Suspend/hibernate not working on Dell XPS 15 9500 laptop

[Trevor Bentley <trevor@xxxxxxxxxx>]

I asked on #xen on Freenode and was requested to post here.

Summary: Under Xen, both suspend and hibernate operations put the laptop 
into some sort of unrecoverable low-power mode, and a force power-off is 
required.
Environment:
 - Dell XPS 15 9500, BIOS 1.0.1 (this is a new 2020 model)
 - Intel i7-10750H
 - Intel i915 + Nvidia GTX 1650 Ti Mobile
 - Arch linux (clean install)
 - Linux kernels 5.7.5, 5.7.6 tested
 - i915 driver loaded, no nvidia drivers loaded (nouveau blacklisted)
 - Xen 4.13.1, 4.14.0-rc3 tested
 - UEFI, GRUB2 bootloader, LUKS-encrypted /boot, /root, and swap 
(unencrypted /efi with GRUB stub loader)
My Xen was built from this Arch AUR: https://aur.archlinux.org/packages/xen/
With these small patches to bump to 4.13.1 using gcc 10: 
https://github.com/sl4mmy/docker-aur-xen
And the same patches manually applied to a git checkout to bump to 
4.14.0-rc3.
In non-Xen boots, both suspend and hibernate work (using 's2idle' for 
mem, unsure about 'deep').
In Xen, with no guests configured, both suspend and hibernate in Dom0 
power down the screen, fans, and keyboard backlight.  Once suspended, 
nothing awakens it.
I tried every combination of /sys/power/pm_test 
(freezer/devices/platform with s2idle and 
core/processors/platform/devices/freezer with deep), and all pm tests 
*succeed*.  It goes into whichever low power mode, waits 5 seconds, and 
fully recovers.
I'm unable to interpret the results when /sys/power/pm_trace is enabled. 
 "hash matches" is printed in the dmesg log, seemingly always in a 
different place/format.  /sys/power/pm_trace_dev_match just reported 
"acpi" the few times I checked it.
Please let me know if you have any suggestions to try, or if you need 
any extra information for debugging.
Thanks,

Trevor