kai/samba.git
14 years agoModified SamDB to accept options like Ldb.
Nadezhda Ivanova [Wed, 27 May 2009 10:49:51 +0000 (13:49 +0300)]
Modified SamDB to accept options like Ldb.

14 years agos3: update manpage as to the new passdb backend default
Björn Jacke [Wed, 27 May 2009 10:01:21 +0000 (12:01 +0200)]
s3: update manpage as to the new passdb backend default

14 years agos3: make passdb backend defaults to tdbsam
Björn Jacke [Mon, 25 May 2009 12:55:04 +0000 (14:55 +0200)]
s3: make passdb backend defaults to tdbsam

14 years agoFix the build of nfs4_acls.c
Volker Lendecke [Sun, 24 May 2009 20:17:58 +0000 (22:17 +0200)]
Fix the build of nfs4_acls.c

14 years agoFix a size_t/int warning
Volker Lendecke [Sun, 24 May 2009 20:16:34 +0000 (22:16 +0200)]
Fix a size_t/int warning

14 years agoFix some nonempty blank lines
Volker Lendecke [Sun, 24 May 2009 19:46:53 +0000 (21:46 +0200)]
Fix some nonempty blank lines

14 years agoGna, how long do I program in C now??? :-)
Volker Lendecke [Wed, 27 May 2009 06:09:23 +0000 (08:09 +0200)]
Gna, how long do I program in C now??? :-)

14 years agoDetect missing 'witch' before detecting missing autoconf
Andrew Bartlett [Mon, 18 May 2009 05:56:31 +0000 (15:56 +1000)]
Detect missing 'witch' before detecting missing autoconf

14 years agos3/docs Add manpage for "map untrusted to domain" parameter
Steven Danneman [Fri, 22 May 2009 23:57:52 +0000 (16:57 -0700)]
s3/docs Add manpage for "map untrusted to domain" parameter

This fixes bug 6352.

14 years agoHandle the krbtgt special case by looking for RID -514
Andrew Bartlett [Tue, 26 May 2009 03:09:57 +0000 (13:09 +1000)]
Handle the krbtgt special case by looking for RID -514

It turns out (seen in MS-SAMR 3.1.1.7.1 for example) that the primary
way the krbtgt account is recognised as special is that RID.  This
should fix issues such as 'password expired' on the kpasswd service.

Andrew Bartlett

14 years agoAdd DOMAIN_RID_KRBTGT define to security.idl
Andrew Bartlett [Tue, 26 May 2009 04:43:49 +0000 (14:43 +1000)]
Add DOMAIN_RID_KRBTGT define to security.idl

14 years agos4:tevent: Increase trace debug level to 50.
Andrew Kroeger [Sun, 24 May 2009 06:48:49 +0000 (01:48 -0500)]
s4:tevent: Increase trace debug level to 50.

The sheer volume of messages generated by tevent when the trace level is set to
10 makes it difficult to debug issues in a level 10 log.  Increasing this to
50 allows tevent tracing to be enabled if needed, but otherwise keeps the extra
chatter out of a level 10 log.

14 years agoAttempt to fix the build on HP/UX
Volker Lendecke [Tue, 26 May 2009 21:37:14 +0000 (23:37 +0200)]
Attempt to fix the build on HP/UX

14 years agoAttempt to fix the build on NetBSD
Volker Lendecke [Tue, 26 May 2009 20:39:50 +0000 (22:39 +0200)]
Attempt to fix the build on NetBSD

14 years agoRevert "s3: fix build on systems with struct stat member st_flags"
Björn Jacke [Tue, 26 May 2009 21:19:00 +0000 (23:19 +0200)]
Revert "s3: fix build on systems with struct stat member st_flags"

for a cleaner and more complete patch that Volker has in the queue :-)

14 years agos3: fix build on systems with struct stat member st_flags
Björn Jacke [Tue, 26 May 2009 20:37:17 +0000 (22:37 +0200)]
s3: fix build on systems with struct stat member st_flags

14 years agos4:ldb_modules: Correct typos.
Andrew Kroeger [Sun, 24 May 2009 22:40:57 +0000 (17:40 -0500)]
s4:ldb_modules: Correct typos.

