udev: Do not disable powersaving for NVIDIA audio devices #149
Conversation
Somehow, disabling power saving for NVIDIA audio device also prevents the dGPU from going into sleep mode. Signed-off-by: Vasiliy Stelmachenok <[email protected]>
|
Happy to confirm that this fixes the issue :) |
Nice, thanks for the quick report and testing!
Yeah, it's a bit upsetting, but as it is... |
|
FWIW I independently found #147 broke D3 for my GPU too - except I have an AMD 7900XT, not an NVIDIA GPU. So it affected AMD GPUs as well... Mentioning this for the changelog, but that's not super important :) |
|
Oh wait -- I didn't actually try this PR, only removed the original file (as I don't use intel audio anyway). The workaround seems to exclude only NVIDIA GPUs, this is wrong for the reason above... I don't have any such hardware to test, but maybe it might cause issues for intel dGPU HDMI audio? |
Can you please share output of |
(EDIT: Ah, sorry, the PR's udev rules do indeed check the vendor - I've probably misunderstood why you wanted to check the See:
It seems like that on my system the 80:1f.3 device is indeed the onboard audio seeing as my only "audio device"s in lspci are this one and the AMD dGPU: |
I see, thank you. Could you also check if the problem is present if change values as in this diff: diff --git a/usr/lib/modprobe.d/20-audio-pm.conf b/usr/lib/modprobe.d/20-audio-pm.conf
index 761a2fa..1a0d1b9 100644
--- a/usr/lib/modprobe.d/20-audio-pm.conf
+++ b/usr/lib/modprobe.d/20-audio-pm.conf
@@ -2,4 +2,4 @@
# running on battery power. This is needed because it prevents audio cracks on
# some hardware. If device starts working on battery power, same named udev rule
# re-enables power saving.
-options snd_hda_intel power_save_controller=N
+options snd_hda_intel power_save=0
diff --git a/usr/lib/udev/rules.d/20-audio-pm.rules b/usr/lib/udev/rules.d/20-audio-pm.rules
index fa4f11c..b0f88a0 100644
--- a/usr/lib/udev/rules.d/20-audio-pm.rules
+++ b/usr/lib/udev/rules.d/20-audio-pm.rules
@@ -2,7 +2,7 @@
# running on battery power. This is needed because it prevents audio cracks on
# some hardware.
SUBSYSTEM=="power_supply", ENV{POWER_SUPPLY_ONLINE}=="0", TEST=="/sys/module/snd_hda_intel", \
- RUN+="/bin/sh -c 'echo Y > /sys/module/snd_hda_intel/parameters/power_save_controller'"
+ RUN+="/bin/sh -c 'echo 0 > /sys/module/snd_hda_intel/parameters/power_save'"
SUBSYSTEM=="power_supply", ENV{POWER_SUPPLY_ONLINE}=="1", TEST=="/sys/module/snd_hda_intel", \
- RUN+="/bin/sh -c 'echo $(grep -w -q 0x10de /sys/module/snd_hda_intel/drivers/*/*/vendor && echo Y || echo N) > /sys/module/snd_hda_intel/parameters/power_save_controller'"
+ RUN+="/bin/sh -c 'echo 10 > /sys/module/snd_hda_intel/parameters/power_save'" |
Applied this and rebooted, the problem seems to be solved. (also double-checked my |
My mistake was that I blindly trusted information on the Arch Wiki, which mentions the power_save_controller parameter without noting that power_save=0 is sufficient to solve the issues, and power_save_controller=N breaks power management completely. Fixes: #149 (comment) Fixes: #147 (comment) Signed-off-by: Vasiliy Stelmachenok <[email protected]>
Thanks, looks like this should be finally fixed. |
|
I'm a bit confused - as far as I understand, this udev rule is supposed to set power_save to 0 when the laptop is charging, and to 10 when running on battery. However, on my system, when the laptop is charging, the output of cat /sys/module/snd_hda_intel/parameters/power_save is 10, and when it's not charging, the output is 0. Isn't this reversed? |
Sorry, I was a bit tired last night, so I didn't notice. But anyway it's already fixed in 81a5ef2. |
Somehow, disabling power saving for NVIDIA audio device also prevents the dGPU from going into sleep mode.
See: #147 (comment)