torture: beginning of a mdssvc RPC service test-suite

Yikes! Most tests fail atm.

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

rpcclient: add some simple commands for Spotlight RPC

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

rpcclient: domain_sid is only needed for SAMR

This allows using rpcclient commands with servers that don't implement LSA
QueryInfoPolicy.

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

lib:torture: add torture_assert_u32_[not_]equal[_goto] macros

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

s3:mdssvc: fix unmarshalling of empty CNID array

len=0 is invalid, len=8 is an empty array, len>8 is an array with members, so
for the len=8 case we must add the empty cnid array.

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

s3:mdssvc: marshalling: fix unpacking empty CNID structure

Pass the correct tag member tag.size to sl_unpack_CNID(), not
tag.length. tag.size is the size actually used in a buffer of size tag.length.

Cf other users of tag.size that already do this correctly, this was only wrong
in this place.

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

s3:mdssvc: marshalling: fix unpacking empty filemeta structure

This is how a correct dump of an empty sl_filemeta_t container should look like:

DALLOC_CTX(#1): {
sl_array_t(#3): {
uint64_t: 0x0023
CNIDs: unkn1: 0x0, unkn2: 0x0
DALLOC_CTX(#0): {
}
sl_filemeta_t(#0): {
}
}
}

This is basically the response from macOS mdssvc for a query that yields no
results: sl_filemeta_t is empty, the CNIDs array as well.

Looking at the raw packet data, the empty sl_filemeta_t container as a size of 8
bytes which fails the following check in sl_unpack_cpx():

        case SQ_CPX_TYPE_FILEMETA:
                ...
if (tag.size < 16) {
        *boom*
                }

Only tag.size=0 is invalid, tag.size=8 denotes an empty container and tag.size>=16
denotes a sl_filemeta_t container with actual content must be unpacked by
calling sl_unpack(). Note that size is always a muliple of 8.

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

s3:mdssvc: marshalling: add a newline to a DEBUG message

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

torture: start of a mdssvc packet (un)marshalling testsuite

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

librpc: add RPC_NDR_MDSSVC

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

s3:wscript: enable Spotlight by default

Now that we have a no-op backend that is always available, we can compile mdssvc
by default.

The new behaviour is:

option not used       Default: build mdsvc with available backends
                      from autodetection
--disable-spotlight   Do not build mdssvc
--enable-spotlight    Build mdssvc and require a real backend
      (currently Tracker, in the future also Elasticsearch)

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

s3:mdssvc: avoid strncpy when marshalling strings

Avoids failure when at O3 level:

  [2082/4232] Compiling source3/rpc_server/mdssvc/marshalling.c

  ==> /builds/samba-team/devel/samba/samba-o3.stderr <==
  In file included from /usr/include/string.h:494,
                   from /usr/include/bsd/string.h:30,
                   from ../../lib/tevent/../replace/replace.h:164,
                   from ../../source3/include/includes.h:23,
                   from ../../source3/rpc_server/mdssvc/marshalling.c:21:
  In function ‘strncpy’,
      inlined from ‘sl_pack_string’ at ../../source3/rpc_server/mdssvc/marshalling.c:493:2,
      inlined from ‘sl_pack_loop’ at ../../source3/rpc_server/mdssvc/marshalling.c:607:13:
  /usr/include/bits/string_fortified.h:106:10: error: ‘__builtin_strncpy’ output
  truncated before terminating nul copying as many bytes from a string as its
  length [-Werror=stringop-truncation]
    106 |   return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
        |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  ../../source3/rpc_server/mdssvc/marshalling.c: In function ‘sl_pack_loop’:
  ../../source3/rpc_server/mdssvc/marshalling.c:458:8: note: length computed here
    458 |  len = strlen(s);
        |        ^~~~~~~~~
  cc1: all warnings being treated as errors

Marshalled strings are not 0 terminated.

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

unittest: workaround dependency problem in test_lib_util_modules

waf somehow screws the dependencies and the module ends up with a bunch of
missing RPC related symbols once an RPC service has special dependencies like
the mdssvc RPC service.

  $ bin/test_lib_util_modules
  test: test_samba_module_probe
  success: test_samba_module_probe
  test: test_samba_module_probe_dummy
  Error loading module '/home/samba/src/bin/modules/rpc/test_dummy_module.so':
  /home/samba/src/bin/modules/rpc/test_dummy_module.so: undefined symbol:
  rpc_service_mode
  failure: test_samba_module_probe_dummy [
  NT_STATUS_IS_OK(status)
  ../../testsuite/unittests/test_lib_util_modules.c:39: error: Failure!
  ]
  test: test_samba_module_probe_slash
  success: test_samba_module_probe_slash

This is currently not noticed as mdssvc is not enabled in selftest, but that is
going to change with a subsequent commit.

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

s3:mdssvc: supposed status field is in fact a fragment indicator

Spotted this in mdssvc response that containied many results for a search
request: if the mdssvc response blob is larger then ~32k, the server fragments
the response in 32k fragments and sets the "fragment" field to 1.

Note that mdssvc implemenets result set "fragmentation" at the result set layer,
not at the marshalled response buffer layer. Therefor mdssvc always sets this
field to 0.

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

s3:mdssvc: add noindex backend

Add a new default backend that, while allowing mdsvc RPC and search queries from
clients, always returns no results.

Shares using this backend will behave the same way as shares on a macOS SMB
server where indexing is disabled.

This change will later also allow us to compile the Spotlight RPC service by
default which is a big step in the direction of adding tests to CI.

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

s3:mdssvc: log inode number in ino_path_map_destr_cb()

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

s3:mdssvc: move some code to a subsystem

We need this in a later patchset when adding a mdssvc client library and for
unit tests.

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

s3:mdssvc: mds_dalloc_dump() -> dalloc_dump()

Move mds_dalloc_dump() to dalloc and rename it to dalloc_dump().

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

s3:mdssvc: add Unicode normalisation

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

s3:mdssvc: new option "spotlight backend"

Currently there's only the tracker backend, but subsequent commits will add
other backends.

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

s3:mdssvc: pass down sharename and SNUM

Not used for now, will be needed in the upcoming Elasticsearch backend.

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

s3-mdssvc: factor out Tracker backend logic

This moves all Tracker backend logic into a modularized component.

This should not result in any change in behaviour, it just paves the way
for adding additional backends. Currently the only available backend is
Gnome Tracker.

slq_destroy_send/recv is not needed anymore as the problem is solved now by
correctly checking if an async Tracker request was cancelled and we got
G_IO_ERROR_CANCELLED in tracker_con_cb() or tracker_query_cb() and avoid using
user_data in that the case.

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

s3:mdssvc: move debug macro to header file

This is going to be used from other parts of the code soon.

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

s3:mdssvc: README.Coding fixes, long lines, modernize debug macros

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

s3:rpc_server: factor out rpc_mdssvc_sources

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

tfork: Free memory in torture test to satisfy sanitizer

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Matthias Dieter Wallnöfer <mdw@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Thu Aug  8 11:33:32 UTC 2019 on sn-devel-184

ldb: Free memory in torture test to satisfy sanitizer

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Matthias Dieter Wallnöfer <mdw@samba.org>

util: Free memory in charset torture test to satisfy sanitizer

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Matthias Dieter Wallnöfer <mdw@samba.org>

ndr: Free memory in torture test to satisfy sanitizer

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Matthias Dieter Wallnöfer <mdw@samba.org>

util: Free memory in genrand torture test to satisfy sanitizer

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Matthias Dieter Wallnöfer <mdw@samba.org>

ldb: Fix mem allocation in torture test to satisfy sanitizer

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Matthias Dieter Wallnöfer <mdw@samba.org>

registry: Free memory at the end of each loop run to prevent mem leak

Found during torture test runs with enable address-sanitizer.

Signed-off-by: Swen Schillig <swen@linux.ibm.com>
Reviewed-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Thu Aug  8 06:44:12 UTC 2019 on sn-devel-184

torture: SMB1 unlink needs delay for a stream's SHARING_VIOLATION

Survives against W2k12R2

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): Thu Aug  8 01:05:38 UTC 2019 on sn-devel-184

smbd: Move handling the 1sec sharing_violation delay into smb1 code

Simplify the flow in open_file_ntcreate, streamline it for SMB2

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

smbd: Add defer_sharing_violation_smb1

This is close to what Windows SMB1 does: Instead of waiting for the
share entry causing the SHARING_VIOLATION to disappear, retry every
200msec up to one second. Windows does it a little differently: Retry
up to 5 times. But up to one second should be close enough.

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

smbd: Get "req->request_time" early in create_file_default()

This is necessary for the following case:

We want to delete a file with an open stream that is not open with
FILE_SHARE_DELETE. In SMB1, we need to defer the sharing violation
reply (we don't do that right now, test to follow). However, when we
move that sharing violation delay to where it belongs, into the outer
layers, only very deep in the nested open_streams_for_delete smb1
sharing violation delay handling call we will hit the sharing
violation in the 1-second retry case. However, that
open_streams_for_delete itself is INTERNAL_OPEN_ONLY and thus not
deferred itself. This means that it will not overwrite
req->request_time at all.

Exec summary: We only have one request_time now, set it properly as
early as possible.

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

smbd: Do not exceed the req's max timeout in setup_poll_open()

This will become important in the next commits when the SMB1 sharing
violation delay will use this. We want to be able to reduce the
timeout to less than 200msec, see the next commits.

