[PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[Navid Emamdoost]

In the impelementation of v3d_submit_cl_ioctl() there are two memory
leaks. One is when allocation for bin fails, and the other is when bin
initialization fails. If kcalloc fails to allocate memory for bin then
render->base should be put. Also, if v3d_job_init() fails to initialize
bin->base then allocated memory for bin should be released.

Fixes: a783a09ee76d ("drm/v3d: Refactor job management.")
Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
---
 drivers/gpu/drm/v3d/v3d_gem.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
index 5d80507b539b..19c092d75266 100644
--- a/drivers/gpu/drm/v3d/v3d_gem.c
+++ b/drivers/gpu/drm/v3d/v3d_gem.c
@@ -557,13 +557,16 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
 
 	if (args->bcl_start != args->bcl_end) {
 		bin = kcalloc(1, sizeof(*bin), GFP_KERNEL);
-		if (!bin)
+		if (!bin) {
+			v3d_job_put(&render->base);
 			return -ENOMEM;
+		}
 
 		ret = v3d_job_init(v3d, file_priv, &bin->base,
 				   v3d_job_free, args->in_sync_bcl);
 		if (ret) {
 			v3d_job_put(&render->base);
+			kfree(bin);
 			return ret;
 		}
 
-- 
2.17.1

Re: [PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[Markus Elfring]

> In the impelementation of v3d_submit_cl_ioctl() there are two memory leaks.

Please avoid another typo also in this change description.


> … If kcalloc fails to allocate memory for bin then
> render->base should be put. Also, if v3d_job_init() fails to initialize
> bin->base then allocated memory for bin should be released.

Will an “imperative mood” be more appropriate for such wordings?
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?id=7d194c2100ad2a6dded545887d02754948ca5241#n151


…
> +++ b/drivers/gpu/drm/v3d/v3d_gem.c
> @@ -557,13 +557,16 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
…
>  		if (ret) {
>  			v3d_job_put(&render->base);
> +			kfree(bin);
…

Can it be helpful to move the added function call before the other
in this if branch (if you prefer to avoid the addition of a jump target here)?

Regards,
Markus

Re: [PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[Markus Elfring]

> …
>> +++ b/drivers/gpu/drm/v3d/v3d_gem.c
>> @@ -557,13 +557,16 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
> …
>>  		if (ret) {
>>  			v3d_job_put(&render->base);
>> +			kfree(bin);
> …
>
> Can it be helpful to move the added function call before the other
> in this if branch (if you prefer to avoid the addition of a jump target here)?

I got into the mood to take another look at these implementation details.
I suggest then to look at the commit 0d352a3a8a1f26168d09f7073e61bb4b328e3bb9
("drm/v3d: don't leak bin job if v3d_job_init fails." from 2019-09-18)
once more.

With which software versions did you perform your source code analysis?

Regards,
Markus

Re: [PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[Daniel Vetter]

On Mon, Oct 21, 2019 at 01:52:49PM -0500, Navid Emamdoost wrote:
> In the impelementation of v3d_submit_cl_ioctl() there are two memory
> leaks. One is when allocation for bin fails, and the other is when bin
> initialization fails. If kcalloc fails to allocate memory for bin then
> render->base should be put. Also, if v3d_job_init() fails to initialize
> bin->base then allocated memory for bin should be released.
> 
> Fixes: a783a09ee76d ("drm/v3d: Refactor job management.")
> Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
> ---
>  drivers/gpu/drm/v3d/v3d_gem.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
> index 5d80507b539b..19c092d75266 100644
> --- a/drivers/gpu/drm/v3d/v3d_gem.c
> +++ b/drivers/gpu/drm/v3d/v3d_gem.c
> @@ -557,13 +557,16 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
>  
>  	if (args->bcl_start != args->bcl_end) {
>  		bin = kcalloc(1, sizeof(*bin), GFP_KERNEL);
> -		if (!bin)
> +		if (!bin) {
> +			v3d_job_put(&render->base);

The job isn't initialized yet, this doesn't work.

