Comment # 10 on bug 1174116 from Vlastimil Babka

(In reply to Michal Svec from comment #9)
> And this seems to have helped.
> The host doesn't crash any more (with the 15.2 stock kernel).

Eh, too bad, it was supposed to catch the culprit, not prevent the bug from
manifesting by changed layout/timing...
Let's try being more specific about the debugging and limit it only to
suspected caches.

What happens if you boot with just this parameter?
slub_nomerge

And then, cat /proc/slabinfo | grep biovec, is the output like this?
gehinom:~ # cat /proc/slabinfo | grep biovec
biovec-128           272    272   2048   16    8 : tunables    0    0    0 :
slabdata     17     17      0
biovec-64            544    544   1024   32    8 : tunables    0    0    0 :
slabdata     17     17      0

Then try booting N times with a special parameter for each biovec-X in the list
above, e.g. in my case, 2 times:

slub_nomerge slub_debug=FU,biovec-64
slub_nomerge slub_debug=FU,biovec-128

Thanks.