s3:smbd: pass down twrp from SMB2_CREATE to filename_convert()

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13455

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

s3:smbd: add twrp args to filename_convert()

All existing callers pass NULL, no change in behaviour.

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13455

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

s3:smbd: add twrp processing to filename_convert_internal()

Not used for now, existing callers pass NULL.

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13455

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

s3:smbd: prepare filename_convert_internal() for twrp

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13455

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

s3:selftest: add a VSS test reading a stream

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13455

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

vfs_shadow_copy2: nicely deal with attempts to open previous version for writing

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13688

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

vfs_shadow_copy2: add shadow_copy2_strip_snapshot_converted

Can be used by callers to determine if a path is in fact pointing at a
file in a snapshot. Will be used in the next commit.

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13688

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

vfs_shadow_copy2: add _already_converted arg to shadow_copy2_strip_snapshot_internal()

Not used for now, all existing callers pass NULL.

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13688

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

s3:script/tests: add a test for VSS write behaviour

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13688

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

s4:torture: add a test-suite for VSS

This test will not be run from the main torture test runner in selftest,
as there we don't pass the required arguments 'twrp_file' and
'twrp_snapshot'.

The test needs a carefully prepared environment with provisioned
snapshot data, so the test will be started from a blackbox test
script. That comes next.

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13688

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

vfs_error_inject: add EBADF error

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13688

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

vfs_error_inject: add pwrite

Bug: https://bugzilla.samba.org/show_bug.cgi?id=13688

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

tests: Rework backup test inheritance to make LP constraints clearer

The backup tests have a special constraint where we always want to use
check_output() over runcmd(). The reason is we need the samba-tool
backup/restore commands executed in a separate process. Otherwise the
global underlying LoadParm can accumulate settings from earlier test
case runs.

We can avoid someone in future inadvertently running runcmd() by
mistake, by simply changing the inheritance so we no longer inherit from
SambaToolCmdTest (so the runcmd functions are no longer present).

The comment explaining this has been moved to the top of the file.

Note that the TestCaseInTempDir inheritance was redundant.
BlackboxTestCase inherits from TestCaseInTempDir (and SambaToolCmdTest
was inheriting from BlackboxTestCase).

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Tim Beale <timbeale@samba.org>
Autobuild-Date(master): Tue Nov 27 06:57:03 CET 2018 on sn-devel-144

tests: Work out DOMSID via samdb rather than environs

Not all testenvs have the DOMSID set as an environment variable.
However, it's easy enough to work out from querying the samdb.

This is a slight change in that we use a source4-generated loadparm
to connect to the DB (self.lp is source3-generated, presumably for
some SMB connection dependency).

This change is so we can run the ntacls_backup tests against a DC with
SMBv1 disabled (the restoredc). Note that currently the tests fail in
the smb.SMB() connection in the setUp(), so we can't run them as part
of autobuild just yet (because we can't known-fail test errors).

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

tests: Run backup tests against restoredc (SMBv1 disabled)

Running the backup tests against the restoredc highlights that the
backup online/rename commands don't work if SMBv1 is disabled. Note that
the offline commands still work because they don't rely on an SMB
connection to the server.

(Note that running the backup tests against the restoredc is probably a
good idea anyway, to prove that there's no limit to the number of times
you can restore a domain from backup, i.e. we support more than just a
one-off restore).

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

python/samba/test: PY3 port samba.tests.domain_backup

The restoredc already runs under python3, so before we can run the
domain_backup tests against the restoredc, we need to make sure they
work under python3.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Noel Power <noel.power@suse.com>
Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

tests: Handle backup command exceptions as test failures, not errors

If the backup command fails (i.e. throws an exception), we want the test
to fail. This makes it easier to mark tests as 'knownfail' (because we
can't knownfail test errors).

In theory, this should just involve updating run_cmd() to catch any
exceptions from the command and then call self.fail().

However, if the backup command fails, it can leave behind files in the
targetdir. Partly this is intentional, as these files may provide clues
to users as to why the command failed. However, in selftest, it causes
the TestCaseInTempDir._remove_tempdir() assertion to fire. Because this
assert actually gets run as part of the teardown, the assertion gets
treated as an error rather than a failure (and so we can't knownfail the
backup tests). To get around this, we remove any files in the tempdir
prior to calling self.fail().

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

tests: Rework backup test_backup_invalid_args test-case

self.create_backup() uses self.run_cmd(), which is a wrapper around
self.check_output(). Rework the code to call the underlying
check_output() function directly instead.