>  			return -ENOMEM;
> +		}
>  
>  		ret = v3d_job_init(v3d, file_priv, &bin->base,
>  				   v3d_job_free, args->in_sync_bcl);
>  		if (ret) {
>  			v3d_job_put(&render->base);

v3d_job_put will call kfree, if you chase the callchain long enough (in
v3d_job_free). So no bug here, this would lead to a double kfree and
crash.
-Daniel

> +			kfree(bin);
>  			return ret;
>  		}
>  
> -- 
> 2.17.1
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

Re: [PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[Navid Emamdoost]

On Tue, Oct 22, 2019 at 4:36 AM Daniel Vetter <daniel@ffwll.ch> wrote:
>
> On Mon, Oct 21, 2019 at 01:52:49PM -0500, Navid Emamdoost wrote:
> > In the impelementation of v3d_submit_cl_ioctl() there are two memory
> > leaks. One is when allocation for bin fails, and the other is when bin
> > initialization fails. If kcalloc fails to allocate memory for bin then
> > render->base should be put. Also, if v3d_job_init() fails to initialize
> > bin->base then allocated memory for bin should be released.
> >
> > Fixes: a783a09ee76d ("drm/v3d: Refactor job management.")
> > Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
> > ---
> >  drivers/gpu/drm/v3d/v3d_gem.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
> > index 5d80507b539b..19c092d75266 100644
> > --- a/drivers/gpu/drm/v3d/v3d_gem.c
> > +++ b/drivers/gpu/drm/v3d/v3d_gem.c
> > @@ -557,13 +557,16 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
> >
> >       if (args->bcl_start != args->bcl_end) {
> >               bin = kcalloc(1, sizeof(*bin), GFP_KERNEL);
> > -             if (!bin)
> > +             if (!bin) {
> > +                     v3d_job_put(&render->base);
>
> The job isn't initialized yet, this doesn't work.
Do you mean we have to release render via kfree() here?

>
> >                       return -ENOMEM;
> > +             }
> >
> >               ret = v3d_job_init(v3d, file_priv, &bin->base,
> >                                  v3d_job_free, args->in_sync_bcl);
> >               if (ret) {
> >                       v3d_job_put(&render->base);
>
> v3d_job_put will call kfree, if you chase the callchain long enough (in
> v3d_job_free). So no bug here, this would lead to a double kfree and
> crash.
Yes, v3d_job_put() takes care of render,

> -Daniel
>
> > +                     kfree(bin);
but how about leaking bin?

> >                       return ret;
> >               }
> >
> > --
> > 2.17.1
> >
>
> --
> Daniel Vetter
> Software Engineer, Intel Corporation
> http://blog.ffwll.ch



-- 
Navid.

Re: [PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[Daniel Vetter]

On Tue, Oct 22, 2019 at 10:53:57PM -0500, Navid Emamdoost wrote:
> On Tue, Oct 22, 2019 at 4:36 AM Daniel Vetter <daniel@ffwll.ch> wrote:
> >
> > On Mon, Oct 21, 2019 at 01:52:49PM -0500, Navid Emamdoost wrote:
> > > In the impelementation of v3d_submit_cl_ioctl() there are two memory
> > > leaks. One is when allocation for bin fails, and the other is when bin
> > > initialization fails. If kcalloc fails to allocate memory for bin then
> > > render->base should be put. Also, if v3d_job_init() fails to initialize
> > > bin->base then allocated memory for bin should be released.
> > >
> > > Fixes: a783a09ee76d ("drm/v3d: Refactor job management.")
> > > Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
> > > ---
> > >  drivers/gpu/drm/v3d/v3d_gem.c | 5 ++++-
> > >  1 file changed, 4 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
> > > index 5d80507b539b..19c092d75266 100644
> > > --- a/drivers/gpu/drm/v3d/v3d_gem.c
> > > +++ b/drivers/gpu/drm/v3d/v3d_gem.c
> > > @@ -557,13 +557,16 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
> > >
> > >       if (args->bcl_start != args->bcl_end) {
> > >               bin = kcalloc(1, sizeof(*bin), GFP_KERNEL);
> > > -             if (!bin)
> > > +             if (!bin) {
> > > +                     v3d_job_put(&render->base);
> >
> > The job isn't initialized yet, this doesn't work.
> Do you mean we have to release render via kfree() here?
> 
> >
> > >                       return -ENOMEM;
> > > +             }
> > >
> > >               ret = v3d_job_init(v3d, file_priv, &bin->base,
> > >                                  v3d_job_free, args->in_sync_bcl);
> > >               if (ret) {
> > >                       v3d_job_put(&render->base);
> >
> > v3d_job_put will call kfree, if you chase the callchain long enough (in
> > v3d_job_free). So no bug here, this would lead to a double kfree and
> > crash.
> Yes, v3d_job_put() takes care of render,
> 
> > -Daniel
> >
> > > +                     kfree(bin);
> but how about leaking bin?

