[PATCH] reset: Fix potential use-after-free in __of_reset_control_get()

[PATCH] reset: Fix potential use-after-free in __of_reset_control_get()

[Geert Uytterhoeven]

Calling of_node_put() decreases the reference count of a device tree
object, and may free some data.

However, the of_phandle_args structure embedding it is passed to
reset_controller_dev.of_xlate() after that, so it may still be accessed.

Move the call to of_node_put() down to fix this.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
 drivers/reset/core.c | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/drivers/reset/core.c b/drivers/reset/core.c
index 225e34c56b94a2e3..bc9df10d31b4bae1 100644
--- a/drivers/reset/core.c
+++ b/drivers/reset/core.c
@@ -496,27 +496,28 @@ struct reset_control *__of_reset_control_get(struct device_node *node,
 			break;
 		}
 	}
-	of_node_put(args.np);
 
 	if (!rcdev) {
-		mutex_unlock(&reset_list_mutex);
-		return ERR_PTR(-EPROBE_DEFER);
+		rstc = ERR_PTR(-EPROBE_DEFER);
+		goto out;
 	}
 
 	if (WARN_ON(args.args_count != rcdev->of_reset_n_cells)) {
-		mutex_unlock(&reset_list_mutex);
-		return ERR_PTR(-EINVAL);
+		rstc = ERR_PTR(-EINVAL);
+		goto out;
 	}
 
 	rstc_id = rcdev->of_xlate(rcdev, &args);
 	if (rstc_id < 0) {
-		mutex_unlock(&reset_list_mutex);
-		return ERR_PTR(rstc_id);
+		rstc = ERR_PTR(rstc_id);
+		goto out;
 	}
 
 	/* reset_list_mutex also protects the rcdev's reset_control list */
 	rstc = __reset_control_get_internal(rcdev, rstc_id, shared);
 
+out:
+	of_node_put(args.np);
 	mutex_unlock(&reset_list_mutex);
 
 	return rstc;
-- 
2.17.1

Re: [PATCH] reset: Fix potential use-after-free in __of_reset_control_get()

[Philipp Zabel]

Hi Geert,

On Mon, 2018-10-08 at 13:14 +0200, Geert Uytterhoeven wrote:
> Calling of_node_put() decreases the reference count of a device tree
> object, and may free some data.
> 
> However, the of_phandle_args structure embedding it is passed to
> reset_controller_dev.of_xlate() after that, so it may still be accessed.
> 
> Move the call to of_node_put() down to fix this.
> 
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
>  drivers/reset/core.c | 15 ++++++++-------
>  1 file changed, 8 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/reset/core.c b/drivers/reset/core.c
> index 225e34c56b94a2e3..bc9df10d31b4bae1 100644
> --- a/drivers/reset/core.c
> +++ b/drivers/reset/core.c
> @@ -496,27 +496,28 @@ struct reset_control *__of_reset_control_get(struct device_node *node,
>  			break;
>  		}
>  	}
> -	of_node_put(args.np);
>  
>  	if (!rcdev) {
> -		mutex_unlock(&reset_list_mutex);
> -		return ERR_PTR(-EPROBE_DEFER);
> +		rstc = ERR_PTR(-EPROBE_DEFER);
> +		goto out;
>  	}
>  
>  	if (WARN_ON(args.args_count != rcdev->of_reset_n_cells)) {
> -		mutex_unlock(&reset_list_mutex);
> -		return ERR_PTR(-EINVAL);
> +		rstc = ERR_PTR(-EINVAL);
> +		goto out;
>  	}
>  
>  	rstc_id = rcdev->of_xlate(rcdev, &args);
>  	if (rstc_id < 0) {
> -		mutex_unlock(&reset_list_mutex);
> -		return ERR_PTR(rstc_id);
> +		rstc = ERR_PTR(rstc_id);
> +		goto out;
>  	}
>  
>  	/* reset_list_mutex also protects the rcdev's reset_control list */
>  	rstc = __reset_control_get_internal(rcdev, rstc_id, shared);
>  
> +out:
> +	of_node_put(args.np);
>  	mutex_unlock(&reset_list_mutex);

Thank you for the patch. I'd like to move of_node_put after mutex_unlock
for symmetry. If you agree, I can switch the two when applying.

