Dev1 Galaxy Forum / Suspend regression in kernel 4.19.0-10

Re: Suspend regression in kernel 4.19.0-10

dummy@example.com (Magnus) — Tue, 11 Aug 2020 15:41:41 +0000

OK. I didn't understand that you config your kernels yourself. I supposed that your kernels was stock kernels.
You can search after kernel change logs on the net to see what differs from 4.19.133 and 4.19.137.

https://cdn.kernel.org/pub/linux/kernel … g-4.19.134
https://cdn.kernel.org/pub/linux/kernel … g-4.19.135
https://cdn.kernel.org/pub/linux/kernel … g-4.19.136
https://cdn.kernel.org/pub/linux/kernel … g-4.19.137

Re: Suspend regression in kernel 4.19.0-10

dummy@example.com (RJVB) — Tue, 11 Aug 2020 08:35:22 +0000

Well, I can assure you that I didn't change anything in my .config when I built the 4.19.133 and .137 kernels; I always copy my latest/current /boot/configXXX file to ./.config before running `make xconfig` for a new kernel, and then do a visual diff afterwards.

I could have run into the same issue after installing my 5.7.14 kernel build, but fortunately I ran into what actually seems like a plausible explanation researching a trivial build failure: don't call `finish_cpu()` when offlining a core. I've added an answer with sources to the stackexchange topic referenced in my OP.

Re: Suspend regression in kernel 4.19.0-10

dummy@example.com (Magnus) — Tue, 11 Aug 2020 08:22:53 +0000

It may not be a bug but a function. ;-) Take a look in the two kernel configuration files - in / boot or in the kernel package - and see under the heading "Power management and ACPI options", or similar, if you see any differences. It is very possible that a flag has been changed from "y" (to) to commented and "is not set" (off). Older hardware seems to be phased out...

Suspend regression in kernel 4.19.0-10

dummy@example.com (RJVB) — Sun, 09 Aug 2020 09:12:48 +0000

Apparently upgrading from kernel 4.19.0-9 to 4.19.0-10 makes suspending the computer impossible on certain hardware:

https://unix.stackexchange.com/question … 19-0-9-ker

I haven't experienced this with 4.19.0-10 but can affirm that the regression still exists in the stock 4.19.0.137 kernel.

Reports of this exact issue are floating around by the handful, but none has a real solution (or explanation) that I could find, other than "upgrade or downgrade to version X". There was a suggestion somewhere it could be related to XHCI (USB-3; I can confirm support of that can be a PITA) but I haven't been able to research that sufficiently (rebuilding the kernel just takes too damn long).