IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
The point of this realloc is that we are not using this array, but
keeping it around to remain a node the talloc tree. We'd prefer to
reduce it to nothing.
Coverity rightly spotted that it was reallocing an array of `struct
ldb_val` to an array of `struct ldb_val *`, which has a different size
and all. But it doesn't matter in this case, because we will never use
it.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15590
Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Jennifer Sutton <josutton@catalyst.net.nz>
(cherry picked from commit e2a74963fb)
We can't just free it, because has the GUID index list as a child, and
these are shared by the new dn list (from the subtransaction we are
committing). But if the dn list is long and the main transaction is
long-lived, we can save a lot of memory by turning this dn list into
an almost empty node in the talloc tree. This returns us to roughly
the situation we had prior to the last commit.
For example, with the repro.sh script on bug 15590 in indexes mode
with 10000 rules, The last 3 commits use this much memory at the end
of an unusually large transaction:
full talloc report on 'struct ldb_context' (total 4012222 bytes in 90058 blocks)
full talloc report on 'struct ldb_context' (total 2405482219 bytes in 90058 blocks)
full talloc report on 'struct ldb_context' (total 4282195 bytes in 90058 blocks)
That is, the last commit increased usage 500 fold, and this commit
brings it back to normal.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15590
Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
(cherry picked from commit 1bf9ede94f)
We don't want to modify the original list, but we can reuse the keys
if we treat them as immutable and don't free them. That makes it a lot
quicker if there are many keys (i.e. where an index is useful) and may
sub-transactions. In particular, it avoids O(n²) talloc_memdups.
A removed comment that says "We have to free the top level index
memory otherwise we would leak", and this will be addressed in the
next commit.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15590
Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
(cherry picked from commit 5f0198d698)
Redaction may be expensive if we end up needing to fetch a security
descriptor to verify rights to an attribute. Checking the search scope
is probably cheaper, so do that first.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15270
Signed-off-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Add a hook, acl_redact_msg_for_filter(), in the aclread module, that
marks inaccessible any message elements used by an LDAP search filter
that the user has no right to access. Make the various ldb_match_*()
functions check whether message elements are accessible, and refuse to
match any that are not. Remaining message elements, not mentioned in the
search filter, are checked in aclread_callback(), and any inaccessible
elements are removed at this point.
Certain attributes, namely objectClass, distinguishedName, name, and
objectGUID, are always present, and hence the presence of said
attributes is always allowed to be checked in a search filter. This
corresponds with the behaviour of Windows.
Further, we unconditionally allow the attributes isDeleted and
isRecycled in a check for presence or equality. Windows is not known to
make this special exception, but it seems mostly harmless, and should
mitigate the performance impact on searches made by the show_deleted
module.
As a result of all these changes, our behaviour regarding confidential
attributes happens to match Windows more closely. For the test in
confidential_attr.py, we can now model our attribute handling with
DC_MODE_RETURN_ALL, which corresponds to the behaviour exhibited by
Windows.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15270
Pair-Programmed-With: Andrew Bartlett <abartlet@samba.org>
Signed-off-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Change all uses of ldb_kv_filter_attrs() to use
ldb_filter_attrs_in_place() instead. This function does less work than
its predecessor, and no longer requires the allocation of a second ldb
message. Some of the work is able to be split out into separate
functions that each accomplish a single task, with a purpose to make the
code clearer.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=15270
Signed-off-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
We should only do this in the LDB_ERR_NO_SUCH_ATTRIBUTE case.
Signed-off-by: Joseph Sutton <josephsutton@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
The LDB filter processing is where the time is spent in the LDB stack
but the timeout event will not get run while this is ongoing, so we
must confirm we have not yet timed out manually.
RN: Ensure that the LDB request has not timed out during filter processing
as the LDAP server MaxQueryDuration is otherwise not honoured.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14694
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
This avoids using "ldb" as the memory context in most cases, and may avoid
a long-term memory leak if future changes cause dn_key not to be freed.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14299
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
We avoid bare if conditions in Samba.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14299
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
We need to ensure we TALLOC_FREE(tmp_ctx) in every error exit.