regards
Philipp

Re: [PATCH] reset: Fix potential use-after-free in __of_reset_control_get()

[Geert Uytterhoeven]

Hi Philipp,

On Mon, Oct 8, 2018 at 2:56 PM Philipp Zabel <p.zabel@pengutronix.de> wrote:
> On Mon, 2018-10-08 at 13:14 +0200, Geert Uytterhoeven wrote:
> > Calling of_node_put() decreases the reference count of a device tree
> > object, and may free some data.
> >
> > However, the of_phandle_args structure embedding it is passed to
> > reset_controller_dev.of_xlate() after that, so it may still be accessed.
> >
> > Move the call to of_node_put() down to fix this.
> >
> > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> > ---
> >  drivers/reset/core.c | 15 ++++++++-------
> >  1 file changed, 8 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/reset/core.c b/drivers/reset/core.c
> > index 225e34c56b94a2e3..bc9df10d31b4bae1 100644
> > --- a/drivers/reset/core.c
> > +++ b/drivers/reset/core.c
> > @@ -496,27 +496,28 @@ struct reset_control *__of_reset_control_get(struct device_node *node,
> >                       break;
> >               }
> >       }
> > -     of_node_put(args.np);
> >
> >       if (!rcdev) {
> > -             mutex_unlock(&reset_list_mutex);
> > -             return ERR_PTR(-EPROBE_DEFER);
> > +             rstc = ERR_PTR(-EPROBE_DEFER);
> > +             goto out;
> >       }
> >
> >       if (WARN_ON(args.args_count != rcdev->of_reset_n_cells)) {
> > -             mutex_unlock(&reset_list_mutex);
> > -             return ERR_PTR(-EINVAL);
> > +             rstc = ERR_PTR(-EINVAL);
> > +             goto out;
> >       }
> >
> >       rstc_id = rcdev->of_xlate(rcdev, &args);
> >       if (rstc_id < 0) {
> > -             mutex_unlock(&reset_list_mutex);
> > -             return ERR_PTR(rstc_id);
> > +             rstc = ERR_PTR(rstc_id);
> > +             goto out;
> >       }
> >
> >       /* reset_list_mutex also protects the rcdev's reset_control list */
> >       rstc = __reset_control_get_internal(rcdev, rstc_id, shared);
> >
> > +out:
> > +     of_node_put(args.np);
> >       mutex_unlock(&reset_list_mutex);
>
> Thank you for the patch. I'd like to move of_node_put after mutex_unlock
> for symmetry. If you agree, I can switch the two when applying.

No objection, thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] reset: Fix potential use-after-free in __of_reset_control_get()

[Philipp Zabel]

On Mon, 2018-10-08 at 15:12 +0200, Geert Uytterhoeven wrote:
> Hi Philipp,
> 
> On Mon, Oct 8, 2018 at 2:56 PM Philipp Zabel <p.zabel@pengutronix.de> wrote:
> > On Mon, 2018-10-08 at 13:14 +0200, Geert Uytterhoeven wrote:
> > > Calling of_node_put() decreases the reference count of a device tree
> > > object, and may free some data.
> > > 
> > > However, the of_phandle_args structure embedding it is passed to
> > > reset_controller_dev.of_xlate() after that, so it may still be accessed.
> > > 
> > > Move the call to of_node_put() down to fix this.
> > > 
> > > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> > > ---
> > >  drivers/reset/core.c | 15 ++++++++-------
> > >  1 file changed, 8 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/drivers/reset/core.c b/drivers/reset/core.c
> > > index 225e34c56b94a2e3..bc9df10d31b4bae1 100644
> > > --- a/drivers/reset/core.c
> > > +++ b/drivers/reset/core.c
> > > @@ -496,27 +496,28 @@ struct reset_control *__of_reset_control_get(struct device_node *node,
[...]
> > >       /* reset_list_mutex also protects the rcdev's reset_control list */
> > >       rstc = __reset_control_get_internal(rcdev, rstc_id, shared);
> > > 
> > > +out:
> > > +     of_node_put(args.np);
> > >       mutex_unlock(&reset_list_mutex);
> > 
> > Thank you for the patch. I'd like to move of_node_put after mutex_unlock
> > for symmetry. If you agree, I can switch the two when applying.
> 
> No objection, thanks!

Applied to reset/next with that change.

regards
Philipp