Commit Graph

81 Commits

Author SHA1 Message Date
Jong Hyuk Choi
420210a2a3 IDL cache locking change (mutex -> rwlock) 2003-04-08 06:57:07 +00:00
Howard Chu
c0ee56c306 Define MINIMUM_SEARCH_STACK_DEPTH=8, search_aliases needs at least this. 2003-04-05 10:05:27 +00:00
Howard Chu
48a9cf3416 Added code for BDB_ALIASES. Please test. 2003-04-05 06:35:16 +00:00
Howard Chu
e8c58b4e7f Major API change - (SLAP_OP_BLOCKS) All request parameters are
consolidated into the Operation structure. All reply parameters
are consolidated into the new SlapReply structure. Most operations
now have identical call signatures... Changes are not #ifdef'd,
revert to -r NO_SLAP_OP_BLOCKS if necessary to back out.
2003-03-30 09:03:54 +00:00
Howard Chu
0be6cb5a67 Deleted bi_database_mutex, use DB_LOCK instead. Changed bdb_locker_id to
lookup the threadctx if no Op * is provided. (bdb_db_cache doesn't know
its thread context when it acquires a locker ID.)
2003-03-29 05:15:48 +00:00
Howard Chu
51e10cf23c Save DB_LOCK structure so it can be released later. 2003-03-26 13:33:51 +00:00
Howard Chu
79f3ea6e80 BDB_HIER was defined by accident, undef
Fix up for SLAP_NVALUES_ON_DISK
2003-03-23 15:27:43 +00:00
Howard Chu
a0c58e5f40 Don't use DB_AUTO_COMMIT when opening a database 2003-03-18 00:04:16 +00:00
Howard Chu
44be38c0f0 ITS#2371, fix BDB_SUBDIRs 2003-03-13 15:53:19 +00:00
Howard Chu
18554e7511 Fix attribute/group to allow TXN to abort on lock failure. Save and
restore op->o_do_not_cache on TXN retry, bdb_group will set it if it
fails to get a lock.
2003-02-26 12:09:10 +00:00
Jong Hyuk Choi
55cf8bbd88 use of pre-assigned locker in bdb_group() / bdb_attribute() 2003-02-07 15:47:32 +00:00
Jong Hyuk Choi
85b1783d5c LDAP Sync Operation (draft-zeilenga-ldup-sync) as a groundwork for an LDAP replication design 2003-02-05 07:37:02 +00:00
Kurt Zeilenga
6939c53170 Happy new year 2003-01-03 20:20:47 +00:00
Howard Chu
349e05ff4d Added config keyword "searchstack" for setting search stack cache depth.
Default is still 16.

