openldap-bugs

openldap-bugs@openldap.org

1 participants
20958 discussions

(ITS#6137) Cancel(pending operation) acts like Abandon
by h.b.furuseth＠usit.uio.no 21 May '09

21 May '09

Full_Name: Hallvard B Furuseth Version: HEAD OS: URL: Submission from: (NULL) (129.240.6.233) Submitted by: hallvard Looking at the code (not sure how to reproduce): Cancel(pending operation) abandons the operation, without responding to either Cancel or the cancelled operation. That code in slapd/cancel.c is copied from the equivalent abandon.c code. Cancel as currently written must wait for the cancelled operation to complete. But an active operation must not wait for a pending one, all threads could block. For now we can make Cancel respond with LDAP_CANNOT_CANCEL: "can't handle Cancel now, try Abandon". Repeating myself, a better fix (RE25?) would be to make the cancelled operation send the Cancel response. Don't how deep surgery that'd need to slapd.

1 0

Re: (ITS#5328) backends sending/setting results
by masarati＠aero.polimi.it 21 May '09

21 May '09

h.b.furuseth(a)usit.uio.no wrote: > Fixed relay_back_has_subordinates(). > I've lost track of what I thought to do with back-ldap. I think you should just ignore it, otherwise all you could do is to to perform a onelevel search for each entry (!) that requests the attribute. We (dunno if it's documented) do not guarantee hasSubordinates is correct, all in all. p.

1 0

(ITS#6136) replicated slapd.d interrupts syncrepl refresh
by mbackes＠symas.com 21 May '09

21 May '09

Full_Name: Matthew Backes Version: 2.4, probably head? OS: any URL: Submission from: (NULL) (12.49.250.67) If you modify an object in cn=config in a multi-master environment with replicated cn=config, other databases on replicating other-masters in the refresh-state will be interrupted when they receive the change. Applying any change to their local cn=config resumes the syncrepl refresh operation.

1 0

(ITS#6135) Changed TLS settings in cn=config require a restart
by hyc＠OpenLDAP.org 20 May '09

20 May '09

Full_Name: Howard Chu Version: 2.4.16/HEAD OS: Solaris 10 URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (76.91.220.157) Submitted by: hyc I thought we already had an ITS for this but didn't find it... slapd's global TLS settings are stored in an SSL context that only gets initialized at startup time. So changes to these settings via cn=config take no effect until the next restart. Changes to the other TLS users (syncrepl, back-ldap, back-meta) take effect immediately, as expected. bconfig.c's config_tls_config needs to check whether slapd is online or not, and reinitialize the global context after these changes if so.

1 0

(ITS#6134) Missing ITS state/directory documentation
by h.b.furuseth＠usit.uio.no 20 May '09

20 May '09

Full_Name: Hallvard B Furuseth Version: OS: URL: http://www.openldap.org/lists/openldap-commit/200902/msg00172.html Submission from: (NULL) (129.240.6.233) Submitted by: hallvard https://www.openldap.org/its/help.html does not document states Active and Partial. From the enclosed commit URL: Active - bug is being worked on Partial - a partial fix was released Also the help page's directory list shows "Archive.*", but the ITS system seems to have no match for that.

1 0

Re: (ITS#5328) backends sending/setting results
by h.b.furuseth＠usit.uio.no 20 May '09

20 May '09

Fixed relay_back_has_subordinates(). I've lost track of what I thought to do with back-ldap. -- Hallvard

