Howard Chu [Sun, 11 Jan 2015 11:37:37 +0000 (11:37 +0000)]
Tweak conditionals for fdatasync hack
Howard Chu [Fri, 9 Jan 2015 11:25:07 +0000 (11:25 +0000)]
ITS#8021 env_sync is invalid in RDONLY env
Howard Chu [Thu, 8 Jan 2015 12:54:14 +0000 (12:54 +0000)]
fdatasync hack, again
Check for ext3/ext4 fs, then check kernel version.
Howard Chu [Thu, 8 Jan 2015 12:17:29 +0000 (12:17 +0000)]
Revert "Note MDB_SAFE_FDATASYNC"
This reverts commit
293d6bb47f7f0d54a45d57ce2be3a9595e806dc0.
Revert "Simpler fdatasync hack"
This reverts commit
9585c012335b2a28fbfdd267bcdeeeea91dc0f08.
Howard Chu [Thu, 8 Jan 2015 11:43:57 +0000 (11:43 +0000)]
Note MDB_SAFE_FDATASYNC
Howard Chu [Thu, 8 Jan 2015 11:42:08 +0000 (11:42 +0000)]
Simpler fdatasync hack
Howard Chu [Thu, 8 Jan 2015 11:28:33 +0000 (11:28 +0000)]
Revert "Fix prev commit for env_sync0"
This reverts commit
985bbbbdd5d64e57f55249ffdeb7c08035b240b2.
Revert "Hack for potential ext3/ext4 corruption issue"
This reverts commit
0018eeb2c3b2239c30def9d47c9d194a4ebf35fe.
Howard Chu [Sun, 21 Dec 2014 00:13:50 +0000 (00:13 +0000)]
Fix prev commit for env_sync0
Howard Chu [Thu, 18 Dec 2014 04:38:53 +0000 (04:38 +0000)]
Hack for potential ext3/ext4 corruption issue
Use regular fsync() if we think this commit grew the DB file.
Howard Chu [Mon, 15 Dec 2014 20:40:59 +0000 (20:40 +0000)]
Fix MDB_APPEND doc
We actually compare the new key against the last key, to prevent
corruption.
Hallvard Furuseth [Sat, 13 Dec 2014 19:35:08 +0000 (20:35 +0100)]
make clean: remove *.lo
Hallvard Furuseth [Mon, 8 Dec 2014 08:56:36 +0000 (09:56 +0100)]
Try to avoid an invalid datafile after failed init
Howard Chu [Mon, 8 Dec 2014 03:21:09 +0000 (03:21 +0000)]
Clarify mdb_dbi_open doc
The concurrency restriction is on threads within a single process.
Multiple processes can use mdb_dbi_open without affecting each other.
Hallvard Furuseth [Fri, 5 Dec 2014 20:00:21 +0000 (21:00 +0100)]
C90-compatible code for SysV semaphores
Leo Yuriev [Fri, 5 Dec 2014 19:30:31 +0000 (19:30 +0000)]
ITS#7987 fix excessive space for single write txn
Howard Chu [Fri, 5 Dec 2014 19:15:41 +0000 (19:15 +0000)]
ITS#7994 fix prev commit
Strip docs promising behaviors that we don't promise to hold.
Strip MDB_txnid_t typedef.
David Barbour [Fri, 5 Dec 2014 17:18:30 +0000 (11:18 -0600)]
ITS#7994 Access to current transaction ID.
I, David Barbour, hereby place the following modifications to OpenLDAP
Software (and only these modifications) into the public domain. Hence,
these modifications may be freely used and/or redistributed for any
purpose with or without attribution and/or other notice.
Hallvard Furuseth [Fri, 5 Dec 2014 17:18:53 +0000 (18:18 +0100)]
Fix SysV semaphores - repair mti_txnid.
Similar to
443a7e40c4280cfd07cbf80c5de62ea072b136e3.
Hallvard Furuseth [Wed, 3 Dec 2014 11:37:55 +0000 (12:37 +0100)]
For ITS#7789: Ensure mapsize >= pages in use.
