lib: Fix memalign_array overflow protection

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Thu Nov 19 10:05:08 CET 2015 on sn-devel-104

idl: Some CC can't find indented #defines

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

idl: Avoid // style comments

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

libreplace: Only check for malloc.h if needed

OpenBSD complains that malloc.h is deprecated on every gcc invocation.
That hides a lot of real warnings. malloc and friends nowadays are
typically defined in stdlib.h, only check there.

We need memalign. On OpenBSD this does not exist, so libreplace replaces
it. The wscript test in libreplace only checks whether memalign can
link. Unfortunately in glibc memalign comes from malloc.h.

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

libreplace: Put the malloc.h check on a line of its own

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

libreplace: Only check malloc.h once

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

Set LD_LIBRARY_PATH during tests.

Without this, tests fail ir libtevent is not installed on the system.

Signed-Off-By: Jelmer Vernooij <jelmer@jelmer.uk>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Wed Nov 18 21:40:26 CET 2015 on sn-devel-104

lib: tevent: Whitespace cleanup.

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 Nov 18 15:54:03 CET 2015 on sn-devel-104

lib: tevent: Fix bug in poll backend - poll_event_loop_poll()

If the (pfd->revents & POLLNVAL) case is triggered,
we do DLIST_REMOVE(ev->fd_events, fde); and then
use fde->next in the loop above.

Save off fde->next for loop interation before
this so we can't use a deleted ->next value.

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

ctdb-scripts: Fix CTDB_DBDIR=tmpfs support

Various scripts (including debug_locks.sh, 00.ctdb, 05.system) need
CTDB_DBDIR to point to the right place... but it doesn't.

Move the rewriting of CTDB_DBDIR to loadconfig() so that it happens
for all scripts.  Have this code set internal variable
CTDB_DBDIR_TMPFS_OPTIONS so that ctdbd_wrapper can do the mount.

This loses the generality that was present in dbdir_tmpfs_start() but
it wasn't being used anyway.  If it is needed in the future then it
will be in the git history.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Wed Nov 18 11:51:54 CET 2015 on sn-devel-104

ctdb: Remove unused ctdb_set_process_name

Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Wed Nov 18 07:12:50 CET 2015 on sn-devel-104

ctdb: Use prctl_set_comment from lib/util

Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

lib/util: Move util_process from util to util-core

Make the common prctl_set_comment function available to both ctdb
builds.

Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>

libcli/smb: Use helper function for finding session

This removes some duplicated code.

Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Wed Nov 18 04:04:17 CET 2015 on sn-devel-104

smbd: Simplify check_access()

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Tue Nov 17 22:58:10 CET 2015 on sn-devel-104

smbd: Use check_access_fsp where appropriate

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

smbd: Factor out check_access_fsp() from check_access()

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

smbd: Refactor check_access()

This is a micro-patch to make the next ones just obvious

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

smbd: Early return from dos_mode_debug_print

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

smbd: Add hex value to dos_mode_debug_print

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

smbd: Use dos_mode_debug_print in get_ea_dos_attribute

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

smbd: Use dos_mode_debug_print in dos_mode_from_sbuf

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

smbd: Use dos_mode_debug_print in dos_mode_msdfs

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

smbd: Streamline dos_mode_debug_print

One line per flag is a bit overkill

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

s4-backupkey: Don't use deprecated data structures

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Tue Nov 17 19:52:33 CET 2015 on sn-devel-104

s4-torture: Rename issuer unique id in backupkey test

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

s3-vfs_snapper: Fix a possible use without init warning

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

s4-param: Make sure newname is not used uninitialized

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

smbd: Move a message_send_all to the cleanupd

message_send_all traverses serverid.tdb, which can be expensive

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Mon Nov 16 17:55:36 CET 2015 on sn-devel-104

smbd: Move cleanupd revalidate to a separate fn

Simple preparation for the next patch...

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

smbd: Move brl_validate to the cleanupd

This walks brlock.tdb, which can be time-consuming.

This adds a new includes.h include. It's too much of a pain for me now to
make locking/proto.h clean to include on its own.

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

smbd: Move serverid_deregister() to the cleanupd

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

smbd: Move messaging_cleanup() to the cleanupd

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

smbd: Move smbprofile_cleanup() to the cleanupd

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

smbprofile: Add dst pid to smbprofile_cleanup

The consolidation will soon be done by a separate process. We need to
avoid the getpid() call in smbprofile_cleanup().

