Re: [Intel-xe] [RFC 3/4] drm/xe: Add throttle reasons sysfs attributes

["Sundaresan, Sujaritha" <sujaritha.sundaresan@intel.com>]

On 11/18/2023 3:11 AM, Rodrigo Vivi wrote:
> On Thu, Nov 16, 2023 at 08:00:42PM +0530, Sujaritha Sundaresan wrote:
>> Add throttle reasons sysfs interface under device/../gt#/freq
>> Currently there is one overall status and eight reasons
>> attributes.
>>
>> The new sysfs structure will have the below layout
>>
>> device/tile<n>/gt<n>
>>                 ├── freq
>>                 │   └── 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      |  12 +
>>   drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c | 266 ++++++++++++++++++++++
>>   drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h |  17 ++
>>   4 files changed, 296 insertions(+)
>>   create mode 100644 drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c
>>   create mode 100644 drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h
>>
>> diff --git a/drivers/gpu/drm/xe/Makefile b/drivers/gpu/drm/xe/Makefile
>> index a1a8847e2ba3..c02b8e9bd72e 100644
>> --- a/drivers/gpu/drm/xe/Makefile
>> +++ b/drivers/gpu/drm/xe/Makefile
>> @@ -64,6 +64,7 @@ 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 cc27fe8fc363..296ba9831634 100644
>> --- a/drivers/gpu/drm/xe/regs/xe_gt_regs.h
>> +++ b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
>> @@ -457,4 +457,16 @@
>>   #define PVC_GT0_PLATFORM_ENERGY_STATUS		XE_REG(0x28106c)
>>   #define PVC_GT0_PACKAGE_POWER_SKU		XE_REG(0x281080)
>>   
>> +#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 MTL_MEDIA_PERF_LIMIT_REASONS		XE_REG(0x138030)
>> +
>>   #endif
>> 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..1b64058d9561
>> --- /dev/null
>> +++ b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.c
>> @@ -0,0 +1,266 @@
>> +// SPDX-License-Identifier: MIT
>> +/*
>> + * Copyright © 2023 Intel Corporation
>> + */
>> +
>> +#include <drm/drm_managed.h>
>> +
>> +#include <regs/xe_reg_defs.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_mmio.h"
>> +
>> +/**
>> + * DOC: Xe GT Throttle
>> + *
>> + * Provides sysfs entries for frequency throttle reasons in GT
>> + *
>> + * device/gt#/throttle/status - Overall status
>> + * device/gt#/throttle/reason_pl1 - Frequency throttle due to PL1
>> + * device/gt#/throttle/reason_pl2 - Frequency throttle due to PL2
>> + * device/gt#/throttle/reason_pl4 - Frequency throttle due to PL4, Iccmax etc.
>> + * device/gt#/throttle/reason_thermal - Frequency throttle due to thermal
>> + * device/gt#/throttle/reason_prochot - Frequency throttle due to prochot
>> + * device/gt#/throttle/reason_ratl - Frequency throttle due to RATL
>> + * device/gt#/throttle/reason_vr_thermalert - Frequency throttle due to VR THERMALERT
>> + * device/gt#/throttle/reason_vr_tdc -  Frequency throttle due to VR TDC
>> + */
>> +
>> +static 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;
>> +}
>> +
>> +static u32 read_status(struct xe_gt *gt)
>> +{
>> +	u32 status = read_perf_limit_reasons(gt) & GT0_PERF_LIMIT_REASONS_MASK;
>> +
>> +	return status;
>> +}
>> +
>> +static u32 read_reason_pl1(struct xe_gt *gt)
>> +{
>> +	u32 pl1 = read_perf_limit_reasons(gt) & POWER_LIMIT_1_MASK;
>> +
>> +	return pl1;
>> +}
>> +
>> +static u32 read_reason_pl2(struct xe_gt *gt)
>> +{
>> +	u32 pl2 = read_perf_limit_reasons(gt) & POWER_LIMIT_2_MASK;
>> +
>> +	return pl2;
>> +}
>> +
>> +static u32 read_reason_pl4(struct xe_gt *gt)
>> +{
>> +	u32 pl4 = read_perf_limit_reasons(gt) & POWER_LIMIT_4_MASK;
>> +
>> +	return pl4;
>> +}
>> +
>> +static u32 read_reason_thermal(struct xe_gt *gt)
>> +{
>> +	u32 thermal = read_perf_limit_reasons(gt) & THERMAL_LIMIT_MASK;
>> +
>> +	return thermal;
>> +}
>> +
>> +static u32 read_reason_prochot(struct xe_gt *gt)
>> +{
>> +	u32 prochot = read_perf_limit_reasons(gt) & PROCHOT_MASK;
>> +
>> +	return prochot;
>> +}
>> +
>> +static u32 read_reason_ratl(struct xe_gt *gt)
>> +{
>> +	u32 ratl = read_perf_limit_reasons(gt) & RATL_MASK;
>> +
>> +	return ratl;
>> +}
>> +
>> +static u32 read_reason_vr_thermalert(struct xe_gt *gt)
>> +{
>> +	u32 thermalert = read_perf_limit_reasons(gt) & VR_THERMALERT_MASK;
>> +
>> +	return thermalert;
>> +}
>> +
>> +static u32 read_reason_vr_tdc(struct xe_gt *gt)
>> +{
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool status = !!read_status(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool pl1 = !!read_reason_pl1(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool pl2 = !!read_reason_pl2(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool pl4 = !!read_reason_pl4(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool thermal = !!read_reason_thermal(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool prochot = !!read_reason_prochot(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool ratl = !!read_reason_ratl(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool thermalert = !!read_reason_vr_thermalert(gt);
>> +
>> +	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 kobject *kobj = &dev->kobj;
>> +	struct xe_gt *gt = kobj_to_gt(kobj);
>> +	bool tdc = !!read_reason_vr_tdc(gt);
>> +
>> +	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 *gt)
>> +{
>> +	struct xe_device *xe = gt_to_xe(gt);
>> +	struct kobject *kobj1, *kobj2;
>> +	int err;
>> +
>> +
>> +	kobj1 = kobject_create_and_add("freq", gt->sysfs);
>> +	if (!kobj1) {
>> +		drm_warn(&xe->drm, "%s failed, err: %d\n", __func__, -ENOMEM);
>> +		return;
>> +	}
> This cannot be right.
> We need to have a xe_freq midlayer now and that creates the freq sysfs.
> a child throttle creating that is strange.
>
>> +
>> +	kobj2 = kobject_create_and_add("throttle", kobj1);
>> +	if (!kobj2) {
>> +		drm_warn(&xe->drm, "%s failed, err: %d\n", __func__, -ENOMEM);
>> +		return;
>> +	}
>> +
>> +	err = sysfs_create_files(kobj2, throttle_attrs);
>> +	if (err) {
>> +		kobject_put(kobj2);
>> +		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, kobj2);
>> +	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..48e22f6b6442
>> --- /dev/null
>> +++ b/drivers/gpu/drm/xe/xe_gt_throttle_sysfs.h
>> @@ -0,0 +1,17 @@
>> +/* SPDX-License-Identifier: MIT */
>> +/*
>> + * Copyright © 2023 Intel Corporation
>                  ^ strange char here

Hmm yes. Not sure what that is. Will fix.

Thanks,

Suja

>> + */
>> +
>> +#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"
>> +
>> +void xe_gt_throttle_sysfs_init(struct xe_gt *gt);
>> +
>> +#endif /* _XE_GT_THROTTLE_SYSFS_H_ */
>> +
>> -- 
>> 2.25.1
>>