Reported by Alex Kolesnik, ABISoft Ltd., www.abisoft.biz. Thanks!
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14299
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
This changes the LDB behaviour in the combination of a SCOPE_ONE search and
an index returning less than 10 results.
After b6b5b5fe35 the list->strict flag
became set to false in all cases, rather than being left to the
value set by the caller.
This changes the ldb_kv_index_dn_one() code to force strict
mode on success instead.
Thanks to Marcus Granér, ICEYE Oy for reporting.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14270
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Stefan Metzmacher <metze@samba.org>
The msg for each database record is allocated on the module context, but
never freed. The module seems like it could be a long-running context (as
the database would normally get repacked by the samba executable).
Even if it's not a proper leak, it shouldn't hurt to cleanup the memory.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14059
Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Autobuild-User(master): Garming Sam <garming@samba.org>
Autobuild-Date(master): Tue Aug 20 04:57:10 UTC 2019 on sn-devel-184
Firstly, with Samba AD this looks a little weird because we log the same
message 5 times (once for every partition). If we log that we're doing
this to records in different partitions, hopefully someone with a little
Samba knowledge can figure out what's going on.
Secondly, the info about what partitions are actually changing might be
useful. E.g. if we hit a fatal error repacking the 3rd partition, and
the transaction doesn't abort properly, then it would be useful to know
what partitions were repacked and which ones weren't.
There doesn't appear to be a useful name for the partition
(ldb_kv->kv_ops->name() doesn't seem any more intelligible to a user),
so just log the first record that we update. We can use that to infer
the partition database).
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14059
Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
The "format 0x26011968" log confused me (and I'm a developer).
We can subtract the base offset from the pack format to get a more
user-friendly number, e.g. v0 (not actually used), v1, v2, etc.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14059
Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Store it on the repack context so that we can log a more informative
message "Repacking from format x to format y".
While this is not really a big deal currently, it could be worth
recording for potential future scenarios (i.e. supporting three or more
pack versions), where upgrades could potentially skip an intermediary
pack format version.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14059
Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
LDB_DEBUG_WARNING gets logged by Samba as level 2, whereas the default
log level for Samba is 0. It's not really fair to the user to change the
format of their database on disk and potentially not tell them.
This patch adds a log with level zero (using a alias define, as this
technically isn't a fatal problem).
BUG: https://bugzilla.samba.org/show_bug.cgi?id=14059
Signed-off-by: Tim Beale <timbeale@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Use iterate_range kv op to skip the index section of the database when
running a full search. Quick local testing showed 18% improved throughput
on a full search with no results on a 50k database. With more results,
improvement is smaller but still noticeable.
Signed-off-by: Aaron Haslett <aaronhaslett@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Autobuild-User(master): Andrew Bartlett <abartlet@samba.org>
Autobuild-Date(master): Fri Aug 2 02:29:42 UTC 2019 on sn-devel-184
The previous code would override the caller with the DB size
estimate rather than allowing the caller to force the bigger size.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
When performing a join the overhead of the sub transactions protecting
key value operations becomes significant. This commit adds a new
"batch_mode" option that disables the sub transactions around key value
operations.
The operation level index cache is also disabled, which means the
overall transaction level index cache can become inconsistent if an
operation fails. To protect against this and other possible on disk
inconsistencies, if any operation fails during a batch_mode
transaction the commit will fail and transaction will be rolled back.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
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): Fri Jun 21 05:45:02 UTC 2019 on sn-devel-184
As a performance enhancement the key value layer maintains a cache of
the index records, which is written to disk as part of a prepare commit.
This patch adds an extra cache at the operation layer to ensure that the
cached indexes remain consistent in the event of an operation failing.
