working directory not writable

[Mark Andrews]

> >> > (a) If the effective uid/gid is 0 (root), how can the access()
> >> >     call be resulting in "access denied"?
> 
> MA>                 access() != eaccess()
> 
> I think your earlier answer was clearer (at least to me),
> when you said the named initially drops most (if not all)
> of root's special privileges.  That explains everything I
> saw, but doesn't change the fact that the access() test in
> named/server.c is being done at the wrong time.  Here's
> the sequence that I see (plus the adjustment I think is
> needed):
> 
>   named starts, running as root w/all privileges
>   root privileges dropped
>   config file loaded
>   cwd to working directory
>   access(".", W_OK)?           <== wrong time for this
>                                    still limited root
>   setgid named/setuid named
>   access(".", W_OK)?           <== correct time for this
> 
> Obviously I left out many things, but the first access()
> call exists in named/server.c and fails when the working
> directory is
> 
>   drwxr-x--- named named   ...etc...
> 
> It works when group write is enabled, because you cannot
> (or do not?) drop the root privilege that makes root behave
> as if it belongs to all groups, as it were.
> 
> This also explains why, later, named has no problem writing
> to the working directory.
> 
> FWIW, it would also be nice if the message at least indicated
> that it was a warning rather than an error in the log, and
> if possible, the other relevant info like, what directory,
> uid/gid that cannot access it, reason access is not allowed.
> Not all sysops like needing to search the sources and then
> *MODIFY* them to determine these things, after all.
> 
> -- 
> Steve Vallière | mailto:bind9 at e-visions.com
> 
> 
	Yep, your correct the access() should be moved after
	the ns_os_changeuser() call.
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark_Andrews at isc.org