The reason we're doing this is we want run_cmd() to catch exceptions and
fail the test (i.e. in the next patch). However, we can't do that because
this test case relies on receiving the exceptions.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

selftest: Designate one testenv as having SMBv1 disabled

We recommend users disable SMBv1 to avoid potential security holes.
However, none of the AD DC testenvs have SMBv1 disabled.

This patch disables SMBv1 on an arbitrarily-chosen testenv (restoredc).

I chose restoredc as we'll want to run the backup tool tests against
this target, and it might be useful to check we can backup a DC if it's
already been restored once.

Note that SMBv2 doesn't support POSIX extensions (only SMBv1 does),
which is why we haven't just disabled SMBv1 on *all* testenvs.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

tests: Use s3 smbclient for testenv smoketests

smbclient4 doesn't support SMBv2 connections, i.e. it won't work against
a DC with SMBv1 disabled. The smoke-test here is that the DC accepts
the connection, so we don't really care what SMB tool we use to connect.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

selftest: Make testenv NTVFS usage match --use-ntvfs

Regardless of whether the testenv uses --use-ntvfs as part of its
provision options, the s4 testenvs all default to using the NTVFS file
server.

It's not particularly obvious that this is happening. The new restore
DCS (restoredc, renamedc, labdc) were all using NTVFS unintentionally.

The problem is the s4 testenvs default to using services '-s3fs +smb".
provision_ad_dc() explicitly overrides this to use s3fs again
(technically it ends up with both in its smb.conf and just uses whatever
comes last).

This patch changes the testenv setup to check for the presence of the
'--use-ntvfs' option and to set the 'server services' config option
appropriately. This way, the provision command and the smb.conf options
should always line up, with respect to NTVFS.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

selftest: Make chgdcpass's NTVFS usage more obvious

The chgdcpass testenv was not passing --use-ntvfs to the provision
command, but it was still actually using NTVFS.

The reason is the smb.conf generated by provision_raw_step1() would
always try to use the s4/NTVFS file server. Because the smb.conf already
existed, this trumped what was passed to the provision command.

This patch doesn't change the chgdcpass file server. It just makes it
more obvious that chgdcpass is using NTVFS.

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

selftest: Be explicit about which testenvs use ntvfs

If a testenv didn't specify any other provision arguments, then it
defaulted to using the NTVFS file server.

This patch makes it explicit, so we just pass through "--use-ntvfs" as
extra provision args in the cases we want.

(Whether all these testenvs really need to use NTVFS or not is another
question, but at least now it's easy to see which testenvs use it).

BUG: https://bugzilla.samba.org/show_bug.cgi?id=13676

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

libads: Align integer types

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Fri Nov 23 20:23:57 CET 2018 on sn-devel-144

libads: Use dom_sid_string_buf

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>

source4 smbd prefork: Add code comments

Add some comments to the prefork code explaining what's going on.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Fri Nov 23 11:40:49 CET 2018 on sn-devel-144

WHATSNEW: prefork restart

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

samba-tool processes: display pre-fork masters and workers

Tag prefork work processes with "(worker 0)", and sort the process list
on server name to get a consistent order.

 Service:                          PID
 --------------------------------------
 cldap_server                     15588
 ...
 ldap_server                      15584
 ldap_server(worker 0)            15627
 ldap_server(worker 1)            15630
 ldap_server(worker 2)            15632
 ldap_server(worker 3)            15634
 nbt_server                       15576
 notify-daemon                    15638
 ...
 samba                                0
 ...
 wrepl_server                     15580

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

source4 dcerpc_server: remove irpc_add_name

Remove the irpc_add_name from dcesrv_sock_accept, as it results in two
identical names being registered for a process.

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

source4 smbd prefork: Cleanup messaging on restart

Clean up names registered in messaging for a terminated process.

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

source4 messaging: clean up terminated processes

Now that the smbd pre-fork process model restarts failed processes rather than
terminating, we end up with names registered to defunct processes.
This patch adds a function to clean up all the names registered to a process.

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

source4 smbd prefork: Add backoff to process restart

Add new smbd.conf variables 'prefork backoff increment' and
'prefork maximum backoff' to control the rate at which failed pre-forked
processes are restarted.

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

source4 smbd prefork: restart on non zero exit code

Restart any pre-fork master or worker process that exits with a non
zero exit code.

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

source4 smbd process: pass the fatal flag to terminate

Pass the fatal flag supplied to task_server_terminate to the process
task_terminate method.  It will be used by the task_terminate methods to
set an appropriate exit code.  The process_prefork model will use a non
zero exit code to indicate that the process should be restarted.

