linux-kernel - Re: [PATCH] drm/amdgpu: allocate large structures dynamically

Open Source and information security mailing list archives

Message-ID: <845a2e12-538c-fb6d-db80-78e8adf1413f@gmail.com>
Date:   Thu, 7 May 2020 08:42:51 +0200
From:   Christian König <ckoenig.leichtzumerken@...il.com>
To:     Joe Perches <joe@...ches.com>, Arnd Bergmann <arnd@...db.de>,
        Alex Deucher <alexander.deucher@....com>,
        Christian König <christian.koenig@....com>,
        "David (ChunMing) Zhou" <David1.Zhou@....com>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>,
        Hawking Zhang <Hawking.Zhang@....com>,
        John Clements <john.clements@....com>
Cc:     Guchun Chen <guchun.chen@....com>, Tao Zhou <tao.zhou1@....com>,
        linux-kernel@...r.kernel.org, amd-gfx@...ts.freedesktop.org,
        dri-devel@...ts.freedesktop.org, Dennis Li <Dennis.Li@....com>
Subject: Re: [PATCH] drm/amdgpu: allocate large structures dynamically

Am 06.05.20 um 21:01 schrieb Joe Perches:
> On Tue, 2020-05-05 at 16:01 +0200, Arnd Bergmann wrote:
>> After the structure was padded to 1024 bytes, it is no longer
>> suitable for being a local variable, as the function surpasses
>> the warning limit for 32-bit architectures:
>>
>> drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:587:5: error: stack frame size of 1072 bytes in function 'amdgpu_ras_feature_enable' [-Werror,-Wframe-larger-than=]
>> int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
>>      ^
>>
>> Use kzalloc() instead to get it from the heap.
> []
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> []
>> @@ -588,19 +588,23 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
>>   		struct ras_common_if *head, bool enable)
>>   {
>>   	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
>> -	union ta_ras_cmd_input info;
>> +	union ta_ras_cmd_input *info;
>>   	int ret;
>>   
>>   	if (!con)
>>   		return -EINVAL;
>>   
>> +        info = kzalloc(sizeof(union ta_ras_cmd_input), GFP_KERNEL);
> Spaces were used for indentation here not a tab.
> It might be useful to run your proposed patches through checkpatch
>
> Is this an actual bug fix as the previous use didn't
> zero unused info members?
>
>> +	if (!info)
>> +		return -ENOMEM;
>> +
>>   	if (!enable) {
>> -		info.disable_features = (struct ta_ras_disable_features_input) {
>> +		info->disable_features = (struct ta_ras_disable_features_input) {
>>   			.block_id =  amdgpu_ras_block_to_ta(head->block),
>>   			.error_type = amdgpu_ras_error_to_ta(head->type),
>>   		};
>>   	} else {
>> -		info.enable_features = (struct ta_ras_enable_features_input) {
>> +		info->enable_features = (struct ta_ras_enable_features_input) {
>>   			.block_id =  amdgpu_ras_block_to_ta(head->block),
>>   			.error_type = amdgpu_ras_error_to_ta(head->type),
>>   		};
>> @@ -609,26 +613,33 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
>>   	/* Do not enable if it is not allowed. */
>>   	WARN_ON(enable && !amdgpu_ras_is_feature_allowed(adev, head));
>>   	/* Are we alerady in that state we are going to set? */
>> -	if (!(!!enable ^ !!amdgpu_ras_is_feature_enabled(adev, head)))
>> -		return 0;
>> +	if (!(!!enable ^ !!amdgpu_ras_is_feature_enabled(adev, head))) {
> And trivia:
>
> The !! uses with bool seem unnecessary and it's probably better
> to make amdgpu_ras_is_feature_enabled to return bool.
>
> Maybe something like:
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 12 ++++++------
>   1 file changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> index 538895cfd862..05c4eaf0ddfa 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> @@ -526,16 +526,16 @@ void amdgpu_ras_parse_status_code(struct amdgpu_device* adev,
>   }
>   
>   /* feature ctl begin */
> -static int amdgpu_ras_is_feature_allowed(struct amdgpu_device *adev,
> -		struct ras_common_if *head)
> +static bool amdgpu_ras_is_feature_allowed(struct amdgpu_device *adev,
> +					  struct ras_common_if *head)
>   {
>   	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
>   
>   	return con->hw_supported & BIT(head->block);
>   }
>   
> -static int amdgpu_ras_is_feature_enabled(struct amdgpu_device *adev,
> -		struct ras_common_if *head)
> +static bool amdgpu_ras_is_feature_enabled(struct amdgpu_device *adev,
> +					  struct ras_common_if *head)
>   {
>   	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
>   
> @@ -560,7 +560,7 @@ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev,
>   	 */
>   	if (!amdgpu_ras_is_feature_allowed(adev, head))
>   		return 0;
> -	if (!(!!enable ^ !!amdgpu_ras_is_feature_enabled(adev, head)))
> +	if (!(enable ^ amdgpu_ras_is_feature_enabled(adev, head)))

And while we are at improving coding style I think that writing this as 
"if (enabled == amdgpu_ras_is_feature_enabled(adev, head))" would be 
much more readable.

Christian.

>   		return 0;
>   
>   	if (enable) {
> @@ -609,7 +609,7 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
>   	/* Do not enable if it is not allowed. */
>   	WARN_ON(enable && !amdgpu_ras_is_feature_allowed(adev, head));
>   	/* Are we alerady in that state we are going to set? */
> -	if (!(!!enable ^ !!amdgpu_ras_is_feature_enabled(adev, head)))
> +	if (!(enable ^ amdgpu_ras_is_feature_enabled(adev, head)))
>   		return 0;
>   
>   	if (!amdgpu_ras_intr_triggered()) {
>
>
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@...ts.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives