Samba 4.0.7 different outcomes i386 vs amd64 accessing [homes]
Dewayne
dewayne.geraghty at heuristicsystems.com.au
Tue Jul 9 00:33:57 MDT 2013
I wonder if anyone can shed any light on a problem where I've built
samba 4.0.7 on an i386 and amd64 machine; and two of the many tests
have different results
The first failure occurs when I try to acquire the xidNumber, using
ldapsearch, which is null; and later when there is a xidNumber,
smbclient correctly accesses a share but not the user's [homes] share.
I've included the url for full build, samba run and test logs files,below.
I'm afraid I'm stuck on how to proceed? Except to use samba4 for
Kerberos/PAC and samba3 for filesharing.
The first test is a failure to acquire a xidNumber
xidNumber failure
------------------------------------------------------------------------
+ /usr/local/samba/bin/samba-tool user create fox f7
User 'fox' created successfully
+ /usr/local/samba/bin/wbinfo --name-to-sid fox
+ SID='S-1-5-21-2002793614-881968167-1911614541-1107 SID_USER (1)'
..
+ /usr/local/samba/bin/ldbsearch -H /usr/local/samba/private/idmap.ldb objectsid=S-1-5-21-2002793614-881968167-1911614541-1107
xidNumber
xidNumber=''
------------------------------------------------------------------------
After all the tests are run, and I perform another ldbsearch I do get a xidNumber,
which succeeds with smbclient when accessing a smb.conf share, but fails to
access a [homes] share; which the i386 does do.
------------------------------------------------------------------------
# /usr/local/samba/bin/ldbsearch -H /usr/local/samba/private/idmap.ldb objectsid=S-1-5-21-2002793614-881968167-1911614541-1107
# record 1
dn: CN=S-1-5-21-2002793614-881968167-1911614541-1107
cn: S-1-5-21-2002793614-881968167-1911614541-1107
objectClass: sidMap
objectSid: S-1-5-21-2002793614-881968167-1911614541-1107
type: ID_TYPE_BOTH
xidNumber: 3000016
distinguishedName: CN=S-1-5-21-2002793614-881968167-1911614541-1107
------------------------------------------------------------------------
However on the amd64, this test fails as follows:
------------------------------------------------------------------------
+ /usr/local/samba/bin/samba-tool domain exportkeytab /tmp/fox.keytab --principal=fox at AS.LAN
+ kinit -t /tmp/fox.keytab fox
+ /usr/local/samba/bin/smbclient -k //t2/shares -c ls
Domain=[AS] OS=[Unix] Server=[Samba 4.0.7]
. D 0 Sat Jul 6 16:20:48 2013
.. D 0 Sat Jul 6 16:20:48 2013
56328 blocks of size 1048576. 10778 blocks available
+ /usr/local/samba/bin/smbclient -k //t2/fox -c ls
Domain=[AS] OS=[Unix] Server=[Samba 4.0.7]
tree connect failed: NT_STATUS_BAD_NETWORK_NAME
+ kdestroy
+ ktutil -v -k /tmp/fox.keytab list
/tmp/fox.keytab:
Vno Type Principal Date Aliases
1 des-cbc-crc fox at AS.LAN 2013-07-09
1 des-cbc-md5 fox at AS.LAN 2013-07-09
1 aes128-cts-hmac-sha1-96 fox at AS.LAN 2013-07-09
1 aes256-cts-hmac-sha1-96 fox at AS.LAN 2013-07-09
1 arcfour-hmac-md5 fox at AS.LAN 2013-07-09
------------------------------------------------------------------------
Further information:
1. The two builds (i386, amd64) are built on the same hardware, and
differ only in their IP and hostnames.
2. Samba was downloaded from http://ftp.samba.org/pub/samba/ on July 2
3. each machine has [home] path /s4/homes
# ls -lrth /s4/homes
total 2
drwxrwxrwx 2 3000016 wheel 512B Jan 1 2010 fox
# ls -lrth /s4/
total 6
drwxr-xr-x 2 root wheel 512B Jan 1 2010 shares
drwxr-xr-x 2 root wheel 512B Feb 4 10:27 kits
drwxrwxrwx 3 root wheel 512B Feb 4 10:55 homes
4. All prerequisites are also built using identical source and make
parameters with the exception that the build of i386 is tuned
for prescott cpu, and amd64 is tuned for core2.
I've included the following files at http://amdg.etowns.org/samba
* Samba build logs for both t4-i386-build-samba4.log.bz2, t5-amd64-build-samba4.log.bz2 - the build process
* Samba test logs t4-i386-test-samba4.log.bz2, t5-amd64-test-samba4.log.bz2 - test results log via "sh -x test-samba4.sh"
* Samba running logs t4-i386-samba-single-d5.log.bz2, t5-amd64-samba-single-d5.log.bz2 - log from "samba -i -M single -d 5"
These machines are specifically for testing Samba4, so if more information is required eg. Debug at 10. I'm happy to provide.
1. Linking sequence diverges at step 3346
< [3346/3799] Linking default/source4/heimdal_build/libhx509-samba4.so
< [3347/3799] Linking default/source4/heimdal_build/libheimbase-samba4.so
< [3348/3799] Linking default/lib/util/libsamba-util.so
---
> [3346/3799] Linking default/nsswitch/libwbclient/libwbclient.so
> [3347/3799] Linking default/lib/util/libsamba-util.so
> [3348/3799] Linking default/lib/socket/libinterfaces.so
2. The i386 build contains 9 lines more than the amd64 - though a diff is 1785 lines.
3. No other observations due to the large number of build differences on the two platforms
2816,2817d2815
< default/librpc/gen_ndr/py_server_id.c: In function 'initserver_id':
< default/librpc/gen_ndr/py_server_id.c:215: warning: overflow in implicit constant conversion
3090,3091d3087
< ../lib/util/time.c: In function 'unix_to_nt_time':
< ../lib/util/time.c:151: warning: comparison is always false due to limited range of data type
3443,3445d3438
< ../source4/torture/smb2/durable_open.c: In function 'test_durable_open_alloc_size':
< ../source4/torture/smb2/durable_open.c:1595: warning: cast from pointer to integer of different size
< ../source4/torture/smb2/durable_open.c:1595: warning: cast from pointer to integer of different size
3477,3478d3469
< ../source4/torture/ndr/samr.c: In function 'samr_changepassworduser3_w2k8r2_out_check':
< ../source4/torture/ndr/samr.c:297: warning: integer constant is too large for 'long' type
Regards, Dewayne.
Sydney, Australia (GMT +10 hours)
More information about the samba-technical
mailing list