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.
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>
Right now we don't do transactions with mutexed tdbs. tdbtorture -m
locks up. I haven't really investigated why that is the case. The lockup
confused me quite a bit until I figured out it works fine as long as it
does not do transactions, which is all we need right now.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andreas Schneider <asn@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>
Try to store a record for which the (circular) freelist does not have
any entry.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
This just walks tdb_find by searching for a nonexistent record
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@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>
Simple bash test for readonly locks on tdbbackup:
1. Running tdbbackup on a database with and without readonly locks enabled.
2. Dump both backups and original.
3. Check all three dumps match.
A binary sample_tdb.tdb file is included for the test because the existing
sample tdbs in lib/tdb/test are either corrupt or empty.
Signed-off-by: Aaron Haslett <aaron.haslett@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
The netcmd 'domain backup offline' command will use the tdbbackup tool but
require readonly locking of tdb databases, otherwise all database access would
be blocked during a backup. This patch adds the option. A backup script
should use this tool with the readonly locks option after taking a transaction
lock on the target database.
Signed-off-by: Aaron Haslett <aaronhaslett@catalyst.net.nz>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Gary Lockyer <gary@catalyst.net.nz>
* Fix build on AIX
* Python3 compatibility fixes
* Use tdb_wipe_all in "erase" command
* Harden allocating the tdb recovery area
* Make sure the hash size fits
* Harden tdb_check_used_record against overflow
* Harden tdb_rec_read
* Handle TDB_NEXT_LOCK_ERR in tdb_traverse_internal
* Fix build warnings
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
BUG: https://bugzilla.samba.org/show_bug.cgi?id=13493
Here is the build error on AIX 7.1.
../../lib/tdb/tools/tdbtool.c:39:12: error: 'disable_lock' redeclared as different kind of symbol
static int disable_lock;
^~~~~~~~~~~~
In file included from /usr/include/sys/gfs.h:24:0,
from /usr/include/sys/vfs.h:27,
from ../../lib/replace/system/filesys.h:48,
from ../../lib/tdb/tools/tdbtool.c:26:
/usr/include/sys/lock_def.h:314:5: note: previous declaration of 'disable_lock' was here
int disable_lock(int,simple_lock_t);
^~~~~~~~~~~~
Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
In py3, iterxxx methods are removed.
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>
This is a lot quicker on large, fragmented databases. tdb_delete can
leave the freelist in a fragmented mess.
Also, it's a lot more robust: I've got a 4GB tdb file that was affected
by the problem fixed with c7211882a79. These databases have large space
at the end that is not part of any record or freelist
entry. tdb_wipe_all converts this space into a freelist entry. One
downside is that with those broken databases (which should not happen
after c7211882a79) have unallocated blocks in their file range after
this operation.
I think the speed advantage outweighs this disadvantage.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Jeremy Allison <jra@samba.org>
Signed-off-by: Noel Power <noel.power@suse.com>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
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
* Add protection against EINTR.
* Truncate the file after expand failure, ENOSPC
* Use posix_fallocate() to expand the file
* Fix GCC compiler warnings
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>
Autobuild-User(master): Stefan Metzmacher <metze@samba.org>
Autobuild-Date(master): Thu Aug 24 21:17:48 CEST 2017 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>
Without this it's very easy to create virtually huge files: ftruncate expands a
file, the pwrites fail with ENOSPC, thus the write fails. The next writer runs
into the same situation, and ftruncate-expands the file even further. tdb_check
will then spend ages reading the 4GB of zeros byte by byte.
Here we hold the freelist lock or are inside a transaction, so it is safe to
cut the file again. Nobody can have used the space that we have tried to
allocate, so we can't have any stray pointers corrupting the database.
Signed-off-by: Volker Lendecke <vl@samba.org>
Reviewed-by: Andrew Bartlett <abartlet@samba.org>
