You are not logged in.
Pages: 1
since some days/updates my Thinkpad T14[amd] does not go to sleep anymore. I haven't pinpointed the exact update that caused this yet. I anyone else having issue putting the system to sleep?
kernel: 6.19.6-arch1-1
amd-ucode: 20260221-1
linux-firmware-amdgpu: 20260221-1
the output of
amd-s2idle testis here: https://0x0.st/PjCC.txt
Offline
same, tons of ppl are facing this issue and seems related to amd-ucode update. May try to rollback see if helped.
Offline
unfortunately rolling back amd-ucode to 20260110-1 did not help
Last edited by transacid (2026-03-08 09:39:18)
Offline
Would be good to see journal output for this. Ideally with a couple of standbys without using amd-s2idle.
You might also want to try enabling wake on lan as suggested in the amd-s2idle report as well as removing any usb devices. I think it's likely to be something else blocking full sleep, but it would be good to make sure they aren't interfering.
Last edited by radiomike (2026-03-08 12:12:38)
Offline
since some days/updates
Next to the journal the most likely culprit would be the 6.19 kernel? Does it happen w/ the LTS one?
Offline
Very similar issue here. My laptop has an AMD Ryzen AI 365 processor and with up-to-date kernel, ucode, and firmware it cannot enter s0i3 state at suspend. Specifically when suspended the screen does turn off but the fan keeps spinning, and the power indicator LED does not blink as it would previously.
I tested multiple kernels, and among them only the linux-lts 6.12.75 (the latest 6.12.y lts kernel. 6.12.68 also works.) and linux-mainline 7.0.0-rc2 kernels are working. None of the linux 6.19.y kernel works. A particularly strange thing is that I tested my old archlinux livecd ISO with kernel 6.18.4 and it does put my laptop into a good s0i3 state, while the downgrade-installed linux 6.18.4 kernel on my current installation does not. I have not yet checked the ucode/firmware version of that livecd ISO but based on these I guess this issue is related to both the kernel and the ucode/firmware.
By comparing your amd-s2idle report, and also some of my gathered reports with the "good" kernels and the "borked" kernels, it seems like the "SMU idlemask" is a key indicator. With the "good" kernels:
amd_pmc: SMU idlemask s0i3: 0xffff9abf However with the "borked" kernels:
amd_pmc: SMU idlemask s0i3: 0xfeff9abf Looking at these masks I guess the difference essentially says that some "device" or "IP block" at bit 24 is not entering some proper sleep state, but sadly what it actually is were not publicly documented per this FreeBSD commit log. Besides, notice that our laptops share the same SMU idlemask with the "borked" kernel, presumably since the processors are both Ryzen AI 300 series using the same SMU implementation.
Last edited by xuwd1 (2026-03-08 18:46:36)
Offline
ok so 6.19.6-arch1-1, amd-ucode-20260309-1 seems to have fixed at least the overall sleep issue. the system was able to sleep. Here is the report. My SMU idlemask looks even different from the two examples above
# dmesg -T | grep -m1 amd_pmc
[Thu Mar 12 13:08:10 2026] amd_pmc: SMU idlemask s0i3: 0xffff9afdThere's another issue with nxp-nci_i2c which I had to blacklist because it interrupts constantly.
Offline
Pages: 1