rpms/libdb
7
0
Fork 0
Code Issues Pull Requests Releases Activity

libdb package is retired on branch c10s for CS-2269

Browse Source
This commit is contained in:
David Fan 2024-06-17 13:19:20 +00:00
parent b5918f0fb3
commit eb2c867951
21 changed files with 1 additions and 3565 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
.gitignore vendored
View File

@ -1,4 +0,0 @@
/db.1.85.tar.gz
/db-5.3.21.tar.gz
/db-5.3.28.tar.gz
/libdb-5.3.28-manpages.tar.gz

49
007-mt19937db.c_license.patch
View File

@ -1,49 +0,0 @@
Description: mt19937db.c license should include the GPL
This file is distributed from upstream Berkeley DB under the Artistic
License (no version specified), althouth it was later released by the
original author under both GPL2+ and BSD.
.
References:
http://web.archive.org/web/20010806225716/http://www.math.keio.ac.jp/matumoto/mt19937int.c
http://web.archive.org/web/20130127064020/http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/CODES/mt19937ar.c
https://bugzilla.redhat.com/show_bug.cgi?format=multiple&id=886838
https://lists.nongnu.org/archive/html/gnu-linux-libre/2010-05/msg00000.html
.
Author: Ruben Rodriguez <ruben@trisquel.info>
--- db-5.3.21/src/crypto/mersenne/mt19937db.c.licensefix
+++ db-5.3.21/src/crypto/mersenne/mt19937db.c
@@ -16,16 +16,27 @@
/* Coded by Takuji Nishimura, considering the suggestions by */
/* Topher Cooper and Marc Rieffel in July-Aug. 1997. */
-/* This library is free software under the Artistic license: */
-/* see the file COPYING distributed together with this code. */
-/* For the verification of the code, its output sequence file */
-/* mt19937int.out is attached (2001/4/2) */
-
/* Copyright (C) 1997, 1999 Makoto Matsumoto and Takuji Nishimura. */
/* Any feedback is very welcome. For any question, comments, */
/* see http://www.math.keio.ac.jp/matumoto/emt.html or email */
/* matumoto@math.keio.ac.jp */
+/* This library is free software; you can redistribute it and/or */
+/* modify it under the terms of the GNU Library General Public */
+/* License as published by the Free Software Foundation; either */
+/* version 2 of the License, or (at your option) any later */
+/* version. */
+/* This library is distributed in the hope that it will be useful, */
+/* but WITHOUT ANY WARRANTY; without even the implied warranty of */
+/* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. */
+/* See the GNU Library General Public License for more details. */
+/* You should have received a copy of the GNU Library General */
+/* Public License along with this library; if not, write to the */
+/* Free Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA */
+/* 02111-1307 USA */
+
+/* This library is free software under the Artistic license: */
+
/* REFERENCE */
/* M. Matsumoto and T. Nishimura, */
/* "Mersenne Twister: A 623-Dimensionally Equidistributed Uniform */

14
checkpoint-opd-deadlock.patch
View File

@ -1,14 +0,0 @@
--- db-5.3.21/src/db/db_cam.c.opd_deadlock 2017-10-31 12:20:54.118979690 +0100
+++ db-5.3.21/src/db/db_cam.c 2017-10-31 12:21:06.828739341 +0100
@@ -868,6 +868,11 @@
flags == DB_PREV || flags == DB_PREV_DUP)) {
if (tmp_rmw && (ret = dbc->am_writelock(dbc)) != 0)
goto err;
+ /* Latch the primary tree page here in order to not deadlock later. */
+ if (cp->page == NULL &&
+ (ret = __memp_fget(mpf, &cp->pgno,
+ dbc->thread_info, dbc->txn, 0, &cp->page)) != 0)
+ goto err;
if (F_ISSET(dbc, DBC_TRANSIENT))
opd = cp->opd;
else if ((ret = __dbc_idup(cp->opd, &opd, DB_POSITION)) != 0)

89
db-1.85-errno.patch
View File

@ -1,89 +0,0 @@
glibc doesn't like errno as the name of a field.
--- db.1.85/hash/hash.h Mon Feb 18 19:12:14 2002
+++ db.1.85/hash/hash.h Mon Feb 18 19:12:20 2002
@@ -103,7 +103,7 @@
BUFHEAD *cpage; /* Current page */
int cbucket; /* Current bucket */
int cndx; /* Index of next item on cpage */
- int errno; /* Error Number -- for DBM
+ int err; /* Error Number -- for DBM
* compatability */
int new_file; /* Indicates if fd is backing store
* or no */
--- db.1.85/hash/hash.c Mon Feb 18 19:12:24 2002
+++ db.1.85/hash/hash.c Mon Feb 18 19:12:44 2002
@@ -505,7 +505,7 @@
else
if (wsize != sizeof(HASHHDR)) {
errno = EFTYPE;
- hashp->errno = errno;
+ hashp->err = errno;
return (-1);
}
for (i = 0; i < NCACHED; i++)
@@ -536,7 +536,7 @@
hashp = (HTAB *)dbp->internal;
if (flag) {
- hashp->errno = errno = EINVAL;
+ hashp->err = errno = EINVAL;
return (ERROR);
}
return (hash_access(hashp, HASH_GET, (DBT *)key, data));
@@ -553,11 +553,11 @@
hashp = (HTAB *)dbp->internal;
if (flag && flag != R_NOOVERWRITE) {
- hashp->errno = errno = EINVAL;
+ hashp->err = errno = EINVAL;
return (ERROR);
}
if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
- hashp->errno = errno = EPERM;
+ hashp->err = errno = EPERM;
return (ERROR);
}
return (hash_access(hashp, flag == R_NOOVERWRITE ?
@@ -574,11 +574,11 @@
hashp = (HTAB *)dbp->internal;
if (flag && flag != R_CURSOR) {
- hashp->errno = errno = EINVAL;
+ hashp->err = errno = EINVAL;
return (ERROR);
}
if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
- hashp->errno = errno = EPERM;
+ hashp->err = errno = EPERM;
return (ERROR);
}
return (hash_access(hashp, HASH_DELETE, (DBT *)key, NULL));
@@ -729,7 +729,7 @@
hashp = (HTAB *)dbp->internal;
if (flag && flag != R_FIRST && flag != R_NEXT) {
- hashp->errno = errno = EINVAL;
+ hashp->err = errno = EINVAL;
return (ERROR);
}
#ifdef HASH_STATISTICS
--- db.1.85/hash/ndbm.c Mon Feb 18 19:12:58 2002
+++ db.1.85/hash/ndbm.c Mon Feb 18 19:13:05 2002
@@ -180,7 +180,7 @@
HTAB *hp;
hp = (HTAB *)db->internal;
- return (hp->errno);
+ return (hp->err);
}
extern int
@@ -190,7 +190,7 @@
HTAB *hp;
hp = (HTAB *)db->internal;
- hp->errno = 0;
+ hp->err = 0;
return (0);
}

12
db-4.5.20-jni-include-dir.patch
View File

@ -1,12 +0,0 @@
diff -up db-4.7.25/dist/configure.ac.jni db-4.7.25/dist/configure.ac
--- db-4.7.25/dist/configure.ac.jni 2008-08-20 14:22:59.000000000 +0200
+++ db-4.7.25/dist/configure.ac 2008-08-20 14:23:39.000000000 +0200
@@ -418,7 +418,7 @@ if test "$db_cv_java" = "yes"; then
AC_PROG_JAVAC
AC_PROG_JAR
AC_PROG_JAVA
- AC_JNI_INCLUDE_DIR
+ JNI_INCLUDE_DIRS="/usr/lib/jvm/java/include /usr/lib/jvm/java/include/linux"
AC_MSG_CHECKING(java version)
case "$JAVA" in

