https://bugzilla.kernel.org/show_bug.cgi?id=216516 --- Comment #32 from kolAflash (kolAflash@xxxxxxxxxxxx) --- Created attachment 301909 --> https://bugzilla.kernel.org/attachment.cgi?id=301909&action=edit HP EliteBook 845 G8: dmidecode and lshw (In reply to Mario Limonciello (AMD) from comment #31) > > Can the power consumption in s2idle be lowered further? > > Once the amd-pmc driver is reporting nearly the whole sleep in the deepest > state, I don't know anything else that can be done from the Linux side. > > External things to Linux that come to mind: > - Anything USB plugged in over the sleep can stay powered and sip some > battery. Unplug these. A good example here is a wireless USB dongle for a > mouse or keyboard. > - Do you use any manageability features like AIM-T? These can keep some > parts of the SOC doing things. If so; you can try to disable those. No external devices connected. I'm currently *experimenting* with additional stuff. This kernel command line options: pcie_aspm=force iommu=pt (looks like iommu=pt may already be the default, but I guess adding it should have no disadvantage) And also setting sysfs options before entering s2idle. powersupersave > /sys/module/pcie_aspm/parameters/policy find /sys/devices/ -type f -name wakeup -exec bash -c 'echo disabled > "{}"' \; > > Is ec_no_wakeup=Y the best solution? If yes, maybe add a quirk to the > kernel. > > If this is to be quirked by the kernel by default then it needs to be tied > to this BIOS version, as hopefully this will be fixed in a future HP BIOS. I guess you'll need dmidecode or lshw for that. So I attached it. Here you can see which BIOS versions contain which AMD SMU Firmware version. https://support.hp.com/us-en/drivers/selfservice/swdetails/hp-elitebook-845-g8-notebook-pc/38492638/model/2100000127/swItemId/ob-294554-1?sku=5Z621EA BIOS: 01.10.00 Rev.A 01.09.10 Rev.A 01.09.00 Rev.A SMU Firmware: 64.61.0 BIOS: 01.08.20 Rev.A 01.08.03 Rev.A SMU Firmware: 64.60.0 BIOS: 01.07.01 Rev.A SMU Firmware: 64.48.0 Do you have internal information if all these SMU Firmware versions are affected by the bug? If yes, I'd say the quirk can be enabled for all of them. I'm wondering if there could be a way to dynamically detect broken EC GPE's. And such an GPE apears, automatically set /sys/module/acpi/parameters/ec_no_wakeup = Y Not for myself, but for other users affected by this bug. Because it seems that not just the EliteBook 845 G8 is affected by this. (see above for the ThinkPad Carbon X1 6th) So an automatic detection may spare a lot of future work. I already thought about doing this in userspace. Systemd service: ========== [Unit] After=suspend.target hybrid-sleep.target suspend-then-hibernate.target [Install] WantedBy=suspend.target hybrid-sleep.target suspend-then-hibernate.target ========== And the service could call a script, which checks if there are two IRQs in the dmesg like this: [ 90.048438] PM: Triggering wakeup from IRQ 9 [ 90.048801] PM: Triggering wakeup from IRQ 1 And if yes, ec_no_wakeup=Y is being set. Or at least there's a message telling the user to enable ec_no_wakeup=Y. If this would be in the kernel, that message could appear in dmesg. But I'd guess a general detection wouldn't be that simple!?... -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug.
