Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1403261 > unrolled thread

[PATCH] drm/mm: avoid possible null pointer dereference

Started byHeinrich Schuchardt <xypron.glpk@gmx.de>
First post2016-05-18 22:20 +0200
Last post2016-05-23 18:00 +0200
Articles 8 — 4 participants

Back to article view | Back to linux.kernel

Contents

  [PATCH] drm/mm: avoid possible null pointer dereference Heinrich Schuchardt <xypron.glpk@gmx.de> - 2016-05-18 22:20 +0200
    Re: [PATCH] drm/mm: avoid possible null pointer dereference Eric Engestrom <eric.engestrom@imgtec.com> - 2016-05-23 12:30 +0200
      Re: [PATCH] drm/mm: avoid possible null pointer dereference Daniel Vetter <daniel@ffwll.ch> - 2016-05-23 15:00 +0200
        Re: [PATCH] drm/mm: avoid possible null pointer dereference Chris Wilson <chris@chris-wilson.co.uk> - 2016-05-23 15:10 +0200
          Re: [PATCH] drm/mm: avoid possible null pointer dereference Eric Engestrom <eric.engestrom@imgtec.com> - 2016-05-23 15:40 +0200
            Re: [PATCH] drm/mm: avoid possible null pointer dereference Chris Wilson <chris@chris-wilson.co.uk> - 2016-05-23 15:50 +0200
              Re: [PATCH] drm/mm: avoid possible null pointer dereference Eric Engestrom <eric.engestrom@imgtec.com> - 2016-05-23 17:40 +0200
                Re: [PATCH] drm/mm: avoid possible null pointer dereference Chris Wilson <chris@chris-wilson.co.uk> - 2016-05-23 18:00 +0200

#1403261 — [PATCH] drm/mm: avoid possible null pointer dereference

FromHeinrich Schuchardt <xypron.glpk@gmx.de>
Date2016-05-18 22:20 +0200
Subject[PATCH] drm/mm: avoid possible null pointer dereference
Message-ID<rAg6K-6L0-7@gated-at.bofh.it>
Do not dereference node before the check if node is NULL.

Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
---
 drivers/gpu/drm/drm_mm.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
index 04de6fd..cb39f45 100644
--- a/drivers/gpu/drm/drm_mm.c
+++ b/drivers/gpu/drm/drm_mm.c
@@ -179,12 +179,14 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
 int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
 {
 	struct drm_mm_node *hole;
-	u64 end = node->start + node->size;
+	u64 end;
 	u64 hole_start;
 	u64 hole_end;
 
 	BUG_ON(node == NULL);
 
+	end = node->start + node->size;
+
 	/* Find the relevant hole to add our node to */
 	drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
 		if (hole_start > node->start || hole_end < end)
-- 
2.1.4

[toc] | [next] | [standalone]

#1405216

FromEric Engestrom <eric.engestrom@imgtec.com>
Date2016-05-23 12:30 +0200
Message-ID<rBVhv-4Bg-11@gated-at.bofh.it>
In reply to#1403261
On Wed, May 18, 2016 at 10:17:19PM +0200, Heinrich Schuchardt wrote:
> Do not dereference node before the check if node is NULL.
> 
> Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> ---
>  drivers/gpu/drm/drm_mm.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> index 04de6fd..cb39f45 100644
> --- a/drivers/gpu/drm/drm_mm.c
> +++ b/drivers/gpu/drm/drm_mm.c
> @@ -179,12 +179,14 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
>  int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
>  {
>  	struct drm_mm_node *hole;
> -	u64 end = node->start + node->size;
> +	u64 end;
>  	u64 hole_start;
>  	u64 hole_end;
>  
>  	BUG_ON(node == NULL);
>  
> +	end = node->start + node->size;
> +
>  	/* Find the relevant hole to add our node to */
>  	drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
>  		if (hole_start > node->start || hole_end < end)
> -- 
> 2.1.4

Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

[toc] | [prev] | [next] | [standalone]

#1405337

FromDaniel Vetter <daniel@ffwll.ch>
Date2016-05-23 15:00 +0200
Message-ID<rBXCF-5ZQ-3@gated-at.bofh.it>
In reply to#1405216
On Mon, May 23, 2016 at 11:27:14AM +0100, Eric Engestrom wrote:
> On Wed, May 18, 2016 at 10:17:19PM +0200, Heinrich Schuchardt wrote:
> > Do not dereference node before the check if node is NULL.
> > 
> > Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> > ---
> >  drivers/gpu/drm/drm_mm.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> > index 04de6fd..cb39f45 100644
> > --- a/drivers/gpu/drm/drm_mm.c
> > +++ b/drivers/gpu/drm/drm_mm.c
> > @@ -179,12 +179,14 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
> >  int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
> >  {
> >  	struct drm_mm_node *hole;
> > -	u64 end = node->start + node->size;
> > +	u64 end;
> >  	u64 hole_start;
> >  	u64 hole_end;
> >  
> >  	BUG_ON(node == NULL);
> >  
> > +	end = node->start + node->size;
> > +
> >  	/* Find the relevant hole to add our node to */
> >  	drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
> >  		if (hole_start > node->start || hole_end < end)
> > -- 
> > 2.1.4
> 
> Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

