alexpeshkoff
6cbefa1f46
Introduced class SharedMemory - base class for everything, using IPC. Related to CORE-3035.
2010-06-25 11:55:11 +00:00
asfernandes
75af3ccf65
Misc/cleanup
2010-06-22 00:53:35 +00:00
alexpeshkoff
b6c0bbcd41
frontported fix for CORE-3021: Segfault in DEV_BUILD
2010-06-17 07:34:15 +00:00
asfernandes
e37d097240
Misc
2010-05-22 02:00:45 +00:00
alexpeshkoff
5d9094a563
Frontported - fixed and improved diagnostics when thread failed to start
2010-05-20 12:07:28 +00:00
alexpeshkoff
88b8a40a04
1. Cleaned up most of SUPERSERVER usage.
...
SUPERSERVER has 4 different meanings in our code:
- this is milticlient server (replaced with Config::getMultiClientServer())
- use shared data cache (replaced with Config::getSharedCache())
- use shared metadata cache (replaced with macro SHARED_METADATA_CACHE)
- database are NOT shared between processes (replaced with getSharedDatabase())
2. Use fb_smp_server as both superclassic and classic binary on posix
(xinetd autodecection is done).
3. Small posix build cleanup.
2010-04-15 14:40:27 +00:00
alexpeshkoff
ee2fd4db95
Frontported fix for CORE-2942: Huge (>10Gb) memory consumption by some fb_inet_server processes
2010-03-25 11:02:05 +00:00
asfernandes
77a2706b73
Misc
2010-03-01 02:14:36 +00:00
alexpeshkoff
e5017f1a12
Extend format of firebird configuration file with a set of additional parameters for any parameter.
...
Use traditional firebird configuration files format in all places (intl, trace, etc.).
Add per-database configuration facility to aliases.conf (currently 15 parameters from firebird.conf).
2010-02-28 18:00:51 +00:00
robocop
00c5d74904
Cleanup.
2009-12-13 13:49:43 +00:00
robocop
676dc419ed
Avoid warning.
2009-12-06 11:27:42 +00:00
asfernandes
e99a07bffd
Remove usage of tdbb (and tdbb->tdbb_status_vector) in the lock manager
2009-12-06 01:34:57 +00:00
asfernandes
96c367b7bc
Cleanup
2009-12-05 18:53:08 +00:00
asfernandes
90dfb9a411
Misc
2009-09-30 01:10:11 +00:00
alexpeshkoff
ecbf4312a1
Implemented CORE-2644: Make lock table expandable on HPUX
2009-09-28 16:03:44 +00:00
asfernandes
57c4670972
Misc
2009-09-27 00:54:42 +00:00
hvlad
bc7f2e683a
Fixed bug CORE-2640 : A regular deadlock may not be detected by the lock manager under some conditions, thus resulting in a hang
2009-09-25 18:28:22 +00:00
robocop
aaf86179d8
Style.
2009-08-21 09:03:10 +00:00
robocop
a301a4daf0
Misc.
2009-08-20 11:46:49 +00:00
asfernandes
398975b940
Misc
2009-08-20 01:32:24 +00:00
dimitr
b3843ebca4
Cleanup. New nbackup internals don't need this code anymore.
2009-08-19 08:07:18 +00:00
alexpeshkoff
4fe226c84d
Much better fix for CORE-2590, thanks to Dmitry Yemanov
2009-08-19 07:49:37 +00:00
robocop
e71bd1400d
Comments.
2009-08-14 07:25:09 +00:00
robocop
022895cbde
Style.
2009-08-14 05:51:21 +00:00
dimitr
3f6dd2bceb
Easy fix for CORE-2591: High mutex wait ratio and degraded performance after some time of performing normally.
...
Let's field tests to show us whether we would need a more complex solution in order to utilize the shared memory better.
Also, changed the allocation granularity from 4 to 8 bytes, as our most common locks are equal or longer than 8 bytes now.
2009-08-13 14:35:21 +00:00
robocop
2e3ce0d23d
Misc.
2009-06-07 09:49:58 +00:00
robocop
c6106603df
Misc.
2009-06-05 10:24:39 +00:00
robocop
51d3c9b756
Let's use more typical #ifdef to test macros instead o #if.
2009-05-29 11:27:41 +00:00
robocop
66b1227b22
Unused params, some formatting, cleanup.
2009-05-01 17:21:36 +00:00
robocop
2d7cb94844
Unused params
...
Some BLR checks
2009-04-28 13:08:04 +00:00
dimitr
774213d76c
Misc.
2009-04-13 08:24:39 +00:00
asfernandes
d136f88092
Misc
2009-04-11 19:41:01 +00:00
robocop
0bf3d11247
Misc.
