kernel panic on jailed sshd - 4.9-release

[Marc G. Fournier]

On Wed, 12 Nov 2003, Oliver Fromme wrote:

> Possibly a program which now triggers a NULLFS problem. For example, it
> might be that some program in 4.8 used standard read/write to access
> files, and it was changed to use mmap() in 4.9, which can be a problem
> with NULLFS.
>
> Of course I could be wrong.  Something else could have caused that
> particular crash.  But still it is a good idea not to use NULLFS if
> possible.  (Particularly in a jail environment if you don't have strict
> control over which programs get executed, such as a user shell box.)

I stop'd using NULLFS myself awhile back, since I too was getting
problems, but figured I was over-tempting fate by using UNIONFS *and*
NULLFS at the same time ...

But, if we can get a core dump from the system in question, at least we
can try and isolate where the issue is ... it might be something that a
simple 2 line patch can fix ...

> If a miracle happened and NULLFS got stable recently, then the
> mount_null(8) manpage should be fixed, because it states just the
> opposite.  ;-)

What I'd love to see is the man page changed so that the BUGS section
indicates in what circumstances NULL/UNIONFS are unstable ... document it
so that ppl don't have to go through trial-n-error each time ...