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

smbd: Implement a cleanup daemon

We do way too much stuff in the parent smbd in remove_child_pid(). In
particular accessing ctdbd is not a good idea when ctdbd is stuck in something.
We've had a case where smbd exited itself with "ctdb timeout" being set to 60
seconds. ctdb was just stuck doing recoveries, and the parent smbd was sitting
in serverid_exists trying to retrieve a record for a child that had exited. Not
good.

This daemon sits there as parent->cleanupd and receives MSG_SMB_NOTIFY_CLEANUP
messages that hold the serverid and exit status of a former child. The next
commits will step by step empty remove_child_pid in the parent and move the
tasks to the helper.

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

build:wafsamba: Waf 1.8 compatible declaration of 'mandatory' configuration tests

The configuration tests raise exceptions by default in later Waf versions,
but the samba tests do not specify whether the errors should be raised or
not. This changes lifts the ambiguity.

Signed-off-by: Thomas Nagy <tnagy@waf.io>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Mon Nov 16 14:50:39 CET 2015 on sn-devel-104

build:wafsamba: Install named.conf only once

The wildcard *.conf already lists named.conf. Adding files
more than once will cause unnecessary rebuilds and raise
errors in later Waf versions.

Signed-off-by: Thomas Nagy <tnagy@waf.io>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

build:wafsamba: Replace Options.is_install by bld.is_install

Options.is_install is a deprecated module variable that is more
verbose than bld.is_install and complicates the Waf 1.8 upgrade.

Signed-off-by: Thomas Nagy <tnagy@waf.io>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

ctdb-recoverd: Factor out recovery master validation

Starting to untangle cluster management, database recovery and public
IP allocation.  This is a non-trivial subset of the cluster management
code that runs in the recovery daemon on all nodes.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Mon Nov 16 11:47:45 CET 2015 on sn-devel-104

ctdb-recmaster: Update capabilities before calling first election

Capabilities are used when computing an election result so having them
up-to-date seems like a good idea.

Also update several instances of an ambiguous comment.

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

ctdb-recoverd: Move VNN map retrieval to where it is needed

The VNN map is only needed on the recovery master, so no need for all
recovery daemons to retrieve it.

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

ctdb-recoverd: Drop explicit check for recovery lock

This is already handled in update_recovery_lock(), which is called
immediately before.

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

ctdb-recoverd: Simplify using TALLOC_FREE()

The only non-obvious part here is dropping the setting of the nodemap
local variable to NULL.  If the following control succeeds then it is
set, otherwise return and it doesn't matter.

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

ctdb-recoverd: Clarify that recmaster is being set on the current node

That is, using CTDB_CURRENT_NODE makes this more obvious.

Also fix incorrect error messages.

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

ctdb-recoverd: Do not sanity check recovery master with local daemon

Each recovery daemon knows who the recmaster is and is in sync with
its local daemon.  The recovery master is running this check so do not
bother checking with its local daemon - both agree that it is the
recovery master.

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

ctdb-recoverd: Don't retrieve recovery master from local daemon

The recovery daemon already knows which node is the master.  This
relies on rec->recmaster being correctly initialised and correctly set
during elections.

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

ctdb-recoverd: Explicitly set initial recovery master to unknown

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

ctdb-recoverd: Do not set recovery master during recovery

Recovery should not do cluster management functions.  Setting the
recovery master should only be done via an election.

Main loop will determine if recovery master is inconsistent across the
cluster and force an election if necessary.

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

ctdb-recoverd: Have recovery daemon remember election result

The recovery daemon pushes knowledge of recovery master election
progress/result to local daemon.  It then retrieves that information
again.

Instead, have the recovery daemon reliably track election
progress/result in rec->recmaster so it doesn't need to be retrieved.
Be careful to maintain consistency by only doing this when the local
daemon has been updated.

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

ctdb-recoverd: Clarify recovery master validation logic

There can be no holes in the nodemap.  Even if a node has been deleted
it will take a slot in the nodemap.  The only exception is that the
nodemap shrinks if nodes are deleted from the end.  That should never
include the master because a node should be shutdown before being
deleted, and an election should already have take place.

To avoid walking off the end of the nodemap nodes array just confirm
that the master node's PNN is a valid index into the array.  No need
to walk through the nodemap.

