Re: [SPF:fail] Re: [PATCH] SASL problems with spnego on 8.0-BETA4

From: George Mamalakis <mamalos_at_eng.auth.gr>
Date: Thu, 25 Feb 2010 13:57:15 +0200
On 25/02/2010 13:42, George Mamalakis wrote:
> I have sent an email to fbsd-stable list with subject:  "openldap 
> client GSSAPI authentication segfaults in fbsd8stable i386" regarding 
> this issue, where I list all my tests on all different machines, and a 
> stack trace of the system where ldapwhoami segfaults. I have received 
> no answer for this topic yet, but I think that if some of you reads 
> it, he may find an answer. At the time of this writing, on fbsd8stable 
> systems (i386) with heimdal/openldap-sasl-client/cyrus-sasl, 
> ldapwhoami and ldapsearch segfault when called without a ticket. 

Moreover,

as stated in this mail (subject:  "openldap client GSSAPI authentication 
segfaults in fbsd8stable i386"), which I have copied in this email along 
with one correction after this paragraph, the problem seems to be with 
gss_release_buffer () from /usr/lib/libgssapi.so.10. Please read my 
comment and little "hack" we tried over this function, on the end of the 
aforementioned email, I think it will help.

Thank you all for your help, once more.


"Dear all,

I am facing many instabilities in FBSD8 with openldap-client and sasl 
authentication (GSSAPI in particular). I have setup an openldap 2.4.1 
server with gssapi support (through cyrus-sasl-2.1.23) on a fbsd8-stable 
amd64 latest sources, in a esxi host. In the same host I have setup two 
fbsd8-stable i386 clients; one has latest sources, the other one is 
installed via the iso-image of January's fbsd snapshot;  on both systems 
openldap client and sasl is installed (all ldap/cyrus versions on all 
hosts mentioned in this email are the same). My laptop has fbsd8-i386 
stable (sources 25 January 2010), and on my laptop I have setup an 
fbsd8-stable i386 (snapshot iso image) on a virtualbox client. Lastly, 
on the esxi host I have setup another fbsd8-stable amd64 system, to act 
as an ldap client (latest sources).

To summarize, and put a label-number on each host, we have:

1 - esxi: fbsd8(latest) amd64 openldap server
2 - esxi: fbsd8(latest) i386 openldap client
3 - esxi: fbsd8(snapshot) i386 openldap client
4 - esxi: fbsd8(latest) amd64 openldap client
5 - laptop: fbsd8(jan 25) i386 openldap client
6 - laptop/vbox: fbsd8(snapshot) i386 openldap client

The openldap server is installed in a jail, and the client is tested in 
the same jail.

Kerberos works on all machines (same /etc/krb5.conf), and ldap as well.

In all machines, line 96 of /usr/bin/krb5-config is changed to read:

lib_flags="$lib_flags -lgssapi -lgssapi_spnego -lgssapi_krb5  -lheimntlm"

instead of:
lib_flags="$lib_flags -lgssapi -lheimntlm"

which was the default, since without these lines I couldn't get gssapi 
authentication to work for cyrus (and spnego). This change was made 
after recommendations given from this very mailing list, and I was very 
happy to see that the ldap server worked as I expected.

On all system, cat /usr/local/etc/openldap/ldap.conf:

BASE    dc=ee,dc=auth,dc=gr
URI ldap://ldap.ee.auth.gr
SASL_MECH    GSSAPI


without kiniting in any client I get the following outcomes when I give 
ldapwhoami (with no arguments):

1:
SASL/GSSAPI authentication started
ldap_sasl_interactive_bind_s: Local error (-2)
     additional info: SASL(-1): generic failure: GSSAPI Error:  
Miscellaneous failure (see text) (unknown mech-code 2 for mech unknown)

which is expected.

