thursday, January 24, 2013 5:05 PM
To: Su, Xuemin
Cc: [email protected]; [email protected]; [email protected]; [email protected]; He, Bo
Subject: Re: [PATCH V3] drm_crtc: check if fb_create return NULL

Quoted 8 lines
Quoted 4 lines

Ah, sorry, never mind, I missed Daniel's comment. The benefit of the BUG_ON() is making it clear what's expected of the drivers.

Do you think this patch is still needed?
Currently I fix a buggy function radeon_user_framebuffer_create() which returns NULL and patch is added to 3.7-stable tree. Do you think it's also needed to do something in drm_mode_addfb()?

Imo it's still useful, just to document the assumptions of the interface. Dave? -Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch