Commit Graph

215 Commits

Author SHA1 Message Date
Hallvard Furuseth
b99cb20483 mdbx: minor fix mdb_page_list() message.
Change-Id: I25697a7123aef182762a91536e54ca118d6a0958
2016-08-25 17:34:07 +03:00
Leo Yuriev
5239bce5e5 mdbx: adds minor assert.
More for ae92d0d671.

Change-Id: I32f8f54ec5888fe794b83d42ed7f5ae0138070d4
2016-08-24 11:52:36 +03:00
Leo Yuriev
ae92d0d671 mdbx: fixes issues detected by PVS-Studio.
See http://www.viva64.com/en/b/0422/

This is Partial for https://github.com/ReOpen/ReOpenLDAP/issues/107

Change-Id: Ic47d3cb1ff88ca0807bdd692a1e01487c70b264c
2016-08-22 14:39:45 +03:00
Leo Yuriev
bc7b9984d3 mdbx: fix copyright timestamps.
1 of 2 for https://github.com/ReOpen/libmdbx/issues/3

Change-Id: I7a0b820e092e7b0064ae22341bebcd015e37b7c7
2016-08-22 13:46:40 +03:00
Leo Yuriev
a306a1eac4 mdbx: fix assertions when debug enabled for various open/sync modes.
No bug is here, but had a wrong condition for assertion-check.

Change-Id: I6db76932e235f8212543fdd160a15cc9299d79e8
2016-08-18 17:09:09 +03:00
Leo Yuriev
7cfcc9ff7a mdbx: check assertions depending on NDEBUG.
Change-Id: I0c8ce98242e7451dd82f8abfbac70b2f4dcfa7a0
2016-08-18 17:09:09 +03:00
Hallvard Furuseth
ac09708bfe mdbx-doc: comment MDB_page, rename mp_ksize.
Change-Id: Ia9d28a76fe0b51ccfc40ba97b913f064b81bc00e
2016-08-04 10:44:39 +03:00
Hallvard Furuseth
85612d9722 mdbx-doc: VALID_FLAGS, mm_last_pg, mt_loose_count.
Change-Id: I6c615244d8b137ac4ed67fb7be70e8ccd4e7cac4
2016-08-04 10:44:33 +03:00
Hallvard Furuseth
f627930bdc mdbx: backport - Refactor mdb_page_get().
Change-Id: I6a44000d954025f87637c8b60da85b210b4df65e
2016-07-26 17:13:06 +03:00
Leo Yuriev
a3a98a7a2e mdbx: fix 'mdb_copy' init warning.
Change-Id: Ifc3d0b565a0dd6d6da40e7fe2be2c9ff0f5458fb
2016-07-20 12:51:52 +03:00
Hallvard Furuseth
dd98ab22da mdbx: backport - Add error MDB_PROBLEM, replace some MDB_CORRUPTED.
When problem is most likely in txn, not on disk.

Change-Id: Ie01d9eb32e8f250f6dd98f3fe84c38ed15593a2e
2016-07-18 18:34:59 +03:00
Hallvard Furuseth
3befcdab01 mdbx: backport - ITS#8209 fix MDB_CP_COMPACT.
Handle errors.  Fix cond_wait condition so mc_new
is the sole control var.  Drop specious cond_waits.
Do not look at 'mo' while copythr writes it.

Preserve DB flags (use metapage#1) when main DB is empty.
Fail if metapage root != actual root in output file.

Some _aligned_malloc() doc seems to think arg NULL = user error.
Don't know if posix_memalign() pointer is defined after failure.

Change-Id: Idfdc118b4848bb96bace0f29db9dcdd710b7a1f4
2016-07-18 18:34:07 +03:00
Leo Yuriev
6d99bb59a8 mdbx: use __GNUC_PREREQ and __GNUC_PREREQ.
Change-Id: Ib1927f6ab2e3397fb2a0e83a73c5c0252bc8a593
2016-06-09 15:23:47 +03:00
Leo Yuriev
9c5d029417 mdbx: clarify copyright and license.
Change-Id: Ib1b9516ce09497559937f749d520719046c5b7e6
2016-06-09 13:21:20 +03:00
Leo Yuriev
0e90ea1d8d mdbx: separate MDBX's features.
Change-Id: I1964ee5d8b8e5fca170b8e955bfc1a6efe25f6c5
2016-06-09 08:54:46 +03:00
Hallvard Furuseth
f5c61018ba mdbx: backport - mdb_env_setup_locks() Plug mutexattr leak on error.
Change-Id: Iac82f957e4961ec964d9d0a75ef0b24a9f68929c
2016-06-09 04:18:55 +03:00
Howard Chu
2806453b54 mdbx: backport - ITS#8339 Solaris 10/11 robust mutex fixes.
Check for PTHREAD_MUTEX_ROBUST_NP definition (this doesn't work
on Linux/glibc because they used an enum). Zero out mutex before
initing.

Change-Id: Ic618a6a72fbd7680dd76d5fd0aef06545dcba994
2016-06-09 03:11:05 +03:00
Leo Yuriev
3deb4577fb mdbx: fix build by clang (missing-field-initializers).
Change-Id: I45b966a6572d0669596106c3c70862dc113fb642
2016-06-06 01:53:00 +03:00
Leo Yuriev
f943bbf91c mdbx: more for ASAN.
Change-Id: I2bd0573cfe4c7822fa8f3cf94db60b684964a40d
2016-05-19 22:05:46 +03:00
Howard Chu
856f30b52c mdbx: backport - ITS#8424 init cursor in mdb_env_cwalk.
Change-Id: I979b6f492c56d1c61ce233727ff058df48f327d1
2016-05-16 17:46:45 +03:00
Leo Yuriev
09d7904317 mdbx: clarify ov-pages copying in cursor_put().
Change-Id: I48ae57579a7d68178ec1857785ffdd6f0c0f7e13
2016-05-10 13:21:38 +03:00
Hallvard Furuseth
1ef0106b51 mdbx: backport - Comment ovpage code in mdb_cursor_put().
Change-Id: I6b3bff87dc49135fdda3e9ce0a505822ab26d3d5
2016-05-10 13:15:05 +03:00
Leo Yuriev
7c3c28239c mdbx: more likely/unlikely for mdb_node_add.
Change-Id: I1f241c6cea55db6fed744ba9f6f5733f6622015f
2016-05-08 02:29:59 +03:00
Leo Yuriev
4d3e349c16 mdbx: don't memcpy when src eq dest.
Workaround for 'destination overlaps source' from Valgrind.

Change-Id: Ib12eacbd224057334311816346ea0ccece2b42cf
2016-05-07 23:44:41 +03:00
Leo Yuriev
8791e3243f mdbx: workaround for pthread_setspecific's memleak.
Change-Id: I8b48548a40bb2fe58db14b8f6944f5a3802ffbb9
2016-05-07 15:21:59 +03:00
Leo Yuriev
14b3afdcff mdbx: notify OOM-handler about end of loop.
Change-Id: I71d66e371df869560801e12b8b06c4f4cbf90e98
2016-05-07 03:52:04 +03:00
Leo Yuriev
f70d6638c5 mdbx: MDBX_ALLOC_KICK for freelist backlog.
Tries to sync only when all pages was allocated,
but not every time reclaiming had stopped at weak checkpoint.

Change-Id: I0b41fd87f1eab4725379a17fc29e16820122467d
2016-05-06 04:53:14 +03:00
Leo Yuriev
65354dcd01 mdbx: lifo-reclaimig for weak-to-steady conversion.
Change-Id: Id078c434cf3a624331c53f6b38ce34e31dc65aa5
2016-05-06 04:50:51 +03:00
Leo Yuriev
f335c13b45 mdbx: clarify mdbx_oomkick() for LMDB-mode.
Change-Id: I6e38a1e2892084f8a5f98af7f947c1c9d2f18b9c
2016-05-06 04:13:29 +03:00
Leo Yuriev
599069b2c4 mdbx: remove needless assert.
Change-Id: I0129a9e2cd8bda4c541d8b610bcf3f887ddcb75b
2016-05-04 22:31:58 +03:00
Leo Yuriev
2dd9638321 mdbx: refine find_oldest() and oom_kick().
Change-Id: I758eba1a93550b687bbec6bf90e5b2d48090af25
2016-05-04 22:31:58 +03:00
Leo Yuriev
50f5475185 mdbx: don't mimics the MDBX's features for MDB.
Change-Id: Ie3865434c741da77b9a285e43b7d6a1d9ec0c5e2
2016-05-04 22:30:11 +03:00
Leo Yuriev
78e6bf5e36 mdbx: refine/speedup mdb_cmp_memn().
Change-Id: I2c62246b26485039f29bc400a4d42f05270a7398
2016-04-27 16:08:08 +03:00
Leo Yuriev
477932e2b3 mdbx: MADV_REMOVE for unallocated space.
Change-Id: I14db843cf6e0dc713e2f302dc5ad86dc405c8d62
2016-04-27 01:52:52 +03:00
Leo Yuriev
745bdd346a mdbx: extra backlog's page for MDB_LIFORECLAIM.
Change-Id: I24d656de96a4bc3a3d4f6409db75868b818a4233
2016-04-26 22:48:08 +03:00
Howard Chu
7ae6c0f768 mdbx: backport - ITS#8412 fix NEXT_DUP after cursor_del.
Change-Id: If0bc63a9d398ae440a7a9ab7f0502b9c9608640e
2016-04-26 15:19:26 +03:00
Leo Yuriev
bfd82b5830 mdbx: refine mdb_env_sync0().
Change-Id: If27986d02b2ca19ba547fe8fa5c8f4a8b9019900
2016-04-20 19:51:29 +03:00
Howard Chu
78f1fbd038 mdbx: backport - ITS#8406 fix xcursors after cursor_del.
Don't leave them uninit'd if they now point at a valid DUP node

Change-Id: Idccf9540686816e3d06e404c9a5bef1bce059bfc
2016-04-20 19:44:01 +03:00
Leo Yuriev
ae324984cb mdbx: rework backlog for freeDB deletion.
Change-Id: Ib535233313e8ddd21c35df46c03e5e298f3587e6
2016-04-20 19:44:01 +03:00
Leo Yuriev
90aa4b154a mdbx: refine mdbx_cmp2int().
Change-Id: I5497db2e88d764e18a1aae1809dc66fd62a5adbe
2016-04-19 11:59:36 +03:00
Leo Yuriev
026a566fc6 mdbx: minor refine madvise order.
Change-Id: Ia6b78cad2aae68f5cb4c8475fbd27f9a9cb44cf2
2016-04-19 11:59:36 +03:00
Leo Yuriev
80e660317f mdbx: prohibit MDB_LIFORECLAIM with just MDB_NOMETASYNC.
Change-Id: Ib605174bf75c473aaf42b9aa9011db29343e4484
2016-04-19 11:59:36 +03:00
Leo Yuriev
bf7f3814f9 mdbx: minor fix non-ascii comment.
Change-Id: Ie9a38e2bef3c8ef91adc3f45818096658c84cd34
2016-04-19 11:59:35 +03:00
Howard Chu
ebf1ae1bdc mdbx: backport - mdb_drop optimization.
If we know there are no sub-DBs and no overflow pages, skip leaf scan.

Change-Id: I3005aaf0f80821f87d8a88c435fda1a52ee0557a
2016-04-13 13:26:26 +03:00
Leo Yuriev
828a5d73ca mdbx: always copy the rest of page (MDB_RESERVE case).
Change-Id: Iebc406767bee98a85ab6efec887ed698ffe59066

> On 30. mars 2016 19:25, Леонид Юрьев wrote:
>
> Why mdb_cursor_put() doesn't copy the rest of page in case MDB_RESERVE?
>
> In other words - why we should copy or not copy an end of page in
> dependence from MDB_RESERVE?

2016-03-30 19:46 GMT+03:00 Howard Chu <hyc@symas.com>:

That logic is backwards because I was preserving existing behavior,
i.e. make the page - including unused portions - look the same with
put() in the child txn as put() in the parent.  So with MDB_RESERVE
the entire page must be copied in case the user peeks at it before
writing to it.  Without MDB_RESERVE, the initial part of the page
body need not be copied since LMDB is about to overwrite it.

I'll comment that, unless that code should just go away.  Don't know
if this is the best behavior or it was just easier to code it than
to decide whether to cater to such obscure user behavior.  But
people do come up with special things do do about overflow pages.
2016-04-13 13:12:47 +03:00
Leo Yuriev
c17cff1d4e check MDB_RESERVE against MDB_DUPSORT.
Change-Id: I7459a92ecea867a5fd6510c828ff65cd551a1e9a
2016-03-30 02:32:59 +03:00
Leo Yuriev
b59f25fd46 mdbx: minor format spaces/tabs.
Change-Id: Ic9cce2e7a2552fb062ab8709121c72a9131be8b6
2016-03-30 02:32:50 +03:00
Howard Chu
e7859934f2 mdbx: backport - ITS#8393 fix MDB_GET_BOTH on non-dup record.
Change-Id: I96028ff6de55915bab98a7d0e2d97072296fc769
2016-03-29 23:03:58 +03:00
Leo Yuriev
33e2b52118 mdbx: fix nasty/stupid mistake in cmp-functions.
Change-Id: If9aff5d9cd46dde614ddaa688113be8be12f3e5d
2016-03-29 23:03:58 +03:00
Leo Yuriev
321bc2b901 mdbx: fix one more minor warning.
Change-Id: I80466714d739fc46aedbedfee0919b64704ffc37
2016-03-25 05:08:35 +03:00