Check new mapsizes against mm_last_pg. Move
mdb_env_init_meta0() so it can set mm_last_pg earlier.
Hallvard Furuseth [Wed, 3 Dec 2014 11:27:53 +0000 (12:27 +0100)]
Fix WIN32 -> _WIN32
Hallvard Furuseth [Mon, 1 Dec 2014 07:59:29 +0000 (08:59 +0100)]
mdb_strerror: Silence warning @ Windows
Hallvard Furuseth [Mon, 1 Dec 2014 07:59:29 +0000 (08:59 +0100)]
Fix comments
Hallvard Furuseth [Mon, 1 Dec 2014 07:59:29 +0000 (08:59 +0100)]
Clean up SysV semaphore code.
Change MDB_LOCK_FORMAT to differ from Posix semaphores/mutexes.
Do not use array/member length 0 for MDB_txninfo.mt2.
Hallvard Furuseth [Mon, 1 Dec 2014 07:59:29 +0000 (08:59 +0100)]
Fix robust mutexes - repair mti_txnid.
Broken by
e3b6c359a935283c21e1fd7d03b790d87d53b933, if
commit() crashes in write_meta before setting mti_txnid.
Hallvard Furuseth [Mon, 1 Dec 2014 07:59:29 +0000 (08:59 +0100)]
Simplify recent changes a bit.
Hallvard Furuseth [Sat, 22 Nov 2014 21:56:31 +0000 (22:56 +0100)]
ITS#7961,#7987 Re-fix txn init.
More fallout from
4d02c741b120786df1b87ee9ed49c1d3f9bc7522:
Don't modify another thread's write txn. Reinit me_txn0 fully.
leo@yuriev.ru [Fri, 17 Oct 2014 22:46:53 +0000 (22:46 +0000)]
ITS#7971 LMDB: clarification in mdb_txn_renew0().
leo@yuriev.ru [Fri, 17 Oct 2014 22:35:41 +0000 (22:35 +0000)]
ITS#7970 LMDB: Critical Heisenbug
Inconsistent reading & SIGSEGV due to the race condition.
leo@yuriev.ru [Fri, 17 Oct 2014 22:22:39 +0000 (22:22 +0000)]
ITS#7969 LMDB: volatile & __synchronize().
Globally shared fields of meta-data are not 'volatile'.
Howard Chu [Wed, 8 Oct 2014 18:25:27 +0000 (19:25 +0100)]
ITS#7959 fix prev commit
fstat returns -1, not an errno. Move code into its own function
for reuse.
Howard Chu [Sat, 4 Oct 2014 20:48:49 +0000 (21:48 +0100)]
ITS#7961 fix txn init
More fallout from
4d02c741b120786df1b87ee9ed49c1d3f9bc7522
Jean-Christophe DUBOIS [Fri, 3 Oct 2014 18:24:25 +0000 (20:24 +0200)]
ITS#7959 Check fstat return value
Signed-off-by: Jean-Christophe DUBOIS <jcd@tribudubois.net>
Leo Yuriev [Thu, 4 Sep 2014 20:19:16 +0000 (00:19 +0400)]
ITS#7841 trivial cleanup
TRIVIA - lmdb: clean testdb-dir while "make test".
Howard Chu [Thu, 2 Oct 2014 21:11:43 +0000 (22:11 +0100)]
ITS#7956 fix compact of empty env
Howard Chu [Thu, 2 Oct 2014 20:58:20 +0000 (21:58 +0100)]
ITS#7955, #7671 fix MDB_PREV_DUP
Howard Chu [Thu, 2 Oct 2014 20:35:15 +0000 (21:35 +0100)]
ITS#7953 .gitignore *.lo
Howard Chu [Thu, 2 Oct 2014 20:33:35 +0000 (21:33 +0100)]
ITS#7952 .gitignore mdb_dump/load
Howard Chu [Wed, 1 Oct 2014 20:00:03 +0000 (21:00 +0100)]
Don't use -fPIC for static lib
Howard Chu [Wed, 1 Oct 2014 19:58:15 +0000 (20:58 +0100)]
Use robust mutexes by default
Making it optional on pthreads just complicates things;
they're always robust on other implementations
Howard Chu [Sun, 20 Jul 2014 18:08:55 +0000 (11:08 -0700)]
Use SysV semaphores instead of POSIX
Since they can cleanup after themselves on process exit
Howard Chu [Sun, 20 Jul 2014 15:10:34 +0000 (08:10 -0700)]
Must do dlist_free before releasing Wmutex
Hallvard Furuseth [Sun, 20 Jul 2014 14:42:04 +0000 (16:42 +0200)]
MDB_ROBUST: Use MDB_PID_T. Drop a wmutex variable.