Sorry, I totally missed that this is bin, no render. Patch looks correct
to me.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

> 
> > >                       return ret;
> > >               }
> > >
> > > --
> > > 2.17.1
> > >
> >
> > --
> > Daniel Vetter
> > Software Engineer, Intel Corporation
> > http://blog.ffwll.ch
> 
> 
> 
> -- 
> Navid.

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

Re: [PATCH] drm/v3d: Fix memory leak in v3d_submit_cl_ioctl

[Daniel Vetter]

On Wed, Oct 23, 2019 at 11:16:01AM +0200, Daniel Vetter wrote:
> On Tue, Oct 22, 2019 at 10:53:57PM -0500, Navid Emamdoost wrote:
> > On Tue, Oct 22, 2019 at 4:36 AM Daniel Vetter <daniel@ffwll.ch> wrote:
> > >
> > > On Mon, Oct 21, 2019 at 01:52:49PM -0500, Navid Emamdoost wrote:
> > > > In the impelementation of v3d_submit_cl_ioctl() there are two memory
> > > > leaks. One is when allocation for bin fails, and the other is when bin
> > > > initialization fails. If kcalloc fails to allocate memory for bin then
> > > > render->base should be put. Also, if v3d_job_init() fails to initialize
> > > > bin->base then allocated memory for bin should be released.
> > > >
> > > > Fixes: a783a09ee76d ("drm/v3d: Refactor job management.")
> > > > Signed-off-by: Navid Emamdoost <navid.emamdoost@gmail.com>
> > > > ---
> > > >  drivers/gpu/drm/v3d/v3d_gem.c | 5 ++++-
> > > >  1 file changed, 4 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
> > > > index 5d80507b539b..19c092d75266 100644
> > > > --- a/drivers/gpu/drm/v3d/v3d_gem.c
> > > > +++ b/drivers/gpu/drm/v3d/v3d_gem.c
> > > > @@ -557,13 +557,16 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
> > > >
> > > >       if (args->bcl_start != args->bcl_end) {
> > > >               bin = kcalloc(1, sizeof(*bin), GFP_KERNEL);
> > > > -             if (!bin)
> > > > +             if (!bin) {
> > > > +                     v3d_job_put(&render->base);
> > >
> > > The job isn't initialized yet, this doesn't work.
> > Do you mean we have to release render via kfree() here?
> > 
> > >
> > > >                       return -ENOMEM;
> > > > +             }
> > > >
> > > >               ret = v3d_job_init(v3d, file_priv, &bin->base,
> > > >                                  v3d_job_free, args->in_sync_bcl);
> > > >               if (ret) {
> > > >                       v3d_job_put(&render->base);
> > >
> > > v3d_job_put will call kfree, if you chase the callchain long enough (in
> > > v3d_job_free). So no bug here, this would lead to a double kfree and
> > > crash.
> > Yes, v3d_job_put() takes care of render,
> > 
> > > -Daniel
> > >
> > > > +                     kfree(bin);
> > but how about leaking bin?
> 
> Sorry, I totally missed that this is bin, no render. Patch looks correct
> to me.
> 
> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Double-checked with Eric and applied to drm-misc-fixes.
-Daniel

> 
> > 
> > > >                       return ret;
> > > >               }
> > > >
> > > > --
> > > > 2.17.1
> > > >
> > >
> > > --
> > > Daniel Vetter
> > > Software Engineer, Intel Corporation
> > > http://blog.ffwll.ch
> > 
> > 
> > 
> > -- 
> > Navid.
> 
> -- 
> Daniel Vetter
> Software Engineer, Intel Corporation
> http://blog.ffwll.ch

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch