[PATCH] drm/amdgpu/pm: Fix potential Spectre v1

[PATCH] drm/amdgpu/pm: Fix potential Spectre v1

[Gustavo A. R. Silva]

idx can be indirectly controlled by user-space, hence leading to a
potential exploitation of the Spectre variant 1 vulnerability.

This issue was detected with the help of Smatch:

drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c:408 amdgpu_set_pp_force_state()
warn: potential spectre issue 'data.states'

Fix this by sanitizing idx before using it to index data.states

Notice that given that speculation windows are large, the policy is
to kill the speculation on the first load and not worry if it can be
completed with a dependent load/store [1].

[1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2

Cc: stable@vger.kernel.org
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
index 15a1192..a446c7c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
@@ -31,7 +31,7 @@
 #include <linux/power_supply.h>
 #include <linux/hwmon.h>
 #include <linux/hwmon-sysfs.h>
-
+#include <linux/nospec.h>
 
 static int amdgpu_debugfs_pm_init(struct amdgpu_device *adev);
 
@@ -403,6 +403,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev,
 			count = -EINVAL;
 			goto fail;
 		}
+		idx = array_index_nospec(idx, ARRAY_SIZE(data.states));
 
 		amdgpu_dpm_get_pp_num_states(adev, &data);
 		state = data.states[idx];
-- 
2.7.4

Re: [PATCH] drm/amdgpu/pm: Fix potential Spectre v1

[Alex Deucher]

On Mon, Jul 23, 2018 at 12:32 PM, Gustavo A. R. Silva
<gustavo@embeddedor.com> wrote:
> idx can be indirectly controlled by user-space, hence leading to a
> potential exploitation of the Spectre variant 1 vulnerability.
>
> This issue was detected with the help of Smatch:
>
> drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c:408 amdgpu_set_pp_force_state()
> warn: potential spectre issue 'data.states'
>
> Fix this by sanitizing idx before using it to index data.states

Is this actually necessary?  We already check that idx is valid a few
lines before:
        if (ret || idx >= ARRAY_SIZE(data.states)) {
                        count = -EINVAL;
                        goto fail;
                }

Alex


>
> Notice that given that speculation windows are large, the policy is
> to kill the speculation on the first load and not worry if it can be
> completed with a dependent load/store [1].
>
> [1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2
>
> Cc: stable@vger.kernel.org
> Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
> index 15a1192..a446c7c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
> @@ -31,7 +31,7 @@
>  #include <linux/power_supply.h>
>  #include <linux/hwmon.h>
>  #include <linux/hwmon-sysfs.h>
> -
> +#include <linux/nospec.h>
>
>  static int amdgpu_debugfs_pm_init(struct amdgpu_device *adev);
>
> @@ -403,6 +403,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev,
>                         count = -EINVAL;
>                         goto fail;
>                 }
> +               idx = array_index_nospec(idx, ARRAY_SIZE(data.states));
>
>                 amdgpu_dpm_get_pp_num_states(adev, &data);
>                 state = data.states[idx];
> --
> 2.7.4
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [PATCH] drm/amdgpu/pm: Fix potential Spectre v1

[Michel Dänzer]

On 2018-07-24 10:53 PM, Alex Deucher wrote:
> On Mon, Jul 23, 2018 at 12:32 PM, Gustavo A. R. Silva
> <gustavo@embeddedor.com> wrote:
>> idx can be indirectly controlled by user-space, hence leading to a
>> potential exploitation of the Spectre variant 1 vulnerability.
>>
>> This issue was detected with the help of Smatch:
>>
>> drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c:408 amdgpu_set_pp_force_state()
>> warn: potential spectre issue 'data.states'
>>
>> Fix this by sanitizing idx before using it to index data.states
> 
> Is this actually necessary?  We already check that idx is valid a few
> lines before:
>         if (ret || idx >= ARRAY_SIZE(data.states)) {
>                         count = -EINVAL;
>                         goto fail;
>                 }

A Spectre attack would be based on idx ending up too large, but the CPU
speculatively executing the following code assuming idx <
ARRAY_SIZE(data.states), and extracting information from the incorrectly
speculated code via side channels.

I'm not sure if that's actually possible in this case, but better safe
than sorry?


