lib: Try nonblocking writes in writev_send()

All callers now use nonblocking sockets, so that we can optimize by
doing early writes

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Stefan Metzmacher <metze@samba.org>
Autobuild-Date(master): Wed Mar 25 10:41:29 UTC 2020 on sn-devel-184

lib: Use optimize_empty in writev_send()

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>

winbind: Set the parent->child sock nonblocking

The parent goes through wb_simple_trans_send(), which deals with short
writes fine.

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>

libsmb: Make sure that the TCP socket is non-blocking

All traffic goes through smbXcli_base.c, and that is prepared to deal
with short writes via the conn->outgoing queue. Instead of making sure
that all callers properly set the socket nonblocking, do it here, so
that we can later optimize sending out data to the server.

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>

lib: Remove unused SOCKET_FLAG_BLOCK

Nobody in the code set this flag, so remove it

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>

lib: Simplify writev_cancel()

We can only reasonably cancel a writev request that is still
queued. Once writing has started, cancel is pointless. Simplify the
if-conditions.

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>

third_party: Update resolv_wrapper to version 1.1.6

Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Stefan Metzmacher <metze@samba.org>
Autobuild-Date(master): Wed Mar 25 08:33:17 UTC 2020 on sn-devel-184

third_party: Update socket_wrapper to version 1.2.4

Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

third_party: Update pam_wrapper to version 1.1.2

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>

vfs: remove SMB_VFS_OPENDIR()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Tue Mar 24 21:23:43 UTC 2020 on sn-devel-184

smbd: use open_internal_dirfsp_at() and OpenDir_fsp() in OpenDir()

This allows consolidating symlink safe processing of pathnames to the low-level
function non_widelink_open() used in the file open codepath via fd_open() and
the new utility function open_internal_dirfsp_at().

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: add open_internal_dirfsp_at()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: reformat OpenDir() function definition

Gives a nicer diff in the next commit. No change in behaviour.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: update smb_Dir_destructor() to cope with fsp->dptr not being set

Currently the only caller of OpenDir_fsp() is dptr_create() which means
fsp->dptr will always be set by dptr_create().

A subsequent commit will add another caller so that fsp->dptr will end up being
NULL.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: remove fdopendir() fallback

In order to get rid of SMB_VFS_OPENDIR, we have to require
fdopendir(). Everybody but macOS seems to have it, so RIP opendir() fallback.

This also prepares for a subsequent commit where we're going to start calling
OpenDir_fsp() from inside OpenDir(). Without removing the fallback this would
result in a recursion.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: simplify non_widelink_open()

As fsp->is_directory is already correctly populated by our callers, we can drop
the complicated and possibly broken logic that relies on O_DIRECTORY being
defined.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_shadow_copy2: use create_internal_dirfsp_at() and SMB_VFS_FDOPENDIR()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: add create_internal_dirfsp_at()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_shadow_copy: use OpenDir() in shadow_copy_get_shadow_copy_data()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_streams_depot: use OpenDir() in walk_streams()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_fruit: use OpenDir() in fruit_disk_free()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_fruit: pass e->d_name directly to fruit_tmsize_do_dirent()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_fruit: let fruit_get_num_bands() take bundle as const

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_fruit: use OpenDir() instead of SMB_VFS_OPENDIR() in fruit_get_num_bands()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_ceph_snapshots: use OpenDir() in ceph_snap_gmt_convert_dir()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_ceph_snapshots: use OpenDir() in ceph_snap_enum_snapdir()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfstest: use OpenDir()

Note that as ReadDirName() returns translated names (in Windows "encoding"), in
cmd_translate_name() test we have to translate back to UNIX "encoding" to check
if the filename matches the user requested path which is also in UNIX
"encoding".