14 years agos4:ldb:modules: Correct typos.
Andrew Kroeger [Tue, 26 May 2009 20:22:13 +0000 (15:22 -0500)]
s4:ldb:modules: Correct typos.

14 years agoFix some nonempty blank lines
Volker Lendecke [Tue, 26 May 2009 19:07:08 +0000 (21:07 +0200)]
Fix some nonempty blank lines

14 years agoIntroduce "struct stat_ex" as a replacement for SMB_STRUCT_STAT
Volker Lendecke [Thu, 14 May 2009 13:34:42 +0000 (15:34 +0200)]
Introduce "struct stat_ex" as a replacement for SMB_STRUCT_STAT

This patch introduces

struct stat_ex {
        dev_t           st_ex_dev;
        ino_t           st_ex_ino;
        mode_t          st_ex_mode;
        nlink_t         st_ex_nlink;
        uid_t           st_ex_uid;
        gid_t           st_ex_gid;
        dev_t           st_ex_rdev;
        off_t           st_ex_size;
        struct timespec st_ex_atime;
        struct timespec st_ex_mtime;
        struct timespec st_ex_ctime;
        struct timespec st_ex_btime; /* birthtime */
        blksize_t       st_ex_blksize;
        blkcnt_t        st_ex_blocks;
};
typedef struct stat_ex SMB_STRUCT_STAT;

It is really large because due to the friendly libc headers playing macro
tricks with fields like st_ino, so I renamed them to st_ex_xxx.

Why this change? To support birthtime, we already have quite a few #ifdef's at
places where it does not really belong. With a stat struct that we control, we
can consolidate the nanosecond timestamps and the birthtime deep in the VFS
stat calls.

At this moment it is triggered by a request to support the birthtime field for
GPFS. GPFS does not extend the system level struct stat, but instead has a
separate call that gets us the additional information beyond posix. Without
being able to do that within the VFS stat calls, that support would have to be
scattered around the main smbd code.

It will very likely break all the onefs modules, but I think the changes will
be reasonably easy to do.

14 years agos3:smbd: remove unused global 'orig_inbuf'
Stefan Metzmacher [Tue, 26 May 2009 13:30:39 +0000 (15:30 +0200)]
s3:smbd: remove unused global 'orig_inbuf'

metze

14 years agos3:pam_smbpass: don't call openlog() or closelog() from pam_smbpass
Björn Jacke [Tue, 26 May 2009 13:40:21 +0000 (15:40 +0200)]
s3:pam_smbpass: don't call openlog() or closelog() from pam_smbpass

Patch from Steve Langasek with tiny fixes by me to make it apply to master.
Also see Debian bug #434372 and bugzilla #4831.

Calling openlog() or closelog() inside a pam module is not good as these
functions are not stackable and no program won't re-do openlog() just because a
pam module might have called closelog().

14 years agogitignore: Ignore additional auto-generated files.
Andrew Kroeger [Mon, 25 May 2009 19:46:08 +0000 (14:46 -0500)]
gitignore: Ignore additional auto-generated files.

Corrected path to tdr_proto.h and added librpc/gen_ndr/{cli,srv}_dcerpc.[ch].

14 years agos3/docs: Fix typo in man idmap_rid.
Karolin Seeger [Tue, 26 May 2009 12:16:10 +0000 (14:16 +0200)]
s3/docs: Fix typo in man idmap_rid.

Karolin

14 years agos3:smbd: move SMB1 specific stuff into a substructure of smbd_server_connection
Stefan Metzmacher [Tue, 26 May 2009 08:48:12 +0000 (10:48 +0200)]
s3:smbd: move SMB1 specific stuff into a substructure of smbd_server_connection

metze

14 years agos3:smbd: add support for SMB2 signing
Stefan Metzmacher [Fri, 22 May 2009 20:58:39 +0000 (22:58 +0200)]
s3:smbd: add support for SMB2 signing

metze

14 years agos3:smbd: return the correct security mode and capabilities in SMB2 Negotitate
Stefan Metzmacher [Fri, 22 May 2009 19:26:03 +0000 (21:26 +0200)]
s3:smbd: return the correct security mode and capabilities in SMB2 Negotitate

metze

14 years agos4:libcli/smb2: remove old dialect revision constants
Stefan Metzmacher [Tue, 26 May 2009 07:37:29 +0000 (09:37 +0200)]
s4:libcli/smb2: remove old dialect revision constants

metze

14 years agos4:smb2srv: We only support SMB 2.002.
Stefan Metzmacher [Tue, 26 May 2009 06:44:27 +0000 (08:44 +0200)]
s4:smb2srv: We only support SMB 2.002.