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

source4 smbd prefork: Restart failed processes

Restart any pre-forked master or worker process that terminated with SIGABRT,
SIGBUS, SIGFPE, SIGILL or SIGSYS

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

source4 smbd prefork: Pass restart information

Pass information about the pre-fork master and worker processes that
will allow them to be restarted.

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

source4 smbd test: prefork process restart

Add tests for the restarting of failed/terminated process, by the
pre-fork process model.

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

replmd: remove unnecessary indent

The previous refactor now means we return early if we don't need to
re-apply isDeleted to the object. The 'else' is redundant and we can
remove it to avoid unnecessary indent.

This patch is basically just a whitespace change. It should not alter
functionality.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Fri Nov 23 08:10:41 CET 2018 on sn-devel-144

replmd: Move logic into new replmd_should_apply_isDeleted() function

It's easier to follow the logic involved here when it's split out into a
separate function.

This patch should not alter the existing logic/functionality.

Note the 'else' case is somewhat redundant, but it avoids excessive
whitespace changes to the function. It'll be tidied up in the next
patch.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: Avoid passing msg to replmd_process_linked_attribute()

We can prevent anyone from inadvertently adding/removing msg->elements[]
in replmd_process_linked_attribute() by just not passing msg into the
function. Currently we only actually need the source DN and a memory
context for reallocating old_el->values.

The warning comment has been moved to a more appropriate place.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

tests: Fix flappiness in DRS tests due to RID Set changing

The test_link_utdv_hwm test case in getnc_exop has started getting
slightly flappy (8 failures in the last 2 weeks). The problem is the
test case creates a new computer, which can occasionally result in a new
RID pool being allocated.

The problem can be reproduced by running the test case repeatedly (it
usually fails after ~250 times).

This patch updates the _check_ctr6() assertion to filter out the 'CN=RID
Set' object, if it happens to be present.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

tests: Add better error for DRS tests

We've got a flappy test hitting this assertion failure, but we can't
tell why it's failing intermittently (probably because we're bumping the
RID-Set, but there's no way to confirm this).

Add some extra debug info if the test assertion fails.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

librpc:ndr: Fix undefined behavior in ndr.c

librpc/ndr/ndr.c:1430 runtime error: left shift of 1 by 31 places cannot
be represented in type 'int'

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
Autobuild-User(master): Gary Lockyer <gary@samba.org>
Autobuild-Date(master): Fri Nov 23 01:23:09 CET 2018 on sn-devel-144

lib:compression: Fix undefined behavior in lzxpress

lib/compression/lzxpress.c:228 runtime error: store to misaligned
address 0x5631d53ca9fe for type 'uint32_t', which requires 4 byte
alignment

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

lib:util: Fix undefined behavior in bitmap.c

lib/util/bitmap.c:77: runtime error: left shift of 1 by 31 places cannot
be represented in type 'int'

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

lib:util: Fix undefined behavior in idtree

lib/util/idtree.c:84 runtime error: left shift of 1 by 31 places cannot
be represented in type 'int'

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

lib:util: Fix undefined behavior in asn1 parser

lib/util/asn1.c:969 runtime error: left shift of negative value -1

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

s3:lib: Fix undefined behavior in messages_dgm

source3/lib/messages_dgm.c:1290:7: runtime error: variable length array
bound evaluates to non-positive value 0

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

lib:crypto: Fix undefined behavior in md4

runtime error: left shift of 145 by 24 places cannot be represented in type 'int'

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

s3:lib: Do not redefine bool and use stdbool.h

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

lib:util: Always include unistd.h for setgroups

This is needed to pass configure checks

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

s3:tests: Initialize pointers with NULL in netdisplay test

Found by Undefined Sanitizer

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

s4:torture: Initialize pointers in libnetapi user test

Found by Undefined Sanitizer.

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

s4:bind_dlz: Only copy the dn once

../source4/dns_server/dlz_bind9.c:1132:4: error: 'dn' may be used uninitialized in this function [-Werror=maybe-uninitialized]
    state->log(ISC_LOG_INFO, "failed to find dnsRecord for %s",
    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
        ldb_dn_get_linearized(dn));
        ~~~~~~~~~~~~~~~~~~~~~~~~~~

Found by Undefined Sanitizer.

Signed-off-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>

utils: Free host realm after processing

In case of error and at the end of processing the list
of realms have to be free'd again.

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Thu Nov 22 11:18:41 CET 2018 on sn-devel-144

utils: Free krb5 context for net_lookup_kdc

In case of an error while processing and at the end of processing
the krb5 conext must be free'd.