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: use canonicalize_connect_path() in create_conn_struct_as_root()

This fix ensures create_conn_struct_as_root() works the same way as
make_connection_snum() used in smbd for SMB sessions. Without this any caller of
create_conn_struct_as_root() will fail to use work on shares if the sharepath
contains a symlink.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: make canonicalize_connect_path() public

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: use OpenDir() in form_junctions()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

smbd: use OpenDir() in count_dfs_links()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_fruit: move fruit_unlink_internal() back into fruit_unlinkat()

I think this makes fruit_unlinkat() easier to read. No change in behaviour.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

vfs_fruit: remove internal directory cleanup logic

Currently fruit does a cleanup run on a directory when it is being removed: it
lists the directory and does an explicit unlink on any found file that is a
genuine AppleDouble file.

This is not realy needed: the ._ AppleDouble sidecar files are either vetoed by
default, so cleanup can be turned on by the built-in "delete veto files"
options. Or, if the user sets "fruit:veto_appledouble", the client will see the
file in a directory list as ordinary files, so the client is responsible for
deleting them first before removing the parent directory.

tl;dr

One caller of SMB_VFS_OPENDIR() less. :)

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>

selftest: Fix string compare in DnsHandler() of dns_hub.py

dns_hub.py:115: SyntaxWarning: "is" with a literal. Did you mean "=="?
  if forwarder is 'ignore':
dns_hub.py:117: SyntaxWarning: "is" with a literal. Did you mean "=="?
  elif forwarder is 'fail':

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Tue Mar 24 14:31:25 UTC 2020 on sn-devel-184

ctdb-recoverd: Avoid dereferencing NULL rec->nodemap

Inside the nested event loop in ctdb_ctrl_getnodemap(), various
asynchronous handlers may dereference rec->nodemap, which will be
NULL.

One example is lost_reclock_handler(), which causes rec->nodemap to be
unconditionally dereferenced in list_of_nodes() via this call chain:

  list_of_nodes()
  list_of_active_nodes()
  set_recovery_mode()
  force_election()
  lost_reclock_handler()

Instead of attempting to trace all of the cases, just avoid leaving
rec->nodemap set to NULL.  Attempting to use an old value is generally
harmless, especially since it will be the same as the new value in
most cases.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14324

Reported-by: Volker Lendecke <vl@samba.org>
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Martin Schwenke <martins@samba.org>
Autobuild-Date(master): Tue Mar 24 01:22:45 UTC 2020 on sn-devel-184

ctdb-daemon: Don't allow attach from recovery if recovery is not active

Neither the recovery daemon nor the recovery helper should attach
databases outside of the recovery process.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-daemon: Remove more unused old client database functions

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Remove old code for creating missing databases

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Create database on nodes where it is missing

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Fetch database name from all nodes where it is attached

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Pass db structure for each database recovery

Instead of db_id and db_flags.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: GET_DBMAP from all nodes

This builds a complete list of databases across the cluster so it can
be used to create databases on the nodes where they are missing.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Replace use of ctdb_dbid_map with local db_list

This will be used to build a merged list of databases from all nodes,
allowing the recovery helper to create missing databases.

It would be possible to also include the db_name field in this
structure but that would cause a lot of churn.  This field is used
locally in the recovery of each database so can continue to live in
the relevant state structure(s).

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-daemon: Respect CTDB_CTRL_FLAG_ATTACH_RECOVERY when attaching databases

This is currently only set by the recovery daemon when it attaches
missing databases, so there is no obvious behaviour change.  However,
attaching missing databases can now be moved to the recovery helper as
long as it sets this flag.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Use CTDB_CTRL_FLAG_ATTACH_RECOVERY to attach during recovery

ctdb_ctrl_createdb() is only called by the recovery daemon, so this is
a safe, temporary change.  This is temporary because
ctdb_ctrl_createdb(), create_missing_remote_databases() and
create_missing_local_databases() will all go away soon.

