kernel panic on 4.9-release with nullfs

[Kovács János]

Hi!

I moved out dev from nullfs mounted part, and sshd starts up!

But now after some minutes the same crash. I made some other
dump:

1: one nullfs-mounted jail at init 6: reboot after panic: null_checkvp
http://turando.felvi.hu/crash/crash_2.tgz   (~430M)
http://turando.felvi.hu/crash/gdb.out.2


2: simply nullfs mounted jails: 2xbind9 (etc,named.conf+zone files,usr);
      oops proxy(etc,usr); mysqld(etc,usr), djbdns (etc,usr)

after starting oops, numvnodes -and also freevnodes- starting to
increment 6/sec, with no connection to this proxy. But after starting
djbdns's dnscache, numvnodes starting to increment by ~200/sec !!
1-2 minutes and at 33695 numvnodes it stops incrementing and
after that (3-4 sec) the system freeze. Till the freeze cow,wire,act
changing on systat -v m1 screen.

(Someone said, that crash can be because the adaptec 2120s and
2GB of ram. Anyone about that?)

http://turando.felvi.hu/crash/crash_3.tgz  (~440M)
http://turando.felvi.hu/crash/gdb.out.3
http://turando.felvi.hu/crash/systat.out.3


3: the same as 2, but only dnscache running over nullfs mount,
nothing else. The numvnodes incrementing to the same value
(to: 33695, ~100/sec) then it freeze in 3-4 sec.

http://turando.felvi.hu/crash/crash_4.tgz  (~230M)
http://turando.felvi.hu/crash/gdb.out.4


I think, the crash_4.tgz can be the most interesting...


KJ.

ps: I'm going out till dec 3 where I can't read emails, but after that I
can continue making crahs_dumps if there is any patch :)