s3: krb5: keytab - The done label can be jumped to with context == NULL.

Ensure we don't crash in this case.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=11959

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Uri Simchoni <uri@samba.org>
Autobuild-User(master): Uri Simchoni <uri@samba.org>
Autobuild-Date(master): Thu Jun  9 13:18:56 CEST 2016 on sn-devel-144

lib: Fix uninitialized read in msghdr_copy

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Wed Jun  8 18:34:27 CEST 2016 on sn-devel-144

ctdb-daemon: Reset push_started flag once DB_PUSH_CONFIRM is done

Once DB_PUSH_START is processed as part of recovery, push_started
flag tracks if there are multiple attempts to send DB_PUSH_START.
In DB_PUSH_CONFIRM, once the record count is confirmed, all information
related to DB_PUSH should be reset.  However, The push_started flag was
not reset when the push_state was reset.

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Wed Jun  8 14:31:52 CEST 2016 on sn-devel-144

ctdb-recoverd: Avoid duplicate recoverd event in parallel recovery

BUG: https://bugzilla.samba.org/show_bug.cgi?id=11956

In do_recovery, after the recovery and takeover is complete, recoverd
event is triggered.  When the parallel database recovery was separated,
ctdb_recovery_helper implemented sending END_RECOVERY control which
causes recoverd event to be triggered.  So when there is parallel database
recovery, recoverd event is triggered twice.

Instead move the call to run_recovered_eventscript() explicitly in
the serial recovery code path.  This avoids the duplication trigger of
recoverd event.

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>

ctdb-system: Remove duplicate functions

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>

ctdb-daemon: Use lib/util functions instead of redefinitions

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>

lib/util: Add a generic definition for set_close_on_exec

Avoid changing function names to smb_set_close_on_exec in ctdb.

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>

ctdb-tests: Re-use set_blocking instead of re-definition

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>

ctdb-tests: Re-use async accept wrapper from async_req

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>

ctdb-scripts: Add eventscript 06.nfs

This generates takeip-pre and releaseip-pre call-out events.

One use is to put NFS into grace before an IP is assigned to an
interface.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-scripts: Move NFS callout-related code to functions file

A second NFS eventscript may be required, so make this code available
to it.

The initialisation code can't be evaluated in the functions file
because service_state_dir isn't yet setup, so put it in a function and
call it with other initialisation code.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

build: Address may be used uninitialized in this function on Ubuntu 10.04

This is not found by modern compilers, but prevents the -Werror -O3 build on Ubuntu 10.04

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Wed Jun  8 08:48:57 CEST 2016 on sn-devel-144

build: Try to work around strict aliasing rules on Ubuntu 10.04

We get cc1: warnings being treated as errors
../lib/util/util_net.c: In function get_socket_port:
../lib/util/util_net.c:921: error: dereferencing pointer sa.106 does break strict-aliasing rules
../lib/util/util_net.c:921: note: initialized from here
../lib/util/util_net.c:925: error: dereferencing pointer sa.107 does break strict-aliasing rules
../lib/util/util_net.c:925: note: initialized from here

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

ctdb-cluster-mutex: Fix #endif decoration

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
Autobuild-User(master): Martin Schwenke <martins@samba.org>
Autobuild-Date(master): Wed Jun  8 04:52:18 CEST 2016 on sn-devel-144

ctdb-recoverd: Release recovery lock on exit

The recovery lock helper must exit when it notices its parent is gone.
However, that can take a few seconds.

The usual way of terminating the recovery daemon is for the main ctdbd
to send it a SIGTERM.  Installing a handler is nice and simple.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recoverd: Add handler for lost recovery lock

If the process holding the recovery lock terminates unexpectedly then
the recovery daemon needs to know that the lock is no longer held.

While here, rename hold_reclock_handler() to take_reclock_handler() so
there is a clear difference between the two handler names.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-cluster-mutex: Register an extra handler for when mutex is lost

Pass NULL if not needed.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-cluster-mutex: ctdb_cluster_mutex() registers handler and private data

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-cluster-mutex: Drop cluster_mutex_handler() ctdb and handle arguments

This makes the API more general.  If they are needed in a handler then
they can be in the private data.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recovery: Wrap private data for reclock test callback

This will allow a simplification of the cluster mutex API, so the
private data can be registered when calling ctdb_cluster_mutex().

