[Nagiosplug-devel] [ nagiosplug-Bugs-741296 ] check_dns.c segfault with -a and NXDOMAIN

[SourceForge.net]

Bugs item #741296, was opened at 2003-05-21 14:28
Message generated for change (Settings changed) made by kdebisschop
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=397597&aid=741296&group_id=29880

Category: None
Group: CVS
>Status: Closed
>Resolution: Fixed
Priority: 5
Submitted By: David Maynard (dpmaynard)
Assigned to: Karl DeBisschop (kdebisschop)
Summary: check_dns.c segfault with -a and NXDOMAIN

Initial Comment:
check_dns.c 1.13 from CVS compiled and run under 
Red Hat 7.1 and testing against bind-8.2.3+patches on 
a Cobalt generates a segmentation fault if the requested 
domain does not exist and the "-a" flag is specified.

The "address" variable is NULL when the code reaches 
line 183 and attempts a strcmp.  The null pointer to 
strcmp generates a segfault.

Arguably, this is a bug in the parsing of the "address" 
variable since it probably shouldn't be a NULL at this 
point in the code.  However, the code for "-a" is 
predicated on STATE_OK, so if something earlier in the 
code indicates STATE_WARNING (e.g., because it 
detects a NXDOMAIN return), then the "-a" check will 
not be performed and Nagios will "miss" the CRITICAL 
error.  (If I tell it to match an expected address, it is 
critical if it doesn't exist.)

As a result, my "fix" (to achieve the results I need), was 
to change the code for the "-a" test to include a check 
for a null address pointer.  My patch attached.

-dpm


----------------------------------------------------------------------

Comment By: Karl DeBisschop (kdebisschop)
Date: 2003-05-31 10:44

Message:
Logged In: YES 
user_id=1671

further testing should my revised patch failed to trap the
condition you had described - I tried to trap the NULL in
the while loop. After testing and rereading your post, I
realized that it was the break in the while() loop that
skipped past the address check.

The code in CVS should not segfault. It will be less
informative than old code in some cases, so I may do a
little more work later. But it should be revady for testing,
in both HEAD and r1_3_0


----------------------------------------------------------------------

Comment By: Karl DeBisschop (kdebisschop)
Date: 2003-05-31 10:00

Message:
Logged In: YES 
user_id=1671

IMO, your thoughts about where the true bug lies are correct.

And we know form experince that other implementations will
segfault on your patch because you try to print a null string.

I take you point about failing to note that the address is
null and move the check up to that point in the code.

This is the patch I have applied.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=397597&aid=741296&group_id=29880