We need to loop over all given dialects and check
if we can find SMB2_DIALECT_REVISION_202.

metze

14 years agos4:libcli/smb2: use new SMB2_DIVELECT_REVISION constants
Stefan Metzmacher [Tue, 26 May 2009 07:38:09 +0000 (09:38 +0200)]
s4:libcli/smb2: use new SMB2_DIVELECT_REVISION constants

Also send them in the order a windows client would
send them (the lowest first).

metze

14 years agos4:libcli/smb2: add some more SMB2 constants
Stefan Metzmacher [Fri, 22 May 2009 20:23:42 +0000 (22:23 +0200)]
s4:libcli/smb2: add some more SMB2 constants

metze

14 years agos3 WHATSNEW: Mention the changes to net
Kai Blin [Tue, 26 May 2009 07:26:56 +0000 (09:26 +0200)]
s3 WHATSNEW: Mention the changes to net

14 years agoMerge branch 'master' of ssh://git.samba.org/data/git/samba
Andrew Tridgell [Tue, 26 May 2009 04:18:16 +0000 (14:18 +1000)]
Merge branch 'master' of ssh://git.samba.org/data/git/samba

14 years agos4:provision: Added ComPartitionSets entry.
Andrew Kroeger [Sun, 24 May 2009 08:37:37 +0000 (03:37 -0500)]
s4:provision: Added ComPartitionSets entry.

Without this entry, opening the COM+ tab under the properties of an OU within
ADUC results in the following error:

"Unable to retrieve all user properties, 0x80072030"

14 years agos4:Added Extended-Rights and subentries.
Andrew Kroeger [Sun, 24 May 2009 19:47:46 +0000 (14:47 -0500)]
s4:Added Extended-Rights and subentries.

Without these entries, using the 'Delegate Control' option in ADUC results in
the following error message in the Delegation of Control Wizard:

"The templates could not be applied.  One or more of the templates is not
applicable.  Click Back and select different templates, and then try again."

14 years agos4:provision: Update DisplaySpecifiers (#5139).
Andrew Kroeger [Fri, 22 May 2009 05:28:36 +0000 (00:28 -0500)]
s4:provision: Update DisplaySpecifiers (#5139).

The classDisplayName attribute controls the actual text displayed to the user
for the top-level menus, so added it to the existing entries.

The attributeDisplayNames attribute contains both the text displayed to the
user and a mapping to the internal directory attribute name for the particular
field, so added these to the existing entries as well.

Added new entries as appropriate to properly complete all menus and labels
within ADUC.

14 years agoDon't use crossRef records to find our own domain
Andrew Bartlett [Tue, 26 May 2009 02:31:39 +0000 (12:31 +1000)]
Don't use crossRef records to find our own domain

A single AD server can only host a single domain, so don't stuff about
with looking up our crossRef record in the cn=Partitions container.
We instead trust that lp_realm() and lp_workgroup() works correctly.

Andrew Bartlett

14 years agoAdd support for sendmsg() in socket_wrapper
Andrew Bartlett [Tue, 26 May 2009 01:43:37 +0000 (11:43 +1000)]
Add support for sendmsg() in socket_wrapper

This is required because the deferred connect code skips the connect()
until sending the packet, but unless we catch this call, the connect()
never happens.

Andrew Bartlett

14 years agoMerge branch 'master' of ssh://git.samba.org/data/git/samba
Andrew Tridgell [Mon, 25 May 2009 23:26:47 +0000 (09:26 +1000)]
Merge branch 'master' of ssh://git.samba.org/data/git/samba

14 years agoadded some more speed tests to tdbtool
Andrew Tridgell [Mon, 25 May 2009 23:24:37 +0000 (09:24 +1000)]
added some more speed tests to tdbtool

This adds 3 simple speed tests to tdbtool, for transaction store,
store and fetch.

