Prev: [tip:x86/gcc46] x86, gcc-4.6: Avoid unused by set variables in rdmsr
Next: udevd / ext4 issue mounting 2.6.35-rc5
From: Dave Airlie on 20 Jul 2010 20:40
On Wed, 2010-07-21 at 01:07 +0400, Alexander Y. Fomichev wrote:
> On Wed, Jul 21, 2010 at 1:00 AM, Alex Deucher <alexdeucher(a)gmail.com> wrote:
> > On Tue, Jul 20, 2010 at 4:13 PM, Alexander Y. Fomichev
> > <git.user(a)gmail.com> wrote:
> >> On Tue, Jul 20, 2010 at 9:37 PM, Alex Deucher <alexdeucher(a)gmail.com> wrote:
> >>> On Mon, Jul 19, 2010 at 5:42 PM, Alexander Y. Fomichev
> >>> <git.user(a)gmail.com> wrote:
> >>>> This patch fix possible NULL pointer dereference when
> >>>> r600_prepare_blit_copy tries to fill dev_priv->blit_vb->file_priv
> >>>> without check of dev_priv->blit_vb. dev_priv->blit_vb should be
> >>>> filled by r600_nomm_get_vb but latest can fail with EAGAIN.
> >>>> Addresses: https://bugzilla.kernel.org/show_bug.cgi?id=16375
> >>>>
> >>>> ---
> >>>> drivers/gpu/drm/radeon/r600_blit.c | 2 ++
> >>>> 1 files changed, 2 insertions(+), 0 deletions(-)
> >>>>
> >>>> diff --git a/drivers/gpu/drm/radeon/r600_blit.c b/drivers/gpu/drm/radeon/r600_blit.c
> >>>> index f4fb88e..0df4a2b 100644
> >>>> --- a/drivers/gpu/drm/radeon/r600_blit.c
> >>>> +++ b/drivers/gpu/drm/radeon/r600_blit.c
> >>>> @@ -541,6 +541,8 @@ r600_prepare_blit_copy(struct drm_device *dev, struct drm_file *file_priv)
> >>>> DRM_DEBUG("\n");
> >>>>
> >>>> r600_nomm_get_vb(dev);
> >>>> + if (!dev_priv->blit_vb)
> >>>> + return;
> >>>
> >>> r600_prepare_blit_copy returns an int so something like this would be better:
> >>> --- a/drivers/gpu/drm/radeon/r600_blit.c
> >>> +++ b/drivers/gpu/drm/radeon/r600_blit.c
> >>> @@ -539,8 +539,10 @@ r600_prepare_blit_copy(struct drm_device *dev,
> >>> struct drm_file *file_priv)
> >>> {
> >>> drm_radeon_private_t *dev_priv = dev->dev_private;
> >>> DRM_DEBUG("\n");
> >>> + int ret = r600_nomm_get_vb(dev);
> >>>
> >>> - r600_nomm_get_vb(dev);
> >>> + if (ret)
> >>> + return ret;
> >>>
> >>> dev_priv->blit_vb->file_priv = file_priv;
> >>>
> >>>
> >>> Alex
> >>>
> >>>>
> >>>> dev_priv->blit_vb->file_priv = file_priv;
> >>>>
> >>>> --
> >>>> 1.7.1.1
> >>>>
> >>>
> >>
> >> i haven't any preferneces, the only thing is - it would be logical
> >> to have every check in common style, so other cases
> >> (r600_blit_copy, r600_blit_swap) should be fixed, nop?
> >
> > Those are void functions so there's nothing to return.
>
> i mean both of them call r600_nomm_get_vb and both of them
> check if (!dev_priv->blit_vb), not return value.I mean would be
> logical to check it the same way everytime r600_nomm_get_vb
> gets called.

I'm going to go with Alex's patch, as none of the other callsites return
an error, and your patch doesn't return value from a function which has
a int return type, causing a warning.

Dave.



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 | 
Pages: 1
Prev: [tip:x86/gcc46] x86, gcc-4.6: Avoid unused by set variables in rdmsr
Next: udevd / ext4 issue mounting 2.6.35-rc5