Applied to drm-misc, thanks.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

[toc] | [prev] | [next] | [standalone]

#1405351

FromChris Wilson <chris@chris-wilson.co.uk>
Date2016-05-23 15:10 +0200
Message-ID<rBXMm-6ir-29@gated-at.bofh.it>
In reply to#1405337
On Mon, May 23, 2016 at 02:56:45PM +0200, Daniel Vetter wrote:
> On Mon, May 23, 2016 at 11:27:14AM +0100, Eric Engestrom wrote:
> > On Wed, May 18, 2016 at 10:17:19PM +0200, Heinrich Schuchardt wrote:
> > > Do not dereference node before the check if node is NULL.
> > > 
> > > Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> > > ---
> > >  drivers/gpu/drm/drm_mm.c | 4 +++-
> > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> > > index 04de6fd..cb39f45 100644
> > > --- a/drivers/gpu/drm/drm_mm.c
> > > +++ b/drivers/gpu/drm/drm_mm.c
> > > @@ -179,12 +179,14 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
> > >  int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
> > >  {
> > >  	struct drm_mm_node *hole;
> > > -	u64 end = node->start + node->size;
> > > +	u64 end;
> > >  	u64 hole_start;
> > >  	u64 hole_end;
> > >  
> > >  	BUG_ON(node == NULL);
> > >  
> > > +	end = node->start + node->size;
> > > +
> > >  	/* Find the relevant hole to add our node to */
> > >  	drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
> > >  		if (hole_start > node->start || hole_end < end)
> > > -- 
> > > 2.1.4
> > 
> > Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

Remove useless check instead?
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

[toc] | [prev] | [next] | [standalone]

#1405372

FromEric Engestrom <eric.engestrom@imgtec.com>
Date2016-05-23 15:40 +0200
Message-ID<rBYfn-6sj-1@gated-at.bofh.it>
In reply to#1405351
On Mon, May 23, 2016 at 02:02:26PM +0100, Chris Wilson wrote:
> On Mon, May 23, 2016 at 02:56:45PM +0200, Daniel Vetter wrote:
> > On Mon, May 23, 2016 at 11:27:14AM +0100, Eric Engestrom wrote:
> > > On Wed, May 18, 2016 at 10:17:19PM +0200, Heinrich Schuchardt wrote:
> > > > Do not dereference node before the check if node is NULL.
> > > > 
> > > > Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> > > > ---
> > > >  drivers/gpu/drm/drm_mm.c | 4 +++-
> > > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> > > > index 04de6fd..cb39f45 100644
> > > > --- a/drivers/gpu/drm/drm_mm.c
> > > > +++ b/drivers/gpu/drm/drm_mm.c
> > > > @@ -179,12 +179,14 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
> > > >  int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
> > > >  {
> > > >  	struct drm_mm_node *hole;
> > > > -	u64 end = node->start + node->size;
> > > > +	u64 end;
> > > >  	u64 hole_start;
> > > >  	u64 hole_end;
> > > >  
> > > >  	BUG_ON(node == NULL);
> > > >  
> > > > +	end = node->start + node->size;
> > > > +
> > > >  	/* Find the relevant hole to add our node to */
> > > >  	drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
> > > >  		if (hole_start > node->start || hole_end < end)
> > > > -- 
> > > > 2.1.4
> > > 
> > > Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
> 
> Remove useless check instead?
> -Chris

I tend to prefer erring on the side of caution and have (too) many checks,
especially simple ones like avoiding null dereferences.

That said, BUG() might be too extreme. I'm not all that familiar with
the code, but it doesn't seem like it would be an unrecoverable failure.
WARN() instead?

[toc] | [prev] | [next] | [standalone]

#1405384

FromChris Wilson <chris@chris-wilson.co.uk>
Date2016-05-23 15:50 +0200
Message-ID<rBYp3-6vy-13@gated-at.bofh.it>
In reply to#1405372
On Mon, May 23, 2016 at 02:38:29PM +0100, Eric Engestrom wrote:
> On Mon, May 23, 2016 at 02:02:26PM +0100, Chris Wilson wrote:
> > On Mon, May 23, 2016 at 02:56:45PM +0200, Daniel Vetter wrote:
> > > On Mon, May 23, 2016 at 11:27:14AM +0100, Eric Engestrom wrote:
> > > > On Wed, May 18, 2016 at 10:17:19PM +0200, Heinrich Schuchardt wrote:
> > > > > Do not dereference node before the check if node is NULL.
> > > > > 
> > > > > Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> > > > > ---
> > > > >  drivers/gpu/drm/drm_mm.c | 4 +++-
> > > > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > > > 
> > > > > diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> > > > > index 04de6fd..cb39f45 100644
> > > > > --- a/drivers/gpu/drm/drm_mm.c
> > > > > +++ b/drivers/gpu/drm/drm_mm.c
> > > > > @@ -179,12 +179,14 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
> > > > >  int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
> > > > >  {
> > > > >  	struct drm_mm_node *hole;
> > > > > -	u64 end = node->start + node->size;
> > > > > +	u64 end;
> > > > >  	u64 hole_start;
> > > > >  	u64 hole_end;
> > > > >  
> > > > >  	BUG_ON(node == NULL);
> > > > >  
> > > > > +	end = node->start + node->size;
> > > > > +
> > > > >  	/* Find the relevant hole to add our node to */
> > > > >  	drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
> > > > >  		if (hole_start > node->start || hole_end < end)
> > > > > -- 
> > > > > 2.1.4
> > > > 
> > > > Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
> > 
> > Remove useless check instead?
> > -Chris
> 
> I tend to prefer erring on the side of caution and have (too) many checks,
> especially simple ones like avoiding null dereferences.

