[Intel-xe] [PATCH v2 3/3] drm/xe/hwmon: Expose power1_max_interval
Gupta, Anshuman
anshuman.gupta at intel.com
Thu Oct 19 06:57:19 UTC 2023
> -----Original Message-----
> From: Nilawar, Badal <badal.nilawar at intel.com>
> Sent: Thursday, October 19, 2023 12:34 AM
> To: intel-xe at lists.freedesktop.org
> Cc: Gupta, Anshuman <anshuman.gupta at intel.com>; Dixit, Ashutosh
> <ashutosh.dixit at intel.com>; andi.shyti at linux.intel.com; Tauro, Riana
> <riana.tauro at intel.com>; Vivi, Rodrigo <rodrigo.vivi at intel.com>
> Subject: [PATCH v2 3/3] drm/xe/hwmon: Expose power1_max_interval
>
> Expose power1_max_interval, that is the tau corresponding to PL1, as a
> custom hwmon attribute. Some bit manipulation is needed because of the
> format of PKG_PWR_LIM_1_TIME in PACKAGE_RAPL_LIMIT register (1.x *
> power(2,y))
>
> v2: Get rpm wake ref while accessing power1_max_interval
> v3: %s/hwmon/xe_hwmon/
> v4:
> - As power1_max_interval is rw attr take lock in read function as well
> - Refine comment about val to fix point conversion (Andi)
> - Update kernel version and date in doc
> v5:
> - Fix review comments (Anshuman)
>
> Acked-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray at intel.com>
> Signed-off-by: Badal Nilawar <badal.nilawar at intel.com>
> ---
> .../ABI/testing/sysfs-driver-intel-xe-hwmon | 9 ++
> drivers/gpu/drm/xe/regs/xe_mchbar_regs.h | 8 +
> drivers/gpu/drm/xe/xe_hwmon.c | 149 +++++++++++++++++-
> 3 files changed, 165 insertions(+), 1 deletion(-)
>
> diff --git a/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon
> b/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon
> index 1a7a6c23e141..8c321bc9dc04 100644
> --- a/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon
> +++ b/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon
> @@ -59,3 +59,12 @@ Contact: intel-xe at lists.freedesktop.org
> Description: RO. Energy input of device in microjoules.
>
> Only supported for particular Intel xe graphics platforms.
> +
> +What: /sys/devices/.../hwmon/hwmon<i>/power1_max_interval
> +Date: October 2023
> +KernelVersion: 6.6
> +Contact: intel-xe at lists.freedesktop.org
> +Description: RW. Sustained power limit interval (Tau in PL1/Tau) in
> + milliseconds over which sustained power is averaged.
> +
> + Only supported for particular Intel xe graphics platforms.
> diff --git a/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h
> b/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h
> index d8ecbe1858d1..519dd1067a19 100644
> --- a/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h
> +++ b/drivers/gpu/drm/xe/regs/xe_mchbar_regs.h
> @@ -22,15 +22,23 @@
> #define PKG_TDP GENMASK_ULL(14, 0)
> #define PKG_MIN_PWR GENMASK_ULL(30,
> 16)
> #define PKG_MAX_PWR GENMASK_ULL(46,
> 32)
> +#define PKG_MAX_WIN GENMASK_ULL(54,
> 48)
> +#define PKG_MAX_WIN_X GENMASK_ULL(54, 53)
> +#define PKG_MAX_WIN_Y GENMASK_ULL(52, 48)
> +
>
> #define PCU_CR_PACKAGE_POWER_SKU_UNIT
> XE_REG(MCHBAR_MIRROR_BASE_SNB + 0x5938)
> #define PKG_PWR_UNIT REG_GENMASK(3, 0)
> #define PKG_ENERGY_UNIT REG_GENMASK(12, 8)
> +#define PKG_TIME_UNIT REG_GENMASK(19,
> 16)
>
> #define PCU_CR_PACKAGE_ENERGY_STATUS
> XE_REG(MCHBAR_MIRROR_BASE_SNB + 0x593c)
>
> #define PCU_CR_PACKAGE_RAPL_LIMIT
> XE_REG(MCHBAR_MIRROR_BASE_SNB + 0x59a0)
> #define PKG_PWR_LIM_1 REG_GENMASK(14,
> 0)
> #define PKG_PWR_LIM_1_EN REG_BIT(15)