1 0

(ITS#6133) back-relay issues
by h.b.furuseth＠usit.uio.no 20 May '09

20 May '09

Full_Name: Hallvard B Furuseth Version: HEAD OS: Linux URL: Submission from: (NULL) (129.240.6.233) Submitted by: hallvard Back-relay issues - fixing some, awaiting confirmation/discussion of others: * The handlers for Abandon, Cancel and connection-init/destroy should not exist, as far as I can tell. They forward the call to the underlying backend database, but that means it receives the call twice: Slapd calls these handlers for all databases anyway (unless one returns success in case of Cancel), since there's no DN on which to dispatch. See abandon.c:fe_op_abandon(), cancel.c:cancel_extop(), and backend.c:backend_connection_init()/backend_connection_destroy(). The Cancel handler seems broken anyway. It sets op->o_cancel = LDAP_CANNOT_CANCEL on the Cancel operation. Presumably it intended the to-be-cancelled operation. I'm "#if 0"ing these out for now in case I've missed something. Can remove the code later. * Fixing RB_ERR (set rs->sr_err) and RB_SEND (send if error) handling: relay_back_select_backend() did not obey (!(fail_mode & RB_ERR)) except when it caught recursion. LDAP_SUCCESS makes no sense combined with these flags. relay_back_operational() could set rs->sr_err = success on failure, which seemed a bit excessive. (Though it should return success.) Removing the RB_ERR flag. relay_back_chk_referrals() can send success on failure. Except the function is not used, so I'm just #if 0ing and commenting it. This also means relay_back_select_backend() without these flags could just as well get SlapReply rs == NULL, which again means rs==NULL could function as a !RB_ERR flag. Not doing that yet, in case the changes below lead to some reason not to. * Fixing relay_back_select_backend() crash when select_backend()==NULL: include servers/slapd/schema/core.schema database relay suffix cn=relay,cn=test overlay rwm rwm-suffixmassage cn=nowhere ldapsearch -b cn=urgle,cn=relay,cn=test -s base * back-relay can be configured to cause infinite recursion. This dumps core with the above search: include servers/slapd/schema/core.schema database relay subordinate suffix cn=relay,cn=test relay cn=relay,cn=test database null suffix cn=test There are some tests to catch that, but half-hearted ones. And some comments about it, like this: /* FIXME: this test only works if there are no overlays, so * it is nearly useless; if made stricter, no nested back-relays * can be instantiated... too bad. */ if ( bd == NULL || bd == op->o_bd ) { return LDAP_SUCCESS; } I don't know how that's worse in this case than in other cases, nor what exactly how that snippet is supposed to work, so I don't know if the suggestion below fixes it and the test can be removed. The 'bd == op->o_bd' part can be removed since relay_back_select_backend() now returns bd==NULL when that happens: (bd->be_private == op->o_bd->be_private) implies (bd == op->o_bd). Anyway, recursion can now be properly caught with op->o_extra. Wrap calls to the underlying backend (e.g. in relay_back_op) in - untested: OpExtraDB oex; oex.oe_db = op->o_bd; oex.oe.oe_key = op->o_bd->be_private; LDAP_SLIST_INSERT_HEAD( &op->o_extra, &oex.oe, oe_next ); ... LDAP_SLIST_REMOVE( &op->o_extra, &oex.oe, OpExtra, oe_next ); and check in relay_back_select_backend() with relay_back_info *ri = (relay_back_info *)op->o_bd->be_private; OpExtra *oex; LDAP_SLIST_FOREACH( oex, &op->o_extra, oe_next ) { if ( oex->oe_key == ri ) <caught recursion, fail and return NULL>; } However, does the comment quoted above mean that it's desirable to allow a few rounds of recursion (a back-relay instance calling itself)? If so, define a config parameter relay-max-recursion with default=0, and define a struct with the OpExtra members + a counter which says how many rounds of recursion the operation has had so far for this database. See the OpExtra comment in slap.h.

1 0

(ITS#6132) slapi_entry_has_children broken
by h.b.furuseth＠usit.uio.no 20 May '09

20 May '09

Full_Name: Hallvard B Furuseth Version: HEAD OS: URL: Submission from: (NULL) (129.240.6.233) Submitted by: hallvard slapi_entry_has_children() does not check for errors, and also calls the database's be_has_subordinates handler even when it does not exist. Finally, when there is no be_has_subordinates to call, it builds and destroys a pblock which isn't used for anyting. I don't know what a pblock is, but it looks like a waste of time. Fixing.

1 0

(ITS#6131) "TLSVerifyClient try" not working with GNU TLS
by subbarao＠computer.org 20 May '09

20 May '09

Full_Name: Kartik Subbarao Version: 2.4.16 OS: Debian 5.0.1 URL: ftp://ftp.openldap.org/incoming/ Submission from: (NULL) (76.99.175.5) When TLSVerifyClient is set to "try", OpenLDAP improperly rejects SSL connections without a client certificate. The problem appears to start with this section of code in tls.c around line 1564: #ifdef HAVE_GNUTLS if ( ld->ld_options.ldo_tls_require_cert != LDAP_OPT_X_TLS_NEVER ) { err = tls_cert_verify( ssl ); if ( err && ld->ld_options.ldo_tls_require_cert != LDAP_OPT_X_TLS_ALLOW ) return err; } #endif tls_cert_verify() calls gnutls_certificate_verify_peers2(), which appears to return error 49 when no client certificate is presented. tls_cert_verify() doesn't seem to distinguish between this case, and the case of an invalid client certificate, returning -1 in both cases.

1 0

Re: (ITS#6129) slapd seg faults on shut-down
by michael＠stroeder.com 17 May '09

17 May '09

Howard Chu wrote: > This appears to be a Cyrus/SASL issue. > > In particular, your trace of thread 1 shows that slapd is shutting down > the SASL library and SASL is trying to tear down libldap. Inside slapd, > the Cyrus SASL library should never be touching libldap. As such, you > should check your SASL configuration, you probably have a plugin enabled > that shouldn't be. You're right. Removing cyrus-sasl-ldap-auxprop solved this problem. Sorry for the noise. Ciao, Michael.

1 0

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

openldap-bugs