After this, in this section of the code j is now invalid.  So use the
master's PNN to index into the nodemap.  This is safe.

In the process, clean up some log messages to avoid saying "Force
reelection".  It's just an "election".

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

WHATSNEW: Document CTDB tunable change

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

ctdb-daemon: Rename EventScriptTimeoutCount to MonitorTimeoutCount

This only applies to monitor events so renaming clarifies this.

Note that this change is not backward compatible.  Users with

  CTDB_SET_EventScriptTimeoutCount=<n>

in their configuration will get failures when starting CTDB but the
cause will be clearly logged.

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

ctdb-daemon: Move script timeout count into monitor state

It is only used by the monitoring code.

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

ctdb-daemon: Reset script timeout count in monitor code

This is the only place it is used.

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

ctdb-daemon: Do not bother printing script timeout count

It is only updated for monitor events, so it is meaningless here.

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

ctdb-doc: Correct documentation for tunables for script timeout

* The defaults for EventScriptTimeout and EventScriptTimeoutCount are
  wrong.

* EventScriptTimeout is the total time for all enabled scripts that
  are run for an event, not a single event script.

* EventScriptTimeoutCount only applies to monitor events.

* EventScriptUnhealthyOnTimeout is obsolete, so remove it.

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

ctdb-ipalloc: Don't consider runstates in the IP takeover code

Checking runstates is unnecessary now that nodes that are not RUNNING
will return no available IP addresses.  I have no idea why I didn't do
it this way originally.

Tweak the test code to cope with this.

Note that this is a backward-incompatible change.  If new and old
versions of CTDB are running together in a cluster and a new node
takes over as recovery master then old nodes will be able to host
public IP addresses before they are in RUNNING runstate.  This is
mitigated by the bias towards recovery master stability in elections.
If it is important that nodes do not host IPs until they are RUNNING
then do not restart nodes running the old version.

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

ctdb-ipalloc: Check for available IPs, not runstate, in takeover run

The available IPs list is now only non-empty for nodes that are in
RUNNING runstate.  So, to avoid running the IP allocation algorithm
when there are no available available IPs, explicitly check for
available IPs rather than checking runstates.

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

ctdb-ipalloc: A VNN can only host IPs if node is in RUNNING runstate

This will allow wonderful simplification (i.e. removal) of some of the
runstate checking in the takeover run code.

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

ctdb-build: Remove ctdb-common-util subsystem

Include system_util.c in ctdb-system subsystem.

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Michael Adam <obnox@samba.org>
Autobuild-User(master): Martin Schwenke <martins@samba.org>
Autobuild-Date(master): Mon Nov 16 03:51:36 CET 2015 on sn-devel-104

ctdb-daemon: Remove unused ctdb_logging.[ch]

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

ctdb-tool: Use new debug level API

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

ctdb-common: Use new debug level API

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

ctdb-tests: Build in common/logging.c

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

ctdb-tests: Replace ctdb_logging.h with common/logging.h

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

ctdb-transport: Replace ctdb_logging.h with common/logging.h

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

ctdb-server: Replace ctdb_logging.h with common/logging.h

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

ctdb-common: Replace ctdb_logging.h with common/logging.h

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

ctdb-client: Replace ctdb_logging.h with common/logging.h

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

ctdb-system: Replace ctdb_logging.h with common/logging.h

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

ctdb-logging: Remove duplicate definitions

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

ctdb-tests: Update test to set numeric debug levels

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

ctdb-logging: Allow numeric specification of debug level

This makes the function compatible with parse_debug().

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

ctdb-logging: Allow sub-string matching for debug level strings

This allows for backward compatibility with the current code where
DEBUG_ERR corresponds to "ERR".

Fix the test that relies on "ERR" instead of "ERROR".

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

ctdb-logging: Fix for loop condition

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

ctdb-logging: Add APIs to convert debug_level to and from integer

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

ctdb-logging: Move debug_extra definition to server/ctdb_logging.c

This variable is used for adding a prefix to log entries from various
child processes.

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

ctdb-daemon: Move ctdb_fork.c to server

These functions are only used in the ctdb daemon code.

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

ctdb-daemon: Move switch_from_server_to_client() to ctdb_daemon.c

This function can only called from ctdb daemon.