2009-04-10 14:28:35 +00:00
alexpeshkoff
c4f37d6f9b
No sense keeping old value of shared memory pointer when file is already unmapped - thanks to Claudio
2009-04-10 10:35:45 +00:00
alexpeshkoff
ab90217604
Better fix for lock files in CORE-2402
2009-04-08 11:30:15 +00:00
alexpeshkoff
1e7898ea22
Fix for CORE-2402: bad permissions on temp files, /tmp is polluted with a lot of files
2009-04-08 11:00:58 +00:00
robocop
5acf561eb6
Get rid of some typedefs.
2009-04-02 08:53:25 +00:00
robocop
d766c4231d
Simplify the call to ISC_event_wait.
2009-02-28 11:47:11 +00:00
dimitr
8576856193
Fixed CORE-2347: "Deadlock. Page type <N> lock conversion denied" error is reported under a concurrent "connect-work-disconnect" kind of load.
2009-02-27 11:33:27 +00:00
asfernandes
d369e26979
1) Fix posix build
...
2) Remove some warnings
2009-02-15 14:45:37 +00:00
dimitr
4c9413f8a9
Minor optimizations.
2009-02-11 08:26:55 +00:00
asfernandes
f4fbdffe07
Misc
2009-01-29 17:44:21 +00:00
dimitr
939f8e9724
If we write the dump into the globally accessible directory, let's have its name suggesting it belongs to the Firebird family :-)
2009-01-29 17:05:53 +00:00
alexpeshkoff
31a437f986
Choose better place for lock_table.dump
2009-01-29 16:54:58 +00:00
dimitr
1b3927b026
1) Globally named per-database shmem regions.
...
2) Slightly reworked monitoring internals.
3) Encapsulated the event manager, many internal changes there.
4) Various cleanup, etc.
2009-01-28 12:27:18 +00:00
robocop
6406e45337
Shut up compiler warnings (unreferenced vars, uninitialized vars, compile-time conditional code, etc.)
2009-01-02 06:36:12 +00:00
robocop
d11c994960
Style.
2008-12-31 05:06:08 +00:00
robocop
e9be920cc0
Shut up some compiler warnings (some warnings should be enabled again in autoconfig_msvc.h IMHO because hiding our possible mistakes won't help us).
2008-12-30 10:52:52 +00:00
robocop
d39690fbff
Style (and let lines grow up to 100 columns).
2008-12-04 09:29:16 +00:00
hvlad
111a770839
Release lock manager's mutex in its destructor
2008-11-27 20:20:38 +00:00
alexpeshkoff
d8ea120ded
AIX port - use FB_ prefix for alignment constants
2008-11-05 11:26:23 +00:00
dimitr
73b50d0fe0
Misc.
2008-10-24 04:49:15 +00:00
alexpeshkoff
e79f55708d
1. Cleanup - use Firebird::Semaphore instead of local events, removed code
...
related with local events from isc_sync.cpp.
2. Cleanup - removed unused parameter 'number of events' from ISC_event_wait()
and two unused parameters from ISC_event_init().
3. Added SYSV-semaphore based implementation of IPC to isc_sync.cpp.
This should fix CORE-2102: Firebird 2.5 does not build on MacOS (Darwin).
4. Fixed use of posic CS without (x)inetd (i.e. fb_inet_server -s).
5. Fixed (to be reviewed by Dmitry) recovery after some process death in lock manager.
2008-10-10 15:58:05 +00:00
alexpeshkoff
796a2e4b23
Fixed CORE-2064: AV when process is closing under high load
2008-09-04 09:58:20 +00:00
alexpeshkoff
97f89cb5b1
Fixed CORE-2062: error remaping lock file in classic server
2008-09-03 12:00:59 +00:00
asfernandes
077f8360a0
Misc
2008-07-31 00:45:43 +00:00
hvlad
c1b161fc2d
Fixed bug CORE-2000 : Lock manager reported false deadlocks under high load
2008-07-28 21:21:21 +00:00
roman-simakov
cf37531005
Add missed que changing in LockManager::setOwnerHandle
2008-07-23 16:14:54 +00:00
robocop
a302401357
Misc.
2008-07-09 12:34:43 +00:00
hvlad
6ce4216ba3
Fixed bug CORE-1984 : Lock manager reported false deadlocks if one of deadlock participants waits with permitted timeout
2008-07-09 07:57:50 +00:00
hvlad
40f3722373
Fixed bug CORE-1985 : Lock manager code may create 100% CPU load for some time
2008-07-09 07:52:50 +00:00
robocop
8a77b30645
Cleanup: get rid of useless third param in ISC_unmap_file
2008-06-26 11:23:40 +00:00
dimitr
f1907db32d
Fixed a deadlock which could occur during the shared memory region remapping. This bug has been introduced by myself along with the LM rework.
2008-06-12 07:34:50 +00:00
asfernandes
7bb0ae3616
Misc
2008-06-11 01:41:42 +00:00
robocop
7d58d35e15
Misc.
2008-06-10 12:10:03 +00:00
robocop
017d04dbfc
Misc.
2008-06-10 10:11:39 +00:00
dimitr
cb2e3f0f61
Misc.
2008-06-10 04:18:50 +00:00
dimitr
19717fa63f
Misc.
