named mystery -- error: dumping master file: master/tmp-wTjhUzoix6

[Alex Goncharov]

,--- Oliver Fromme (Mon, 1 Sep 2008 15:31:07 +0200 (CEST)) ----*
| Alex Goncharov wrote:
|  > [...]
|  > After this change, every time I restart `named', the ownership of the
|  > `master' directory is changed to `bind' -- and this is what I want:
|  > user `bind', I would think, should be allowed to write to this
|  > directory.
| 
| No, it shouldn't.  It's a security matter.  If there's an
| exploitable bug in BIND, an attacker could manipulate your
| master zone files.  That's why the bind user should *not*
| be able to write to your master directory.

OK, I am ready to accept this point of view and make it my starting
point again (I tried, in the past).

| There's no reason that the named process needs write access
| to the master directory.  If you use dynamic zone updates,
| you should use the "dynamic" directory for those zones,
| which is writable by bind.

I just tried a simplistic change:

 a. Changed "type master" to "type dynamic" in named.conf.

 b. cp master/* dynamic

Starting `named' with this I get:

   /etc/namedb/named.conf:358: 'dynamic' unexpected

How do I use the `dynamic' directory? (If you know the answer -- I'll
do more reading later.)

OTOH, I see this example at
`http://www.boran.com/security/sp/bind9_20010430.html#BM4_setting_jail_permission'

--------------------
zone "test2.com" {
type master;
file "test2.com";
allow-update { updaters; };
};
--------------------

Which is:

  a. Close enough to what I have, in my original `named.conf', before
     a `dynamic' change attempt.

  b. Implies that updating a master zone is not such an unusual idea.

Any comments on this?

-- Alex -- alex-goncharov at comcast.net --