Re: [Intel-xe] [RFC][PATCH 1/1] drm/xe: Add throttle reason sysfs attributes

[Matthew Brost <matthew.brost@intel.com>]

On Tue, Aug 22, 2023 at 10:28:25AM +0530, Sujaritha Sundaresan wrote:
> Add throttle reasons sysfs interface under device/../gt#/
> Currently there is one overall status and eight reasons
> attributes.
> 
> The new sysfs structure will have the below layout
> 
> device/<dev/gt
>              ├── gt0
>              │   └── throttle
>              │       ├── <throttle_reasons>
>              │
>              │
>              ├── gtN
>              │   └── throttle
>              │       ├── <throttle_reasons>
> 
> Signed-off-by: Sujaritha Sundaresan <sujaritha.sundaresan@intel.com>
> ---
>  drivers/gpu/drm/xe/Makefile                   |   1 +
>  drivers/gpu/drm/xe/regs/xe_gt_regs.h          |  13 +
>  .../drm/xe/regs/xe_gt_throttle_sysfs_types.h  |  36 +++
>  drivers/gpu/drm/xe/xe_gt.c                    |   3 +
>  drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c     | 282 ++++++++++++++++++
>  drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h     |  17 ++
>  .../gpu/drm/xe/xe_gt_throttle_sysfs_types.h   |  35 +++
>  drivers/gpu/drm/xe/xe_gt_types.h              |   4 +
>  8 files changed, 391 insertions(+)
>  create mode 100644 drivers/gpu/drm/xe/regs/xe_gt_throttle_sysfs_types.h
>  create mode 100644 drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c
>  create mode 100644 drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h
>  create mode 100644 drivers/gpu/drm/xe/xe_gt_throttle_sysfs_types.h
> 
> diff --git a/drivers/gpu/drm/xe/Makefile b/drivers/gpu/drm/xe/Makefile
> index 550cdfed729e..eb3ccd90b2fa 100644
> --- a/drivers/gpu/drm/xe/Makefile
> +++ b/drivers/gpu/drm/xe/Makefile
> @@ -63,6 +63,7 @@ xe-y += xe_bb.o \
>  	xe_gt_mcr.o \
>  	xe_gt_pagefault.o \
>  	xe_gt_sysfs.o \
> +	xe_gt_throttle_sysfs.o \
>  	xe_gt_tlb_invalidation.o \
>  	xe_gt_topology.o \
>  	xe_guc.o \
> diff --git a/drivers/gpu/drm/xe/regs/xe_gt_regs.h b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
> index baa00557f114..8416d483e7c3 100644
> --- a/drivers/gpu/drm/xe/regs/xe_gt_regs.h
> +++ b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
> @@ -400,4 +400,17 @@
>  #define XEHPC_BCS5_BCS6_INTR_MASK		XE_REG(0x190118)
>  #define XEHPC_BCS7_BCS8_INTR_MASK		XE_REG(0x19011c)
>  
> +#define GT0_PERF_LIMIT_REASONS			XE_REG(0x1381a8)
> +#define   GT0_PERF_LIMIT_REASONS_MASK		0xde3
> +#define   PROCHOT_MASK				REG_BIT(0)
> +#define   THERMAL_LIMIT_MASK			REG_BIT(1)
> +#define   RATL_MASK				REG_BIT(5)
> +#define   VR_THERMALERT_MASK			REG_BIT(6)
> +#define   VR_TDC_MASK				REG_BIT(7)
> +#define   POWER_LIMIT_4_MASK			REG_BIT(8)
> +#define   POWER_LIMIT_1_MASK			REG_BIT(10)
> +#define   POWER_LIMIT_2_MASK			REG_BIT(11)
> +#define   GT0_PERF_LIMIT_REASONS_LOG_MASK	REG_GENMASK(31, 16)
> +#define MTL_MEDIA_PERF_LIMIT_REASONS		XE_REG(0x138030)
> +
>  #endif
> diff --git a/drivers/gpu/drm/xe/regs/xe_gt_throttle_sysfs_types.h b/drivers/gpu/drm/xe/regs/xe_gt_throttle_sysfs_types.h
> new file mode 100644
> index 000000000000..b71865f467ce
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/regs/xe_gt_throttle_sysfs_types.h
> @@ -0,0 +1,36 @@
> +/* SPDX-License-Identifier: MIT */
> +/*
> + * Copyright © 2023 Intel Corporation
> + */
> +
> +#ifndef _XE_GT_THROTTLE_SYSFS_TYPES_H_
> +#define _XE_GT_THROTTLE_SYSFS_TYPES_H_
> +
> +#include <linux/types.h>
> +
> +/**
> + * struct xe_gt_throttle - A struct that contains frequency throttle reasons in gt
> + */
> +struct xe_gt_throttle {
> +	/** @status: status */
> +	bool status;
> +	/** @reason_pl1: */
> +	bool reason_pl1;
> +	/** @reason_pl2: */
> +	bool reason_pl2;
> +	/** @reason_pl4: */
> +	bool reason_pl4;
> +	/** @reason_thermal: */