On my laptop this shows transactions costing about 10ms

14 years agos3:dbwrap_tool: add listkeys operation
Michael Adam [Mon, 25 May 2009 22:47:15 +0000 (00:47 +0200)]
s3:dbwrap_tool: add listkeys operation

Michael

14 years agos3:dbwrap_tool: remove superfluous command mapping
Michael Adam [Mon, 25 May 2009 22:26:39 +0000 (00:26 +0200)]
s3:dbwrap_tool: remove superfluous command mapping

Michael

14 years agos3:dbwrap_tool: add "erase" opearation
Michael Adam [Mon, 25 May 2009 21:27:28 +0000 (23:27 +0200)]
s3:dbwrap_tool: add "erase" opearation

Michael

14 years agonet: Use samba default command line arguments.
Kai Blin [Thu, 14 May 2009 09:39:01 +0000 (11:39 +0200)]
net: Use samba default command line arguments.

Attention:

The meaning of the -N flag changed.
To get the old meaning for net groupmap set, use the long option --ntname
The long option for using kerberos changed from --kerberos to --use-kerberos

net rpc commands will now prompt for a password if none is given.

As a benefit, net will now accept an authentication file like other samba
command line tools. So no need to specify the password on the command line in
scripts anymore.

This should fix bug #6357

Signed-off-by: Kai Blin <kai@samba.org>
14 years agos3 ctags: ignore all proto.h files for tag generation
Kai Blin [Wed, 20 May 2009 10:07:07 +0000 (12:07 +0200)]
s3 ctags: ignore all proto.h files for tag generation

14 years agos3:winbind_util: remove trailing spaces
Michael Adam [Mon, 25 May 2009 09:55:51 +0000 (11:55 +0200)]
s3:winbind_util: remove trailing spaces

Michael

14 years agos3:dbwrap_ctdb: fix some function header comments
Michael Adam [Mon, 25 May 2009 19:59:40 +0000 (21:59 +0200)]
s3:dbwrap_ctdb: fix some function header comments

Michael

14 years agosource3/utils/log2pcaphex.c(main): fixed file descriptors leak.
Slava Semushin [Fri, 22 May 2009 18:10:05 +0000 (01:10 +0700)]
source3/utils/log2pcaphex.c(main): fixed file descriptors leak.

One of leaks found by cppcheck:
[./source3/utils/log2pcaphex.c:367]: (error) Resource leak: out

14 years agosource{3,4}/torture/smbiconv.c(main): fixed file descriptor leak.
Slava Semushin [Sat, 23 May 2009 13:51:53 +0000 (20:51 +0700)]
source{3,4}/torture/smbiconv.c(main): fixed file descriptor leak.

File descriptor leaks only when we use file instead of stdout.

Found by cppcheck:
[./source3/torture/smbiconv.c:219]: (error) Resource leak: out
[./source4/torture/smbiconv.c:211]: (error) Resource leak: out

14 years agonsswitch/winbind_nss_aix.c(fill_grent): fixed memory leak.
Slava Semushin [Sat, 23 May 2009 14:02:40 +0000 (21:02 +0700)]
nsswitch/winbind_nss_aix.c(fill_grent): fixed memory leak.

Found by cppcheck:
[./nsswitch/winbind_nss_aix.c:241]: (error) Memory leak: result

14 years agos3-selftest: fix typo.
Günther Deschner [Mon, 25 May 2009 13:55:26 +0000 (15:55 +0200)]
s3-selftest: fix typo.

Guenther

14 years agos3:winbind:idmap_ldap: warn about duplicate SID->XID mappings (bug #6387)
Michael Adam [Mon, 25 May 2009 09:54:43 +0000 (11:54 +0200)]
s3:winbind:idmap_ldap: warn about duplicate SID->XID mappings (bug #6387)

With the current infrastructure, we should not return error on
duplicate mappings but just warn instead (because an error would
trigger the attempt to create yet another mapping).

Michael

14 years agos3:winbind:idmap_ldap: warn about duplicate XID->SID mappings (bug #6387)
Michael Adam [Mon, 25 May 2009 09:29:14 +0000 (11:29 +0200)]
s3:winbind:idmap_ldap: warn about duplicate XID->SID mappings (bug #6387)

With the current infrastructure, we should not return error on
duplicate mappings but just warn instead (because an error would
trigger the attempt to create yet another mapping).

