TL;DR: It does seem that the kernel was holding things back to the
point that things weren't working. :-(
On 6/8/19 12:01 PM, Grant Taylor wrote:
I'm having problems with newly compiled modules (zfs (et al.) and vbox
(et al.)) and kernel after doing two "emerge -DuNe @world"s.
I've managed to straighten things out. I'm now running a 4.14.120
kernel, with gcc 8.3.0, binutils 2.31.1-r6, and zfs 0.8.1.
I got to a point that I could successfully do the following:
emerge -DuNe @world --exclude sys-fs/zfs --exclude sys-fs/zfs-kmod
--exclude sys-kernel/spl
I had configured gcc and binutils to use 6.4.0 and 2.30-r4 respectively.
At that point I upgraded from 4.9.76 to 4.9.177 successfully. Then I
upgraded to 4.14.120 successfully.
After I got to 4.14.120 successfully, I did another emerge … @world …
(above command) just to make sure that everything is copacetic.
Then I stepped up to gcc 8.3.0, rebuilt some test modules. After that I
stepped binutils (and libs) up to 2.31.1-r6 and rebuilt some test
modules. Everything seemed to work. All tests pass. All modules load
and unload properly.
So I did another emerge … @world … and everything seems to be completely
stable.
I'm guessing I don't need to do all the emerge … @world … but this
system is idle and I can start it in the morning and it finishes
recompiling just shy of 900 packages by the end of the day, including
emerge --depclean && revdep-rebuild.
So, I think that this system is now in a much better state than it was a
week ago.
The only thing I can think of is that the 4.9.76 kernel was far enough
back and before all the Specter / Meltdown / Retpoline fiasco that
things were no longer working like they should. This machine had been
staying back at 4.9.<something> because OvS didn't want to go above that
at the time this system was last updated. Now OvS is happy with
4.14.<something>. (That's why I didn't go to a newer kernel.)
Thank you everybody for the suggestions and comments.