Howard Chu [Wed, 16 Jul 2014 16:24:45 +0000 (09:24 -0700)]
Free write mutex earlier in txn_reset0
Hallvard Furuseth [Wed, 16 Jul 2014 14:15:58 +0000 (07:15 -0700)]
Add MDB_ROBUST
Howard Chu [Thu, 25 Sep 2014 18:25:56 +0000 (19:25 +0100)]
Android also only has memalign
Howard Chu [Sat, 20 Sep 2014 06:24:32 +0000 (07:24 +0100)]
new date
Howard Chu [Fri, 19 Sep 2014 00:45:43 +0000 (01:45 +0100)]
ITS#7943 reinit txn flags
When reusing the main write txn
Howard Chu [Thu, 18 Sep 2014 00:22:02 +0000 (01:22 +0100)]
Plug leak in
4d02c741b120786df1b87ee9ed49c1d3f9bc7522
Free writer txn on env_close
Howard Chu [Tue, 16 Sep 2014 20:54:07 +0000 (21:54 +0100)]
Silence some clang warnings
Howard Chu [Mon, 15 Sep 2014 18:48:32 +0000 (19:48 +0100)]
Release 0.9.14
Quanah Gibson-Mount [Sat, 13 Sep 2014 21:21:17 +0000 (16:21 -0500)]
Change retry to num times 60. Testing shows that on a known dataset, this has the same growth behavior as 2.4.39, while num times 20 resulted in significant growth.
Howard Chu [Tue, 9 Sep 2014 18:44:23 +0000 (19:44 +0100)]
More for me_pgoldest
Leave it in main env struct, it doesn't need to be rolled back
with nested txns.
Howard Chu [Tue, 9 Sep 2014 18:40:05 +0000 (19:40 +0100)]
Remember oldest reader txnid
Don't walk the readers table unless we really need to.
Howard Chu [Thu, 4 Sep 2014 09:40:11 +0000 (10:40 +0100)]
Use a single write txn
Howard Chu [Sun, 24 Aug 2014 20:18:03 +0000 (21:18 +0100)]
ITS#7913 cleanup
we already checked for data == NULL, drop unnecessary if
Howard Chu [Fri, 22 Aug 2014 00:40:20 +0000 (17:40 -0700)]
More for MIPS
Move CACHEFLUSH, avoid unnecessary if()
Howard Chu [Mon, 18 Aug 2014 03:44:53 +0000 (20:44 -0700)]
Partially revert
ada26306992babeccad0c5443fbd76e28dd24cd6
Of course, MacOSX only has posix_memalign, and only in recent versions.
(Older versions could use valloc instead. That precludes ever changing
the LMDB pagesize to be different from the OS pagesize.) Not all that
crucial since it only affects mdb_env_copyfd1.
Hallvard Furuseth [Sat, 16 Aug 2014 23:50:32 +0000 (01:50 +0200)]
Microoptimize loose pages
Hallvard Furuseth [Sat, 16 Aug 2014 22:29:12 +0000 (00:29 +0200)]
Fix loose pages in mdb_freelist_save().
Leaving them in dirty_list caused breakage. Instead merge them
into me_pghead at the end, when no more pages will be allocated.
Hallvard Furuseth [Sat, 16 Aug 2014 09:34:44 +0000 (11:34 +0200)]
Split out mdb_midl_xmerge, add MDB_IDL_ALLOCLEN
Howard Chu [Mon, 11 Aug 2014 18:01:20 +0000 (11:01 -0700)]
ITS#7917
Howard Chu [Mon, 11 Aug 2014 18:00:11 +0000 (11:00 -0700)]
ITS#7917 fix mdb_dbi_open
Don't let dummy go out of scope
Howard Chu [Thu, 7 Aug 2014 18:06:53 +0000 (11:06 -0700)]
Just use memalign, not posix_memalign
Everything has it. Solaris doesn't have posix_memalign.
