Re: [PATCH 2/6] drm/nouveau/gsp: add RUSD telemetry support

[lyude@redhat.com]

Some lil nitpicks

On Wed, 2026-07-15 at 01:14 +0400, Mohamed Ahmed wrote:
> +	case NVKM_GSP_RUSD_ROWREMAP_HIST_MAX ... NVKM_GSP_RUSD_ROWREMAP_FAILURE: {
> +		RUSD_MEM_ROW_REMAP remap;
> +
> +		ret = r570_rusd_read_section(gsp,
> +					     offsetof(NV00DE_SHARED_DATA, memRowRemap),
> +					     sizeof(remap), &remap);
> +		if (ret)
> +			return ret;
> +
> +		switch (item) {
> +		case NVKM_GSP_RUSD_ROWREMAP_HIST_MAX:
> +			*val = remap.info.histogramMax;
> +			break;
> +		case NVKM_GSP_RUSD_ROWREMAP_HIST_HIGH:
> +			*val = remap.info.histogramHigh;
> +			break;
> +		case NVKM_GSP_RUSD_ROWREMAP_HIST_PARTIAL:
> +			*val = remap.info.histogramPartial;
> +			break;
> +		case NVKM_GSP_RUSD_ROWREMAP_HIST_LOW:
> +			*val = remap.info.histogramLow;
> +			break;
> +		case NVKM_GSP_RUSD_ROWREMAP_HIST_NONE:
> +			*val = remap.info.histogramNone;
> +			break;
> +		case NVKM_GSP_RUSD_ROWREMAP_CORRECTABLE:
> +			*val = remap.info.correctableRows;
> +			break;
> +		case NVKM_GSP_RUSD_ROWREMAP_UNCORRECTABLE:
> +			*val = remap.info.uncorrectableRows;
> +			break;
> +		case NVKM_GSP_RUSD_ROWREMAP_PENDING:
> +			*val = remap.info.isPending;
> +			break;
> +		default: /* NVKM_GSP_RUSD_ROWREMAP_FAILURE */
> +			*val = remap.info.hasFailureOccurred;
> +			break;

Maybe this should be an explicit case instead of default, so we can
drm_WARN_ON_ONCE unknown values?

> +		}
> +		return 0;
> +	}
> +	case NVKM_GSP_RUSD_PCIE_GEN ... NVKM_GSP_RUSD_PCIE_UNSUPPORTED_REQUESTS: {
> +		RUSD_PCIE_DATA pcie;
> +
> +		int idx = item - NVKM_GSP_RUSD_PCIE_GEN;
> +
> +		ret = r570_rusd_read_section(gsp,
> +					     offsetof(NV00DE_SHARED_DATA, pciBusData),
> +					     sizeof(pcie), &pcie);
> +		if (ret)
> +			return ret;
> +
> +		*val = pcie.info.data[idx];
> +		return 0;
> +	}
> +	default:
> +		return -EINVAL;
> +	}
> +
> +	/* A reading of 0 mW is indistinguishable from "not supported on this
> +	 * board" (e.g. module power on Ampere/Hopper).
> +	 */
> +	if (!mw)
> +		return -ENODATA;
> +
> +	*val = (s64)mw * 1000;
> +	return 0;
> +}
> +
> +const struct nvkm_rm_api_rusd
> +r570_rusd = {
> +	.init = r570_rusd_init,
> +	.resume = r570_rusd_resume,
> +	.read = r570_rusd_read,
> +};
> diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/rm.h b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/rm.h
> index a9af94adf9ef..1338b4a88f95 100644
> --- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/rm.h
> +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/rm/rm.h
> @@ -130,6 +130,12 @@ struct nvkm_rm_api {
>  			void (*fini)(struct r535_gr *);
>  		} scrubber;
>  	} *gr;
> +
> +	const struct nvkm_rm_api_rusd {
> +		int (*init)(struct nvkm_gsp *);
> +		void (*resume)(struct nvkm_gsp *);
> +		int (*read)(struct nvkm_gsp *, enum nvkm_gsp_rusd_item, s64 *);
> +	} *rusd;
>  };
>  
>  extern const struct nvkm_rm_impl r535_rm_tu102;
> @@ -188,4 +194,5 @@ extern const struct nvkm_rm_api_gr r570_gr;
>  int r570_gr_gpc_mask(struct nvkm_gsp *, u32 *mask);
>  int r570_gr_tpc_mask(struct nvkm_gsp *, int gpc, u32 *mask);
>  extern const struct nvkm_rm_api_engine r570_ofa;
> +extern const struct nvkm_rm_api_rusd r570_rusd;
>  #endif