> +#define PKG_PWR_LIM_1_TIME REG_GENMASK(23,
> 17)
> +#define PKG_PWR_LIM_1_TIME_X REG_GENMASK(23,
> 22)
> +#define PKG_PWR_LIM_1_TIME_Y REG_GENMASK(21,
> 17)
>
> #endif /* _XE_MCHBAR_REGS_H_ */
> diff --git a/drivers/gpu/drm/xe/xe_hwmon.c
> b/drivers/gpu/drm/xe/xe_hwmon.c index c87d028f1955..661d5269fb5b
> 100644
> --- a/drivers/gpu/drm/xe/xe_hwmon.c
> +++ b/drivers/gpu/drm/xe/xe_hwmon.c
> @@ -38,6 +38,7 @@ enum xe_hwmon_reg_operation {
> #define SF_CURR 1000 /* milliamperes */
> #define SF_VOLTAGE 1000 /* millivolts */
> #define SF_ENERGY 1000000 /* microjoules */
> +#define SF_TIME 1000 /* milliseconds */
>
> struct xe_hwmon_energy_info {
> u32 reg_val_prev;
> @@ -50,6 +51,7 @@ struct xe_hwmon {
> struct mutex hwmon_lock; /* For rw attributes */
> int scl_shift_power;
> int scl_shift_energy;
> + int scl_shift_time;
> struct xe_hwmon_energy_info ei; /* Energy info for
> energy1_input */
> };
>
> @@ -239,6 +241,149 @@ xe_hwmon_energy_get(struct xe_hwmon
> *hwmon, long *energy)
> hwmon->scl_shift_energy);
> }
>
> +static ssize_t
> +xe_hwmon_power1_max_interval_show(struct device *dev, struct
> device_attribute *attr,
> + char *buf)
> +{
> + struct xe_hwmon *hwmon = dev_get_drvdata(dev);
> + u32 x, y, x_w = 2; /* 2 bits */
> + u64 r, tau4, out;
> +
> + xe_device_mem_access_get(gt_to_xe(hwmon->gt));
> +
> + mutex_lock(&hwmon->hwmon_lock);
> +
> + xe_hwmon_process_reg(hwmon, REG_PKG_RAPL_LIMIT,
> + REG_READ32, &r, 0, 0);
> +
> + mutex_unlock(&hwmon->hwmon_lock);
> +
> + xe_device_mem_access_put(gt_to_xe(hwmon->gt));
> +
> + x = REG_FIELD_GET(PKG_PWR_LIM_1_TIME_X, r);
> + y = REG_FIELD_GET(PKG_PWR_LIM_1_TIME_Y, r);
> +
> + /*
> + * tau = 1.x * power(2,y), x = bits(23:22), y = bits(21:17)
> + * = (4 | x) << (y - 2)
> + *
> + * Here (y - 2) ensures a 1.x fixed point representation of 1.x
> + * As x is 2 bits so 1.x can be 1.0, 1.25, 1.50, 1.75
> + *
> + * As y can be < 2, we compute tau4 = (4 | x) << y
> + * and then add 2 when doing the final right shift to account for units
> + */
> + tau4 = ((1 << x_w) | x) << y;
> +
> + /* val in hwmon interface units (millisec) */
> + out = mul_u64_u32_shr(tau4, SF_TIME, hwmon->scl_shift_time +
> x_w);
> +
> + return sysfs_emit(buf, "%llu\n", out); }
> +
> +static ssize_t
> +xe_hwmon_power1_max_interval_store(struct device *dev, struct
> device_attribute *attr,
> + const char *buf, size_t count)
> +{
> + struct xe_hwmon *hwmon = dev_get_drvdata(dev);
> + u32 x, y, rxy, x_w = 2; /* 2 bits */
> + u64 tau4, r, max_win;
> + unsigned long val;
> + int ret;
> +
> + ret = kstrtoul(buf, 0, &val);
> + if (ret)
> + return ret;
> +
> + /*
> + * Max HW supported tau in '1.x * power(2,y)' format, x = 0, y = 0x12
> + * The hwmon->scl_shift_time default of 0xa results in a max tau of
> 256 seconds
"Add . full stop above"
> + * For DG2, PVC PKG_PWR_SKU does not show valid PKG_MAX_WIN
> so going with default
I think we need something like below, you can format that as code comment.
"Ideally we should have read PKG_MAX_WIN from PKG_PWR_SKU register but our discrete GPU
does not provide the correct value, therefore using the default constant value, future discrete GPU
should fix that and we should read the PKG_MAX_WIN from PKG_PWR_SKU "
With that.
Reviewed-by: Anshuman Gupta <anshuman.gupta at intel.com>
" "
> + * For future platforms this limitation may get addressed.
