Hi Rafael,
To help with testing the "ACPI: scan: Defer enumeration of devices with _DEP lists"
patch I've added it (the entire set) to my local 5.10 based tree about a month ago.
Approx. 2 weeks ago I noticed that sometimes the kernel build from my local tree
would not boot on a Lenovo Yoga Tablet 2 1051L machine. Reverting those changes
fixed the problem that the device was sometimes not booting.
When it did not boot, it would hang at:
Starting udev Coldplug all Devices...
And FWIW there was an oops about an unhandled kernel pagefault pointing to a sysfs
directory entry reading function (I thought I saved the oops but it seems
that I did not).
I started debugging this but I did not really reach any conclusions before
going on vacation. Fast forward to today when I build a 5.11-rc1 kernel and
booted that 10 times on the Lenovo Yoga Tablet 2 1051L machine without issues...
So this is really just a FYI, I'm pretty sure that I did not botch cherry-picking
the patches on top of 5.10, do you know if there is a (somewhat) isolated patch which
is necessary for the splitting of the scanning of the ACPI root into 2 steps which
I might have missed ?
Anyways this seems to be resolved, but I did have some ideas on how to pinpoint
the culprit with the "backport" to 5.10, so if you get any bug-reports esp. ones
about hanging at "Starting udev Coldplug all Devices..." I will get back to
my 5.10 backport (I've a tag for it) and restart debugging on it. For now I'm
happy to report that it is resolved with 5.11-rc1.
Regards,
Hans
