[Date Prev][Date Next]
RE: Index corruption and crash in back-ldbm (ITS#2348)
Thanks, your bug analysis is very helpful. The actual main bug in the 2.1
code is that k2 was cont_alloc'd twice, and should not have been. The extra
cont_free should be deleted as well. And you're right about returning without
free'ing all the instances. I'm working on a patch for 2.1, not sure if 2.0
will be patched or not.
-- Howard Chu
Chief Architect, Symas Corp. Director, Highland Sun
Symas: Premier OpenSource Development and Support
> -----Original Message-----
> From: owner-openldap-bugs@OpenLDAP.org
> [mailto:owner-openldap-bugs@OpenLDAP.org]On Behalf Of
> Sent: Wednesday, March 05, 2003 1:38 AM
> To: openldap-its@OpenLDAP.org
> Subject: Index corruption and crash in back-ldbm (ITS#2348)
> Full_Name: Villy Kruse
> Version: 2.0.27
> OS: RH6.2 linux
> URL: ftp://ftp.openldap.org/incoming/Villy-Kruse-030305.patch
> Submission from: (NULL) (188.8.131.52)
> There seems still to be a problem wiht back-ldbm where the index is
> corrupted after heavy insert and delete activity. The bug is very
> hard to reproduce, but when the index has been corrupted, running
> slapindex repeatedly will also cause a index corruption every time.
> This depends on the sequence the keys are added to the various index
> dbm files. The problem occurs when a inserting an item in the middle
> of a full block and the last item is deleted from the current block to
> make room and inserted into the next block.
> I have made a patch which contains some more details of the
> bug and how I
> tried to fix it.