> + */
> +#define PKG_MAX_WIN_DEFAULT 0x12ull
> +
> + /*
> + * val must be < max in hwmon interface units. The steps below are
> + * explained in xe_hwmon_power1_max_interval_show()
> + */
> + r = FIELD_PREP(PKG_MAX_WIN, PKG_MAX_WIN_DEFAULT);
> + x = REG_FIELD_GET(PKG_MAX_WIN_X, r);
> + y = REG_FIELD_GET(PKG_MAX_WIN_Y, r);
> + tau4 = ((1 << x_w) | x) << y;
> + max_win = mul_u64_u32_shr(tau4, SF_TIME, hwmon->scl_shift_time
> + x_w);
> +
> + if (val > max_win)
> + return -EINVAL;
> +
> + /* val in hw units */
> + val = DIV_ROUND_CLOSEST_ULL((u64)val << hwmon->scl_shift_time,
> +SF_TIME);
> +
> + /*
> + * Convert val to 1.x * power(2,y)
> + * y = ilog2(val)
> + * x = (val - (1 << y)) >> (y - 2)
> + */
> + if (!val) {
> + y = 0;
> + x = 0;
> + } else {
> + y = ilog2(val);
> + x = (val - (1ul << y)) << x_w >> y;
> + }
> +
> + rxy = REG_FIELD_PREP(PKG_PWR_LIM_1_TIME_X, x) |
> +REG_FIELD_PREP(PKG_PWR_LIM_1_TIME_Y, y);
> +
> + xe_device_mem_access_get(gt_to_xe(hwmon->gt));
> +
> + mutex_lock(&hwmon->hwmon_lock);
> +
> + xe_hwmon_process_reg(hwmon, REG_PKG_RAPL_LIMIT,
> REG_RMW32, (u64 *)&r,
> + PKG_PWR_LIM_1_TIME, rxy);
> +
> + mutex_unlock(&hwmon->hwmon_lock);
> +
> + xe_device_mem_access_put(gt_to_xe(hwmon->gt));
> +
> + return count;
> +}
> +
> +static SENSOR_DEVICE_ATTR(power1_max_interval, 0664,
> + xe_hwmon_power1_max_interval_show,
> + xe_hwmon_power1_max_interval_store, 0);
> +
> +static struct attribute *hwmon_attributes[] = {
> + &sensor_dev_attr_power1_max_interval.dev_attr.attr,
> + NULL
> +};
> +
> +static umode_t xe_hwmon_attributes_visible(struct kobject *kobj,
> + struct attribute *attr, int index) {
> + struct device *dev = kobj_to_dev(kobj);
> + struct xe_hwmon *hwmon = dev_get_drvdata(dev);
> + int ret = 0;
> +
> + xe_device_mem_access_get(gt_to_xe(hwmon->gt));
> +
> + if (attr == &sensor_dev_attr_power1_max_interval.dev_attr.attr)
> + ret = xe_hwmon_get_reg(hwmon, REG_PKG_RAPL_LIMIT) ?
> attr->mode : 0;
> +
> + xe_device_mem_access_put(gt_to_xe(hwmon->gt));
> +
> + return ret;
> +}
> +
> +static const struct attribute_group hwmon_attrgroup = {
> + .attrs = hwmon_attributes,
> + .is_visible = xe_hwmon_attributes_visible, };
> +
> +static const struct attribute_group *hwmon_groups[] = {
> + &hwmon_attrgroup,
> + NULL
> +};
> +
> static const struct hwmon_channel_info *hwmon_info[] = {
> HWMON_CHANNEL_INFO(power, HWMON_P_MAX |
> HWMON_P_RATED_MAX | HWMON_P_CRIT),
> HWMON_CHANNEL_INFO(curr, HWMON_C_CRIT), @@ -555,6
> +700,7 @@ xe_hwmon_get_preregistration_info(struct xe_device *xe)
> REG_READ32, &val_sku_unit, 0, 0);
> hwmon->scl_shift_power = REG_FIELD_GET(PKG_PWR_UNIT,
> val_sku_unit);
> hwmon->scl_shift_energy =
> REG_FIELD_GET(PKG_ENERGY_UNIT, val_sku_unit);
> + hwmon->scl_shift_time = REG_FIELD_GET(PKG_TIME_UNIT,
> val_sku_unit);
> }
>
> /*
> @@ -601,7 +747,8 @@ void xe_hwmon_register(struct xe_device *xe)
> /* hwmon_dev points to device hwmon<i> */
> hwmon->hwmon_dev =
> devm_hwmon_device_register_with_info(dev, "xe", hwmon,
>
> &hwmon_chip_info,
> - NULL);
> +
> hwmon_groups);
> +
> if (IS_ERR(hwmon->hwmon_dev)) {
> drm_warn(&xe->drm, "Failed to register xe hwmon (%pe)\n",
> hwmon->hwmon_dev);
> xe->hwmon = NULL;
> --
> 2.25.1
More information about the Intel-xe
mailing list