https://bugzilla.kernel.org/show_bug.cgi?id=216516 --- Comment #25 from Mario Limonciello (AMD) (mario.limonciello@xxxxxxx) --- > So about 2.5 % of power (1.3398 Wh) was consumed. > If the system was in perfect s2idle, this should have been less than 1% for > 23 minutes. > Bug 2.5 % pretty much matches about 700 seconds of "bad" s2idle after the > initial 700 "good" s2idle seconds in deepest state Yup so what we're seeing here is that basically after an event comes in from the EC waking up the SOC, it never goes back to sleep properly. I suspect there is some underlying firmware issue leading to that behavior. Can you please do the following: 1) add right after the usleep_range() you added before this: amd_pmc_idlemask_read(pdev, pdev->dev, NULL); 2) Add https://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git/commit/?h=bleeding-edge&id=cb3e7d624c3ff34a300587929c82af7364cf5c09 That will get us some more debugging information at least when this happens. > I guess that the deepest state has been interrupted by the remnant of what > completely wakes the system if i8042 and LNXSYBUS:00/PNP0C0D:00 are not > disabled (see above). I think you're right. The underlying firmware issue is probably the same as https://bugzilla.kernel.org/show_bug.cgi?id=215770 As a W/A can you have a try with acpi.ec_no_wakeup=1 on your kernel command line? That should stop EC from issuing any wakeups and triggering this other firmware bug. -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug.