Howard Chu [Mon, 4 Aug 2014 18:40:24 +0000 (11:40 -0700)]
More for MIPS
Do a cacheflush after writing data pages, separate from meta's cacheflush
Howard Chu [Mon, 4 Aug 2014 16:26:35 +0000 (09:26 -0700)]
ITS#7793 more for branch key update
Howard Chu [Mon, 4 Aug 2014 12:13:29 +0000 (05:13 -0700)]
ITS#7793 doc update, again
Howard Chu [Mon, 4 Aug 2014 11:57:53 +0000 (04:57 -0700)]
ITS#7793 update branch key if needed
Howard Chu [Sat, 2 Aug 2014 17:33:35 +0000 (10:33 -0700)]
More for MIPS
Paranoia for
3630066843b7ca6b2cd12911d3e2fe3314cd4549 do the
cacheflush before setting mti_txnid.
Howard Chu [Thu, 31 Jul 2014 16:59:52 +0000 (09:59 -0700)]
ITS#7913 mdb_cursor_get(*_DUP) cleanup
FIRST_DUP/LAST_DUP should succeed even on non-duplicate values
PREV_DUP should reset cursor's EOF flag.
Howard Chu [Sun, 27 Jul 2014 01:16:02 +0000 (18:16 -0700)]
Fix MIPS cache coherency on Linux
MIPS chips require manual control of on-chip caches. The cacheflush
syscall being used here only exists on MIPS Linux, other OSs will
require revisiting.
Howard Chu [Thu, 24 Jul 2014 11:15:45 +0000 (04:15 -0700)]
Update date
Howard Chu [Thu, 24 Jul 2014 11:13:55 +0000 (04:13 -0700)]
more for mdb_cursor_count
return MDB_NOTFOUND if the cursor is empty
Hallvard Furuseth [Wed, 9 Jul 2014 19:38:45 +0000 (21:38 +0200)]
Fix mdb_copy & co when MDB pagesize < OS pagesize
Howard Chu [Thu, 24 Jul 2014 00:31:34 +0000 (17:31 -0700)]
ITS#7789 doc updates
Hallvard Furuseth [Thu, 24 Jul 2014 00:24:31 +0000 (17:24 -0700)]
ITS#7789 Fix mdb_env_write_meta().
Always write mapsize = max(other metapage:mapsize, env:mapsize).
Howard Chu [Thu, 24 Jul 2014 00:10:41 +0000 (17:10 -0700)]
Revert ITS#7789 commits, let's try this again.
Revert "ITS#7789 Fix resize vs MDB_NOMETASYNC, and a comment."
Revert "ITS#7789 more"
Revert "Revert "ITS#7789 update RESIZED errmsg text""
Revert "ITS#7789 update RESIZED errmsg text"
Revert "ITS#7789 persist mapsize changes"
This reverts commit
02285aca58f5629547263ba09e7dd685dcf6b4b5.
This reverts commit
38e71c68de2ad61fb0fd72fd40214fd9a94479b9.
This reverts commit
a5e4eecb1e59e9645113bb4c63d06a23a2c4bdf9.
This reverts commit
d8e18551c3f03d5205f4dc8d1711e494d3814b17.
This reverts commit
5926e54bba7c976e0837221efb20cbff384207b8.
Howard Chu [Tue, 22 Jul 2014 01:20:31 +0000 (18:20 -0700)]
Fix env_open mode bits
Howard Chu [Mon, 21 Jul 2014 15:33:59 +0000 (08:33 -0700)]
Fix mdb_cursor_count, reject uninit'd cursor
Howard Chu [Mon, 21 Jul 2014 15:11:33 +0000 (08:11 -0700)]
Silence prototype warnings
Howard Chu [Mon, 14 Jul 2014 16:53:53 +0000 (09:53 -0700)]
Typo
Howard Chu [Mon, 14 Jul 2014 16:51:16 +0000 (09:51 -0700)]
ITS#7896 fix for mdb_env_map
Don't pass non-NULL map address unless MDB_FIXEDMAP was used.
