Comment 3 for bug 1788499

Revision history for this message
Stefan Bader (smb) wrote :

@Andreas, well the 4.15.0-32 kernel was the security update to handle Foreshadow/L1TF. To mitigate that CPU bug there had been substantial changes to how non-present pages are represented in the page tables. For suspend/resume many parts have to play together. With Xen, there is the hypervisor which is a kernel on its own (and I would guess it also hooks into some BIOS functions visible to guests). The hypervisor loads and runs the kernel for the dom0 guest. This is the linux kernel and dom0 is a paravirtualized guest with special privileges.

So bottom line, its really hard to figure out what goes wrong. Especially without any visual clue. I assume there is a triple fault or the like (which causes the reboot) when trying to bring the virtual CPUs of dom0 back. So the screen is still blanked and I do not think any of the peripheral devices have been re-activated, yet.

One step to verify that it is indeed the change from 4.15.0-31 to 4.15.0-32 would be to select the older kernel at boot via grub. If -31 works but -32 not then research can likely be concentrated on upstream hypervisor changes which mention acessing page table entries of guests.