2:
SASL/GSSAPI authentication started
Segmentation fault: 11 (core dumped)

which is not rational at all

3:
SASL/GSSAPI authentication started
Segmentation fault: 11 (core dumped)

4:
SASL/GSSAPI authentication started
ldap_sasl_interactive_bind_s: Local error (-2)
     additional info: SASL(-1): generic failure: GSSAPI Error:  
Miscellaneous failure (see text) (unknown mech-code 2 for mech unknown)

which is the same as 1 (as expected)

5:
SASL/GSSAPI authentication started
Segmentation fault: 11 (core dumped)

6:
SASL/GSSAPI authentication started
Segmentation fault: 11 (core dumped)

if I kinit to mamalos, ldapwhoami returns:

1:
SASL/GSSAPI authentication started
SASL username: mamalos_at_EE.AUTH.GR
SASL SSF: 56
SASL data security layer installed.
dn:uid=mamalos,ou=people,dc=ee,dc=auth,dc=gr

which is super!

2:
SASL/GSSAPI authentication started
Segmentation fault: 11 (core dumped)

which is dramatic.

3:
SASL/GSSAPI authentication started
Segmentation fault: 11 (core dumped)

4:
SASL/GSSAPI authentication started
ldap_sasl_interactive_bind_s: Local error (-2)
     additional info: SASL(-1): generic failure: GSSAPI Error:  
Miscellaneous failure (see text) (unknown mech-code 2529638919 for mech 
unknown)

which is very strange, since mech-code seems unnaturally large.

5:
SASL/GSSAPI authentication started
SASL username: mamalos_at_EE.AUTH.GR
SASL SSF: 56
SASL data security layer installed.
dn:uid=mamalos,ou=people,dc=ee,dc=auth,dc=gr

which is super, but without kinit the same command segfaulted on this 
machine

6:
SASL/GSSAPI authentication started
SASL username: mamalos_at_EE.AUTH.GR
SASL SSF: 56
SASL data security layer installed.
dn:uid=mamalos,ou=people,dc=ee,dc=auth,dc=gr

which is the exact same behavior as 5 above.

All this means that there is no single pattern!!!!!!!

If I gdb ldapwhoami in the clients that segfault, I get:

GNU gdb 6.1.1 [FreeBSD]
Copyright 2004 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you 
are
welcome to change it and/or distribute copies of it under certain 
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i386-marcel-freebsd"...(no debugging symbols 
found)...
(gdb) run -d0
Starting program: /usr/local/bin/ldapwhoami -d0
(no debugging symbols found)...(no debugging symbols found)...(no 
debugging symbols found)...(no debugging symbols found)...(no debugging 
symbols found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...(no debugging symbols found)...(no debugging symbols 
found)...SASL/GSSAPI authentication started

Program received signal SIGSEGV, Segmentation fault.
0x2831e187 in free () from /lib/libc.so.7
(gdb) where
#0  0x2831e187 in free () from /lib/libc.so.7
#1  0x2850fb82 in gss_release_buffer () from /usr/lib/libgssapi.so.10
#2  0x2850f552 in gss_release_name () from /usr/lib/libgssapi.so.10
#3  0x2850bea9 in gss_init_sec_context () from /usr/lib/libgssapi.so.10
#4  0x283f9abf in gssapi_client_mech_step () from 
/usr/local/lib/sasl2/libgssapiv2.so.2
#5  0x280e84b1 in sasl_client_step () from /usr/local/lib/libsasl2.so.2
#6  0x28443100 in ?? ()
#7  0x00000000 in ?? ()
#8  0x00000000 in ?? ()
#9  0xbfbfe968 in ?? ()
#10 0xbfbfe954 in ?? ()
#11 0xbfbfe964 in ?? ()
#12 0x28445860 in ?? ()
#13 0x280e83fe in sasl_client_step () from /usr/local/lib/libsasl2.so.2
#14 0xbfbfe8a8 in ?? ()
#15 0x280e9135 in sasl_client_start () from /usr/local/lib/libsasl2.so.2
#16 0x00000000 in ?? ()
#17 0x00000000 in ?? ()
#18 0xbfbfe968 in ?? ()
#19 0xbfbfe954 in ?? ()
#20 0xbfbfe964 in ?? ()
#21 0xd7a3b2da in ?? ()
#22 0x283abad8 in ?? () from /lib/libc.so.7
#23 0x00000000 in ?? ()
#24 0x283ab730 in __stderrp () from /lib/libc.so.7
#25 0xbfbfe878 in ?? ()
#26 0x2838c764 in vfprintf () from /lib/libc.so.7
Previous frame inner to this frame (corrupt stack?)

