ndc: unix socket ops failing

[Mark_Andrews at isc.org]

	Fixed in BIND 8.2.1.

	See also the SECURITY NOTE in src/README.

	Mark

> 
> Hope this isn't too far out of the scope of this group...even though 
> I'm running into this problem while trying to bring up BIND 8.2, it's 
> probably actually a Solaris problem....  
> 
>     coke:/var/dns> uname -a
>     SunOS coke 5.6 Generic_105181-05 sun4m sparc SUNW,SPARCstation-20
>     coke:/var/dns> head named.run
>     Debug level 1
>     Version = named 8.2 Tue Jun  8 15:44:20 PDT 1999
> 	    dlc at coke:/usr/local/src/BIND/8.2/src/bin/named
> 
> ndc, when talking to named via /etc/ndc, works only partially and 
> causes named to crash with a SIGABRT after a reload.  Seems to work 
> OK, however, when I give it the PIDfile location with -p.  Trace at 
> level 1 gives:
> 
>     ctl_wrtimeout: /etc/ndc: write timeout, closing
>     evDeselectFD(fd 6, mask 0x2)
>     ctl_srvr.c:660: REQUIRE(sess->state == writing) failed
> 
> Oddly, I was able to start this debug trace with the socket-mode ndc 
> session, but "reload" from the same ndc session caused the above crash.
> 
> After it's crashed, I still have this socket:
> 
>     coke:/etc> ls -al ndc
    srw-------   1 root            0 Aug 11 22:08 ndc
> 
> Because of my unfamiliarity with unix-mode sockets, I'm not sure where to go 
> next with the troubleshooting.  Any ideas?
> 
> Thanks.
>     
> 
> -- 
> David Carmean                                           <dlc at netapp.com>
>   PGP fingerprint =  B1 57 EB A8 1D B9 87 86  5F 5C 51 A4 F2 5E ED FD
> 
> 	My God, it's full of Cars!
> 
> 
--
Mark Andrews, Internet Software Consortium
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org