> +	bool reason_thermal;
> +	/** @reason_prochot */
> +	bool reason_prochot;
> +	/** @reason_ratl */
> +	bool reason_ratl;
> +	/** @reason_vr_thermalert */
> +	bool reason_vr_thermalert;
> +	/** @reason_vr_tdc */
> +	bool reason_vr_tdc;
> +};
> +
> +#endif /* _XE_GT_THROTTLE_SYSFS_TYPES_H_ */
> +
> diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c
> index 13320af4ddd3..03be2c2980d4 100644
> --- a/drivers/gpu/drm/xe/xe_gt.c
> +++ b/drivers/gpu/drm/xe/xe_gt.c
> @@ -24,6 +24,7 @@
>  #include "xe_gt_pagefault.h"
>  #include "xe_gt_printk.h"
>  #include "xe_gt_sysfs.h"
> +#include "xe_gt_throttle_sysfs.h"
>  #include "xe_gt_tlb_invalidation.h"
>  #include "xe_gt_topology.h"
>  #include "xe_guc_exec_queue_types.h"
> @@ -442,6 +443,8 @@ int xe_gt_init(struct xe_gt *gt)
>  
>  	xe_gt_sysfs_init(gt);
>  
> +	xe_gt_throttle_sysfs_init(&gt->throttle);

This is the wrong place, xe_gt_sysfs_init should call xe_gt_throttle_sysfs_init.

Matt

> +
>  	err = gt_fw_domain_init(gt);
>  	if (err)
>  		return err;
> diff --git a/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c
> new file mode 100644
> index 000000000000..7db67b5fc138
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c
> @@ -0,0 +1,282 @@
> +// SPDX-License-Identifier: MIT
> +/*
> + * Copyright © 2023 Intel Corporation
> + */
> +
> +#include <drm/drm_managed.h>
> +
> +#include <regs/xe_gt_regs.h>
> +#include "xe_device.h"
> +#include "xe_gt.h"
> +#include "xe_gt_sysfs.h"
> +#include "xe_gt_throttle_sysfs.h"
> +#include "xe_guc_pc.h"
> +#include "xe_mmio.h"
> +
> +/**
> + * DOC: Xe GT Throttle
> + *
> + * Provides sysfs entries for frequency throttle reasons in GT
> + *
> + * device/gt#/throttle/status - Overall status
> + * device/gt#/throttle/throttle_reason_pl1 - Frequency throttle due to PL1
> + * device/gt#/throttle/throttle_reason_pl2 - Frequency throttle due to PL2
> + * device/gt#/throttle/throttle_reason_pl4 - Frequency throttle due to PL4, Iccmax etc.
> + * device/gt#/throttle/throttle_reason_thermal - Frequency throttle due to thermal
> + * device/gt#/throttle/throttle_reason_prochot - Frequency throttle due to prochot
> + * device/gt#/throttle/throttle_reason_ratl - Frequency throttle due to RATL
> + * device/gt#/throttle/throttle_reason_vr_thermalert - Frequency throttle due to VR THERMALERT
> + * device/gt#/throttle/throttle_reason_vr_tdc -  Frequency throttle due to VR TDC
> + */
> +
> +static struct xe_gt_throttle *dev_to_throttle(struct device *dev)
> +{
> +	struct kobject *kobj = &dev->kobj;
> +
> +	return &kobj_to_gt(kobj->parent)->throttle;
> +}
> +
> +static struct xe_gt *throttle_to_gt(struct xe_gt_throttle *throttle)
> +{
> +	return container_of(throttle, struct xe_gt, throttle);
> +}
> +
> +u32 read_perf_limit_reasons(struct xe_gt *gt)
> +{
> +	u32 reg;
> +	if (xe_gt_is_media_type(gt))
> +		reg = xe_mmio_read32(gt, MTL_MEDIA_PERF_LIMIT_REASONS);
> +	else
> +		reg = xe_mmio_read32(gt, GT0_PERF_LIMIT_REASONS);
> +
> +	return reg;
> +}
> +
> +u32 xe_read_status(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 status = read_perf_limit_reasons(gt) & GT0_PERF_LIMIT_REASONS_MASK;
> +
> +	return status;
> +}
> +
> +u32 xe_read_reason_pl1(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 pl1 = read_perf_limit_reasons(gt) & POWER_LIMIT_1_MASK;
> +
> +	return pl1;
> +}
> +
> +u32 xe_read_reason_pl2(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 pl2 = read_perf_limit_reasons(gt) & POWER_LIMIT_2_MASK;
> +
> +	return pl2;
> +}
> +
> +u32 xe_read_reason_pl4(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 pl4 = read_perf_limit_reasons(gt) & POWER_LIMIT_4_MASK;
> +
> +	return pl4;
> +}
> +
> +u32 xe_read_reason_thermal(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 thermal = read_perf_limit_reasons(gt) & THERMAL_LIMIT_MASK;