I built openldap and cyrus-sasl on this machine from sources (not 
ports), and (after a long time of trying to find out how to run 
configure successfully in both installations) the outcome was exactly 
the same (meaning segfaults). So, one of my admins wrote a c program 
that overrides gss_release_buffer returning always 0 (what is expected 
after normal operation) and compiled it as a library. The program code 
is nothing more than just:

int gss_release_buffer(void *a, void *b) {
   return 0;
}

we ld_preloaded the library, and when we ran ldapwhoami the outcome was:
SASL/GSSAPI authentication started
ldap_sasl_interactive_bind_s: Local error (-2)
     additional info: SASL(-1): generic failure: GSSAPI Error:  
Miscellaneous failure (see text) (unknown mech-code 2529638919 for mech 
unknown)

When we ran it with no kerberos tickets, we got:

SASL/GSSAPI authentication started
ldap_sasl_interactive_bind_s: Local error (-2)
     additional info: SASL(-1): generic failure: GSSAPI Error:  
Miscellaneous failure (see text) (unknown mech-code 2 for mech unknown)

The exact same errors as the aforementioned client 4 (esxi, 
amd64)!!!!!!!!!!!!!

What on earth is happening?!?!!?!?!

Now one can easily see that there is a definite problem regarding memory 
freeing, and after overcoming that the mech-code 2529638919 implies that 
some segment in memory is overwritten by some "random" value, so 
mech-code returns the number 2529638919 instead of a number of 
marginality 1.

What is more definite, is that openldap doesn't work out-of-the-box if 
gssapi support is required, and behaves randomly in different 
architectures/virtualHosts/platforms.

The problem may have been something related to line 96 in 
/usr/bin/krb5-config... I don't know.

What is sure, is that I am having second thoughts on using fbsd as my 
openldap/heimdal server for my setup, which would be quite disappointing 
for me, since I am using fbsd for the last many-many years, on all my 
laptops and servers. The only "good part" is that the only machine that 
works seamlessly (until now, at least) is the heimdal/ldap server.

Thank you all in advance and I hope that we will find an answer to all 
this."

"On 11/02/2010 20:45, George Mamalakis wrote:
> 4:
> SASL/GSSAPI authentication started
> ldap_sasl_interactive_bind_s: Local error (-2)
>     additional info: SASL(-1): generic failure: GSSAPI Error:  
> Miscellaneous failure (see text) (unknown mech-code 2529638919 for 
> mech unknown)
>
> which is very strange, since mech-code seems unnaturally large. 
This problem has been resolved. I had an issue with my /etc/hosts file, 
where the name of the ldap server could not be resolved correctly (via 
the gssapi library I assume), and openldap client gave me this reply 
(both ldap server and heimdal server had the same IP (two jails on the 
same host)). After changing the order in which the host and its IP 
appeared in /etc/hosts the problem stopped (which is still strange, 
since ldapwhoami -D 'blabla' -W worked ok, even with the old /etc/hosts). "


I still believe that the aforementioned behavior on the amd64 build is 
still not the one to be expected, something must be overriden if the 
client won't get the result it expects from the kdc.

-- 
George Mamalakis

IT Officer
Electrical and Computer Engineer (Aristotle Un. of Thessaloniki),
MSc (Imperial College of London)

Department of Electrical and Computer Engineering
Faculty of Engineering
Aristotle University of Thessaloniki

phone number : +30 (2310) 994379
Received on Thu Feb 25 2010 - 10:57:25 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:01 UTC