Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Message-ID: <006e29037a9314fb286a5d6d84905bce99cef6c7.camel@perches.com>
Subject: Re: [PATCH] drm/amdgpu: allocate large structures dynamically
From:   Joe Perches <joe@perches.com>
To:     Arnd Bergmann <arnd@arndb.de>,
        Alex Deucher <alexander.deucher@amd.com>,
        Christian =?ISO-8859-1?Q?K=F6nig?= <christian.koenig@amd.com>,
        "David (ChunMing) Zhou" <David1.Zhou@amd.com>,
        David Airlie <airlied@linux.ie>,
        Daniel Vetter <daniel@ffwll.ch>,
        Hawking Zhang <Hawking.Zhang@amd.com>,
        John Clements <john.clements@amd.com>
Cc:     Tao Zhou <tao.zhou1@amd.com>, Guchun Chen <guchun.chen@amd.com>,
        Dennis Li <Dennis.Li@amd.com>, amd-gfx@lists.freedesktop.org,
        dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org
Date:   Wed, 06 May 2020 12:01:41 -0700
In-Reply-To: <20200505140208.284473-1-arnd@arndb.de>
References: <20200505140208.284473-1-arnd@arndb.de>
Content-Type: text/plain; charset="ISO-8859-1"
User-Agent: Evolution 3.36.1-2 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

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)))
 		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()) {