Bug 1505647

Summary: NSCD not properly caching netgroup
Product: Red Hat Enterprise Linux 7 Reporter: DJ Delorie <dj>
Component: glibcAssignee: DJ Delorie <dj>
Status: CLOSED ERRATA QA Contact: Sergey Kolosov <skolosov>
Severity: medium Docs Contact: Vladimír Slávik <vslavik>
Priority: medium    
Version: 7.5CC: alanm, al, ashankar, codonell, cww, dj, fweimer, jwright, mcermak, mnewsome, pfrankli, qe-baseos-tools-bugs, vslavik
Target Milestone: rcKeywords: Patch
Target Release: ---   
Hardware: Unspecified   
OS: Linux   
Whiteboard:
Fixed In Version: glibc-2.17-237.el7 Doc Type: Bug Fix
Doc Text:
`nscd` netgroups cache properly handles expiration Previously, when the `nscd` service was used to cache LDAP requests for netgroups, some entries in the cache might not expire correctly. As a consequence, the `nscd` cache did not reflect the service database correctly e.g. LDAP. The cache expiry code has been fixed, and changes in netgroups propagate to the cache correctly again.
Story Points: ---
Clone Of: 1436335 Environment:
Last Closed: 2018-10-30 09:36:35 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On: 1436335    
Bug Blocks: 1477664, 1505884    

Comment 2 Carlos O'Donell 2018-02-27 17:27:25 UTC
Description of problem:
NSCD netgroup cache failure. Persistent cache inspection and logs show that setnetgrent() and innetgr() cache separate GETNETGRENT records in the cache. When the entry corresponding to innetgr() expires, on reload the netgroup entry has 0 results and further calls to innetgr always return 0 even if it should return 1. Debugging nscd, I found that innetgr adds the GETNETGRENT record to cache with the group string not null-terminated.

Version-Release number of selected component (if applicable):
glibc-2.17-196.el7

How reproducible:
always

Steps to Reproduce:
1. set up netgroups provided by ldap / nslcd
2. perform innetgr() with triplet that returns 1
3. wait for nscd debug log to show expire and reload of GETNETGRENT from innetgr()
4. perform innetgr() again, it returns 0

Actual results:
innetgr always returns 0 after GETNETGRENT record from previous innetgr() expires

Expected results:
innetgr should return 1 if entry is in netgroup or 0 if it is not

Additional info:
I've tested attached patch and it seems to work. One cache record exists from both setnetgrent() and innetgr(), and when it exires it reloads without issue. This has a side benefit of fewer type 19 records in nscd cache and fewer ldap lookup.

The proposed fix by Al Heisner <al> is:

--- netgroupcache.c	2017-03-27 12:49:39.833083147 -0400
+++ netgroupcache.c.patched	2017-03-27 12:49:25.797939129 -0400
@@ -484,7 +484,7 @@
 {
   const char *group = key;
   key = (char *) rawmemchr (key, '\0') + 1;
-  size_t group_len = key - group - 1;
+  size_t group_len = key - group; //include null terminator in size
   const char *host = *key++ ? key : NULL;
   if (host != NULL)
     key = (char *) rawmemchr (key, '\0') + 1;
---

This has to go upstream.

Comment 12 errata-xmlrpc 2018-10-30 09:36:35 UTC
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHSA-2018:3092