> +
> +	return thermal;
> +}
> +
> +u32 xe_read_reason_prochot(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 prochot = read_perf_limit_reasons(gt) & PROCHOT_MASK;
> +
> +	return prochot;
> +}
> +
> +u32 xe_read_reason_ratl(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 ratl = read_perf_limit_reasons(gt) & RATL_MASK;
> +
> +	return ratl;
> +}
> +
> +u32 xe_read_reason_vr_thermalert(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 thermalert = read_perf_limit_reasons(gt) & VR_THERMALERT_MASK;
> +
> +	return thermalert;
> +}
> +
> +u32 xe_read_reason_vr_tdc(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	u32 tdc = read_perf_limit_reasons(gt) & VR_TDC_MASK;
> +
> +	return tdc;
> +}
> +
> +static ssize_t status_show(struct device *dev,
> +			   struct device_attribute *attr,
> +			   char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool status = !!xe_read_status(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", status);
> +}
> +static DEVICE_ATTR_RO(status);
> +
> +static ssize_t reason_pl1_show(struct device *dev,
> +			       struct device_attribute *attr,
> +			       char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool pl1 = !!xe_read_reason_pl1(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", pl1);
> +}
> +static DEVICE_ATTR_RO(reason_pl1);
> +
> +static ssize_t reason_pl2_show(struct device *dev,
> +			       struct device_attribute *attr,
> +			       char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool pl2 = !!xe_read_reason_pl2(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", pl2);
> +}
> +static DEVICE_ATTR_RO(reason_pl2);
> +
> +static ssize_t reason_pl4_show(struct device *dev,
> +			       struct device_attribute *attr,
> +			       char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool pl4 = !!xe_read_reason_pl4(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", pl4);
> +}
> +static DEVICE_ATTR_RO(reason_pl4);
> +
> +static ssize_t reason_thermal_show(struct device *dev,
> +				   struct device_attribute *attr,
> +				   char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool thermal = !!xe_read_reason_thermal(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", thermal);
> +}
> +static DEVICE_ATTR_RO(reason_thermal);
> +
> +static ssize_t reason_prochot_show(struct device *dev,
> +				   struct device_attribute *attr,
> +				   char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool prochot = !!xe_read_reason_prochot(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", prochot);
> +}
> +static DEVICE_ATTR_RO(reason_prochot);
> +
> +static ssize_t reason_ratl_show(struct device *dev,
> +				struct device_attribute *attr,
> +				char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool ratl = !!xe_read_reason_ratl(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", ratl);
> +}
> +static DEVICE_ATTR_RO(reason_ratl);
> +
> +static ssize_t reason_vr_thermalert_show(struct device *dev,
> +					 struct device_attribute *attr,
> +					 char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool thermalert = !!xe_read_reason_vr_thermalert(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", thermalert);
> +}
> +static DEVICE_ATTR_RO(reason_vr_thermalert);
> +
> +static ssize_t reason_vr_tdc_show(struct device *dev,
> +				  struct device_attribute *attr,
> +				  char *buff)
> +{
> +	struct xe_gt_throttle *throttle = dev_to_throttle(dev);
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	bool tdc = !!xe_read_reason_vr_tdc(&gt->throttle);
> +
> +	return sysfs_emit(buff, "%u\n", tdc);
> +}
> +static DEVICE_ATTR_RO(reason_vr_tdc);
> +
> +static const struct attribute *throttle_attrs[] = {
> +	&dev_attr_status.attr,
> +	&dev_attr_reason_pl1.attr,
