2017-03-30 10:27:28

by Vladis Dronov

[permalink] [raw]
Subject: [PATCH] kernel: drm/vmwgfx: limit the number of mip levels in vmw_gb_surface_define_ioctl()

The 'req->mip_levels' parameter in vmw_gb_surface_define_ioctl() is
a user-controlled 'uint32_t' value which is used as a loop count limit.
This can lead to a kernel lockup and DoS. Add check for 'req->mip_levels'.

References:
https://bugzilla.redhat.com/show_bug.cgi?id=1437431
Signed-off-by: Vladis Dronov <[email protected]>
---
drivers/gpu/drm/vmwgfx/vmwgfx_surface.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
index b445ce9..b30824b 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
@@ -1281,6 +1281,10 @@ int vmw_gb_surface_define_ioctl(struct drm_device *dev, void *data,
if (req->multisample_count != 0)
return -EINVAL;

+ if (req->mip_levels > DRM_VMW_MAX_SURFACE_FACES *
+ DRM_VMW_MAX_MIP_LEVELS)
+ return -EINVAL;
+
if (unlikely(vmw_user_surface_size == 0))
vmw_user_surface_size = ttm_round_pot(sizeof(*user_srf)) +
128;
--
2.9.3


2017-03-31 10:47:32

by Vladis Dronov

[permalink] [raw]
Subject: Re: [PATCH] kernel: drm/vmwgfx: limit the number of mip levels in vmw_gb_surface_define_ioctl()

This flaw was assigned an id CVE-2017-7346 by MITRE:

http://seclists.org/oss-sec/2017/q1/696

Best regards,
Vladis Dronov | Red Hat, Inc. | Product Security Engineer

----- Original Message -----
From: "Vladis Dronov" <[email protected]>
To: "VMware Graphics" <[email protected]>, "Sinclair Yeh" <[email protected]>, "Thomas Hellstrom" <[email protected]>, "David Airlie" <[email protected]>, [email protected], [email protected]
Cc: "Vladis Dronov" <[email protected]>
Sent: Thursday, March 30, 2017 12:27:12 PM
Subject: [PATCH] kernel: drm/vmwgfx: limit the number of mip levels in vmw_gb_surface_define_ioctl()

The 'req->mip_levels' parameter in vmw_gb_surface_define_ioctl() is
a user-controlled 'uint32_t' value which is used as a loop count limit.
This can lead to a kernel lockup and DoS. Add check for 'req->mip_levels'.

References:
https://bugzilla.redhat.com/show_bug.cgi?id=1437431
Signed-off-by: Vladis Dronov <[email protected]>

2017-03-31 15:07:28

by Sinclair Yeh

[permalink] [raw]
Subject: Re: [PATCH] kernel: drm/vmwgfx: limit the number of mip levels in vmw_gb_surface_define_ioctl()

Hi Vladis,


On Thu, Mar 30, 2017 at 12:27:12PM +0200, Vladis Dronov wrote:
> The 'req->mip_levels' parameter in vmw_gb_surface_define_ioctl() is
> a user-controlled 'uint32_t' value which is used as a loop count limit.
> This can lead to a kernel lockup and DoS. Add check for 'req->mip_levels'.
>
> References:
> https://urldefense.proofpoint.com/v2/url?u=https-3A__bugzilla.redhat.com_show-5Fbug.cgi-3Fid-3D1437431&d=DwIBAg&c=uilaK90D4TOVoH58JNXRgQ&r=HaJ2a6NYExoV0cntAYcoqA&m=5yR87BuuU86aoAjCveInxh6jvgOyumqIHQhTs0xLo38&s=tWQs7vwNLgD_b2RWMddVtusEKh9FQTIF5rKFOWudslE&e=
> Signed-off-by: Vladis Dronov <[email protected]>
> ---
> drivers/gpu/drm/vmwgfx/vmwgfx_surface.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
> index b445ce9..b30824b 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
> @@ -1281,6 +1281,10 @@ int vmw_gb_surface_define_ioctl(struct drm_device *dev, void *data,
> if (req->multisample_count != 0)
> return -EINVAL;
>
> + if (req->mip_levels > DRM_VMW_MAX_SURFACE_FACES *
> + DRM_VMW_MAX_MIP_LEVELS)
> + return -EINVAL;
> +

Here, the check should be "> DRM_VMW_MAX_MIP_LEVELS" because req->mip_levels
is only for one layer.

Also, as long as we can doing a check, I would suggest we check for 0 as
well.

Sinclair

2017-04-04 09:43:34

by Vladis Dronov

[permalink] [raw]
Subject: Re: [PATCH] kernel: drm/vmwgfx: limit the number of mip levels in vmw_gb_surface_define_ioctl()

Hello, Sinclair!

> Here, the check should be "> DRM_VMW_MAX_MIP_LEVELS" because req->mip_levels
> is only for one layer.

Got it, thanks!

> Also, as long as we can doing a check, I would suggest we check for 0 as
> well.

Do you mean a check for "req->mip_levels > 0" or for "req->mip_levels >= 0" ?

I glimpsed thru the code and I do not see problems with req->mip_levels being 0,
surely I may be mistaking.

Best regards,
Vladis Dronov | Red Hat, Inc. | Product Security Engineer


----- Original Message -----
From: "Sinclair Yeh" <[email protected]>
To: "Vladis Dronov" <[email protected]>
Cc: "VMware Graphics" <[email protected]>, "Thomas Hellstrom" <[email protected]>, "David Airlie" <[email protected]>, [email protected], [email protected]
Sent: Friday, March 31, 2017 5:07:12 PM
Subject: Re: [PATCH] kernel: drm/vmwgfx: limit the number of mip levels in vmw_gb_surface_define_ioctl()

Hi Vladis,


On Thu, Mar 30, 2017 at 12:27:12PM +0200, Vladis Dronov wrote:
> The 'req->mip_levels' parameter in vmw_gb_surface_define_ioctl() is
> a user-controlled 'uint32_t' value which is used as a loop count limit.
> This can lead to a kernel lockup and DoS. Add check for 'req->mip_levels'.
>
> References:
> https://urldefense.proofpoint.com/v2/url?u=https-3A__bugzilla.redhat.com_show-5Fbug.cgi-3Fid-3D1437431&d=DwIBAg&c=uilaK90D4TOVoH58JNXRgQ&r=HaJ2a6NYExoV0cntAYcoqA&m=5yR87BuuU86aoAjCveInxh6jvgOyumqIHQhTs0xLo38&s=tWQs7vwNLgD_b2RWMddVtusEKh9FQTIF5rKFOWudslE&e=
> Signed-off-by: Vladis Dronov <[email protected]>
> ---
> drivers/gpu/drm/vmwgfx/vmwgfx_surface.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
> index b445ce9..b30824b 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
> @@ -1281,6 +1281,10 @@ int vmw_gb_surface_define_ioctl(struct drm_device *dev, void *data,
> if (req->multisample_count != 0)
> return -EINVAL;
>
> + if (req->mip_levels > DRM_VMW_MAX_SURFACE_FACES *
> + DRM_VMW_MAX_MIP_LEVELS)
> + return -EINVAL;
> +

Here, the check should be "> DRM_VMW_MAX_MIP_LEVELS" because req->mip_levels
is only for one layer.

Also, as long as we can doing a check, I would suggest we check for 0 as
well.

Sinclair