On 29.07.2025 12:03, Andy Mindful wrote:
Kernel-version: 6.8+ (confirmed in Fedora kernel 6.8.4-200.fc38.x86_64 and above) Regression-from: 6.7.11 (working) to 6.8.4+ (regressed) Affected-hardware: Lenovo ThinkPad X1 Carbon Gen 10 (Intel Alder Lake platform) Affected-subsystems: TTY/Console, Power Management, WWAN Modem (Intel XMM7560, iosm driver), Kernel DRM/i915 stack Description: We are reporting a significant regression introduced in Linux kernel version 6.8 (specifically confirmed in Fedora kernel 6.8.4-200.fc38.x86_64 and subsequent versions), affecting Lenovo ThinkPad X1 Carbon Gen 10 laptops. This regression manifests primarily as tty lockups and complete loss of WWAN modem functionality after hibernate cycle. Problematic Behavior:
(..)
3. WWAN Modem (Intel XMM7560, `iosm` driver): The modem fails to reinitialize after hibernation, consistently showing "msg timeout" errors in logs. The modem works correctly after a cold boot but completely fails to recover after resuming from hibernate. Rescanning PCI devices or reloading the `iosm` module does not resolve the issue. Attempts to manually remove the device via `/sys/bus/pci/devices/.../remove` result in a system hang.
Do you have this commit in your test kernel? https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=0b6f6593aa8c3a05f155c12fd0e7ad33a5149c31 With this commit you should *not* see any "msg timeout" on resume from hibernation. Either the modem will reinitialize successfully OR it will be completely dead and the iosm driver will refuse to re-bind to its PCI device in this case. The root cause of these issues seems to be that some platforms cut the power to this modem on S3 and/or S4 entry without abstracting this power control using proper power management interfaces. Some platforms won't even restore modem power on resume until a device-specific reset ACPI method is called. This means the modem device PCI configuration space data won't be restored correctly on resume and so the device will be effectively dead until reboot. ModemManager has a whole issue dedicated to tracking power management issues in a very similar Intel XMM7360 modem on various platforms: https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/992 AFAIK the major difference between XMM7360 and XMM7560 is that XMM7360 speaks proprietary XMMRPC protocol while XMM7560 speaks MBIM, but this should not make a difference here. Also, this is probably not a regression, just something that has never worked correctly (I mean the WWAN modem part). If it indeed used to work then a kernel log from a hibernate/restore cycle on a working kernel version would be nice to have.
Best regards, Andriy
Thanks, Maciej
