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.
With locking.tdb now based on g_lock.c code, we change locking.tdb a
lot more often. I have a customer case where LDX tortures smbd very
hard with 800+ concurrent connections, which now completely falls over
where 4.12 still worked fine. Some debugging showed a thundering herd
on fcntl locking.tdb index 48 (TDB_SEQNUM_OFS). We still use fcntl for
the seqnum, back when we converted the chainlocks to mutexes we did
not consider it to be a problem. Now it is, but all we need to do with
the SEQNUM is to increment it, so an __atomic_add_fetch() of one is
sufficient.
I've taken a look at the C11 standard atomics, but I could not figure
out how to use them properly, to me they seem more general to be
initialized first etc. All we need is a X86 "lock incl 48(%rax)" to be
emitted, and the gcc __atomic_add_fetch seems to do this.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Solaris Studio compiler 12.4 is pedantic about prototypes in headers having
the external visibility declarations too. It throws errors like:
redeclaration must have the same or more restrictive linker scoping: ...
Signed-off-by: Bjoern Jacke <bjacke@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
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 Jan 23 20:41:46 UTC 2020 on sn-devel-184
When you set
in tdbtorture.c to make it more similar to locking.tdb use,
bin/tdbtorture -m -n 1 -l 100000 -s
becomes twice as fast. This is a pretty extreme case, but all other
tests that I did improve significantly as well.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is common between both implementations of tdb_oob(). It's
faster if we don't have to dereference function pointers.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
tdb_oob() will become a public function encapsulating the pointer
dereferences.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Fix the ubsan warning
lib/tdb/common/tdb.c:184:9: runtime error: null pointer passed as
argument 2, which is declared to never be null"
memcmp call now guarded by a length check.
memcmp returns zero when called with a zero length parameter.
Signed-off-by: Gary Lockyer <gary@catalyst.net.nz>
Reviewed-by: Noel Power <npower@samba.org>
Autobuild-User(master): Noel Power <npower@samba.org>
Autobuild-Date(master): Mon Jul 1 14:50:54 UTC 2019 on sn-devel-184
Fixes:
lib/tdb/common/rescue.c:299:2: warning: Null pointer passed as an argument to a 'nonnull' parameter <--[clang]
qsort(found.arr, found.num, sizeof(found.arr[0]), cmp_key);
^ ~~~~~~~~~
Signed-off-by: Noel Power <noel.power@suse.com>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
Fixes:
lib/tdb/common/transaction.c:354:2: warning: Null pointer passed as an argument to a 'nonnull' parameter <--[clang]
memcpy(tdb->transaction->blocks[blk] + off, buf, len);
&
Signed-off-by: Noel Power <noel.power@suse.com>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
Fixes:
lib/tdb/common/lock.c:933:6: warning: Dereference of null pointer <--[clang]
if (tdb->allrecord_lock.count) {
^~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Noel Power <noel.power@suse.com>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
Fixes
lib/tdb/common/freelistcheck.c:96:3: warning: Value stored to 'last_ptr' is never read <--[clang]
last_ptr = rec_ptr;
^ ~~~~~~~
Signed-off-by: Noel Power <noel.power@suse.com>
Reviewed-by: Gary Lockyer gary@catalyst.net.nz
Autobuild-User(master): Noel Power <npower@samba.org>
Autobuild-Date(master): Tue Jun 11 13:31:01 UTC 2019 on sn-devel-184
The call to tdb_repack() inside tdb_transaction_commit()
is an optimization, not part of the transaction itself,
so failing due to lock or other errors isn't a fatal error
that should cause the caller to think the transaction was
a failure by returning -1.
The tdb transaction itself has finished and been committed
onto stable storage via fsync and all locks released at the
point tdb_repack() is called.
tdb_repack() is only called here as it's a convenient point
to attempt to reduce tdb fragmentation without having to add
a timer call to repack in all users of tdb.
This causes lock ordering issues in Samba, showing up as:
ldb: ltdb: tdb(../private/sam.ldb.d/DC=SAMBA2008R2,DC=EXAMPLE,DC=COM.ldb): tdb_transaction_prepare_commit: failed to upgrade hash locks: Locking error
This is because Samba has multiple tdb databases open, and the lock order between them
is important.
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13952
Signed-off-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
tdb->max_dead_records is "int", as is the corresponding parameter to
tdb_set_max_dead(). Not that a signed variable makes any sense, but
this is old code and tdb_set_max_dead() is a public API which we
should not change for this.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Tue Nov 6 21:52:32 CET 2018 on sn-devel-144
This is a prerequisite to allow gencache to run on a non-transactioned
database with mutexes.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This is a lightweight readonly traverse of a single chain, see the
comment in the header file.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Before this patch we had 3 loops walking a hash chain to delete
records:
tdb_do_delete() to find the predecessor of the record that was to be
deleted. tdb_count_dead(), the name says it all and tdb_purge_dead()
to give back all dead records from a chain to the freelist.
This patch introduces tdb_trim_dead that walks a hash chain just
once. While it does so it counts the number of dead records, and all
records beyond tdb->max_dead_records are moved to the freelist.
Normal record deletion now works by always marking a record as dead in
step 1 and then calling tdb_trim_dead. This is made safe against
circular chains by doing the slow chain walk only in the case when we
did not delete a dead record during our walk.
It changes our dynamics a bit:
When deleting a record with non-zero max_dead_records, now we always
leave that number of records around when deleting, doing a blocking
lock on the freelist when we found too many dead records.
Previously when exceeding max_dead_records we wiped all dead records
to start accumulating them from scratch, assuming we could lock the
freelist in a nonblocking fashion.
The net effect for an uncontended freelist is the same: In
tdb_allocate() we still completely hand over all dead records to the
freelist when we could lock it, it just happens later than without
this patch.
This means for a lightly loaded system we will potentially leave more
dead records around in databases like locking.tdb. However, on a
heavily loaded system we become more predictable: If the freelist is
so heavily contended that across many deletes we can't get hold of it,
previously we accumulated more dead records than max_dead_records
would allow. This is a really lowlevel tradeoff that is likely hard to
measure, but to me becoming more deterministic without sacrificing too
much parallelism (we keep more dead records around) is worth trying.
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 Oct 30 02:48:38 CET 2018 on sn-devel-144
The next commit will change the handling of dead records, removing the
"tdb_do_delete" function. As traverses should not happen in normal
operations, dead records from them should be rare, and relying on
traverses to remove them is a very bad idea IMHO.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This has not been activated by default for ages and can be very
inefficient. With check_merge_with_left_record() we have an
alternative that will merge freelist records while we walk it
anyway. This has reduced fragmentation significantly
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@samba.org>
We can't really do the full check while the freelist is modified on the
fly. As long as we don't merge any freelist entries, we should be good
to apply this check.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This captures the tdb_rescue protection against circular hash chains
with a slow pointer updated only on every other record traverse
If a hash chain has a loop, eventually the next_ptr
will cycle around and be identical to the 'slow' pointer.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Many of those warnings are difficult to fix, but this one was easy :-)
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 Mar 22 07:21:44 CET 2018 on sn-devel-144
This should be significantly faster than pwriting.
openbsd doesn't have posix_fallocate, so we do need the fallback. Also, it
might have weird failure modes, so we keep the old code in place except for
posix_fallocate returning success or ENOSPC.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Autobuild-User(master): Jeremy Allison <jra@samba.org>
Autobuild-Date(master): Thu Aug 24 05:38:49 CEST 2017 on sn-devel-144
More README.Coding, but I need "ret" in the next commit as well :-)
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