Note that this doesn't cause a change in behaviour.  The main daemon
will still only defer attaches from non-recoverd processes during
recovery.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-protocol: Add control flag CTDB_CTRL_FLAG_ATTACH_RECOVERY

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-daemon: Remove unused old client database functions

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-daemon: Fix database attach deferral logic

Commit 3cc230b5eeca749ab68d19cfda969f72c269f1f6 says:

  Dont allow clients to connect to databases untile we are well past
  and through the initial recovery phase

It is unclear what this commit was attempting to do.  The commit
message implies that more attaches should be deferred but the code
change adds a conjunction that causes less attaches to be deferred.
In particular, no attaches will be deferred after startup is complete.
This seems wrong.

To implement what seems to be stated in the commit message an "or"
needs to be used so that non-recovery daemon attaches are deferred
either when in recovery or before startup is complete.  Making this
change highlights that attaches need to be allowed during the
"startup" event because this is when smbd is started.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Refactor banning a node into separate computation

If a node is marked for banning, confirm that it's not become inactive
during the recovery.  If yes, then don't ban the node.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Don't trust nodemap obtained from local node

It's possible to have a node stopped, but recovery master not yet
updated flags on the local ctdb daemon when recovery is started.  So do
not trust the list of active nodes obtained from the local node.  Query
the connected nodes to calculate the list of active nodes.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Consolidate node state

This avoids passing multiple arguments to async computation.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

ctdb-recovery: Fetched vnnmap is never used, so don't fetch it

New vnnmap is constructed using the information from all the connected
nodes.  So there is no need to fetch the vnnmap from recovery master.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14294

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

s3: libsmbclient.h: add missing time.h include

A recent change added a struct with timespec members in the public
libsmbclient header. This type is defined in time.h which was not
included thus making users of libsmbclient not build properly.

    /.../libsmbclient.h:158:18: error: field 'btime_ts' has incomplete type
      struct timespec btime_ts;

Fixes: bf13fe0f222 ("s3: libsmbclient: Add internal/external structures needed for readdirplus.")
Signed-off-by: Aurelien Aptel <aaptel@suse.com>
Reviewed-by: Noel Power <npower@samba.org>
Autobuild-User(master): Noel Power <npower@samba.org>
Autobuild-Date(master): Mon Mar 23 20:47:44 UTC 2020 on sn-devel-184

s4-librpc: Simplify bytes or unicode input checking in python GUID() bindings

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <npower@samba.org>

py3: Remove #define IsPy3BytesOrString(pystr)

This was

  (PyUnicode_Check(pystr) || PyBytes_Check(pystr))

This allows us to end the use of Python 2/3 compatability macros.

The one caller will be simplified in the next commit

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <nopower@samba.org>

py3: Remove #define IsPy3Bytes PyBytes_Check

This allows us to end the use of Python 2/3 compatability macros.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <npower@samba.org>

py3: Remove #define PyInt_FromLong PyLong_FromLong

This allows us to end the use of Python 2/3 compatability macros.

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

py3: Remove #define PyInt_AsLong PyLong_AsLong

This allows us to end the use of Python 2/3 compatability macros.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <npower@samba.org>

py3: Remove #define PyInt_Type PyLong_Type

This allows us to end the use of Python 2/3 compatability macros.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <npower@samba.org>

pidl: Remove reference to PyInt_Type from error string when we wanted a sensible Long

PyInt_Type is no longer a thing in Python3, we will remove the compatability
reference shortly.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <npower@samba.org>

pidl: Remove duplicate "if (PyLong_Check($cvar)" clauses

Since we moved to Python3, these have been dead code, and it is clearer
now that we have removed the compatability define.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <npower@samba.org>

py3: Remove #define PyInt_Check PyLong_Check

This will allow us to remove some unused code in the PIDL-generated
python bindings.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Noel Power <npower@samba.org>

autobuild: Merge samba-simpleserver into samba-nt4

This saves CI resources (20mins build time) at the expense of making
the samba-nt4 job take 45min (5 min longer).

