Bug#900248: nvidia-driver: update to 390.59 breaks direct rendering

[Dean Loros]

Well---I'm still picking up the pieces here.  Not sure what happened,
but that did not work for me & for some reason, I can't revert it back
or boot the older 4.16-1. everything just hangs. I'm in my backup
system, looking at the kernel logs. Last entry starts:

Jun 23 21:22:52 debian kernel: [    0.000000] Linux version
4.16.0-2-amd64 (debian-kernel at lists.debian.org) (gcc version 7.3.0
(Debian 7.3.0-23)) #1 SMP Debian 4.16.16-2 (2018-06-22)

Jun 23 21:22:52 debian kernel: [    0.000000] Command line:
BOOT_IMAGE=/boot/vmlinuz-4.16.0-2-amd64
root=UUID=c18c39f1-0e2c-4c89-a534-e3c8414bb273 ro cgroup_enable=memory
swapaccount=1 elevator=noop nvidia-current.NVreg_EnablePCIeGen3=1
slab_common.usercopy_fallback=y

Normal boot until GDM "should" come up, then I get:

Jun 23 21:23:00 debian kernel: [   28.992343] nvidia-uvm: Loaded the
UVM driver in 8 mode, major device number 242
Jun 23 21:23:00 debian kernel: [   29.383969] resource sanity check:
requesting [mem 0x000c0000-0x000fffff], which spans more than PCI Bus
0000:00 [mem 0x000c0000-0x000dffff window]
Jun 23 21:23:00 debian kernel: [   29.384003] caller
_nv001169rm+0xe3/0x1d0 [nvidia] mapping multiple BARs
Jun 23 21:23:02 debian kernel: [   30.747951] show_signal_msg: 14
callbacks suppressed
Jun 23 21:23:02 debian kernel: [   30.747953] gnome-shell[1370]:
segfault at 20 ip 00007f2ee7714aed sp 00007fff4654cb00 error 4 in
libmutter-2.so.0.0.0[7f2ee7617000+165000]

Last line repeats about 50 times with slightly different addresses
[7f2ee7617000+165000] is a typical example--the 165000 stays constant.

There is also a note about updating runlevels during this period.

I get the same result with my older 4.16-1 kernel now...no combination
of workaround(s) seem to have any effect. Thoughts welcome.