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

WHATSNEW: Add async SMB2 flush and new aio parameter

Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Sat Nov 14 00:39:17 CET 2015 on sn-devel-104

docs: Update doc for 'strict sync' parameter for async SMB2 flush

Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

s3: smbd: Change aio_pending_size static variable to a new "aio max threads" smb.conf parameter.

Removes accessor functions as now this parameter is set
under user control in smb.conf. Default is 100.

Note that this doesn't limit the number of outstanding
aio requests, it just causes them to go onto the
pthreadpool queue.

Now we need to prioritize pthreadpool pipe replies
ahead of incoming SMB2 requests, but that's a patch
for another day.

Based on ideas from Volker.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

s3: smbd: Remove checks causing fallback to sync on pread/pwrite/fsync.

Rely on pthreadpool queueing instead of falling back.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

s3: smbd: Remove --with-aio-support. We no longer would ever prefer POSIX-RT aio, use pthread_aio instead.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>

docs: correct the name of the idmap config group of parameters

"idmap config DOMAIN : OPTION" is the pattern used.

Signed-off-by: Michael Adam <obnox@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Fri Nov 13 19:55:33 CET 2015 on sn-devel-104

ctdb: Fix the O3 developer build

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Fri Nov 13 12:42:22 CET 2015 on sn-devel-104

selftest: Use strict sync = yes

This enables the codepaths calling fsync for FLUSH requests during
selftest.

Signed-off-by: Christof Schmitt <cs@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Christof Schmitt <cs@samba.org>
Autobuild-Date(master): Fri Nov 13 04:47:00 CET 2015 on sn-devel-104

smbd: Issue fsync for SMB2 FLUSH asynchronously

SMB2 FLUSH mainly calls fsync and there is already code in place to
handle fsync asynchronously, so use the asynchronous code path for SMB2
FLUSH. This avoids a SMB2 FLUSH stalling other requests processing.

Signed-off-by: Christof Schmitt <cs@samba.org>
Signed-off-by: Jeremy Allison <jra@samba.org>

s3: smbd: Remove outstanding_aio_calls from globals.

Access via functions only.

Signed-off-by: Jeremy Allison <jra@samba.org>
Signed-off-by: Christof Schmitt <cs@samab.org>

s3: smbd: Remove aio_pending_size from globals.

Access via functions only.

Signed-off-by: Jeremy Allison <jra@samba.org>
Signed-off-by: Christof Schmitt <cs@samba.org>

s4-auth: Fix some debugging and crash in error cases

Guenther

Signed-off-by: Günther Deschner <gd@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Fri Nov 13 01:43:36 CET 2015 on sn-devel-104

smbd: Remove dead code

94f0716fff has removed the dcelogin_atmost_once variable.

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

ctdb-recoverd: Reload remote IPs as part of takeover run

This is currently done before each IP takeover run, so just factor it
in.

ctdb_reload_remote_public_ips() becomes static.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Thu Nov 12 09:28:45 CET 2015 on sn-devel-104

ctdb-recoverd: Move ctdb_reload_remote_public_ips() to ctdb_takeover.c

This will help to untangle known and available public IP lists from
the CTDB context.

verify_remote_ip_allocation() needs a forward declaration.

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

ctdb-recoverd: Remote IP validation can't cause a takeover run

Remote IP validation is only called when a takeover run is about to
happen anyway, so don't bother flagging one.  Given that a takeover
run isn't being triggered, also drop the test that checks if takeover
runs are disabled.  These are the only uses of the rec argument, so
drop it.

One possible further simplification would be to remove this function
because it doesn't accomplish anything.  However, it is worth leaving
it as a reminder that remote IP validation should be done properly at
some time in the future.

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

ctdb-recoverd: Drop culprit argument from ctdb_reload_remote_public_ips()

It is only used by the caller to print a message that includes the
culprit.  However, ctdb_reload_remote_public_ips() already prints
perfectly good messages and they include the culprit.

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

ctdb-recoverd: Trigger takeover run after rebalance timeout

No need to do it immediately.  It will happen in less than a second.

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

ctdb-recoverd: Remove unnecessary assignments of need_takeover_run

do_takeover_run() unsets this if it succeeds.

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

ctdb-recoverd: Do not run recovery-related events around IP takeover

This is not a recovery, so do not run "startrecovery and "recovered"
events.  There are other IP takeover runs where these are not run.

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