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recoverd: Simplify reclock handler

Do the interesting work outside the handler.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recoverd: Recovery lock handle should be in recovery deamon context

This shouldn't be in the CTDB context.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-cluster-mutex: Pass a talloc context to allocate the handle off

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recoverd: No need to reset reclock handler

It won't be called more than once by the cluster mutex code.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-cluster-mutex: Don't call the supplied hander more than once

After the first activity on the file descriptor, ignore any subsequent
activity.  Single-shot handlers are easier to write.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recoverd: Fix buggy function return on memory allocation failure

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recoverd: Don't expose internal cluster mutex status

Just expose whether the lock was taken.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-daemon: Rename recovery lock file to just recovery lock

It isn't necessarily a file.

Don't bother changing the control, since it doesn't pervade the code.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-daemon: Drop function ctdb_set_recovery_lock_file()

Setting the recovery lock file at startup can be done more simply.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-protocol: CTDB_CONTROL_SET_RECLOCK_FILE is obsolete

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-protocol: Drop support for SET_RECLOCK

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-client: Remove support for SET_RECLOCK

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recovery: Don't update recovery lock from daemon

It can't change after startup.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recovery: Don't sync recovery lock across cluster

Support for updating the recovery lock is being removed because it
isn't possible to recover from failure.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-recovery: Consistency check reclock in start recovery control

If the recovery lock setting is not consistent with that of the
recovery master then abort.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-tool: Drop support for "ctdb setreclock" command

The recovery lock can not be reliably updated at run-time.  If it
fails to update on some nodes then split-brain protection is gone and
there is no reasonable way to repair the situation.  CTDB will have to
be restarted on all nodes.  So, if this feature is being used to avoid
scheduling an outage then an outage will have to be scheduled just in
case!

To update the recovery lock, shut down CTDB on all nodes, reconfigure
the recovery lock and start CTDB again.

Those that *really* want to be able to change the recovery lock at
run-time can still do so.  Set CTDB_RECOVERY_LOCK to point to a script
and this script can then be modified at run-time.  However, please
don't report bugs if bad things happen...

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

ctdb-tests: Replace "ctdb setrelock" test with "ctdb getreclock" test

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

libsmb: Fix two CIDs for NULL dereference

This whole area is a known-to-be-broken mess, but this patch should fix
the immediate crash

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Tue Jun  7 18:31:30 CEST 2016 on sn-devel-144

rpc_server: Fix CID 1362565 Improper use of negative value

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

lib: Fix CID 1362566 Dereference null return value

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

lib: Add accept_send/recv

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

lib: Move poll_funcs to lib/

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

lib/util: Expose few more subsystems for standalone ctdb build

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Volker Lendecke <vl@samba.org>

lib: Move msghdr to lib/util/

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

lib/poll_funcs: Build as SAMBA_SUBSYSTEM

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Volker Lendecke <vl@samba.org>

repl: Avoid use-after-free when working with the working_schema

The original schema must live as long as the working_schema
as the working_schema starts as a shallow-copy of schema.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
BUG: https://bugzilla.samba.org/show_bug.cgi?id=11953

Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Tue Jun  7 14:33:39 CEST 2016 on sn-devel-144

selftest: Add a reverse variation to ReplicateMoveObject3

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Samba <garming@catalyst.net.nz>

selftest: Assert replPropertyMetaData values before and after replication

This covers renames, addition of attributes, and the delete.

We also confirm the results via DRS.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Fix rename and RDN handling for replPropertyMetaData

This matches Windows 2012R2, which both has the RDN not sorted last and has it updated with the local
invocation_id and a local version.

The RDN attribute, unlike name, is not replicated over DRS, so the impact for interopability extends only to
the incorrect RDN values that we were finding with dbcheck (values that did not match the name values).

