All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/xe/hwmon: Expose powerX_cap_max
@ 2026-11-19 10:40 Karthik Poosa
  2026-11-19 10:47 ` sashiko-bot
  0 siblings, 1 reply; 3+ messages in thread
From: Karthik Poosa @ 2026-11-19 10:40 UTC (permalink / raw)
  To: intel-xe
  Cc: rodrigo.vivi, anshuman.gupta, badal.nilawar, raag.jadav,
	riana.tauro, sk.anirban, mallesh.koujalagi, soham.purkait,
	ravi.kishore.koppuravuri, Karthik Poosa

Add a read-only power<i>_cap_max attribute for the card and package
channels. It reports the PL2 (burst) power limit that firmware sets at
boot, which is the highest value userspace can write to power<i>_cap.
The attribute is present only on platforms using mailbox power limits
where PL2 is enabled at boot.

Signed-off-by: Karthik Poosa <karthik.poosa@intel.com>
---
 .../ABI/testing/sysfs-driver-intel-xe-hwmon   | 24 +++++++++++++++++++
 drivers/gpu/drm/xe/xe_hwmon.c                 | 11 +++++++--
 2 files changed, 33 insertions(+), 2 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon b/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon
index 0da739d9a816..097902fd8346 100644
--- a/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon
+++ b/Documentation/ABI/testing/sysfs-driver-intel-xe-hwmon
@@ -296,6 +296,30 @@ Description:	RW. Package burst (PL2) power limit in microwatts.
 
 		Only supported for particular Intel Xe graphics platforms.
 
+What:		/sys/bus/pci/drivers/xe/.../hwmon/hwmon<i>/power1_cap_max
+Date:		November 2026
+KernelVersion:	7.4
+Contact:	intel-xe@lists.freedesktop.org
+Description:	RO. Card default burst (PL2) power limit in microwatts.
+
+		This is the firmware-provided default limit and the maximum
+		value accepted for power1_cap.
+
+		Only supported for particular Intel Xe graphics platforms where
+		PL2 is enabled.
+
+What:		/sys/bus/pci/drivers/xe/.../hwmon/hwmon<i>/power2_cap_max
+Date:		November 2026
+KernelVersion:	7.4
+Contact:	intel-xe@lists.freedesktop.org
+Description:	RO. Package default burst (PL2) power limit in microwatts.
+
+		This is the firmware-provided default limit and the maximum
+		value accepted for power2_cap.
+
+		Only supported for particular Intel Xe graphics platforms where
+		PL2 is enabled.
+
 What:		/sys/bus/pci/drivers/xe/.../hwmon/hwmon<i>/power1_cap_interval
 Date:		May 2025
 KernelVersion:	6.15
diff --git a/drivers/gpu/drm/xe/xe_hwmon.c b/drivers/gpu/drm/xe/xe_hwmon.c
index 515b288f18e5..023a317a443e 100644
--- a/drivers/gpu/drm/xe/xe_hwmon.c
+++ b/drivers/gpu/drm/xe/xe_hwmon.c
@@ -865,8 +865,9 @@ static const struct hwmon_channel_info hwmon_temp_info = {
 static const struct hwmon_channel_info * const hwmon_info[] = {
 	&hwmon_temp_info,
 	HWMON_CHANNEL_INFO(power, HWMON_P_MAX | HWMON_P_RATED_MAX | HWMON_P_LABEL | HWMON_P_CRIT |
-			   HWMON_P_CAP,
-			   HWMON_P_MAX | HWMON_P_RATED_MAX | HWMON_P_LABEL | HWMON_P_CAP),
+			   HWMON_P_CAP | HWMON_P_CAP_MAX,
+			   HWMON_P_MAX | HWMON_P_RATED_MAX | HWMON_P_LABEL | HWMON_P_CAP |
+			   HWMON_P_CAP_MAX),
 	HWMON_CHANNEL_INFO(curr, HWMON_C_LABEL, HWMON_C_CRIT | HWMON_C_LABEL),
 	HWMON_CHANNEL_INFO(in, HWMON_I_INPUT | HWMON_I_LABEL, HWMON_I_INPUT | HWMON_I_LABEL),
 	HWMON_CHANNEL_INFO(energy, HWMON_E_INPUT | HWMON_E_LABEL, HWMON_E_INPUT | HWMON_E_LABEL),
@@ -1423,6 +1424,8 @@ xe_hwmon_power_is_visible(struct xe_hwmon *hwmon, u32 attr, int channel)
 	struct xe_mmio *mmio = xe_root_tile_mmio(hwmon->xe);
 