The new maximum job time is 1:15 by samba-o3 on the different hosts.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Mon Mar 23 14:45:59 UTC 2020 on sn-devel-184

selftest: Run smb2.compound_find against filesrever only, not nt4_dc and ad_dc

This is a slow test of SMB2 protocol features, it needs not to be repeated.  It only
got run against both by accident at the original merge and then extended to
run on the [compount_find] share (with smbd:find async delay usec = 10000)
but really does not need to be run three times.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

autobuild: Merge the samba-ktest-heimdal and samba-fileserver jobs

This avoids a full compile of Samba just to test Kerberos with a system Heimdal
while still providing an environment to test other fileserver features.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

.gitlab-ci.yml: Move the short samba-xc job into "others"

This job is over in 8mins total and much of that is spent setting up.

It can safely be combined with the others in "others"

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

build: Require --without-ad-dc for --without-ads

Building an AD DC while setting --without-ads makes no sense and just wastes compile time on our build hosts.

To allow samba-nt4 to build --without-ad-dc we set rpc.spoolss.notify
(which is built on the NTVFS fileserver for the callbacks) to run in
the ad_member environment rather than nt4_dc and ad_dc.

This is also just more realistic in any case.

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

s3: Remove tevent_wait code. The last user was fsp->deferred_close.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Mon Mar 23 12:06:45 UTC 2020 on sn-devel-184

s3: VFS: Remove deferred_close from struct files_struct.

Bump up the VFS number to 43. Samba 4.13 will ship with that.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: smbd: Remove all references to fsp->deferred_close.

We are now free to remove it from struct files_struct.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: smbd: Convert async SMB2 close to use the wait_queue idiom.

Use the same mechanism to wait for aio on a handle used by
SMB1 reply_tdis(), reply_ulogoffX(), reply_exit(), reply_close()
and SMB2 tree disconnect.

This removes the last user of fsp->deferred_close, allowing
us to remove it.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: smbd: Remove old async versions of SMB1 reply_close().

Use the common pattern to wait for aio.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: smbd: Add async internals of reply_close().

Waits until all aio requests on the closing fsps
before returning to the client.

Slightly modified version of the existing async
reply_close code, updated to use the wait_queue
pattern standardized in reply_tdis, reply_ulogoffX
and reply_exit.

Done this way (commented out) so it is a clean
diff and it's clear what is being added.

The next commit will remove the old version.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: smbd: In asnyc do_smb1_close() use the utility function meant for async SMB1 replies.

Don't use the raw call.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: smbd: Update reply_close() to modern DBG_ coding style.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: smbd: In SMB1 reply_close() rename all uses of 'struct smb_request' to smb1req.

Will make further changes easier to see, as we now use
req for tevent_req structs.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

idl/drsblobs: do not overwrite number of schedules == 1

If the struct has zero or two schedules, that is what it has, and we
should let that be.

Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Sun Mar 22 06:19:51 UTC 2020 on sn-devel-184

dsdb: Allow delete (directly and over DRS) of an object with a link to itself

Previously this would fail with Unsupported critical extension 1.3.6.1.4.1.7165.4.3.2

Reported by Alexander Harm.  Many thanks for helping make Samba better
and for your patience with patches and providing debugging information.

REF: https://lists.samba.org/archive/samba/2020-February/228153.html
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14306

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

dsdb: Add test for the case of a link pointing back at its own object

This type of object was not possible to delete in Samba without first removing
the link.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14306

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

selftest: Add test for dangling backlink to ourself, a missing and a real object

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14306

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

selftest: Add test for dangling backlinks to objects that do not exist

BUG: https://bugzilla.samba.org/show_bug.cgi?id=14306

Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

ldb build: Remove some PEP8 warnings from wscript

Fix indentation of list members and fix lines > 79 characters to remove
PEP8 warnings.

Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

ldb tests: Confirm lmdb free list handling

Add cmocka tests to confirm lmdb's handling of the free list.

As a result of lmdb's MVCC (Multiversion Concurrency Control) long
running read transactions or stale readers (read transactions where the
process exited without ending the transaction) can cause the database to
run out of space.

Items in the free list are only reused when they would not be visible in
a read transaction.  So long running read transactions prevent entries
in the free list being reused, and the database can run out of space.

Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Remove dead code

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Samuel Cabrero <scabrero@samba.org>
Autobuild-Date(master): Fri Mar 20 17:11:28 UTC 2020 on sn-devel-184

s3:rpc_server: Remove unused RPC module init and shutdown callbacks

The setup function registers the endpoint server and RPC core routines
initialize and shutdown it.

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Do not generate and build s3 RPC server code

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Do not include s3 autogenerated headers

Prototype is generated by the server compat parser.

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Remove api_struct

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Remove dead code and unused struct members

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Remove s3 rpc server loop

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Remove call to s3 rpc services shutdown code

The rpc_<service>_shutdown function unregisters the legacy api_struct
from the local dispatch table, which is not longer used as local
dispatching is done through dcesrv_interface and will be removed in
following commits.

The dcesrv_shutdown_registered_ep_servers will unregister endpoint servers.

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Switch to core dcerpc server loop

This commit finally switches the RPC server implementation.

At the same we have to do other related changes to keep code compiling
and test environments running.

First avoid moving the session_info into the allocated pipes_struct memory
context as it is owned now by the core RPC server, and the s3compat pidl
compiler will update the pipes_struct session_info before dispatching
the call with dcesrv_call->auth_state->session_info.

Also, fix a segfault in the endpoint mapper daemon when it tries to delete
the endpoints previously registered over a NCALRPC connection.

If we have:

rpc_server : epmapper = external
rpc_server : lsarpc = external
rpc_daemon : epmd = fork
rpc_daemon : lsasd = fork

The sequence is:

* The endpoint mapper starts (start_epmd in source3/smbd/server.c)
* The lsarpc daemon starts (start_lsasd in source3/smbd/server.c)
  * The lsarpc daemon creates the sockets and registers its endpoints
    (rpc_ep_register in source3/rpc_server/lsasd.c)
  * The endpoint registration code opens a NCALRPC connection to the
    endpoint mapper daemon (ep_register in source3/librpc/rpc/dcerpc_ep.c)
    and keeps it open to re-register if the endpoint mapper daemon dies
    (rpc_ep_register_loop in source3/rpc_server/rpc_ep_register.c)
* When the endpoint mapper daemon accepts a NCALRPC connection it sets a
  termination function (srv_epmapper_delete_endpoints)
* Suppose the lsarpc daemon exits. The NCALRPC connection termination
  function is called.
* The termination function tries to delete all endpoints registered by that
  connection by calling _epm_Delete
* _epm_Delete calls is_privileged_pipe which access to
  pipes_struct->session_info.

As the call to _epm_Delete occurs outside of the PIDL generated code,
the pipes_stuct->session_info is NULL. This commit also sets
pipes_struct->session_info from the dcerpc_connection before calling
_epm_Delete. As the core rpc server supports security context multiplexing we
need to pass the dcesrv_connection to the termination function and let the
implementation pick a auth context. In the case of the endpoint mapper
the termination function has to pick one of type NCALRPC_AS_SYSTEM to
check if the connection is privileged and delete the endpoints
registered by the connection being closed.

Finally, the samba.tests.dcerpc.raw_protocol testsuite passes against
the ad_member environment.

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Update winspool generated server stub

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:winbindd: Dispatch RPC calls through interface local handler

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Dispatch local calls through interfaces local handler

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

pidl:NDR/ServerCompat: Generate local dispatching function

To be used in rpcint_binding_handle, prepare to remove legacy api_struct
and S3 pidl generated code.

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

pidl:NDR/ServerCompat: Initialize and allocate out vars

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

pidl:NDR/ServerCompat: Add dispatch, reply, pull and push functions to header

Will be used by winspool to forward selected opnums to spoolss.

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

pidl:NDR/ServerCompat: Retrieve and setup pipes struct before dispatch

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

s3:rpc_server: Store dcesrv context and endpoint in ncacn_conn

Signed-off-by: Samuel Cabrero <scabrero@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>