Bug 1139361
Summary: | RHEL6.6 ssh as admin on ipa server fails after 389-ds-base update | ||||||||
---|---|---|---|---|---|---|---|---|---|
Product: | Red Hat Enterprise Linux 6 | Reporter: | Scott Poore <spoore> | ||||||
Component: | 389-ds-base | Assignee: | Noriko Hosoi <nhosoi> | ||||||
Status: | CLOSED DUPLICATE | QA Contact: | Sankar Ramalingam <sramling> | ||||||
Severity: | unspecified | Docs Contact: | |||||||
Priority: | unspecified | ||||||||
Version: | 6.6 | CC: | jgalipea, jhrozek, ksiddiqu, lkrispen, nkinder, nsoman, rmeggins, xdong | ||||||
Target Milestone: | rc | Keywords: | TestBlocker | ||||||
Target Release: | --- | ||||||||
Hardware: | Unspecified | ||||||||
OS: | Unspecified | ||||||||
Whiteboard: | |||||||||
Fixed In Version: | Doc Type: | Bug Fix | |||||||
Doc Text: | Story Points: | --- | |||||||
Clone Of: | Environment: | ||||||||
Last Closed: | 2014-09-11 18:43:16 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: | |||||||||
Bug Blocks: | 990143, 1109379 | ||||||||
Attachments: |
|
Description
Scott Poore
2014-09-08 18:19:55 UTC
Created attachment 935431 [details]
dirsrv/ logs
Fyi, marking testblocker because it was preventing ipa-replica-install from successfully passing the ipa-replica-conncheck. Hi Scott, I think you had no problem with 389-ds-base-1.2.11.15-42. 1.2.11.15-43 was built with this patch on top of 1.2.11.15-42. * Fri Sep 5 2014 Noriko Hosoi <nhosoi> - 1.2.11.15-43 - Release 1.2.11.15-43 - Resolves: #1112702 - Broken dereference control with the FreeIPA 4.0 ACIs (#47885) Looking into the access log in /var/log/dirsrv/slapd-TESTRELM-TEST, it ends with ABANDON. The timestamp matches "Sep 8 13:07:20 rhel6-1 sshd[3351]: pam_sss(sshd:account): Access denied for user admin: 4 (System error)" (although, since the DS access logs are buffered, the timestamp is not very accurate.) [08/Sep/2014:13:07:20 -0500] conn=11 op=16 SRCH base="cn=accounts,dc=testrelm,dc=test" scope=2 filter="(&(objectClass=ipaHost)(fqdn=rhel6-1.testrelm.test))" attrs="objectClass cn fqdn serverHostName memberOf ipaSshPubKey ipaUniqueID" [08/Sep/2014:13:07:20 -0500] conn=11 op=16 RESULT err=0 tag=101 nentries=1 etime=0 notes=P [08/Sep/2014:13:07:20 -0500] conn=11 op=17 SRCH base="fqdn=rhel6-1.testrelm.test,cn=computers,cn=accounts,dc=testrelm,dc=test" scope=0 filter="(objectClass=*)" attrs="objectClass cn memberOf ipaUniqueID" [08/Sep/2014:13:07:20 -0500] conn=11 op=17 RESULT err=0 tag=101 nentries=1 etime=0 notes=P [08/Sep/2014:13:07:20 -0500] conn=11 op=18 ABANDON targetop=NOTFOUND msgid=17 No other errors nor operation failures are reported in the log files. I wonder why the DS received ABANDON request. And deref (fixed by the 47885 patch) is involved in the operations there? Ludwig, do you have any idea? no. The change from .42 to .43 seems to be only the fix for #47885 and it changes the behaviour of the deref plugin (as requested by sssd). Now we have the message: Sep 8 13:07:20 rhel6-1 sssd[be[testrelm.test]]: dereference processing failed : No such file or directory So to me it looks like sssd is nothandling the change, but we need someone from sssd to look into it what is the IPA version you are using ? [root@rhel6-1 ~]# rpm -q ipa-server sssd ipa-server-3.0.0-42.el6.x86_64 sssd-1.11.6-4.el6.x86_64 Created attachment 935745 [details]
/var/log/sssd dir
I thought the issue fixed in #47885 was only visible with the new acis in IPA 4.0. So the question is if .43 is needed in that env ? Chances are this is caused by the same SSSD bug as 1139044. Scott, can you re-test with an older SSSD version? 1.11.6-25 or older would do. sssd-1.11.6-4.el6.x86_64 is the version I had on the server where I saw the problem. Is that too old? FYI, I tried with sssd-1.11.6-25.el6.x86_64 and see the same issue still. So the DS behaviour changed to something SSSD doesn't expect and I'm not sure which part behaves correctly. It's apparently fallout of the dereference changes. We are dereferencing a memberof attribute of a host object that has no memberof attribute. With the old version, whenever we searched and used the deref control, we got the deref control back in the reply. That's not the case with the new version and the SSSD dereference code relies on getting the control back. Which of the two is the expected behavior? Here is the diff between the ldapsearches with DS-42 and 43:
diff new-version old-version
12c12,13
< krbLastSuccessfulAuth: 20140910090956Z
---
> control: 1.3.6.1.4.1.4203.666.5.16 false MIQAAAAA
> krbLastSuccessfulAuth: 20140910090146Z
The krbLastSuccessfulAuth changed b/c I was authenticating as the host.
(In reply to Jakub Hrozek from comment #14) > So the DS behaviour changed to something SSSD doesn't expect and I'm not > sure which part behaves correctly. It's apparently fallout of the > dereference changes. > > We are dereferencing a memberof attribute of a host object that has no > memberof attribute. > > With the old version, whenever we searched and used the deref control, we > got the deref control back in the reply. That's not the case with the new > version and the SSSD dereference code relies on getting the control back. > > Which of the two is the expected behavior? This is part of the change, when removing entries returned with the deref control because there is no access I noticed that the list could be empty and then did not return the control - maybe this is incorrect. Will try to check the RFC and ask my team. I can revert this change and always return the control, even if it is empty. But reading this part of the spec:
>>
The control value may contain dereference attribute values without
any dereferenced attribute values, as detailed in Section 2.3. The
control semantics does not specify whether this is a consequence of a
dangling link or of the application of access restrictions on the
values of the attributes to be dereferenced.
<<
it should also be ok to return references to entries without access.
This issue is a regression caused by the fix for bug 1112702. Closing this as a duplicate, as we'll handle fixing this in the original bug. *** This bug has been marked as a duplicate of bug 1112702 *** |