Finally, we always force the RDN to match the name attribute, which avoids issues
in dbcheck where these diverge.  As such, we can finally remove dbcheck as a
flapping test, last re-added in e4bab3a8282d263eb2391bc7e8a6fd64ae068935

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Fix incorrect sorting of replPropertyMetaData with RDN last

Per tests against Windows 2012R2 the RDN is not sorted last and is
instead sorted normally with all the other elements.

The RDN attribute, unlike name, is not replicated over DRS, so this
has no interopability impact.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

dsdb: Show initial replicated modify as well as resolved modify in repl_meta_data

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

selftest: Add more tests to cover attribute changes vs DN renames

This covers a bug where unrelated attribute changes would reverse a rename

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

dsdb: Use replmd_replPropertyMetaData1_new_should_be_taken in replmd_replicated_apply_search_callback()

This is the primary handler for renames

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

dsdb: Use replmd_replPropertyMetaData1_new_should_be_taken in replmd_replicated_apply_merge()

This is the main handler for attribute conflicts

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

dsdb: Use replmd_replPropertyMetaData1_new_should_be_taken in replmd_replicated_handle_rename()

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Use replmd_replPropertyMetaData1_new_should_be_taken in replmd_op_possible_conflict_callback()

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Add new helper function replmd_replPropertyMetaData1_new_should_be_taken()

This will allow the test for "name" and the actual DN to be consistent,
and so avoids dbcheck errors when CN and name do not match the DN

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

selftest: Do not scan the full DB to confirm a specific DN in dbcheck

This avoids a full DB scan and therefore reduces the test time taken
when we just modified the cn=administrator record.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

selftest: Run the krb5.kdc test on a more selective basis

The previous tests would take 20mins, the new set of tests take around 7 mins and still cover
the important combinations, given that it is the same KDC code in each environment

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

selftest: These replication tests are now OK after we fixed all the replication bugs

We have run this test a number of times, and these tests do not flap
in those tests

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Tue Jun  7 01:45:08 CEST 2016 on sn-devel-144

selftest: Add tests to show that we can not create duplicate schema entries

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Mon Jun  6 20:32:54 CEST 2016 on sn-devel-144

samldb: Make checks for schema attributes much more strict

This avoids corrupting Samba when invalid schema is imported

Signed-off-by: Andrew Bartlett <abartlet@samba.org>

Remove the try/catch from urgent_replication.py

This meant that for ages, the duplicate OID was unnoticed, and when the syntax
was corrected recently, this caused the test to run, and so cause trouble

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

selftest: Assert that name, the RDN attribute and actual RDN are in sync

This allows us to catch such errors here, rather than just on dbcheck later

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

selftest: Add another test case to replica_sync test

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

repl: Do not report all replication failures at level 0

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

selftest/drs: Show we return the correct 3 objects for DRSUAPI_EXOP_FSMO_RID_ALLOC

This does not depend on DRSUAPI_DRS_GET_ANC.

This test is not new, but it was not previously being run.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

selftest: initial version of new repl_move test

This tests complex rename and modify combinations in a way that
demonstrated a number of replication failures, due to incorrect
handling in Samba when the parent of the record changes.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

dsdb: Give the objectGUID ahead of LDIF dump of replicated changes

This can help isolate which object this is when the object is involved
in a rename.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Simplify replmd_op_possible_conflict_callback behaviour

The previous behaviour of this code was to overwrite the req->callback of the original, failed request.

This is a problem for many reasons - including that ldb_module_done() may already have been
called on that pointer.

The correct pattern is to create a new request, and to call ldb_module_done() on the parent
request (the one in ar->req) not in this one, in the error case.

We use the passed in new callback either as the callback to call on success, or
as the callback to the ADD request.  We overwrite it with replmd_op_name_modify_callback
in the rename remote case, as before, but no longer modify req->callback as
this will not be used again.

This is less tricky and a little simpler to follow, as we also remove the
now unused handling for RENAME, which is in a separate routine now

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Split rename case out of replmd_op_possible_conflict_callback

This avoids running this code path, originally written for the add case, in
a semi-async manner in the rename case, which caused both bugs and complexity.