Michael

14 years agos3-samr: Fix Bug #6372, usermanager only displaying 1024 groups and aliases.
Günther Deschner [Mon, 25 May 2009 12:05:18 +0000 (14:05 +0200)]
s3-samr: Fix Bug #6372, usermanager only displaying 1024 groups and aliases.

This is now also verified with the RPC-SAMR-LARGE-DC test.

Guenther

14 years agos3-selftest: enable RPC-SAMR-LARGE-DC against Samba3.
Günther Deschner [Mon, 25 May 2009 12:03:16 +0000 (14:03 +0200)]
s3-selftest: enable RPC-SAMR-LARGE-DC against Samba3.

This will fail for alias creation as nss_wrapper does not yet wrap around
libnss_winbind.

Guenther

14 years agos4-smbtorture: add RPC-SAMR-LARGE-DC test.
Günther Deschner [Mon, 25 May 2009 11:08:58 +0000 (13:08 +0200)]
s4-smbtorture: add RPC-SAMR-LARGE-DC test.

This rather simple test creates 4500 objects on a domain controller and checks
the enum calls for the correct number of results.

Guenther

14 years agos4-smbtorture: rename test_EnumDomain{Users,Groups,Aliases} in RPC-SAMR.
Günther Deschner [Fri, 22 May 2009 17:04:25 +0000 (19:04 +0200)]
s4-smbtorture: rename test_EnumDomain{Users,Groups,Aliases} in RPC-SAMR.

Guenther

14 years agos4-smbtorture: re-work test_Create{User,Group,Alias} a little.
Günther Deschner [Thu, 21 May 2009 16:12:29 +0000 (18:12 +0200)]
s4-smbtorture: re-work test_Create{User,Group,Alias} a little.

Guenther

14 years agos3-pamsmbpass: copy _pam_get_item and _pam_get_data from pam_winbind.
Günther Deschner [Fri, 22 May 2009 15:56:37 +0000 (17:56 +0200)]
s3-pamsmbpass: copy _pam_get_item and _pam_get_data from pam_winbind.

Guenther

14 years agos3-rpcclient: use get_domain_handle() fn in enum domain users & groups.
Günther Deschner [Fri, 22 May 2009 14:48:01 +0000 (16:48 +0200)]
s3-rpcclient: use get_domain_handle() fn in enum domain users & groups.

Guenther

14 years agoAttempt to fix a debian build problem
Volker Lendecke [Mon, 25 May 2009 10:36:30 +0000 (12:36 +0200)]
Attempt to fix a debian build problem

14 years agos3/docs: Fix typos.
Karolin Seeger [Mon, 25 May 2009 08:50:23 +0000 (10:50 +0200)]
s3/docs: Fix typos.

Thanks to Oota Toshiya <t-oota at dh.jp.nec.com> for reporting!

Karolin

14 years agofixed interpretation of ACB_PWNOTREQ
Andrew Tridgell [Mon, 25 May 2009 05:23:54 +0000 (15:23 +1000)]
fixed interpretation of ACB_PWNOTREQ

This bit actually means that we should ignore the minimum password
length field for this user. It doesn't mean that the password should
be seen as empty

14 years agofixed the client side password change code
Andrew Tridgell [Mon, 25 May 2009 03:40:52 +0000 (13:40 +1000)]
fixed the client side password change code

The client side code was not falling back to older routines correctly
as it didn't check for the operation range error appropriately. It
also used the old rpc semantics.

14 years agocope with lanman auth being disabled in old password change code
Andrew Tridgell [Mon, 25 May 2009 03:39:56 +0000 (13:39 +1000)]
cope with lanman auth being disabled in old password change code

When lanman auth is disabled and a user calls a password change
method that requires it we should give NT_STATUS_NOT_SUPPORTED

14 years agoTALLOC_FREE happily lives with a NULL ptr. Tim, please check!
Volker Lendecke [Sun, 24 May 2009 20:13:07 +0000 (22:13 +0200)]
TALLOC_FREE happily lives with a NULL ptr. Tim, please check!

Thanks,

Volker

14 years agoFix a race condition in winbind leading to a panic
Volker Lendecke [Sun, 24 May 2009 16:57:13 +0000 (18:57 +0200)]
Fix a race condition in winbind leading to a panic

In winbind, we do multiple events in one select round. This needs fixing, but
as long as we're still using it, for efficiency reasons we need to do that.

