AMD Ryzen 9 3900XT - Slackware Current64 - kernel 5.9.3-burdi64 - VM's fail to start

[burdi01]

Some weeks ago my multi-boot Intel i7-6700 system died on me. From backups I was able to clone things on the AMD Ryzen 9 3900XT replacement. Since then when VB is started under Slackware Current64 all VM's fail to start, whereas when VB is started under Xubuntu 20.04.1 these VM's start as expected.

The failure occurs very early in the start sequence: the big "starting arrow" remains shown and the system freezes all over. Sometimes I can start another user session via <Alt+Fx>. Very often a hard restart is required. The Slackware instance initially ran my own kernel 5.8.14-burdi64, the Xubuntu one kernel 5.4.0-48-generic. During the testing these kernels were upped. At the moment they are 5.9.3-burdi64 and 5.4.0-52-generic respectively. The symptoms however stayed the same.
I started out with VB 6.1.14-140239. Nowadays I test with VB 6.1.16-140961, a preliminary test with VB 6.1.17-141128 shows no discernable effects.
It should be noted that both the Slackware and Xubuntu instances via symlinks refer to the same VM's.

The Slackware "error" VBox.log ends with:

GUI: UIDesktopWidgetWatchdog::sltHandleHostScreenAvailableGeometryCalculated:[[BR]]
 Screen 0 work area is actually resized to: 0x0 x 1920x1173

while the Xubuntu "ok" one contains:

GUI: UIDesktopWidgetWatchdog::sltHandleHostScreenAvailableGeometryCalculated:[[BR]]
 Screen 0 work area is actually resized to: 0x25 x 1920x1150

The i7-6700 had embedded video, so used the i915 driver. My Ryzen has a discrete video card -- an ASUS Radeon RX 550 PH-RX550-2G-EVO one: -- lspci: [AMD/ATI] Lexa PRO [Radeon 540/540X/550/550X / RX 540X/550/550X] (rev c7). supported by the amdgpu driver.

To isolate the problem I did the following tests:
-- I cloned the Slackware instance to an i5 laptop with embedded video: the VM's there started and ran without problems;
-- Back on the AMD box I replaced my 5.9.3-burdi64 kernel with Slackware's stock 5.4.74 one: the VM's started and ran without problems;
-- I upped the Xubuntu kernel to the 5.9.2-050902-generic one from the kernel PPA: the VM's started and ran without problems;
-- I replaced the Xubuntu VB 6.1.16-140961 .deb implementation with the .run one: the VM's started and ran without problems;
-- Back on my Slackware instance I replaced the 5.4.74 kernel with the Xubuntu 5.9.2-050902-generic one: the VM's started and ran without problems.

As said when the problem occurs one often cannot do much more than hard resetting the system. By formally repeating these tests and collecting as much info as possible I found that for the "error" case the syslog reports an oops "unable to handle page fault for address", the call trace of which starts within vboxdrv.

The Slackware the VBox.logs and the /var/log/messages file as collected are:
-- VBox-59S.log: the "error" with my 5.9.3-burdi64 kernel;
-- VBox-54S.log: the "ok" with Slackware's stock 5.4.74 kernel;
-- VBox-59U.log: the "ok" with the Xubuntu 5.9.2-050902-generic kernel;
-- messages: searching for "=========" will give you the "error" and "ok" sections.

It is likely that something in the configuration of my 5.9.3-burdi64 kernel triggers the problem. But which setting is the culprit? My config is 187.4 kB and the Xubuntu one is 248.8 kB. So comparing the two is an almost hopeless endevour ...

Any hints would be appreciated.
:D