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

smbd: Make setup_poll_open() return bool

This is for callers who don't want to call open_was_deferred()
afterwards

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

smbd: Make setup_kernel_oplock_poll_open() more general

This is a generic "open retry without locking.tdb waiter" loop: Take
the specific timeouts as parameters.

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

smbd: Avoid exit_server() in setup_kernel_oplock_poll_open()

Failure to postpone a request is not really fatal: We just don't retry
as wanted but return an error to the client that might have resolved
itself after a few seconds. From my point of view such a spurious and
rare error, which is highly unlikely anyway does not justify to kill
that client's connection.

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

smbd: Remove access check on SHARING_VIOLATION

This piece of code predates our user-space access checks, which we
nowadays always do in open_file()

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

s3: VFS: Make setxattr return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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): Wed Aug  7 17:59:25 UTC 2019 on sn-devel-184

s3: VFS: Make removexattr return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make chflags return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make rmdir return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make mkdir return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make mknod return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make ntimes return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make chown return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make chmod return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s3: VFS: Make unlink return errno = EROFS on a shadow copy path.

smbd has no business modifying a shadow copy filesystem, it should be read-only.

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

s4:librpc/rpc: Use generic roh_connect_channel_send/recv

The HTTP connection code is common to in and out channels.

Signed-off-by: Samuel Cabrero <scabrero@suse.de>
Reviewed-by: Ralph Boehme <slow@samba.org>
Autobuild-User(master): Ralph Böhme <slow@samba.org>
Autobuild-Date(master): Wed Aug  7 14:12:40 UTC 2019 on sn-devel-184

s4:librpc/rpc: Add roh_connect_channel_send/recv

Generic function to connect a ROH channel. The use comes later to avoid
duplicated code.

Signed-off-by: Samuel Cabrero <scabrero@suse.de>
Reviewed-by: Ralph Boehme <slow@samba.org>

s4:lib/http: move to the toplevel

This is going to be used from the s3 RPC server soon...

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

s4:lib/http: add support for http POST

Even though GET would work as well, only adding POST, as that's the only method
that's going to be exersized in code and tests (RPC mdssvc elasticsearch
backend).

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

s4:lib/http: use http_conn in http_send_request_send() and http_read_response_send()

Works, tested with

  $ bin/smbtorture -W RIVERSIDE --realm=RIVERSIDE.SITE -s /dev/null \
    -U Administrator%Passw0rd \
    ncacn_http:10.10.11.164[HttpProxy=10.10.11.164:593,HttpUseTls=false,HttpAuthOption=basic,HttpConnectOption=UseHttpProxy] \
    rpc.epmapper.epmapper.Lookup_simple

I get an ACCESS_DENIED error, but I get it over HTTP. :)

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

s4:librpc/rpc: use http_connect_send/recv in ncacn_http

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

s4:lib/http: remove stream and send_queue from state

They're not used, so remove them.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

s4:lib/http: move struct http_read_response_state to the implementation

This is only used by the implemementation of http_send_request_send/recv, no
need to have this in the header file.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

s4:lib/http: move struct http_send_request_state to the implementation

This is only used by the implemementation of http_send_request_send/recv, no
need to have this in the header file.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

s4:lib/http: add http_[dis]connect_send and recv

Not used for now, that comes later.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Samuel Cabrero <scabrero@suse.de>

charset: add tests for Unicode NFC <-> NFD conversion

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Wed Aug  7 07:25:39 UTC 2019 on sn-devel-184

charset: add support for Unicode normalisation with libicu

This adds a direct conversion hook using libicu to perform NFC <-> NFD
conversion on UTF8 strings. The defined charset strings are "UTF8-NFC" and
"UTF8-NFD", to convert from one to the other the caller calls smb_iconv_open()
with the desired source and target charsets, eg

  smb_iconv_open("UTF8-NFD", "UTF8-NFC");

for converting from NFC to NFD.

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

torture: add torture_assert_errno_equal_goto()

Signed-off-by: Ralph Boehme <slow@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>

Add fuzzing binary for oLschema2ldif

Use the oLschema2ldif library functions introduced in commit
0c7c44a284a26790081c000f5b8f4ed32f9f21d7 to implement a fuzzing utility.

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

Add fuzzing binary for tiniparser

The "tiniparser_load" function is made into a wrapper for the newly
added "tiniparser_load_stream" function which accepts a FILE pointer.
This way no actual files have to be opened for fuzzing (memfd_create(2)
isn't readily available on all systems yet).

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

Add fuzzing support to build system

LibFuzzer, Honggfuzz and other programs implement simple interfaces for
fuzzing appropriately prepared code. Samba contains quite a lot of
parsing code, often a good target for fuzzing.