 	switch (attr) {
+	case hwmon_power_cap_max:
+		return hwmon->pl2_on_boot[channel] ? 0444 : 0;
 	case hwmon_power_max:
 	case hwmon_power_cap:
 		if (hwmon->xe->info.has_mbx_power_limits) {
@@ -1486,6 +1489,10 @@ static int
 xe_hwmon_power_read(struct xe_hwmon *hwmon, u32 attr, int channel, long *val)
 {
 	switch (attr) {
+	case hwmon_power_cap_max:
+		*val = mul_u32_u32(REG_FIELD_GET(PWR_LIM_VAL, hwmon->pl2_on_boot[channel]),
+				   SF_POWER) >> hwmon->scl_shift_power;
+		return 0;
 	case hwmon_power_max:
 	case hwmon_power_cap:
 		return xe_hwmon_power_max_read(hwmon, attr, channel, val);
-- 
2.34.1


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] drm/xe/hwmon: Expose powerX_cap_max
  2026-11-19 10:40 [PATCH] drm/xe/hwmon: Expose powerX_cap_max Karthik Poosa
@ 2026-11-19 10:47 ` sashiko-bot
  2026-11-19 15:24   ` Poosa, Karthik
  0 siblings, 1 reply; 3+ messages in thread
From: sashiko-bot @ 2026-11-19 10:47 UTC (permalink / raw)
  To: Karthik Poosa, intel-xe

Thank you for your contribution! Sashiko AI review found no new issues
in this patch, but noticed 1 related pre-existing issue it does not
introduce:
- [High] drm/xe/hwmon: power limit clamped to 0 in
  xe_hwmon_power_max_write() [1]
--

commit 26f9457c44502cd969a84eaf5b2ee961f51b7a3e
Author: Karthik Poosa <karthik.poosa@intel.com>

drm/xe/hwmon: Expose powerX_cap_max

This commit adds a read-only power<i>_cap_max sysfs attribute on platforms
with mailbox power limits to report the default PL2 (burst) power limit
set by firmware at boot.

> diff --git a/drivers/gpu/drm/xe/xe_hwmon.c b/drivers/gpu/drm/xe/xe_hwmon.c
> index defe8aaca7c11..8e0d295aabc88 100644
> --- a/drivers/gpu/drm/xe/xe_hwmon.c
> +++ b/drivers/gpu/drm/xe/xe_hwmon.c
[ ... ]
> @@ -1423,6 +1424,8 @@ xe_hwmon_power_is_visible(struct xe_hwmon *hwmon, u32 attr, int channel)
>  	struct xe_mmio *mmio = xe_root_tile_mmio(hwmon->xe);
>  
>  	switch (attr) {
> +	case hwmon_power_cap_max:
> +		return hwmon->pl2_on_boot[channel] ? 0444 : 0;

[Severity: High]
This problem wasn't introduced by this patch, but can a pcode read failure
during probe cause xe_hwmon_power_max_write() to clamp user power limits
to 0 [1]?

If reading the initial power limits in xe_hwmon_get_preregistration_info()
fails, pl1_on_boot[] and pl2_on_boot[] remain zero while hwmon registration
still proceeds:

drivers/gpu/drm/xe/xe_hwmon.c:xe_hwmon_get_preregistration_info() {
	if (hwmon->xe->info.has_mbx_power_limits) {
		if (xe_hwmon_pcode_read_power_limit(hwmon, PL1_HWMON_ATTR, CHANNEL_CARD,
						    &hwmon->pl1_on_boot[CHANNEL_CARD]) |
		    ...
		    xe_hwmon_pcode_read_power_limit(hwmon, PL2_HWMON_ATTR, CHANNEL_PKG,
						    &hwmon->pl2_on_boot[CHANNEL_PKG])) {
			drm_warn(&hwmon->xe->drm,
				 "Failed to read power limits, check GPU firmware !\n");
		} else {
}

If the subsequent pcode read in xe_hwmon_power_is_visible() succeeds, the
writable powerX_max and powerX_cap attributes are exposed, and any value
userspace writes is clamped to the zeroed pl1_on_boot[] or pl2_on_boot[]
entry in xe_hwmon_power_max_write(), programming a 0W power limit into
firmware and severely throttling the GPU.