What can happen is the following: We have outgoing data pending for a client,
thus

state->fd_event.flags == EVENT_FD_WRITE

Now a new client comes in, we go through the list of clients to find an idle
one. The detection for idle clients in remove_idle_client does not take the
pending data into account. We close the socket that has pending outgoing data,
the accept(2) one syscall later gives us the same socket.

In new_connection(), we do a setup_async_read, setting up a read fde. The
select from before however had found the socket (that we had already closed!!)
to be writable. In rw_callback we only want to see a readable flag, and we
panic in the SMB_ASSERT(flags == EVENT_FD_READ).

Found using

bin/smbtorture //127.0.0.1/tmp -U% -N 500 -o 2 local-wbclient

Volker

14 years agouse epoll for local-wbclient test
Volker Lendecke [Wed, 20 May 2009 12:18:23 +0000 (14:18 +0200)]
use epoll for local-wbclient test

14 years agoDon't limit the number of retries in wb_trans.
Volker Lendecke [Sun, 24 May 2009 11:22:26 +0000 (13:22 +0200)]
Don't limit the number of retries in wb_trans.

This is better done with a tevent_req_set_endtime the caller should issue.

14 years agoDon't set a timeout deep inside wb_connect
Volker Lendecke [Sun, 24 May 2009 11:18:34 +0000 (13:18 +0200)]
Don't set a timeout deep inside wb_connect

14 years agoChange async_connect to use connect instead of getsockopt to get the error
Volker Lendecke [Sun, 24 May 2009 11:14:12 +0000 (13:14 +0200)]
Change async_connect to use connect instead of getsockopt to get the error

On my Linux box, this is definitely the more reliable strategy with unix domain
sockets, and according to my tests it also works correctly with TCP sockets.

14 years agoDo queueing in wbclient.c
Volker Lendecke [Fri, 22 May 2009 20:30:09 +0000 (22:30 +0200)]
Do queueing in wbclient.c

The _trigger fn must know about wbc_context, while we were waiting in the
queue the fd might have changed

14 years agoFix closed_fd(): select returning 0 means no fd listening
Volker Lendecke [Fri, 22 May 2009 17:29:46 +0000 (19:29 +0200)]
Fix closed_fd(): select returning 0 means no fd listening

14 years agoFix wb_simple_trans queueing
Volker Lendecke [Sun, 10 May 2009 08:49:39 +0000 (10:49 +0200)]
Fix wb_simple_trans queueing

14 years agoAdd "err_on_readability" to writev_send
Volker Lendecke [Sat, 23 May 2009 14:10:54 +0000 (16:10 +0200)]
Add "err_on_readability" to writev_send

A socket where the other side has closed only becomes readable. To catch
errors early when sitting in a pure writev, we need to also test for
readability.

14 years agoAllow NULL queue to writev_send
Volker Lendecke [Sun, 10 May 2009 08:49:18 +0000 (10:49 +0200)]
Allow NULL queue to writev_send

14 years agoEnsure we return NT_STATUS_FILE_IS_A_DIRECTORY on a posix open on a
Jeremy Allison [Fri, 22 May 2009 22:55:27 +0000 (15:55 -0700)]
Ensure we return NT_STATUS_FILE_IS_A_DIRECTORY on a posix open on a
directory name.
Jeremy.

14 years agoTest that POSIX open of a directory returns NT_STATUS_FILE_IS_A_DIRECTORY (ERRDOS...
Jeremy Allison [Fri, 22 May 2009 22:21:55 +0000 (15:21 -0700)]
Test that POSIX open of a directory returns NT_STATUS_FILE_IS_A_DIRECTORY (ERRDOS, EISDIR).
Jeremy.

14 years agos3:smbd: implement SMB2 Tree Disconnect
Stefan Metzmacher [Fri, 22 May 2009 10:42:24 +0000 (12:42 +0200)]
s3:smbd: implement SMB2 Tree Disconnect

metze

14 years agos3:smbd: implement SMB2 Tree Connect
Stefan Metzmacher [Fri, 15 May 2009 09:50:20 +0000 (11:50 +0200)]
s3:smbd: implement SMB2 Tree Connect

For now this only checks if the share is present or not.

metze

14 years agos3:smbd: SMB2 session ids are 64bit...
Stefan Metzmacher [Fri, 22 May 2009 09:06:54 +0000 (11:06 +0200)]
s3:smbd: SMB2 session ids are 64bit...