With this change the build system is amended to support building fuzzing
binaries (added in later changes).

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

smbd: Assert that INTERNAL_OPEN_ONLY never gets real oplocks

Slightly simplify assumptions in the code

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 Aug  6 23:06:41 UTC 2019 on sn-devel-184

smbd: Remove deferred_open_record_create()

This hides a talloc off the NULL context: The caller needs to make sure
this is put on a real talloc context later. Make that more
obvious. Also, it passes down a boolean flag, making its purpose a bit
opaque to the caller sites.

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

smbd: Remove "delayed_for_oplocks" from deferred_open_record_create()

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

smbd: Remove "deferred_open_record->delayed_for_oplocks"

Nobody used this

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

smbd: Remove "id" from deferred_open_record_create()

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

smbd: Remove "deferred_open_record->id"

Nobody used this

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

smbd: Clean up leases.tdb if set_share_mode() fails

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

smbd: Factor out remove_lease_if_stale()

remove_lease_if_stale() does not have the check

if (e == e2) {
/* Not ourselves. */
continue;
}

that remove_share_mode_lease() had. However, remove_share_mode_lease()
has already set e->op_type=NO_OPLOCK, so that the

if (e->op_type != LEASE_OPLOCK) {
continue;
}

statement has the same effect.

Why? The next commit will need it for proper error path cleanup.

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

smbd: Inline remove_oplock_under_lock() into its only caller

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

smbd: Make remove_oplock_under_lock static

We have support for nested get_share_mode_lock calls, so we can avoid
this additional function.

It's one more talloc/free per close, but I hope this can't be
measurable. Our open/close path is pretty expensive anyway.

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

smbd: Slightly simplify set_share_mode()

The fsp carries all required information also for leases. There's no
need to pass that as additional parameters

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

smbd: Move fsp_client_guid() to locking/

Yes, this adds another peek from locking/ back into smbd/proto.h, but
locking/locking.c does the same already.

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

smbd: Slightly optimize reply_open_and_X()

Well, this is just a drive-by patch: We don't use "mtime" if we
exit early. So it's not really a worthwhile optimization, to me
it's more a code clarity thing.

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

smbd: Remove "share_access" from fcb_or_dos_open()

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

smbd: Remove "share_access" from dup_file_fsp()

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

smbd: Remove "share_access" from files_struct

Nobody used this (except vfs_gpfs, which did not need it really). If
you *really* need this, you can always look in locking.tdb, but this
should never happen in any hot code path, as no runtime decisions are
made on the share access after the open is done.

Bump VFS interface number to 42.

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

vfs_gpfs: Avoid "fsp->share_access" in vfs_gpfs_kernel_flock()

This was in vfs_gpfs.c from day one, no idea why both "share_mode" and
"fsp->share_access" were used as if they were two different things.

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

smbd: Pass share_access/access_mask explicitly to grant_fsp_oplock_type()

Why? While restructuring open_file_ntcreate() I found the data flow for
these values confusing: grant_fsp_oplock_type() depends on
fsp->access_mask, which changes its value inside
open_file_ntcreate(). I find the data flow easier to follow if it
happens in explicit variables.

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

smbd: Pass share_access/access_mask explicitly to set_share_mode()

Makes the interface more obvious to me. Also, I want to remove
fsp->share_access, which is not really used anywhere after the fsp has
been fully established.

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

smbd: Fix a typo

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

smbd: Fix a typo

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

smbd: Enhance debugging in push_deferred_open_message_smb()

Real timestamps are easier to read

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

smbd: Use timeval_str_buf() in defer_open()

Also make this a hires timestamp for better debugging.

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

smbd: Fix a signed/unsigned warning

For whatever reason, "st_size" in "struct stat" is an off_t, which is a
signed integer. Negative sizes don't really make sense, so this cast
should be safe.

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

smbd: Fix a signed/unsigned comparison

In the 2nd for-loop we need a signed int as we are comparing to >=0.

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

smbd: Remove a false comment

Looking at the code we do respect the kernel_flock failure.

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

torture3: Don't duplicate set_blocking()

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

lib: Remove a duplicate prototype for set_blocking()

This is now in lib/util/blocking.h

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

build: Only check rpc/xdr.h in tirpc if required

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

vfs: Fix the nfs4acl build on FreeBSD

FreeBSD needs rpc/types.h before rpc/xdr.h

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

build: Correctly detect rpc/xdr.h on FreeBSD

FreeBSD needs rpc/types.h included before rpc/xdr.h

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

wscript: Fix a typo

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

winbind: Fix the clang build

clang complains that lm_resp and nt_resp is used uninitialized. This
is true for the "goto done;" in line 2644. This directly calls
log_authentication without having initialized those two blobs.

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