2008-06-09 09:10:55 +00:00
dimitr
1c885a3679
Worked around the compiler warnings.
2008-06-09 08:11:14 +00:00
dimitr
ec1d5a91b4
1) Encapsulated the lock manager and integrated it into the engine.
...
2) Replaced "early" (before LM) checkouts with "late" (inside LM) ones.
3) Removed the mutex protection that's no longer necessary.
Note: the new LM contains some future code (currently commented out) which will be enabled and tested after the Alpha release.
2008-06-07 09:08:16 +00:00
dimitr
92fb3f0b7a
Misc.
2008-05-21 07:32:39 +00:00
dimitr
6180c37e8d
Extra protection to avoid races at the server shutdown.
2008-03-26 15:01:02 +00:00
dimitr
3d47e0c21e
Added thread ID to the owner block (mostly for debugging purposes).
2008-03-25 10:39:29 +00:00
dimitr
9e1d388535
Fixed unexpected crashed in purge_process().
2008-03-21 07:36:11 +00:00
dimitr
83cebeea26
Fixed server crash during memory remapping.
2008-03-18 11:34:50 +00:00
dimitr
ff090e6a7b
Cleanup.
2008-03-12 16:42:51 +00:00
asfernandes
6f91a0f18d
Misc
2008-03-05 02:51:42 +00:00
dimitr
0a3826d5e6
Defer the process block cleanup till its death. This is dumber but more robust. It avoids races at LCK_fini() and fixes a rare deadlock there.
2008-03-04 19:07:55 +00:00
asfernandes
c979d91901
Misc
2008-02-19 02:34:55 +00:00
dimitr
dc4390981c
Sync declaration with definition.
2008-02-14 07:33:50 +00:00
dimitr
b7a575e43f
Cleanup.
2008-02-13 16:47:34 +00:00
dimitr
49d23b6d22
Major changes in the lock manager, targeted to allow mixed in-process/cross-process working modes. Namely, they are:
...
1) Separate process and owner blocks. Multiplex blocking signals via single blocking thread.
2) Enable direct in-process AST delivery.
3) Drop PREVENT_OWNER_STARVATION code as not compatible with the new LM structure.
4) Synchronize LM using its own mutex instead of SCH.
5) Implement wakeup'n'remap code for waiting owners.
6) Add more debugging points for memory remaps.
7) Cleanup and style.
Hopefully, this attempt is more POSIX-friendly than my previous one.
2008-02-11 17:30:23 +00:00
robocop
83dc6bde8f
Constness
2008-02-02 17:07:37 +00:00
dimitr
44be88ad53
Cleanup. singnal_owner() is able to handle the direct call itself.
2008-01-31 13:39:30 +00:00
dimitr
01df1923ec
Misc.
2008-01-30 10:34:28 +00:00
dimitr
132176a475
Major synchronization changes. Description of the commit will be posted to fb-devel.
2008-01-26 12:52:51 +00:00
alexpeshkoff
39896cb955
Fixed CORE-1671: atexit() calls in client libraries cause segfaults
...
and CORE-1079: Every attach of fbclient/fbembed library leaks 64KB of memory
To do it:
- created new template GlobalPtr, controlling access to destructors of global objects.
- applied it to global objects.
Also some related cleanup.
2008-01-23 15:52:40 +00:00
dimitr
956da93e27
Cleanup.
2008-01-17 13:40:35 +00:00
dimitr
83a070b17a
Misc.
2008-01-17 13:10:28 +00:00
alexpeshkoff
bfdf145879
Fixing merge
2008-01-16 13:07:46 +00:00
dimitr
1dd06bb484
The merge continued.
2008-01-16 07:40:12 +00:00
hvlad
2c76551d48
Correct hash function of lock manager. I don't think that using last four bytes of key is a good hash :)
2007-09-25 20:28:20 +00:00
dimitr
dd853cc9e8
Fixed an occasional infinite loop in DEV_BUILD.
2007-08-28 08:38:22 +00:00
alexpeshkoff
d34e4014f7
Bacported fix for CORE-1418: Races when starting and quickly shutting down blocking AST thread
2007-08-20 13:28:01 +00:00
alexpeshkoff
1f77b68e2b
1. Fixed AV when fb_lock_mgr exits.
...
2. Made timeout in LOCK_manager() 10 times smaller - we are not in year 1993.
2007-08-10 08:59:04 +00:00
asfernandes
613faffafb
Misc
2007-08-02 02:00:37 +00:00
robocop
affeac235f
Cleanup.
2007-07-29 02:32:41 +00:00
asfernandes
6beb5e28bd
Misc
2007-07-26 01:23:18 +00:00
alexpeshkoff
b3031d19c2
Added one more check in DEV_BUILD
2007-07-25 13:26:13 +00:00
skidder
04b9ec818b
NBAK synchronization rework. This should fix CORE-1151 and other Nbackup stability issues
2007-04-24 14:05:46 +00:00