Add test to test for index corruption in a failed modify.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Pair-Programmed-With: Andrew Bartlett <abartlet@samba.org>
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
The callers will soon have two possible parents for this pointer, so
we need to remove this check, which was added out of caution given
the rather strange pattern of putting an active memory pointer into a
TDB (as a hash map).
That is, the only callers that did call this with "true" would
have to call this with "false", so just remove the complexity.
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Add test exercising the sub/nested transactions wrapping the key value
operations.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Use the nested transaction support added to the key value back ends to
make key value operations atomic. This will ensure that rename
operation failures, which delete the original record and add a new
record, leave the database in a consistent state.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Add limited nested transaction support to the back ends to make the key value
operations atomic (for those back ends that support nested transactions).
Note: that only the lmdb backend currently supports nested transactions.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Fixes csbuild error.
Error: COMPILER_WARNING:
lib/ldb/ldb_key_value/ldb_kv_search.c: scope_hint: In
function ‘search_func’ lib/ldb/ldb_key_value/ldb_kv_search.c:308:47:
warning: unused parameter ‘ldb_kv’ [-Wunused-parameter]
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andreas Schneider <asn@samba.org>
Autobuild-User(master): Andreas Schneider <asn@cryptomilk.org>
Autobuild-Date(master): Thu Jun 13 08:27:12 UTC 2019 on sn-devel-184
Fixes csbuild error.
Error: COMPILER_WARNING:
lib/ldb/ldb_key_value/ldb_kv_index.c: scope_hint: In function
‘ldb_kv_index_add1’
lib/ldb/ldb_key_value/ldb_kv_index.c:2601:15: warning: comparison of
integer expressions of different signedness: ‘int’ and ‘unsigned int’
[-Wsign-compare]
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andreas Schneider <asn@samba.org>
Fixes csbuild error.
Error: COMPILER_WARNING:
lib/ldb/ldb_key_value/ldb_kv_index.c: scope_hint: In function
‘ldb_kv_index_traverse_store’
lib/ldb/ldb_key_value/ldb_kv_index.c:779:60: warning: unused parameter
‘tdb’ [-Wunused-parameter]
By tagging the tdb_context parameter in ldb_kv_index_traverse_store as
_UNUSED_
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andreas Schneider <asn@samba.org>
Fixes csbuild error.
Error: COMPILER_WARNING:
lib/ldb/ldb_key_value/ldb_kv_index.c: scope_hint: In function
‘ldb_kv_key_dn_from_idx’
lib/ldb/ldb_key_value/ldb_kv_index.c:545:15: warning: comparison of
integer expressions of different signedness: ‘int’ and ‘unsigned int’
[-Wsign-compare]
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Andreas Schneider <asn@samba.org>
For TDB databases, toggling GUID indexing mode will also toggle
pack format version 2. This provides a convenient downgrade path for
Samba databases, but the process doesn't work for MDB databases because
GUID indexing cannot be disabled when the MDB backend is used. This patch
addresses that corner case by providing support for a pack_format_override
option which will force the database to use pack format version 2.
Signed-off-by: Aaron Haslett <aaronhaslett@catalyst.net.nz>
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Pair-Programmed-With: Andrew Bartlett <abartlet@samba.org>
VERY IMPORTANT PATCH
Now that we have a new packing format, we need to enable it by repacking
the database. We've decided to link all new database features together,
so once GUID indexing is enabled, the database will be repacked with
version 2 format. Repacking is done following the same iterate pattern as
reindexing.
Signed-off-by: Aaron Haslett <aaronhaslett@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Only enable ORDERED_INTEGER and index_format_fn functionality if GUID
indexing is enabled. This is in line with the rest of ldb_kv which binds
the new ORDERED_INTEGER to GUID indexed databases, and allows a practical
way to create the old index format (by disabling the GUID index).
Signed-off-by: Aaron Haslett <aaronhaslett@catalyst.net.nz>
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Garming Sam <garming@catalyst.net.nz>
Pair-Programmed-With: Andrew Bartlett <abartlet@samba.org>