Howard Chu [Wed, 9 Jul 2014 22:03:54 +0000 (15:03 -0700)]
More for cursor EOF
Must also set in sub-cursor if deleting entire subDB
Howard Chu [Wed, 9 Jul 2014 19:05:37 +0000 (12:05 -0700)]
Zero-init copy2 writebuf
Hallvard Furuseth [Wed, 9 Jul 2014 18:18:08 +0000 (20:18 +0200)]
MDB_MAXKEYSIZE = 0 (dynamic value) when MDB_DEVEL
Hallvard Furuseth [Wed, 9 Jul 2014 18:17:13 +0000 (20:17 +0200)]
MDB_DEVEL doxygen fix
Hallvard Furuseth [Wed, 9 Jul 2014 18:16:33 +0000 (20:16 +0200)]
ITS#7789 Fix resize vs MDB_NOMETASYNC, and a comment.
Howard Chu [Wed, 9 Jul 2014 17:22:25 +0000 (10:22 -0700)]
Windows fixes
Always set the filesize when opening for writes. Otherwise
can't use backups from mdb_copy.
Howard Chu [Wed, 9 Jul 2014 16:31:49 +0000 (09:31 -0700)]
Windows compat
We mainly use Win32 functions, so cannot just use C-runtime strerror
to return error messages. We have to use Win32-specific messages.
Unfortunately, we document the API to return C-runtime error codes,
so we can't just switch to all Win32 error codes.
Howard Chu [Wed, 9 Jul 2014 16:16:03 +0000 (09:16 -0700)]
Fix loading of printable dump
Off-by-one in end check, would complain on valid input.
Howard Chu [Wed, 9 Jul 2014 13:54:32 +0000 (06:54 -0700)]
ITS#7789 more
persist size changes from env_open() as well. This was the
original behavior.
Howard Chu [Wed, 9 Jul 2014 13:17:54 +0000 (06:17 -0700)]
Revert "ITS#7789 update RESIZED errmsg text"
This reverts commit
d8e18551c3f03d5205f4dc8d1711e494d3814b17.
And partially reverts mapsize check. Only mapsize increases are relevant.
Howard Chu [Wed, 9 Jul 2014 12:13:10 +0000 (05:13 -0700)]
ITS#7825 more
Set read txn's dbiseqs to env's. Set child txn's dbiseqs to parent's.
Simplify DBI_CHANGED test, no need to check md_name. No-op dbi_close
of already closed handle.
Howard Chu [Wed, 9 Jul 2014 04:40:11 +0000 (21:40 -0700)]
Updates #7789, #7825
Howard Chu [Wed, 9 Jul 2014 04:31:48 +0000 (21:31 -0700)]
ITS#7825 more
only update dbiseq in dbi_open on write txns
Howard Chu [Wed, 9 Jul 2014 04:22:07 +0000 (21:22 -0700)]
ITS#7825 tweak prev commit
dbiseqs are only checked in write txns so omit from read txns
Howard Chu [Wed, 9 Jul 2014 04:02:40 +0000 (21:02 -0700)]
ITS#7825 add MDB_BAD_DBI error code
Use DBI sequence numbers to detect DBIs being closed while in use.
Howard Chu [Wed, 9 Jul 2014 03:05:31 +0000 (20:05 -0700)]
ITS#7789 update RESIZED errmsg text
Howard Chu [Wed, 9 Jul 2014 02:06:45 +0000 (19:06 -0700)]
ITS#7789 persist mapsize changes
Write decreases too, not just increases. Check for any size change
that was not requested by this process.
Howard Chu [Tue, 8 Jul 2014 23:27:26 +0000 (16:27 -0700)]
Update LMDB version date/info
Howard Chu [Tue, 8 Jul 2014 23:27:03 +0000 (16:27 -0700)]
ITS#7892 Fix MacOSX section name