diff mbox

drm: Prevent early NULL dereference of fb_helper->fb

Message ID 20170719152450.19712-1-chris@chris-wilson.co.uk (mailing list archive)
State New, archived
Headers show

Commit Message

Chris Wilson July 19, 2017, 3:24 p.m. UTC
Prior to the fbdev being asynchronously configured on boot, the
fb_helper->fb may be unset.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101767
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Thierry Reding <treding@nvidia.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
---
This may not be necessary with the recent async rework? One hopes.
---
 drivers/gpu/drm/drm_fb_helper.c | 3 +++
 1 file changed, 3 insertions(+)

Comments

Daniel Vetter July 20, 2017, 8:26 a.m. UTC | #1
On Wed, Jul 19, 2017 at 04:24:50PM +0100, Chris Wilson wrote:
> Prior to the fbdev being asynchronously configured on boot, the
> fb_helper->fb may be unset.
> 
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101767
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Thierry Reding <treding@nvidia.com>
> Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
> ---
> This may not be necessary with the recent async rework? One hopes.

It shouldn't be possible to call into check_var (or any other fb_ops
callback) before we register the framebuffer, and we don't do that before
the fb is real.

The bug description also says that after several suspend/resume cycles
does it blow up like that, so very unlikely fbdev isn't set up yet. And
the NULL pointer isn't really NULL.

There's something going wrong for sure, but I don't think it's this.
-Daniel

> ---
>  drivers/gpu/drm/drm_fb_helper.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> index 42090fe00ef9..a44b6b434fa0 100644
> --- a/drivers/gpu/drm/drm_fb_helper.c
> +++ b/drivers/gpu/drm/drm_fb_helper.c
> @@ -1502,6 +1502,9 @@ int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
>  	if (var->pixclock != 0 || in_dbg_master())
>  		return -EINVAL;
>  
> +	if (!fb) /* too early, fbdev is not yet setup */
> +		return -EINVAL;
> +
>  	/*
>  	 * Changes struct fb_var_screeninfo are currently not pushed back
>  	 * to KMS, hence fail if different settings are requested.
> -- 
> 2.13.3
>
Chris Wilson July 20, 2017, 8:52 a.m. UTC | #2
Quoting Daniel Vetter (2017-07-20 09:26:32)
> On Wed, Jul 19, 2017 at 04:24:50PM +0100, Chris Wilson wrote:
> > Prior to the fbdev being asynchronously configured on boot, the
> > fb_helper->fb may be unset.
> > 
> > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101767
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > Cc: Thierry Reding <treding@nvidia.com>
> > Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
> > ---
> > This may not be necessary with the recent async rework? One hopes.
> 
> It shouldn't be possible to call into check_var (or any other fb_ops
> callback) before we register the framebuffer, and we don't do that before
> the fb is real.
> 
> The bug description also says that after several suspend/resume cycles
> does it blow up like that, so very unlikely fbdev isn't set up yet. And
> the NULL pointer isn't really NULL.
> 
> There's something going wrong for sure, but I don't think it's this.

It was a stab in the dark as it looked like fb was the pointer behind
the crash. (And I have an inherent distrust of this code running before
everything is setup, and we know from experience that we can suspend
before it is setup.) We can rule out *var, so use-after-free of fb?
-Chris
diff mbox

Patch

diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
index 42090fe00ef9..a44b6b434fa0 100644
--- a/drivers/gpu/drm/drm_fb_helper.c
+++ b/drivers/gpu/drm/drm_fb_helper.c
@@ -1502,6 +1502,9 @@  int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
 	if (var->pixclock != 0 || in_dbg_master())
 		return -EINVAL;
 
+	if (!fb) /* too early, fbdev is not yet setup */
+		return -EINVAL;
+
 	/*
 	 * Changes struct fb_var_screeninfo are currently not pushed back
 	 * to KMS, hence fail if different settings are requested.