[PATCH] drm: fix a use-after-free when GPU acceleration disabled

[Michel Dänzer]

On Die, 2013-05-21 at 14:23 +0800, Huacai Chen wrote:
> When GPU acceleration is disabled, drm_vblank_cleanup() will free the
> vblank-related data, such as vblank_refcount, vblank_inmodeset, etc.
> But we found that drm_vblank_post_modeset() may be called after the
> cleanup, which use vblank_refcount and vblank_inmodeset. And this will
> cause a kernel panic.
> 
> Fix this by return immediately if dev->num_crtcs is zero. This is the
> same thing that drm_vblank_pre_modeset() does.
> 
> Call trace of a drm_vblank_post_modeset() after drm_vblank_cleanup():
> [   62.628906] [<ffffffff804868d0>] drm_vblank_post_modeset+0x34/0xb4
> [   62.628906] [<ffffffff804c7008>] atombios_crtc_dpms+0xb4/0x174
> [   62.628906] [<ffffffff804c70e0>] atombios_crtc_commit+0x18/0x38
> [   62.628906] [<ffffffff8047f038>] drm_crtc_helper_set_mode+0x304/0x3cc
> [   62.628906] [<ffffffff8047f92c>] drm_crtc_helper_set_config+0x6d8/0x988
> [   62.628906] [<ffffffff8047dd40>] drm_fb_helper_set_par+0x94/0x104
> [   62.628906] [<ffffffff80439d14>] fbcon_init+0x424/0x57c
> [   62.628906] [<ffffffff8046a638>] visual_init+0xb8/0x118
> [   62.628906] [<ffffffff8046b9f8>] take_over_console+0x238/0x384
> [   62.628906] [<ffffffff80436df8>] fbcon_takeover+0x7c/0xdc
> [   62.628906] [<ffffffff8024fa20>] notifier_call_chain+0x44/0x94
> [   62.628906] [<ffffffff8024fcbc>] __blocking_notifier_call_chain+0x48/0x68
> [   62.628906] [<ffffffff8042d990>] register_framebuffer+0x228/0x260
> [   62.628906] [<ffffffff8047e010>] drm_fb_helper_single_fb_probe+0x260/0x314
> [   62.628906] [<ffffffff8047e2c4>] drm_fb_helper_initial_config+0x200/0x234
> [   62.628906] [<ffffffff804e5560>] radeon_fbdev_init+0xd4/0xf4
> [   62.628906] [<ffffffff804e0e08>] radeon_modeset_init+0x9bc/0xa18
> [   62.628906] [<ffffffff804bfc14>] radeon_driver_load_kms+0xdc/0x12c
> [   62.628906] [<ffffffff8048b548>] drm_get_pci_dev+0x148/0x238
> [   62.628906] [<ffffffff80423564>] local_pci_probe+0x5c/0xd0
> [   62.628906] [<ffffffff80241ac4>] work_for_cpu_fn+0x1c/0x30
> [   62.628906] [<ffffffff802427c8>] process_one_work+0x274/0x3bc
> [   62.628906] [<ffffffff80242934>] process_scheduled_works+0x24/0x44
> [   62.628906] [<ffffffff8024515c>] worker_thread+0x31c/0x3f4
> [   62.628906] [<ffffffff802497a8>] kthread+0x88/0x90
> [   62.628906] [<ffffffff80206794>] kernel_thread_helper+0x10/0x18
> 
> Signed-off-by: Huacai Chen <chenhc at lemote.com>
> Signed-off-by: Binbin Zhou <zhoubb at lemote.com>
> Cc: <stable at vger.kernel.org>
> ---
>  drivers/gpu/drm/drm_irq.c |    6 +++++-
>  1 files changed, 5 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c
> index c798eea..c7a1348 100644
> --- a/drivers/gpu/drm/drm_irq.c
> +++ b/drivers/gpu/drm/drm_irq.c
> @@ -981,7 +981,7 @@ EXPORT_SYMBOL(drm_vblank_off);
>   */
>  void drm_vblank_pre_modeset(struct drm_device *dev, int crtc)
>  {
> -       /* vblank is not initialized (IRQ not installed ?) */
> +       /* vblank is not initialized (IRQ not installed ?), or has been freed */
>         if (!dev->num_crtcs)
>                 return;
>         /*
> @@ -1003,6 +1003,10 @@ void drm_vblank_post_modeset(struct drm_device *dev, int crtc)
>  {
>         unsigned long irqflags;
>  
> +       /* vblank is not initialized (IRQ not installed ?), or has been freed */
> +       if (!dev->num_crtcs)
> +               return;
> +
>         if (dev->vblank_inmodeset[crtc]) {
>                 spin_lock_irqsave(&dev->vbl_lock, irqflags);
>                 dev->vblank_disable_allowed = 1;

The comment changes seem superfluous to me, but either way:

Reviewed-by: Michel Dänzer <michel.daenzer at amd.com>


-- 
Earthling Michel Dänzer           |                   http://www.amd.com
Libre software enthusiast         |          Debian, X and DRI developer