increased power usage after upgrade from kinetic to lunar

I can also confirm this one also on a HP 15s-fq4xxx series laptop
I tested this on the kinetic kernel 5.19.

Status changed to 'Confirmed' because the bug affects multiple users.

The laptop has Realtek RTL8822CE 802.11ac PCIe Wireless Network Adapter. I notice that when I use rfkill to block the wlan, that package state c8 is being used. When I rfkill unblock, the best state achieved is once again c3. Kernel version is 6.2.0-26.

So the problem could be the Realtek driver.

kernel bisects have revealed that prior to 5.19.0-10.10, c3 is the best power save level achieved and post 5.19.0-12.12, c8 is achieved.

Inspecting the commits between -10.10 & -12.12 indicates that 3 commits are relevant:
UBUNTU: SAUCE: vmd: fixup bridge ASPM by driver name instead
UBUNTU: SAUCE: PCI/ASPM: Enable LTR for endpoints behind VMD
UBUNTU: SAUCE: PCI/ASPM: Enable ASPM for links under VMD domain.

These 3 commits are also applied to 5.15 and 6.2.

So it seems that these patches do not completely explain the observed behaviour.

Further to the comment above about the Realtek wifi, I tried using an old D-Link GWL-G122 USB wifi (uses rt73usb driver) while the built-in wifi was blocked. It worked ok and the laptop could achieve package c8 while pinging via the usb wifi.

So, the kernel can work with wifi and achieve package c8 as you would expect. I'm still suspicious of the Realtek driver.

I tried the 6.4.0-1.1 kernel and it achieves package c8.

But, the power usage by the Realtek driver has greatly increased under 6.4. Using "top" with each kernel booted for a similar amount of time, 6.2 has cpu usage for irq/154-rtw88_pci of 0:02:92 whilst under 6.4 cpu usage for irq/155-rtw88_pci was 0:24:10.

I tried booting with various mainline kernels and determined that the power usage increase was introduced between 6.4-rc5 and 6.4-rc6. Inspection of the rc6 changelog indicated that this commit was a good chance:
mantic git repo 3918dd0177ee ("wifi: rtw88: correct PS calculation for SUPPORTS_DYNAMIC_PS", 2023-05-27).

So, I created a patch from this commit and applied to lunar git repo. For the patch to apply successfully I also had to apply in the order below:
f0e741e4ddbc ("wifi: rtw88: add bitmap for dynamic port settings", 2023-04-14)
ccf73f6e69c0 ("wifi: rtw88: add port switch for AP mode", 2023-04-14)

Building and booting the kernel worked. Kernel 6.2 now achieves package c8 and also exhibits the same CPU increase in top for irq/154-rtw88. I have only been using the laptop to type up this comment and in that time top shows 0:15:59 cpu time. CPU% is toggling between 0.7% & 1.0%. This seems high when compared to and intel wifi on another laptop.

Is there any interest in applying these commits to hwe-6.2 for jammy ?

Or should I close the bug and create another for the cpu usage ?

lunar only has a few months to go but enabling hwe-6.2 to achieve package c8 would benefit LTS users.

upstream bug raised regarding CPU increase for irq/154-rtw88 - https://bugzilla.kernel.org/show_bug.cgi?id=217828

I see that commit "wifi: rtw88: correct PS calculation for SUPPORTS_DYNAMIC_PS" has been picked up in the update to upstream stable 6.1.34 and is in the master-next branch of the lunar git repo. i'll test when the next kernel is built.

Confirm that 6.2.0-34.34 can reach package C8

6.2.0-34.34 has been released