We only grand ids up to 0x0000000000FFFFFF,
because that's what our idtree implementation can handle.
But also 16777215 sessions on one tcp connection should be enough:-)

metze

14 years agotsocket: allow empty vectors at the end for tstream_writev()/readv()
Stefan Metzmacher [Fri, 22 May 2009 10:28:17 +0000 (12:28 +0200)]
tsocket: allow empty vectors at the end for tstream_writev()/readv()

metze

14 years agos3:winbind:idmap_ldap: fix a crash bug in idmap_ldap_unixids_to_sids (#6387)
Michael Adam [Fri, 22 May 2009 09:58:00 +0000 (11:58 +0200)]
s3:winbind:idmap_ldap: fix a crash bug in idmap_ldap_unixids_to_sids (#6387)

This fixes a crash bug hit when multiple mappings were found by
the ldap search. This crash was caused by an ldap asssertion
in ldap_next_entry because was set to NULL in each iteration.

The corresponding fix was applied to the idmap_ldap_sids_to_unixids()
by Jerry in 2007 (b066668b74768d9ed547f16bf7b6ba6aea5df20a).

This fixes the crash part of bug #6387.

There is a logic part, too:
The problem currently only occurs when multiple mappings are found
for one given unixid. Now winbindd does not crash any more but
it does not correctly handle this situation. It just returns the
last mapping from the ldap search results.
This needs fixing.

Michael

14 years agos3:smbd: implement SMB2 Logoff
Stefan Metzmacher [Fri, 15 May 2009 09:40:19 +0000 (11:40 +0200)]
s3:smbd: implement SMB2 Logoff

metze

14 years agoDon't steal when we know the ptr will be null. Thanks to Simo for
Jeremy Allison [Fri, 22 May 2009 01:48:17 +0000 (18:48 -0700)]
Don't steal when we know the ptr will be null. Thanks to Simo for
pointing this out.
Jeremy.

14 years agoRevert the last two commits (fix for #6386). The actual problem
Jeremy Allison [Fri, 22 May 2009 01:37:36 +0000 (18:37 -0700)]
Revert the last two commits (fix for #6386). The actual problem
was a bug in ldb in 3.2 which could return a freed pointer on
ret != LDAP_SUCCESS. The main thing we must ensure is that we
never talloc_steal until we know LDAP_SUCCESS was returned.
Jeremy.

14 years agoEnsure all possible uses of indirection through res are checked after
Jeremy Allison [Fri, 22 May 2009 01:00:54 +0000 (18:00 -0700)]
Ensure all possible uses of indirection through res are checked after
an ldb_search.
Jeremy.

14 years agoAttempt to fix bug #6386 - Samba Panic triggered by Sophos Control Centre.
Jeremy Allison [Fri, 22 May 2009 00:27:25 +0000 (17:27 -0700)]
Attempt to fix bug #6386 - Samba Panic triggered by Sophos Control Centre.
Don't indirect a potentially null pointer.
Jeremy.

14 years agoDetect tight loop in tdb_find()
Jim McDonough [Thu, 21 May 2009 20:26:26 +0000 (16:26 -0400)]
Detect tight loop in tdb_find()

14 years agos3 torture: Fix warning
Tim Prouty [Thu, 21 May 2009 19:17:53 +0000 (12:17 -0700)]
s3 torture: Fix warning

14 years agos3 onefs: Fix invalid argument from the unix_convert smb_filename struct patch
Tim Prouty [Thu, 21 May 2009 19:17:33 +0000 (12:17 -0700)]
s3 onefs: Fix invalid argument from the unix_convert smb_filename struct patch

14 years agos3:smbd: we want to get the next command offset and not set it...
Stefan Metzmacher [Thu, 21 May 2009 14:17:53 +0000 (16:17 +0200)]
s3:smbd: we want to get the next command offset and not set it...

This should also fix the build on some hosts.

metze

14 years agos3-build: fix the build of ntlm_auth. Bo Yang, please check.
Günther Deschner [Thu, 21 May 2009 10:50:20 +0000 (12:50 +0200)]
s3-build: fix the build of ntlm_auth. Bo Yang, please check.

