Description of problem: nscd segfaults Version-Release number of selected component (if applicable): 2.3.6-4 How reproducible: Steps to Reproduce: 1.Start with default /etc/nscd.conf 2. 3. Actual results: Segmentation fault Expected results: Should start correctly. Additional info: In nscd_conf.c cnt is used twice where that is not correct (cnt is not even initialized at this point). The variable idx should be used instead.
Created attachment 133436 [details] Replace two occurrences of cnt with idx.
I have this same problem in Fedora Core 4 with nscd-2.3.6-3 on an x86_64 using nss_ldap. (LDAP server is running on another machine.) Other similar bugs suggested removing the persistent lines from the nscd.conf file, but that didn't help me. It still segfaults at random times. Have you actually changed those variables and did it make any difference? I want to know because I need this fixed. In fact, I think the severity needs to be elevated to high because when nscd dies, you can't log onto the system, no personal web pages can be located, and incoming mail is sent back with a "user unknown" error. That's pretty serious. I see this bug in a few other places, and it seems to be only an x86_64 problem. I have other i386 FC4 systems with similar software configurations running with no problems. Sep 1 15:31:02 kernel: nscd[25257]: segfault at 00002aab401fee25 rip 0000555555561aaa rsp 00000000401ffa80 error 4 Sep 1 16:31:48 kernel: nscd[26936]: segfault at 00002aab401fe085 rip 00002aaaab1a0750 rsp 00000000401fe5c8 error 4 Sep 2 06:48:09 kernel: nscd[19870]: segfault at 00002aab401fe8c1 rip 00002aaaae654324 rsp 00000000401ff200 error 4
Sorry, but this is on i386. Nscd segfaults immediately.
Fix uploaded to: http://people.redhat.com/jakub/glibc/2.3.6-5/ Whether the Fedora Legacy project chooses to issue it officially or not is their decision.
I would suggest that indeed, this is a critical bug-fix, and that since Jakub Jelinek supplied .src.rpm for us to use, we ought to use what he provided. I will go ahead and submit Jakub's .src.rpm package for source QA here, formally. Note that Jakub's glibc-2.3.6-5.src.rpm package should also fix Bug #188761 (New glibc (2.3.6-3) breaks nis+). Thanks a bunch, Jakub!
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Jakub, would you please verify that your .src.rpm has this as its sha1sum? Thanks! Download location: http://people.redhat.com/jakub/glibc/2.3.6-5/glibc-2.3.6-5.src.rpm $ sha1sum glibc-2.3.6-5.src.rpm 1e2bb1f51172944a598dc79ef755c9913b764ace glibc-2.3.6-5.src.rpm -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (GNU/Linux) iD8DBQFE//XJxou1V/j9XZwRAomlAKD1IxiwCT7uAgOMtZJ/3oFMg5uAawCgyMPX ETE/mhwpkX6SNPdbS8iSe3Q= =CFEd -----END PGP SIGNATURE-----
Yes, that's the correct sha1sum. But, I might have a few further changes in the next week or so (e.g. the glibc-nscd-addinitgroups.patch patch has a bug), together with another FC-5 glibc testing update.
Hi Jakub, Noticed that you pushed a new glibc for FC-5. Were there other changes you had in mind to do here to glibc-2.3.6-5 for FC-4? If not, we will try using the glibc-2.3.6-5 that you proposed. Thanks! -David
Created attachment 137287 [details] glibc-2.4-{10,11} diff Yeah, I meant to add the glibc-2.4-{10,11} changes too. Attaching diff from CVS. Basically: 1) glibc-nscd-addinitgroups.patch patch fixed 2) glibc-rh198762.patch renamed to glibc-rh184086.patch 3) 10 other patches added (not sure 100% if all are applicable and/or apply cleanly to the FC-4 tree) I'll try to verify that and check it into CVS hopefully this Friday.
This is still a bug in FC5 with the following versions nscd-2.4-11 glibc-2.4-11 Oct 3 07:23:49 silver kernel: nscd[15245]: segfault at 00002aab40bff13e rip 00002aaaaeb9f025 rsp 0000000040c00100 error 4
Entering a new Fedora Core 5 bug for Comment #10.
Fedora Core 4 is now completely unmaintained. These bugs can't be fixed in that version. If the issue still persists in current Fedora Core, please reopen. Thank you, and sorry about this.
Where was the "new Fedora Core 5 bug for Comment #10" created? The only thing that I can find is Bug #190431 ...
I was blind. Bug #209881 seems to the FC5 version of this.