-- 
Earthling Michel Dänzer               |               http://www.amd.com
Libre software enthusiast             |             Mesa and X developer

Re: [PATCH] drm/amdgpu/pm: Fix potential Spectre v1

[Alex Deucher]

On Mon, Jul 30, 2018 at 5:55 AM, Michel Dänzer <michel@daenzer.net> wrote:
> On 2018-07-24 10:53 PM, Alex Deucher wrote:
>> On Mon, Jul 23, 2018 at 12:32 PM, Gustavo A. R. Silva
>> <gustavo@embeddedor.com> wrote:
>>> idx can be indirectly controlled by user-space, hence leading to a
>>> potential exploitation of the Spectre variant 1 vulnerability.
>>>
>>> This issue was detected with the help of Smatch:
>>>
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c:408 amdgpu_set_pp_force_state()
>>> warn: potential spectre issue 'data.states'
>>>
>>> Fix this by sanitizing idx before using it to index data.states
>>
>> Is this actually necessary?  We already check that idx is valid a few
>> lines before:
>>         if (ret || idx >= ARRAY_SIZE(data.states)) {
>>                         count = -EINVAL;
>>                         goto fail;
>>                 }
>
> A Spectre attack would be based on idx ending up too large, but the CPU
> speculatively executing the following code assuming idx <
> ARRAY_SIZE(data.states), and extracting information from the incorrectly
> speculated code via side channels.
>
> I'm not sure if that's actually possible in this case, but better safe
> than sorry?

Yeah, I'm not sure.  I guess this can't hurt.

Alex

Re: [PATCH] drm/amdgpu/pm: Fix potential Spectre v1

[Christian König]

Am 30.07.2018 um 22:14 schrieb Alex Deucher:
> On Mon, Jul 30, 2018 at 5:55 AM, Michel Dänzer <michel@daenzer.net> wrote:
>> On 2018-07-24 10:53 PM, Alex Deucher wrote:
>>> On Mon, Jul 23, 2018 at 12:32 PM, Gustavo A. R. Silva
>>> <gustavo@embeddedor.com> wrote:
>>>> idx can be indirectly controlled by user-space, hence leading to a
>>>> potential exploitation of the Spectre variant 1 vulnerability.
>>>>
>>>> This issue was detected with the help of Smatch:
>>>>
>>>> drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c:408 amdgpu_set_pp_force_state()
>>>> warn: potential spectre issue 'data.states'
>>>>
>>>> Fix this by sanitizing idx before using it to index data.states
>>> Is this actually necessary?  We already check that idx is valid a few
>>> lines before:
>>>          if (ret || idx >= ARRAY_SIZE(data.states)) {
>>>                          count = -EINVAL;
>>>                          goto fail;
>>>                  }
>> A Spectre attack would be based on idx ending up too large, but the CPU
>> speculatively executing the following code assuming idx <
>> ARRAY_SIZE(data.states), and extracting information from the incorrectly
>> speculated code via side channels.
>>
>> I'm not sure if that's actually possible in this case, but better safe
>> than sorry?
> Yeah, I'm not sure.  I guess this can't hurt.

Well is idx actually controlable by userspace in an IOCTL? I guess the 
answer is no.

On the other hand the array_index_nospec() macro makes the overhead 
absolute negligible.

So I agree that we should be better safe than sorry.

Christian.

>
> Alex

Re: [PATCH] drm/amdgpu/pm: Fix potential Spectre v1

[Alex Deucher]

On Tue, Jul 31, 2018 at 2:46 AM, Christian König
<christian.koenig@amd.com> wrote:
> Am 30.07.2018 um 22:14 schrieb Alex Deucher:
>>
>> On Mon, Jul 30, 2018 at 5:55 AM, Michel Dänzer <michel@daenzer.net> wrote:
>>>
>>> On 2018-07-24 10:53 PM, Alex Deucher wrote:
>>>>
>>>> On Mon, Jul 23, 2018 at 12:32 PM, Gustavo A. R. Silva
>>>> <gustavo@embeddedor.com> wrote:
>>>>>
>>>>> idx can be indirectly controlled by user-space, hence leading to a
>>>>> potential exploitation of the Spectre variant 1 vulnerability.
>>>>>
>>>>> This issue was detected with the help of Smatch:
>>>>>
>>>>> drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c:408 amdgpu_set_pp_force_state()
>>>>> warn: potential spectre issue 'data.states'
>>>>>
>>>>> Fix this by sanitizing idx before using it to index data.states
>>>>
>>>> Is this actually necessary?  We already check that idx is valid a few
>>>> lines before:
>>>>          if (ret || idx >= ARRAY_SIZE(data.states)) {
>>>>                          count = -EINVAL;
>>>>                          goto fail;
>>>>                  }
>>>
>>> A Spectre attack would be based on idx ending up too large, but the CPU
>>> speculatively executing the following code assuming idx <
>>> ARRAY_SIZE(data.states), and extracting information from the incorrectly
>>> speculated code via side channels.
>>>
>>> I'm not sure if that's actually possible in this case, but better safe
>>> than sorry?
>>
>> Yeah, I'm not sure.  I guess this can't hurt.
>
>
> Well is idx actually controlable by userspace in an IOCTL? I guess the
> answer is no.
>
> On the other hand the array_index_nospec() macro makes the overhead absolute
> negligible.
>
> So I agree that we should be better safe than sorry.

Ok.  Applied.

Thanks,

Alex