Re: slapd crash with passwd database, and a patch (ITS#1467)

[Kurt@OpenLDAP.org]

Fixed in HEAD, thanks.

At 02:14 PM 2001-11-27, brad@vecna.com wrote:
>Full_Name: Brad Thompson
>Version: 2.0.14 and 2.0.18
>OS: Debian Linux (Woody)
>URL: ftp://ftp.openldap.org/incoming/
>Submission from: (NULL) (18.187.6.228)
>
>
>When I run slapd like:
>
>  slapd -d 1023 -f ~/ldap/slapd.conf -h ldap://localhost:1234
>
>And ~/ldap/slapd.conf contains:
>
>  include /usr/local/etc/openldap/schema/core.schema
>  database passwd
>  suffix "dc=vecna,dc=com"
>  access to *
>    by anonymous read
>    by * read
>
>And I search it with:
>  ldapsearch -x -H ldap://localhost:1234 -b 'dc=vecna,dc=com' 'objectclass=*'
>
>It segfaults.  Here is a backtrace:
>
>#0  0x0805ff47 in acl_get (a=0x80b3f58, count=0xbf5ff868, be=0x80b3ed8,
>    op=0x80bb708, e=0x80c2568, desc=0x80bb890, nmatch=10, matches=0xbf5ff870)
>    at acl.c:258
>#1  0x0805fcd4 in access_allowed (be=0x80b3ed8, conn=0x4029797c, op=0x80bb708,
>    e=0x80c2568, desc=0x80bb890, val=0x0, access=ACL_SEARCH) at acl.c:195
>#2  0x0805ef31 in test_presence_filter (be=0x80b3ed8, conn=0x4029797c,
>    op=0x80bb708, e=0x80c2568, desc=0x80bb890) at filterentry.c:245
>#3  0x0805eb68 in test_filter (be=0x80b3ed8, conn=0x4029797c, op=0x80bb708,
>    e=0x80c2568, f=0x80bb830) at filterentry.c:93
>#4  0x0806f571 in passwd_back_search (be=0x80b3ed8, conn=0x4029797c,
>    op=0x80bb708, base=0x80bb640 "dc=vecna,dc=com",
>    nbase=0x80bb658 "DC=VECNA,DC=COM", scope=2, deref=0, slimit=0, tlimit=0,
>    filter=0x80bb830, filterstr=0x80bb860 "(objectClass=*)", attrs=0x0,
>    attrsonly=0) at search.c:134
>#5  0x080509cd in do_search (conn=0x4029797c, op=0x80bb708) at search.c:278
>#6  0x0804f6bc in connection_operation (arg_v=0x80bb6a0) at connection.c:831
>#7  0x08075522 in ldap_int_thread_pool_wrapper (pool=0x80a1a28) at tpool.c:379
>#8  0x40156eca in pthread_start_thread () from /lib/libpthread.so.0
>#9  0x40156f11 in pthread_start_thread_event () from /lib/libpthread.so.0
>
>Here is a patch to fix it.  I don't understand the code that well, so
>it is possible it is the wrong fix, but it fixes the segfault and my
>test case behaves properly.
>
>--- search.c.orig       Tue Nov 27 17:04:19 2001
>+++ search.c    Tue Nov 27 16:35:59 2001
>@@ -84,6 +84,8 @@
>                        e = (Entry *) ch_calloc(1, sizeof(Entry));
>                        e->e_attrs = NULL;
>                        e->e_dn = ch_strdup( base );
>+                       e->e_ndn = ch_strdup( base );
>+                       (void) dn_normalize( e->e_ndn );
> 
>                        /* Use the first attribute of the DN
>                        * as an attribute within the entry itself.