Re: [RFC PATCH 04/18] drm/panthor: Add 64-bit GPU_ID decoding for v15 GPUs

[Boris Brezillon]

On Thu, 28 May 2026 16:05:32 +0100
Karunika Choo <karunika.choo@xxxxxxx> wrote:

> Mali v15 exposes a 64-bit GPU_ID register with a different field layout
> from earlier GPUs.
> 
> Add the register definitions and decoding helpers for the new format,
> read GPU_WIDE_ID when the compatibility value indicates a v15 GPU, and
> populate both the cached GPU_ID fields and the uAPI gpu_id_hi field.
> 
> This allows userspace and the driver to identify v15 GPUs correctly.
> 
> Signed-off-by: Karunika Choo <karunika.choo@xxxxxxx>
> ---
>  .../drm/panthor/panthor_gpu_discover_regs.h   | 17 +++++++++
>  drivers/gpu/drm/panthor/panthor_hw.c          | 38 ++++++++++++++-----
>  include/uapi/drm/panthor_drm.h                | 10 +++++
>  3 files changed, 56 insertions(+), 9 deletions(-)
>  create mode 100644 drivers/gpu/drm/panthor/panthor_gpu_discover_regs.h
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_gpu_discover_regs.h b/drivers/gpu/drm/panthor/panthor_gpu_discover_regs.h
> new file mode 100644
> index 000000000000..54bb104902ee
> --- /dev/null
> +++ b/drivers/gpu/drm/panthor/panthor_gpu_discover_regs.h
> @@ -0,0 +1,17 @@
> +/* SPDX-License-Identifier: GPL-2.0 or MIT */
> +/* Copyright 2026 ARM Limited. All rights reserved. */
> +
> +#ifndef __PANTHOR_GPU_DISCOVER_REGS_H__
> +#define __PANTHOR_GPU_DISCOVER_REGS_H__
> +
> +#define GPU_WIDE_ID					0x0
> +#define   GPU_WIDE_COMPAT				0xF
> +#define   GPU_WIDE_ARCH_MAJOR(x)			(((x) & GENMASK(63, 56)) >> 56)
> +#define   GPU_WIDE_ARCH_MINOR(x)			(((x) & GENMASK(55, 48)) >> 48)
> +#define   GPU_WIDE_ARCH_REV(x)				(((x) & GENMASK(47, 40)) >> 40)
> +#define   GPU_WIDE_PROD_MAJOR(x)			(((x) & GENMASK(39, 32)) >> 32)
> +#define   GPU_WIDE_VER_MAJOR(x)				(((x) & GENMASK(23, 16)) >> 16)
> +#define   GPU_WIDE_VER_MINOR(x)				(((x) & GENMASK(15, 8)) >> 8)
> +#define   GPU_WIDE_VER_STATUS(x)			((x) & GENMASK(7, 0))
> +
> +#endif /* __PANTHOR_GPU_DISCOVER_REGS_H__ */
> diff --git a/drivers/gpu/drm/panthor/panthor_hw.c b/drivers/gpu/drm/panthor/panthor_hw.c
> index 3570e2889584..833d10cbb2d6 100644
> --- a/drivers/gpu/drm/panthor/panthor_hw.c
> +++ b/drivers/gpu/drm/panthor/panthor_hw.c
> @@ -9,6 +9,7 @@
>  #include "panthor_device.h"
>  #include "panthor_device_io.h"
>  #include "panthor_gpu.h"
> +#include "panthor_gpu_discover_regs.h"
>  #include "panthor_gpu_regs.h"
>  #include "panthor_hw.h"
>  #include "panthor_pwr.h"
> @@ -297,18 +298,37 @@ static int panthor_hw_bind_device(struct panthor_device *ptdev)
>  static int panthor_hw_gpu_id_init(struct panthor_device *ptdev)
>  {
>  	struct panthor_gpu_id *gpu_id = &ptdev->gpu_id;
> -	ptdev->gpu_info.gpu_id = gpu_read(ptdev->iomem, GPU_ID);
> +	u32 gpu_id32 = gpu_read(ptdev->iomem, GPU_ID);
>  
> -	if (!ptdev->gpu_info.gpu_id)
> +	if (!gpu_id32)
>  		return -ENXIO;
>  
> -	gpu_id->arch.major = GPU_ARCH_MAJOR(ptdev->gpu_info.gpu_id);
> -	gpu_id->arch.minor = GPU_ARCH_MINOR(ptdev->gpu_info.gpu_id);
> -	gpu_id->arch.rev = GPU_ARCH_REV(ptdev->gpu_info.gpu_id);
> -	gpu_id->prod_major = GPU_PROD_MAJOR(ptdev->gpu_info.gpu_id);
> -	gpu_id->ver.major = GPU_VER_MAJOR(ptdev->gpu_info.gpu_id);
> -	gpu_id->ver.minor = GPU_VER_MINOR(ptdev->gpu_info.gpu_id);
> -	gpu_id->ver.status = GPU_VER_STATUS(ptdev->gpu_info.gpu_id);
> +	ptdev->gpu_info.gpu_id = gpu_id32;
> +
> +	gpu_id->arch.major = GPU_ARCH_MAJOR(gpu_id32);
> +	gpu_id->arch.minor = GPU_ARCH_MINOR(gpu_id32);
> +	gpu_id->arch.rev = GPU_ARCH_REV(gpu_id32);
> +	gpu_id->prod_major = GPU_PROD_MAJOR(gpu_id32);
> +	gpu_id->ver.major = GPU_VER_MAJOR(gpu_id32);
> +	gpu_id->ver.minor = GPU_VER_MINOR(gpu_id32);
> +	gpu_id->ver.status = GPU_VER_STATUS(gpu_id32);
> +
> +	if (GPU_ARCH_MAJOR(gpu_id32) == GPU_WIDE_COMPAT) {
> +		u64 gpu_id64 = gpu_read64(ptdev->iomem, GPU_WIDE_ID);