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

librpc: Free krb5 context on error

If the call to krb5_cc_resolve() fails and processing is aborted,
the krb5 conext must be free'd before return.

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

s3: Free principal if smb_krb5_principal_get_realm() fails

If smb_krb5_principal_get_realm() fails, procesing is aborted and
resources have to be free'd. In this context free'ing the principal
was missing.

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

README.Coding: Approve DBG statements using dom_sid_str_buf

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

smbspool: Free kerberos context on error

If processing cannot continue and the function returns
the kerberos context must be free'd.

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

libads: set proper ads_keytab_flush() return code on error

The return code was left on success when the calls to
ads_get_machine_kvno() or ads_clear_service_principal_names()
failed and the processing had to be aborted.

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

libsmb: Make cli_notify_send cancellable

Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Volker Lendecke <vl@samba.org>
Autobuild-Date(master): Wed Nov 21 20:50:13 CET 2018 on sn-devel-144

libsmb: Make cli_notify_send/recv smb2-aware

smb2-awareness only existed in the sync wrapper

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

libsmb: Introduce cli_smb2_notify_send/recv

This only existed as a sync wrapper where the underlying function could do
send/recv already.

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

libcli: Make smb2cli_notify_send cancellable

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

libsmb: Make cli_trans_send cancellable

Required to cancel SMB1 notify

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

libsmb: Make cli_trans a normal subreq function

A lot of "state" variables, but still I don't think the additional complexity
is justified by saving a few lines. Make this a completely regular tevent_req
"req/subreq" function.

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

logger: use color automatically for a tty

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Wed Nov 21 10:46:20 CET 2018 on sn-devel-144

samba-tool ldapcmp: print DNs on predictable order

Rather than unstable hash order. Ideally we'd do them in proper DN order.

Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: make code pythonic

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: promote re object to global

Then we can reuse the re obj.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: use set instead of list to compare attrs

This will simplify the logic and improve performance.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: fix wrong way for string copy

Two mistakes here:

- res[:-1] will copy but lost the last char
- string is immutable in python, there is no need to copy it explicitly

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: pass --skip-missing-dn to LDAPBase

This option has default value False, and was actually not passed down from cli
to LDAPBase. However, LDAPBase.__init__ has default value True for it.

After the change, a few tests using ldapcmp are affected.
Add --skip-missing-dn explicitly to keep the behavior consistent,
otherwise test will fail.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: avoid modifying data while looping on dict

Just define another dict for return value, seems no need to modify
original dict.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: rm unused global var summary

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: use set instead of list to find missing DNs

This simplify the logic and improve performance a lot.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: avoid list comprehension in for loop

The list comprehension will repeat for each item.
For large database, this make the command freeze.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: add choices arg to --view option

So we don't need to validate ourselves.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: add choices arg to --scope option

So we don't need to validate ourselves.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: rename __eq__ to diff

This method actually changed both objects and print info.
__eq__ is not a proper name and is not designed for this case.
Rename to diff.

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

netcmd/ldapcmp: fix typo for Bundle

Bundel -> Bundle

Signed-off-by: Joe Guo <joeg@catalyst.net.nz>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: Make replmd_process_linked_attribute() mem dependencies clearer

This patch should not alter functionality - it is just making memory
assumptions used in replmd_process_linked_attribute() clearer.

When adding/removing msg->elements we have to take care, as this will
invalidate things like the parsed-DN array or old ldb_message_element
pointers. This has always been the case (i.e. f6bc4c08b19f5615a49),
however, now we need to take even more care, as the msg being modified
is re-used and split across 2 different functions.

Add more code comments to highlight this. We can also free
pdn_list/old_el to prevent them being incorrectly used after realloc.
It seems appropriate to also add a sanity-check that the tmp_ctx alloc
succeeds (which all the other memory hangs off).

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Tim Beale <timbeale@samba.org>
Autobuild-Date(master): Wed Nov 21 05:31:10 CET 2018 on sn-devel-144

replmd: Avoid redundant dsdb_get_deleted_objects_dn() checks

Quite a bit of time was spent in dsdb_get_deleted_objects_dn()
processing during either a join (~9%) or a full-sync (~28%).

The problem is we're *always* doing the dsdb_get_deleted_objects_dn()
call for each object, regardless of whether it's actually deleted or
not. i.e. we were doing an expensive query and a lot of the time just
ignoring the query result.