This does create a deal of duplicated code, but it is easier to follow because
there are no longer special cases for ADD and RENAME in the "common" code and
the behaviour of ldb_module_done() and the callbacks is well defined and expected

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

repl_meta_data: Give more information on replication rename behaviour

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

repl_meta_data: Fail to replicate over local objects not NC_HEAD with a all-zero parentGUID

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Give more errors in operational module when building the parentGUID

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Clearly fail to replicate objects not NC_HEAD with a all-zero parentGUID

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

repl: Enforce that we have parent objects for all replicated objects

The creating of replicated objects without their parent object allows database corruption as they can end up under
the wrong object.  We need to re-try the replication with the DRSUAPI_DRS_GET_ANC flag
set to get the objects in tree order.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Move operational below repl_meta_data so we can query parentGUID

This avoids re-adding the same code in repl_meta_data or making a shared subroutine

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

Add yet another error code when forking an smbd and ctdb is not there. We can see NT_STATUS_CONNECTION_REFUSED in the logs upon such a rare case.

Signed-off-by: Peter Somogyi <psomogyi@hu.ibm.com>
Reviewed-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Martin Schwenke <martin@meltin.net>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Mon Jun  6 16:10:19 CEST 2016 on sn-devel-144

ctdb-packaging: Remove tevent-unix-util public library

BUG: https://bugzilla.samba.org/show_bug.cgi?id=11946

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Volker Lendecke <vl@samba.org>

lib/util: Avoid splitting tevent-unix-util as public library

BUG: https://bugzilla.samba.org/show_bug.cgi?id=11946

Commit 670db6ac1d678babd25dd82c4467c0f094cfabc5 split tevent-util public
library to create tevent-unix-util public library for standalone ctdb
use.  This created a public library dependency between samba and ctdb
for packaging.

Bundle tevent_unix.c in public library tevent-util as before.  However,
to avoid the dependencies for packaging, standalone ctdb build will
build tevent-util as a private library with only tevent_unix.c

This simplifies any new subsystems (or libraries) which need tevent-util
and are linked in both samba and ctdb.

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Volker Lendecke <vl@samba.org>

dsdb: Cache our local invocation_id at the start of each request

This avoids fetching it over and over again

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Mon Jun  6 12:25:14 CEST 2016 on sn-devel-144

repl: Retry replication of the schema on WERR_DS_DRA_SCHEMA_MISMATCH

This makes us replicate the schema, and then go back to what we asked to replicate
originally, when the schema changes.  This should make the replication much more
robust after schema changes

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dbcheck: Find and fix a missing Deleted Objects container

Older Samba versions could delete this.  This patch tries very hard
to put back the original object, with the original GUID, so that
if another replica has the correct container, that we just merge
rather than conflict.

The existing "wrong dn" check can then put any deleted objects
under this container correctly.

Pair-programmed-with: Garming Sam <garming@catalyst.net.nz>
Pair-programmed-with: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

selftest: Only mark runtime dbcheck as flapping

The dbcheck-oldrelease.sh based tests do not flap

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

repl: Remove duplicated delete of sAMAccountType

This causes the version number in replPropertyMetaData to be
incorrectly bumped twice, because it is implied by not being in the
list of attributes to be preserved.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Allow "cn" to be missing on schema import

This avoids a segfault when we remove the duplication of this value from dsdb_convert_object_ex()

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

selftest: Make replica_sync test pass against Windows 2012R2

The tests that invoke --local will not pass, naturally, but otherwise it works

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

samba-tool domain join: Set drsuapi.DRSUAPI_DRS_GET_ANC during initial repl

This is needed so that we get parents before children.