[1] https://sashiko.dev/#/bug/linux-9f11cfb5-f9c5-490e-879e-791ecaae33e5

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261119104029.3066660-1-karthik.poosa@intel.com?part=1

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] drm/xe/hwmon: Expose powerX_cap_max
  2026-11-19 10:47 ` sashiko-bot
@ 2026-11-19 15:24   ` Poosa, Karthik
  0 siblings, 0 replies; 3+ messages in thread
From: Poosa, Karthik @ 2026-11-19 15:24 UTC (permalink / raw)
  To: sashiko-reviews, intel-xe


On 19-11-2026 16:17, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found no new issues
> in this patch, but noticed 1 related pre-existing issue it does not
> introduce:
> - [High] drm/xe/hwmon: power limit clamped to 0 in
>    xe_hwmon_power_max_write() [1]
> --
>
> commit 26f9457c44502cd969a84eaf5b2ee961f51b7a3e
> Author: Karthik Poosa <karthik.poosa@intel.com>
>
> drm/xe/hwmon: Expose powerX_cap_max
>
> This commit adds a read-only power<i>_cap_max sysfs attribute on platforms
> with mailbox power limits to report the default PL2 (burst) power limit
> set by firmware at boot.
>
>> diff --git a/drivers/gpu/drm/xe/xe_hwmon.c b/drivers/gpu/drm/xe/xe_hwmon.c
>> index defe8aaca7c11..8e0d295aabc88 100644
>> --- a/drivers/gpu/drm/xe/xe_hwmon.c
>> +++ b/drivers/gpu/drm/xe/xe_hwmon.c
> [ ... ]
>> @@ -1423,6 +1424,8 @@ xe_hwmon_power_is_visible(struct xe_hwmon *hwmon, u32 attr, int channel)
>>   	struct xe_mmio *mmio = xe_root_tile_mmio(hwmon->xe);
>>   
>>   	switch (attr) {
>> +	case hwmon_power_cap_max:
>> +		return hwmon->pl2_on_boot[channel] ? 0444 : 0;
> [Severity: High]
> This problem wasn't introduced by this patch, but can a pcode read failure
> during probe cause xe_hwmon_power_max_write() to clamp user power limits
> to 0 [1]?
>
> If reading the initial power limits in xe_hwmon_get_preregistration_info()
> fails, pl1_on_boot[] and pl2_on_boot[] remain zero while hwmon registration
> still proceeds:
>
> drivers/gpu/drm/xe/xe_hwmon.c:xe_hwmon_get_preregistration_info() {
> 	if (hwmon->xe->info.has_mbx_power_limits) {
> 		if (xe_hwmon_pcode_read_power_limit(hwmon, PL1_HWMON_ATTR, CHANNEL_CARD,
> 						    &hwmon->pl1_on_boot[CHANNEL_CARD]) |
> 		    ...
> 		    xe_hwmon_pcode_read_power_limit(hwmon, PL2_HWMON_ATTR, CHANNEL_PKG,
> 						    &hwmon->pl2_on_boot[CHANNEL_PKG])) {
> 			drm_warn(&hwmon->xe->drm,
> 				 "Failed to read power limits, check GPU firmware !\n");
> 		} else {
> }
>
> If the subsequent pcode read in xe_hwmon_power_is_visible() succeeds, the
> writable powerX_max and powerX_cap attributes are exposed, and any value
> userspace writes is clamped to the zeroed pl1_on_boot[] or pl2_on_boot[]
> entry in xe_hwmon_power_max_write(), programming a 0W power limit into
> firmware and severely throttling the GPU.
Power-limit reads during probe are expected to be available and pass at 
probe time.
If they are unavailable, we should not create the corresponding entries.
I think we should update |xe_hwmon_power_is_visible()| to use 
|pl1_on_boot[]| or |pl2_on_boot[]| accordingly.

>
> [1] https://sashiko.dev/#/bug/linux-9f11cfb5-f9c5-490e-879e-791ecaae33e5
>

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-11-19 15:24 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-19 10:40 [PATCH] drm/xe/hwmon: Expose powerX_cap_max Karthik Poosa
2026-11-19 10:47 ` sashiko-bot
2026-11-19 15:24   ` Poosa, Karthik

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.