Source: linux-signed-amd64 Severity: important Tags: upstream Subject: linux-image-6.18.15+deb13-amd64: suspend via s2idle fails to reach S0i3 on ThinkPad T14s Gen 4 AMD, while 6.12.74 works
Package: linux-image-6.18.15+deb13-amd64 Version: 6.18.15-1~bpo13+1 Severity: important Tags: upstream Dear Debian Kernel Team, I am reporting what appears to be a suspend regression in the Debian backports kernel on a Lenovo ThinkPad T14s Gen 4 AMD (type 21F8CTO1WW). Summary ------- On linux-image-6.18.15+deb13-amd64 (6.18.15-1~bpo13+1), suspend enters s2idle but fails to reach S0i3 / deep low-power idle. The laptop "suspends", but the fan keeps running and power draw remains high. On the same machine, the Debian security kernel linux-image-6.12.74+deb13+1-amd64 (6.12.74-2) works correctly: s2idle reaches S0i3 successfully and the machine suspends properly. System ------ - Model: Lenovo ThinkPad T14s Gen 4 AMD (21F8CTO1WW) - CPU: AMD Ryzen 7 PRO 7840U - BIOS before testing: 1.25 - BIOS updated during debugging via fwupd to: 1.29 - mem_sleep: [s2idle] - The platform exposes S0/S4/S5 only; no "deep"/S3 is available Regression comparison --------------------- Working kernel: - linux-image-6.12.74+deb13+1-amd64 - version 6.12.74-2 Failing kernel: - linux-image-6.18.15+deb13-amd64 - version 6.18.15-1~bpo13+1 Both kernels were tested on the same machine and firmware. Behavior on failing kernel (6.18.15-1~bpo13+1) ----------------------------------------------- Relevant journal excerpts from the failing boot: Linux version 6.18.15+deb13-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.18.15-1~bpo13+1 ACPI: PM: (supports S0 S4 S5) Low-power S0 idle used by default for system suspend systemd-sleep: Performing sleep operation 'suspend'... PM: suspend entry (s2idle) amd_pmc AMDI0009:00: Last suspend didn't reach deepest state systemd-sleep: System returned from sleep operation 'suspend' PM: suspend exit During testing on 6.18, AMD PMC debugfs showed: - S0ix Residency Time: 0 - Last S0i3 Status: Unknown/Fail Observed user-visible effect: - laptop appears suspended - but fan keeps spinning - opening the lid resumes the machine Behavior on working kernel (6.12.74-2) -------------------------------------- On 6.12.74+deb13+1-amd64, the same machine suspends correctly. AMD PMC debugfs on the working 6.12 kernel reports: - Last S0i3 Status: Success - Time in S0i3: 9852905 us - Residency Time: 9852905 This strongly suggests a regression between the two Debian kernel versions. What was already tested ----------------------- - BIOS updated from 1.25 to 1.29: no fix for 6.18 behavior - acpi.ec_no_wakeup=1: no fix for 6.18 behavior - Suspend tested with and without USB-C dock attached - Suspend tested with dock-side USB/Thunderbolt wake settings adjusted - The separate lid-close/logind issue was fixed independently and is unrelated Potentially relevant hardware path ---------------------------------- This platform includes AMD Pink Sardine USB4/Thunderbolt controllers: - 0000:c5:00.5 - 0000:c5:00.6 While debugging 6.18, amd_pmc debug output suggested USB4 activity remained significant when suspend failed, but the regression remains reproducible even after disconnecting the dock. Installed package versions -------------------------- - linux-image-6.18.15+deb13-amd64: 6.18.15-1~bpo13+1 - linux-image-6.12.74+deb13+1-amd64: 6.12.74-2 - amd64-microcode: 3.20250311.1 - firmware-linux / firmware-linux-nonfree: 20260110-1~bpo13+1 If needed, I can provide additional logs, but the key point is: - 6.18 backports fails to reach S0i3 on suspend - 6.12 on the same machine succeeds Thanks. -- System Information: Debian Release: 13.4 APT prefers stable-updates APT policy: (500, 'stable-updates'), (500, 'stable-security'), (500, 'stable') Architecture: amd64 (x86_64) Kernel: Linux 6.12.74+deb13+1-amd64 (SMP w/16 CPU threads; PREEMPT) Locale: LANG=C.UTF-8, LC_CTYPE=C.UTF-8 (charmap=UTF-8) (ignored: LC_ALL set to C.UTF-8), LANGUAGE not set Shell: /bin/sh linked to /usr/bin/dash Init: systemd (via /run/systemd/system) LSM: AppArmor: enabled