> +	&dev_attr_reason_pl2.attr,
> +	&dev_attr_reason_pl4.attr,
> +	&dev_attr_reason_thermal.attr,
> +	&dev_attr_reason_prochot.attr,
> +	&dev_attr_reason_ratl.attr,
> +	&dev_attr_reason_vr_thermalert.attr,
> +	&dev_attr_reason_vr_tdc.attr,
> +	NULL
> +};
> +
> +static void gt_throttle_sysfs_fini(struct drm_device *drm, void *arg)
> +{
> +	struct kobject *kobj = arg;
> +
> +	sysfs_remove_files(kobj, throttle_attrs);
> +	kobject_put(kobj);
> +}
> +
> +void xe_gt_throttle_sysfs_init(struct xe_gt_throttle *throttle)
> +{
> +	struct xe_gt *gt = throttle_to_gt(throttle);
> +	struct xe_device *xe = gt_to_xe(gt);
> +	struct kobject *kobj;
> +	int err;
> +
> +	kobj = kobject_create_and_add("throttle", gt->sysfs);
> +	if (!kobj) {
> +		drm_warn(&xe->drm, "%s failed, err: %d\n", __func__, -ENOMEM);
> +		return;
> +	}
> +
> +	err = sysfs_create_files(kobj, throttle_attrs);
> +	if (err) {
> +		kobject_put(kobj);
> +		drm_warn(&xe->drm, "failed to register throttle sysfs, err: %d\n", err);
> +		return;
> +	}
> +
> +	err = drmm_add_action_or_reset(&xe->drm, gt_throttle_sysfs_fini, kobj);
> +	if (err)
> +		drm_warn(&xe->drm, "%s: drmm_add_action_or_reset failed, err: %d\n",
> +			 __func__, err);
> +}
> +
> +
> diff --git a/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h
> new file mode 100644
> index 000000000000..809213c3bba1
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h
> @@ -0,0 +1,17 @@
> +/* SPDX-License-Identifier: MIT */
> +/*
> + * Copyright © 2023 Intel Corporation
> + */
> +
> +#ifndef _XE_GT_THROTTLE_SYSFS_H_
> +#define _XE_GT_THROTTLE_SYSFS_H_
> +
> +#include <drm/drm_managed.h>
> +
> +#include "xe_device.h"
> +#include "xe_gt.h"
> +#include "xe_gt_throttle_sysfs_types.h"
> +
> +void xe_gt_throttle_sysfs_init(struct xe_gt_throttle *throttle);
> +
> +#endif /* _XE_GT_THROTTLE_SYSFS_H_ */
> diff --git a/drivers/gpu/drm/xe/xe_gt_throttle_sysfs_types.h b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs_types.h
> new file mode 100644
> index 000000000000..acc23ce8d0fc
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs_types.h
> @@ -0,0 +1,35 @@
> +/* SPDX-License-Identifier: MIT */
> +/*
> + * Copyright © 2023 Intel Corporation
> + */
> +
> +#ifndef _XE_GT_THROTTLE_SYSFS_TYPES_H_
> +#define _XE_GT_THROTTLE_SYSFS_TYPES_H_
> +
> +#include <linux/types.h>
> +
> +struct xe_gt;
> +
> +struct xe_gt_throttle {
> +	/** @status: overall throttling status */
> +	bool status;
> +	/** @pl1: throttling reason is PL1 */
> +	bool pl1;
> +	/** @pl2: throttling reason is PL2 */
> +	bool pl2;
> +	/** @pl4: throttling reason is PL4 */
> +	bool pl4;
> +	/** @thermal: throttling reason is thermal */
> +	bool thermal;
> +	/** @prochot: throttling reason is PROCHOT */
> +	bool prochot;
> +	/** @ratl: throttling reason is running average thermal limit */
> +	bool ratl;
> +	/** @thermalert: throttling reason is vr thermalert */
> +	bool thermalert;
> +	/** @tdc: throttling reason is vr tdc */
> +	bool tdc;
> +};
> +
> +#endif /* _XE_GT_THROTTLE_SYSFS_TYPES_H_ */
> +
> diff --git a/drivers/gpu/drm/xe/xe_gt_types.h b/drivers/gpu/drm/xe/xe_gt_types.h
> index 35b8c19fa8bf..c55815d1114a 100644
> --- a/drivers/gpu/drm/xe/xe_gt_types.h
> +++ b/drivers/gpu/drm/xe/xe_gt_types.h
> @@ -8,6 +8,7 @@
>  
>  #include "xe_force_wake_types.h"
>  #include "xe_gt_idle_sysfs_types.h"
> +#include "xe_gt_throttle_sysfs_types.h"
>  #include "xe_hw_engine_types.h"
>  #include "xe_hw_fence_types.h"
>  #include "xe_reg_sr_types.h"
> @@ -298,6 +299,9 @@ struct xe_gt {
>  	/** @sysfs: sysfs' kobj used by xe_gt_sysfs */
>  	struct kobject *sysfs;
>  
> +	/** @throttle: frequency throttling reasons in GT */
> +	struct xe_gt_throttle throttle;
> +
>  	/** @mocs: info */
>  	struct {
>  		/** @uc_index: UC index */
> -- 
> 2.25.1
>