If it's not a deleted object we're dealing with, we can just return
early and skip the unnecessary processing.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: Minimize get_parsed_dns_trusted() calls during replication

When a group has 10,000+ links, get_parsed_dns_trusted() can be costly
(simply the talloc calls alone are expensive). Instead of re-generating
the pdn_list for every single link attribute, we can change to only
re-generate it when we really need to.

When we add a new link, it reallocates old_el->values, and so we need to
recreate the pdn_list because all the memory pointers will have changed.
However, in the other cases, where we're simply updating the existing
link value (or ignoring the update, if it's already applied), we can
continue using the same pdn_list (rather than re-parsing it again).

This would generally only save time with a full-sync - it won't really
help with the join case (because every link processed results in a
realloc).

On a DB with 5000 users, this makes a full-sync about ~13% faster.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: Pass old_el into replmd_process_linked_attribute()

We should only need to lookup the msg attribute once per source object.
The old_el->values may change due to link-processing, but old_el itself
should not.

This is not aimed at improving performance, but we need to change how
old_el is used before we can change pdn_list (which is more costly
processing-wise).

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: Remove some redundant code

At first glance, this code seemed completely unnecessary. However, it
was added (by commit f6bc4c08b19f5615) for a valid reason: adding the
whenChanged/uSNChanged attributes to the message can cause msg->elements
to be reallocated, which means the old_el pointer (which points to
msg->elements memory) can be out of date.

whenChanged/uSNChanged now get added to the msg last, just before the DB
modify operation. So old_el can no longer become out of date within
replmd_process_link_attribute(), so re-fetching it is now redundant.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: Move where we update the usnChanged/whenChanged

Move this closer to where the source object actually gets modified.

The main reason to do this is that adding fields can cause the
msg->elements to be reallocated, which will invalidate all the
old_el and pdn_list pointers which are derived from the msg.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: Only modify the object if it actually changed

Commit 775054afbe1512 reworked replmd_process_link_attribute() so that
we batch together DB operations for the same source object. However, it
was possible that the object had not actually changed at all, e.g.
- link was already processed by critical-objects-only during join, or
- we were doing a full-sync and processing info that was already
  up-to-date in our DB.

In these cases we modified the object anyway, even though nothing had
changed. This patch fixes it up, so we check that the object has
actually changed before modifying the DB.

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

replmd: replmd_process_link_attribute() returns type of change made

In order to share work across related link attribute updates, we need
replmd_process_link_attribute() to let the caller know what actually
changed.

This patch adds an extra return type that'll be used in the next patch.
What we're interested in is: the update was ignored (i.e. it's old news),
a new link attribute was added (because this affects the overall
msg/element memory), and an existing link attribute was modified (due to
how links are actually stored, this includes deleting the link, as in
reality it simply involves setting the existing link to 'inactive').

Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

source4 samr: Tidy DBG_WARNING calls

Move the calls to GUID_buf_string and dom_sid_str_buf into the
coresponding DBG_WARNING call, instead of using an intermediate variable.
While this violates the coding guidelines, doing this makes the code less
cluttred and means the functions are only called if the debug message is
printed.

Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Wed Nov 21 01:50:11 CET 2018 on sn-devel-144

s4-samr: Use GUID_buf_string() in dcesrv_samr_EnumDomainUsers()

This avoids memory allocation.

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

s4-samr: Use dom_sid_split_rid() to get the RID in dcesrv_samr_EnumDomainUsers

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

source4 samr: cache samr_EnumDomainUsers results

Add a cache of GUID's that matched the last samr_EnunDomainUsers made on a
domain handle.  The cache is cleared if resume_handle is zero, and when the
final results are returned to the caller.

The existing code repeated the database query for each chunk requested.

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

tests samr: Extra tests for samr_EnumDomainUserss

Add extra tests to test the content returned by samr_EnumDomainUsers,
and tests for the result caching added in the following commit.

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

s4-samr: Use GUID_buf_string() in dcesrv_samr_EnumDomainGroups()

This avoids memory allocation

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

s4-samr: Use dom_sid_split_rid() to get the RID in dcesrv_samr_EnumDomainGroups

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

source4 samr: cache samr_EnumDomainGroups results

Add a cache of GUID's that matched the last samr_EnunDomainGroups made on a
domain handle.  The cache is cleared if resume_handle is zero, and when the
final results are returned to the caller.

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

test samr: Extra tests for samr_EnumDomainGroups

Add extra tests to test the content returned by samr_EnumDomainGroups,
and tests for the result caching added in the following commit.

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

s4-samr: Use dom_sid_split_rid() to get the RID in dcesrv_samr_QueryDisplayInfo

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