I know we currently assume that DISCOVER regs on v15 and GPU regs on
older gens are starting at zero, but I'd prefer to make this explicit
with something like:

		/* DISCOVER register block always starts at offset
		 * zero, so does the GPU register block on older
		 * GPUs.
		 */
		void __iomem *discover = ptdev->iomem;
		u64 gpu_id64 = gpu_read64(discover, GPU_WIDE_ID);
> +		if (!gpu_id64)
> +			return -ENXIO;
> +
> +		ptdev->gpu_info.gpu_wide_id = gpu_id64;
> +		ptdev->gpu_info.gpu_id = 0;
> +
> +		gpu_id->arch.major = GPU_WIDE_ARCH_MAJOR(gpu_id64);
> +		gpu_id->arch.minor = GPU_WIDE_ARCH_MINOR(gpu_id64);
> +		gpu_id->arch.rev = GPU_WIDE_ARCH_REV(gpu_id64);
> +		gpu_id->prod_major = GPU_WIDE_PROD_MAJOR(gpu_id64);
> +		gpu_id->ver.major = GPU_WIDE_VER_MAJOR(gpu_id64);
> +		gpu_id->ver.minor = GPU_WIDE_VER_MINOR(gpu_id64);
> +		gpu_id->ver.status = GPU_WIDE_VER_STATUS(gpu_id64);
> +	}

I'd probably move the !has_wide_id() logic to an else branch instead of
assigning unconditionally, and then fixing up if has_wide_id() is true:

	if (GPU_ARCH_MAJOR(gpu_id32) == GPU_WIDE_COMPAT) {
		u64 gpu_id64 = gpu_read64(ptdev->iomem, GPU_WIDE_ID);
		if (!gpu_id64)
			return -ENXIO;

		ptdev->gpu_info.gpu_wide_id = gpu_id64;
		gpu_id->arch.major = GPU_WIDE_ARCH_MAJOR(gpu_id64);
		gpu_id->arch.minor = GPU_WIDE_ARCH_MINOR(gpu_id64);
		gpu_id->arch.rev = GPU_WIDE_ARCH_REV(gpu_id64);
		gpu_id->prod_major = GPU_WIDE_PROD_MAJOR(gpu_id64);
		gpu_id->ver.major = GPU_WIDE_VER_MAJOR(gpu_id64);
		gpu_id->ver.minor = GPU_WIDE_VER_MINOR(gpu_id64);
		gpu_id->ver.status = GPU_WIDE_VER_STATUS(gpu_id64);
	} else {
		ptdev->gpu_info.gpu_id = gpu_id32;
		gpu_id->arch.major = GPU_ARCH_MAJOR(gpu_id32);
		gpu_id->arch.minor = GPU_ARCH_MINOR(gpu_id32);
		gpu_id->arch.rev = GPU_ARCH_REV(gpu_id32);
		gpu_id->prod_major = GPU_PROD_MAJOR(gpu_id32);
		gpu_id->ver.major = GPU_VER_MAJOR(gpu_id32);
		gpu_id->ver.minor = GPU_VER_MINOR(gpu_id32);
		gpu_id->ver.status = GPU_VER_STATUS(gpu_id32);
	}


>  
>  	return 0;
>  }
> diff --git a/include/uapi/drm/panthor_drm.h b/include/uapi/drm/panthor_drm.h
> index 0e455d91e77d..04fc9f133152 100644
> --- a/include/uapi/drm/panthor_drm.h
> +++ b/include/uapi/drm/panthor_drm.h
> @@ -373,6 +373,16 @@ struct drm_panthor_gpu_info {
>  
>  	/** @gpu_features: Bitmask describing supported GPU-wide features */
>  	__u64 gpu_features;
> +
> +	/** @gpu_wide_id: 64-bit GPU_ID for v15 onwards. */

Specify that it's zero on v14- HW, and also update the gpu_id doc to
mention it will be zero on v15+.

> +	__u64 gpu_wide_id;
> +#define DRM_PANTHOR_WIDE_ARCH_MAJOR(x)		(((x) >> 56) & 0xff)
> +#define DRM_PANTHOR_WIDE_ARCH_MINOR(x)		(((x) >> 48) & 0xff)
> +#define DRM_PANTHOR_WIDE_ARCH_REV(x)		(((x) >> 40) & 0xff)
> +#define DRM_PANTHOR_WIDE_PRODUCT_MAJOR(x)	(((x) >> 32) & 0xff)
> +#define DRM_PANTHOR_WIDE_VERSION_MAJOR(x)	(((x) >> 16) & 0xff)
> +#define DRM_PANTHOR_WIDE_VERSION_MINOR(x)	(((x) >> 8) & 0xff)
> +#define DRM_PANTHOR_WIDE_VERSION_STATUS(x)	((x) & 0xff)
>  };
>  
>  /**