linux-kernel - Re: [PATCH RFC RFT] drm/msm: adreno: attach the GMU device to a driver

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <c8058713-b126-461b-8ae6-19c4574a8105@oss.qualcomm.com>
Date: Fri, 31 Oct 2025 02:59:40 +0530
From: Akhil P Oommen <akhilpo@....qualcomm.com>
To: Neil Armstrong <neil.armstrong@...aro.org>
Cc: linux-arm-msm@...r.kernel.org, dri-devel@...ts.freedesktop.org,
        freedreno@...ts.freedesktop.org, linux-kernel@...r.kernel.org,
        Rob Clark <robin.clark@....qualcomm.com>, Sean Paul <sean@...rly.run>,
        Konrad Dybcio <konradybcio@...nel.org>,
        Dmitry Baryshkov <lumag@...nel.org>,
        Abhinav Kumar <abhinav.kumar@...ux.dev>,
        Jessica Zhang <jessica.zhang@....qualcomm.com>,
        Marijn Suijten <marijn.suijten@...ainline.org>,
        David Airlie <airlied@...il.com>, Simona Vetter <simona@...ll.ch>
Subject: Re: [PATCH RFC RFT] drm/msm: adreno: attach the GMU device to a
 driver

On 10/22/2025 6:14 PM, Neil Armstrong wrote:
> Due to the sync_state is enabled by default in pmdomain & CCF since v6.17,
> the GCC and GPUCC sync_state would stay pending, leaving the resources in
> full performance:
> gcc-x1e80100 100000.clock-controller: sync_state() pending due to 3d6a000.gmu
> gpucc-x1e80100 3d90000.clock-controller: sync_state() pending due to 3d6a000.gmu
> 
> In order to fix this state and allow the GMU to be properly
> probed, let's add a proper driver for the GMU and add it to
> the MSM driver components.
> 
> Only the proper GMU has been tested since I don't have
> access to hardware with a GMU wrapper.
> 
> Signed-off-by: Neil Armstrong <neil.armstrong@...aro.org>
> ---
>  drivers/gpu/drm/msm/adreno/a6xx_gmu.c      | 354 ++++++++++++++---------------
>  drivers/gpu/drm/msm/adreno/a6xx_gpu.c      |   6 -
>  drivers/gpu/drm/msm/adreno/a6xx_gpu.h      |   3 -
>  drivers/gpu/drm/msm/adreno/adreno_device.c |   4 +
>  drivers/gpu/drm/msm/adreno/adreno_gpu.h    |   4 +
>  drivers/gpu/drm/msm/msm_drv.c              |  16 +-
>  6 files changed, 192 insertions(+), 195 deletions(-)
> 
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
> index fc62fef2fed8..6e7c3e627509 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
> @@ -1859,11 +1859,14 @@ void a6xx_gmu_sysprof_setup(struct msm_gpu *gpu)
>  	pm_runtime_put(&gpu->pdev->dev);
>  }
>  
> -void a6xx_gmu_remove(struct a6xx_gpu *a6xx_gpu)
> +static void a6xx_gmu_unbind(struct device *dev, struct device *master, void *data)
>  {

I feel we should keep gmu and gmu_wrapper implementations separate. It
is already overloaded. How about adding a separate gmu_wrapper_bind_ops
and keep it in the match data?

> -	struct adreno_gpu *adreno_gpu = &a6xx_gpu->base;
> +	struct platform_device *pdev = to_platform_device(dev);
> +	struct msm_drm_private *priv = dev_get_drvdata(master);
> +	struct msm_gpu *gpu = priv->gpu;

<< snip >>

>  static inline uint32_t get_wptr(struct msm_ringbuffer *ring)
>  {
> diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
> index 7e977fec4100..0618da7e8b40 100644
> --- a/drivers/gpu/drm/msm/msm_drv.c
> +++ b/drivers/gpu/drm/msm/msm_drv.c
> @@ -998,18 +998,30 @@ static const struct of_device_id msm_gpu_match[] = {
>  	{ },
>  };
>  
> +static const struct of_device_id msm_gmu_match[] = {
> +	{ .compatible = "qcom,adreno-gmu" },
> +	{ .compatible = "qcom,adreno-gmu-wrapper" },
> +	{ },
> +};
> +
>  static int add_gpu_components(struct device *dev,
>  			      struct component_match **matchptr)
>  {
> -	struct device_node *np;
> +	struct device_node *np, *gmu;
>  
>  	np = of_find_matching_node(NULL, msm_gpu_match);
>  	if (!np)
>  		return 0;
>  
> -	if (of_device_is_available(np) && adreno_has_gpu(np))
> +	if (of_device_is_available(np) && adreno_has_gpu(np)) {
>  		drm_of_component_match_add(dev, matchptr, component_compare_of, np);
>  
> +		gmu = of_find_matching_node(NULL, msm_gmu_match);

Instead of this, we can probably use the gmu phandle from "qcom,gmu"
property? That is quicker and also doesn't assume that there is only a
single GPU.

> +		if (of_device_is_available(gmu))
> +			drm_of_component_match_add(dev, matchptr, component_compare_of, gmu);
> +		of_node_put(gmu);
I think you missed the recently added headless support. Please check
separate_gpu_kms modparam and msm_gpu_probe().

-Akhil

> +	}
> +
>  	of_node_put(np);
>  
>  	return 0;
> 
> ---
> base-commit: 211ddde0823f1442e4ad052a2f30f050145ccada
> change-id: 20251022-topic-adreno-attach-gmu-to-driver-e47025fd7ebb
> 
> Best regards,