Added IDL caching, modified from ITS#2182. Uses "idlcachesize" config
keyword. Code is #ifdef'd, currently defined, with default cache of 0.
2002-12-10 20:33:49 +00:00
Howard Chu
d4a422329b From ITS#2183, use a long-lived IDL stack per thread. 2002-12-10 17:59:21 +00:00
Kurt Zeilenga
4e223d588b Reorder header inclusion to resolve ITS#2204. 2002-11-26 20:22:24 +00:00
Jong Hyuk Choi
8074294f1d LCUP persistent search code drop 2002-10-25 17:51:30 +00:00
Howard Chu
104f8cc409 Simpler fix for NO_THREADS 2002-09-19 06:34:53 +00:00
Howard Chu
ff9e580d33 Updated BDB_HIER for current API 2002-09-19 06:08:12 +00:00
Howard Chu
919274432b Fixes for NO_THREADS
Use a per-env locker ID
	Always set lock_detect, since we allow slapadd etc. concurrently
Also removed unused lock_detect_task code. lockdetect config keyword only
needs <policy>, no <seconds> argument.
2002-09-19 01:13:27 +00:00
Howard Chu
35e8b8325d Unifdef BDB_IDL_MULTI. 2002-09-10 04:20:45 +00:00
Howard Chu
5ff509f2d3 Fix previous commit, BDB_REUSE_LOCKERS was inside a BDB version #ifdef
but is not version-dependent.
2002-08-25 11:06:35 +00:00
Howard Chu
d6449b1d57 Experimental code that uses one locker ID per thread. Seems to work OK,
is enabled by default. #undef BDB_REUSE_LOCKERS in back-bdb.h to disable.
Probably needs to be disabled when built with NO_THREADS.
2002-08-24 08:11:08 +00:00
Howard Chu
3b9f4a82ee Added support for BDB 4.1.17 2002-08-22 04:00:06 +00:00
Jong Hyuk Choi
7341dc5db6 redesign of back-bdb locking :
cache entry is locked by Berkeley DB lock primitives.
2002-05-31 20:49:19 +00:00
Jong Hyuk Choi
60b1ee1ba9 Patch for ITS# 1643 2002-04-19 21:41:32 +00:00
Jong Hyuk Choi
efa7b93b23 concurrent cache access: separation of cache lock (reader/writer) and lru lock (mutex) 2002-04-15 16:40:04 +00:00
Kurt Zeilenga
6b29def1c8 switch to openldap-data directory 2002-02-02 06:26:06 +00:00
Kurt Zeilenga
1bf1621efe Add cache txn_prepare code based upon submission (ITS#1575) by
Jong Hyuk Choi <jongchoi@us.ibm.com>.  Needs a bit more work
(GID should be globally unique).
2002-02-02 03:28:32 +00:00
Kurt Zeilenga
b57c11f4bd BDB_ALIASES and BDB_SUBENTRIES cleanup 2002-01-29 04:53:47 +00:00
Kurt Zeilenga
98a2c60c3a Delete CDB (no transactions) support 2002-01-28 22:57:50 +00:00
Kurt Zeilenga
14662be692 Add whoami extended operation.
Add no-op control (needs backend implementation)
Updated modify password extended option API
Kludged control infrastructure to support frontend only controls
2002-01-28 20:25:30 +00:00
Howard Chu
713e6beb8d Entry caching. Based on ITS#1545 from Jong Hyuk Choi, jongchoi@us.ibm.com 2002-01-25 07:19:01 +00:00
Howard Chu
6d58126c34 Added compatibility macros for BDB3/4 lock_get, txn_id. 2002-01-18 13:27:25 +00:00
Howard Chu
8b969479f4 Turn on BDB_IDL_MULTI by default 2002-01-13 22:26:29 +00:00
Kurt Zeilenga
0e2af54a3f Update copyright statements 2002-01-04 21:17:25 +00:00
Howard Chu
dd59babf43 More cleanup 2001-12-29 10:25:19 +00:00
Howard Chu
efe8ac1bf9 Revert previous commit. Checked in by mistake. 2001-12-10 12:25:24 +00:00
Howard Chu
ef0b308bea Changed backglue configuration. Added noSubordinates arg to select_backend
to deal with glued subordinates.
2001-12-10 12:09:40 +00:00
Howard Chu
2721df21d4 Changed be_nsuffix from char ** to struct berval **, removed nsufflen 2001-12-08 05:17:31 +00:00
Howard Chu
505a8693e6 Some fixes for BDB_IDL_MULTI. Experimental back-hdb code. 2001-12-07 12:38:25 +00:00
Howard Chu
1d9818a8da Re-introduce BDB_ID2ENTRY_PAGESIZE. Even if the index databases can get
by on 4K pages, the id2entry database still needs more (16K).
2001-12-07 07:07:55 +00:00
Howard Chu
66266f9c25 Bare minimum changes to support BDB 4.0 2001-12-07 05:05:00 +00:00
Howard Chu
a19a5d8c15 When using BDB_IDL_MULTI, there are no longer any issues with storing
large IDLs. Concurrency is improved by using as small a page size as
possible. So, set the pagesize back to the default 4096 in this case.
2001-12-07 01:40:08 +00:00
Howard Chu
859e283834 Experimental index management. Uncomment BDB_IDL_MULTI in back-bdb.h
to activate. Needs to check limits in bdb_idl_insert_key, will get to
that later.
2001-12-06 13:20:18 +00:00
Howard Chu
295e14bd54 Renamed BDB_ID2ENTRY_PAGESIZE to just BDB_PAGESIZE; set it on each database
instead of just the id2entry db. It helps. I also found that tweaking the
environment (set_lg_bsize 2MB; set_cachesize 2MB) helps but those can be
taken care of in a DB_CONFIG file. Tweaked the bdb_bt_compare function; it
really only needs to be set on little-endian machines. (On big-endian machines
a lexical sort gives the same result as an integer sort.) Moved the final
checkpoint back to the dbenv_close, I think this leaves a cleaner log file.
2001-11-28 20:48:06 +00:00
Howard Chu
24d1ab848c Added experimental support for DB_DIRTY_READ. 2001-11-28 03:11:04 +00:00
Howard Chu
5c7a7998a9 Set the id2entry database pagesize to BDB_ID2ENTRY_PAGESIZE to avoid
overflow/fragmentation. (This is now 16K vs default 4K.) It turns out
that the entries' on-disk format is quite space-inefficient, storing
4 bytes per pointer or integer when typically >50% of those bytes are
zero. Oh well. It's about a 2:1 space increase over ldbm now, vs 4:1
before when all the entries were overflowing the 4K pages.
2001-11-27 12:36:52 +00:00
Howard Chu
b45133c958 Eliminated nextid database. id2entry database is now maintained in numerical
order, so the lastid is always the same as the last entry's ID. This is
an incompatible db file change.
2001-11-27 10:15:23 +00:00
Howard Chu
323689da57 Added "dbnotxn" config keyword. If present, back-bdb uses DB_INIT_CDB
(Concurrent Data Store mode) instead of DB_INIT_TXN. Faster, but tends
to impede writers.
2001-11-27 02:35:20 +00:00