BUG(node == NULL) is no more informative than hitting a GFP with *node.

> That said, BUG() might be too extreme. I'm not all that familiar with
> the code, but it doesn't seem like it would be an unrecoverable failure.
> WARN() instead?

It's a programming error, just as would be passing in mm == NULL. Mark up
the function as requiring non-NULL parameters.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

[toc] | [prev] | [next] | [standalone]

#1405462

FromEric Engestrom <eric.engestrom@imgtec.com>
Date2016-05-23 17:40 +0200
Message-ID<rC07w-7GV-19@gated-at.bofh.it>
In reply to#1405384
On Mon, May 23, 2016 at 02:46:59PM +0100, Chris Wilson wrote:
> On Mon, May 23, 2016 at 02:38:29PM +0100, Eric Engestrom wrote:
> > On Mon, May 23, 2016 at 02:02:26PM +0100, Chris Wilson wrote:
> > > On Mon, May 23, 2016 at 02:56:45PM +0200, Daniel Vetter wrote:
> > > > On Mon, May 23, 2016 at 11:27:14AM +0100, Eric Engestrom wrote:
> > > > > On Wed, May 18, 2016 at 10:17:19PM +0200, Heinrich Schuchardt wrote:
> > > > > > Do not dereference node before the check if node is NULL.
> > > > > > 
> > > > > > Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> > > > > > ---
> > > > > >  drivers/gpu/drm/drm_mm.c | 4 +++-
> > > > > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > > > > 
> > > > > > diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> > > > > > index 04de6fd..cb39f45 100644
> > > > > > --- a/drivers/gpu/drm/drm_mm.c
> > > > > > +++ b/drivers/gpu/drm/drm_mm.c
> > > > > > @@ -179,12 +179,14 @@ static void drm_mm_insert_helper(struct drm_mm_node *hole_node,
> > > > > >  int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
> > > > > >  {
> > > > > >  	struct drm_mm_node *hole;
> > > > > > -	u64 end = node->start + node->size;
> > > > > > +	u64 end;
> > > > > >  	u64 hole_start;
> > > > > >  	u64 hole_end;
> > > > > >  
> > > > > >  	BUG_ON(node == NULL);
> > > > > >  
> > > > > > +	end = node->start + node->size;
> > > > > > +
> > > > > >  	/* Find the relevant hole to add our node to */
> > > > > >  	drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
> > > > > >  		if (hole_start > node->start || hole_end < end)
> > > > > > -- 
> > > > > > 2.1.4
> > > > > 
> > > > > Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
> > > 
> > > Remove useless check instead?
> > > -Chris
> > 
> > I tend to prefer erring on the side of caution and have (too) many checks,
> > especially simple ones like avoiding null dereferences.
> 
> BUG(node == NULL) is no more informative than hitting a GFP with *node.

Good point.

> 
> > That said, BUG() might be too extreme. I'm not all that familiar with
> > the code, but it doesn't seem like it would be an unrecoverable failure.
> > WARN() instead?
> 
> It's a programming error, just as would be passing in mm == NULL. Mark up
> the function as requiring non-NULL parameters.
> -Chris

Using `__attribute__((nonnull))`? I approve of that idea, as I find
those attributes extremely useful and try to get them used as much as
possible at $DAYJOB.

I have noticed however that this particular attribute isn't used much in
the kernel (4 times that I could find with a simple grep), to the point
of not even having a `__nonnull`. Any reason you know of?

Cheers,
  Eric

[toc] | [prev] | [next] | [standalone]

#1405474

FromChris Wilson <chris@chris-wilson.co.uk>
Date2016-05-23 18:00 +0200
Message-ID<rC0qS-7NU-17@gated-at.bofh.it>
In reply to#1405462
On Mon, May 23, 2016 at 04:35:53PM +0100, Eric Engestrom wrote:
> On Mon, May 23, 2016 at 02:46:59PM +0100, Chris Wilson wrote:
> > It's a programming error, just as would be passing in mm == NULL. Mark up
> > the function as requiring non-NULL parameters.
> 
> Using `__attribute__((nonnull))`? I approve of that idea, as I find
> those attributes extremely useful and try to get them used as much as
> possible at $DAYJOB.
> 
> I have noticed however that this particular attribute isn't used much in
> the kernel (4 times that I could find with a simple grep), to the point
> of not even having a `__nonnull`. Any reason you know of?

Not that I know of. Even more confusing is its use under tools but not
being added to the main kernel.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web