Guenther

14 years agos4-selftest: adding RPC-SAMR-USERS-PRIVILEGES to knownfail list.
Günther Deschner [Thu, 21 May 2009 10:15:07 +0000 (12:15 +0200)]
s4-selftest: adding RPC-SAMR-USERS-PRIVILEGES to knownfail list.

Samba4 cannot pass this test currently as in Samba4 (unlike Samba3)
the LSA and SAMR account are stored in the same db.
Once you delete a SAMR user the LSA privilege account is deleted
at the same time (which is wrong).

Guenther

14 years agos3: ignore EPIPE error when winbind finally writes to wb client because client might...
Bo Yang [Thu, 21 May 2009 19:22:52 +0000 (03:22 +0800)]
s3: ignore EPIPE error when winbind finally writes to wb client because client might have already closed the socket

Signed-off-by: Bo Yang <boyang@samba.org>
14 years agos3: Fix onlinestatus msg to return status of all domain instead of omitting trusted...
Bo Yang [Thu, 21 May 2009 18:12:59 +0000 (02:12 +0800)]
s3: Fix onlinestatus msg to return status of all domain instead of omitting trusted domains

Signed-off-by: Bo Yang <boyang@samba.org>
14 years agos3: set winbindd request flags in ntlm_auth to make it contact trusted domain when...
Bo Yang [Thu, 21 May 2009 18:03:32 +0000 (02:03 +0800)]
s3: set winbindd request flags in ntlm_auth to make it contact trusted domain when krb5 auth is enabled

Signed-off-by: Bo Yang <boyang@samba.org>
14 years agos3: Fix request flags in wbinfo when perform krb5 authentication
Bo Yang [Thu, 21 May 2009 17:39:03 +0000 (01:39 +0800)]
s3: Fix request flags in wbinfo when perform krb5 authentication

Signed-off-by: Bo Yang <boyang@samba.org>
14 years agoMake cli_posix_open() and cli_posix_mkdir() async.
Jeremy Allison [Thu, 21 May 2009 01:31:36 +0000 (18:31 -0700)]
Make cli_posix_open() and cli_posix_mkdir() async.
Jeremy.

14 years agos3: Change unix_convert (and its callers) to use struct smb_filename
Tim Prouty [Tue, 7 Apr 2009 20:39:57 +0000 (13:39 -0700)]
s3: Change unix_convert (and its callers) to use struct smb_filename

This is the first of a series of patches that change path based
operations to operate on a struct smb_filename instead of a char *.
This same concept already exists in source4.

My goals for this series of patches are to eventually:

1) Solve the stream vs. posix filename that contains a colon ambiguity
   that currently exists.
2) Make unix_convert the only function that parses the stream name.
3) Clean up the unix_convert API.
4) Change all path based vfs operation to take a struct smb_filename.
5) Make is_ntfs_stream_name() a constant operation that can simply
   check the state of struct smb_filename rather than re-parse the
   filename.
6) Eliminate the need for split_ntfs_stream_name() to exist.

My strategy is to start from the inside at unix_convert() and work my
way out through the vfs layer, call by call.  This first patch does
just that, by changing unix_convert and all of its callers to operate
on struct smb_filename.  Since this is such a large change, I plan on
pushing the patches in phases, where each phase keeps full
compatibility and passes make test.

The API of unix_convert has been simplified from:

NTSTATUS unix_convert(TALLOC_CTX *ctx,
      connection_struct *conn,
      const char *orig_path,
      bool allow_wcard_last_component,
      char **pp_conv_path,
      char **pp_saved_last_component,
      SMB_STRUCT_STAT *pst)
to:

NTSTATUS unix_convert(TALLOC_CTX *ctx,
      connection_struct *conn,
      const char *orig_path,
      struct smb_filename *smb_fname,
      uint32_t ucf_flags)

Currently the smb_filename struct looks like:

struct smb_filename {
       char *base_name;
       char *stream_name;
       char *original_lcomp;
       SMB_STRUCT_STAT st;
};

One key point here is the decision to break up the base_name and
stream_name.  I have introduced a helper function called
get_full_smb_filename() that takes an smb_filename struct and
allocates the full_name.  I changed the callers of unix_convert() to
subsequently call get_full_smb_filename() for the time being, but I
plan to eventually eliminate get_full_smb_filename().