We need this to ensure that we always know the correct parent for a
new child object, rather than just trusting the DN string

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Only search the provided partition for the object GUID

We know which NC this is in, so save searching all the backends.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

repl: Pass in the full partition DN to dsdb_replicated_objects_convert()

When we were processing an EXOP, we would pass in a DN specific to that
operation, but this stopped repl_meta_data from finding the parent object

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

repl: Allow GetNCChanges DRSUAPI_EXOP_REPL_OBJ to succeed against a deleted object

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>

dsdb: Remove incorrect RDN attid check in replmd_replPropertyMetaDataCtr1_verify

On windows, the RDN is not sorted last.  Additionally, the check should have been
against dsdb_attribute_get_attid(), not just ->attributeID_id

By including this in earlier versions of Samba, we allow a backport of Samba
databases from Samba 4.5, where this will be sorted in the same way
as Windows 2012R2.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz
BUG: https://bugzilla.samba.org/show_bug.cgi?id=11842

selftest: Use random OIDs from under the Samba OID arc

The urgent_replication.py test used the OID of uid, and this caused
other tests to fail

The other random OIDs should have been from under our arc, not under
iso.member-body.us

We split up the range a little to avoid some of the birthday paradox,
in the tests that create multiple OIDs.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz

ctdb-recovery: Update timeout and number of retries during recovery

The timeout RecoverTimeout (default 120) is used for control messages
sent during the recovery.  If any of the nodes does not respond to any
of the recovery control messages for RecoverTimeout seconds, then it
will cause a failure of recovery of a database.  Recovery helper will
retry the recovery for a database 5 times.

In the worst case, if a database could not be recovered within 5 attempts,
a total of 600 seconds would have passed.  During this time period other
timeouts will be triggered causing unnecessary failures as follows:

1. During the recovery, even though recoverd is processing events,
   it does not send a ping message to ctdb daemon.  If a ping message is
   not received for RecdPingTimeout (default 60) seconds, then ctdb will
   count it as unresponsive recovery daemon.  If the recovery daemon
   fails for RecdFailCount (default 10) times, then ctdb daemon will
   restart recovery daemon.  So after 600 seconds, ctdb daemon will
   restart recovery daemon.

2. If ctdb daemon stays in recovery for RecoveryDropAllIPs (default 120),
   then it will drop all the public addresses.  This will cause all
   SMB client to be disconnected unnecessarily.  The released public
   addresses will not be taken over till the recovery is complete.

To avoid dropping of IPs and restarting recovery daemon during a delayed
recovery, adjust RecoverTimeout to 30 seconds and limit number of
retries for recovering a database to 3.  If we don't hear from a node
for more than 25 seconds, then the node is considered disconnected.
So 30 seconds is sufficient timeout for controls during recovery.

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
Autobuild-User(master): Martin Schwenke <martins@samba.org>
Autobuild-Date(master): Mon Jun  6 08:49:15 CEST 2016 on sn-devel-144

samba_spnupdate: do not interpret failure count as unix error code

Signed-off-by: Bob Campbell <bobcampbell@catalyst.net.nz>
Pair-programmed-with: Garming Sam <garming@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Autobuild-User(master): Douglas Bagnall <dbagnall@samba.org>
Autobuild-Date(master): Fri Jun  3 11:02:26 CEST 2016 on sn-devel-144

samba_dnsupdate: do not interpret failure count as unix error code

Signed-off-by: Bob Campbell <bobcampbell@catalyst.net.nz>
Pair-programmed-with: Garming Sam <garming@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>

Efficient xattr handling for VxFS Signed-off-by: Abhidnya Joshi <Abhidnya.Joshi@veritas.com>

Reviewed-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Fri Jun  3 07:26:34 CEST 2016 on sn-devel-144

lib: replace: snprintf - Fix length calculation for hex/octal 64-bit values.

Prevents truncation due to buffer size being too small.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=11947

Signed-off-by: Lorinczy Zsigmond <lzsiga@freemail.c3.hu>
Reviewed-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>
Autobuild-User(master): Michael Adam <obnox@samba.org>
Autobuild-Date(master): Fri Jun  3 03:48:58 CEST 2016 on sn-devel-144

Improve help wording for samba-tool domain provision as per https://lists.samba.org/archive/samba-technical/2016-April/113740.html

BUG: https://bugzilla.samba.org/show_bug.cgi?id=11894
Signed-off-by: Peter C. Kelly <peter+sambabugzilla@pck.co.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Thu Jun  2 20:07:22 CEST 2016 on sn-devel-144

build: Build less of Samba when building --without-ntvfs-fileserver

We would build, but not use, many components of the NTVFS file server
even when we asked not to.  They would then consume disk, but not be
of any use

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>