14
db-4.6.21-1.85-compat.patch
View File

@ -1,14 +0,0 @@
diff -up db-5.1.19/dist/Makefile.in.185compat db-5.1.19/dist/Makefile.in
--- db-5.1.19/dist/Makefile.in.185compat 2010-08-27 17:08:03.000000000 +0200
+++ db-5.1.19/dist/Makefile.in 2010-09-10 10:02:32.974640425 +0200
@@ -193,8 +193,8 @@ libtso_major= $(libtcl_base)-$(LIBMAJOR)
# local libraries, for example. Do that by adding -I options to the DB185INC
# line, and -l options to the DB185LIB line.
##################################################
-DB185INC= -c @CFLAGS@ -I$(topdir) @CPPFLAGS@
-DB185LIB=
+DB185INC= -c @CFLAGS@ -I$(srcdir) -I$(srcdir)/db.1.85/PORT/linux/include @CPPFLAGS@
+DB185LIB= ${srcdir}/db.1.85/PORT/linux/libdb.a
##################################################
# Performance Event Monitoring definitions

183
db-5.3.21-memp_stat-upstream-fix.patch
View File

@ -1,183 +0,0 @@
diff -r -u db-5.3.21_orig/src/mp/mp_stat.c db-5.3.21/src/mp/mp_stat.c
--- db-5.3.21_orig/src/mp/mp_stat.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_stat.c 2015-05-19 15:07:09.000000000 +0800
@@ -87,6 +87,13 @@
u_int32_t i;
uintmax_t tmp_wait, tmp_nowait;
+ /*
+ * The array holding the lengths related to the buffer allocated for *fspp.
+ * The first element of the array holds the number of entries allocated.
+ * The second element of the array holds the total number of bytes allocated.
+ */
+ u_int32_t fsp_len[2];
+
dbmp = env->mp_handle;
mp = dbmp->reginfo[0].primary;
@@ -193,32 +200,53 @@
if (fspp != NULL) {
*fspp = NULL;
- /* Count the MPOOLFILE structures. */
- i = 0;
- len = 0;
- if ((ret = __memp_walk_files(env,
- mp, __memp_count_files, &len, &i, flags)) != 0)
- return (ret);
+ while (*fspp == NULL) {
+ /* Count the MPOOLFILE structures. */
+ i = 0;
+ /*
+ * Allow space for the first __memp_get_files() to align the
+ * structure array to uintmax_t, DB_MPOOL_STAT's most
+ * restrictive field. [#23150]
+ */
+ len = sizeof(uintmax_t);
+ if ((ret = __memp_walk_files(env,
+ mp, __memp_count_files, &len, &i, flags)) != 0)
+ return (ret);
+
+ if (i == 0)
+ return (0);
+
+ /*
+ * Copy the number of DB_MPOOL_FSTAT entries and the number of
+ * bytes allocated for them into fsp_len. Do not count the space
+ * reserved for allignment.
+ */
+ fsp_len[0] = i;
+ fsp_len[1] = len - sizeof(uintmax_t);
- if (i == 0)
- return (0);
- len += sizeof(DB_MPOOL_FSTAT *); /* Trailing NULL */
+ len += sizeof(DB_MPOOL_FSTAT *); /* Trailing NULL */
- /* Allocate space */
- if ((ret = __os_umalloc(env, len, fspp)) != 0)
- return (ret);
+ /* Allocate space */
+ if ((ret = __os_umalloc(env, len, fspp)) != 0)
+ return (ret);
- tfsp = *fspp;
- *tfsp = NULL;
-
- /*
- * Files may have been opened since we counted, don't walk
- * off the end of the allocated space.
- */
- if ((ret = __memp_walk_files(env,
- mp, __memp_get_files, &tfsp, &i, flags)) != 0)
- return (ret);
+ tfsp = *fspp;
+ *tfsp = NULL;
+ /*
+ * Files may have been opened since we counted, if we walk off
+ * the end of the allocated space specified in fsp_len, retry.
+ */
+ if ((ret = __memp_walk_files(env,
+ mp, __memp_get_files, &tfsp, fsp_len, flags)) != 0) {
+ if (ret == DB_BUFFER_SMALL) {
+ __os_ufree(env, *fspp);
+ *fspp = NULL;
+ tfsp = NULL;
+ } else
+ return (ret);
+ }
+ }
*++tfsp = NULL;
}
@@ -286,28 +314,35 @@
* for the text file names.
*/
static int
-__memp_get_files(env, mfp, argp, countp, flags)
+__memp_get_files(env, mfp, argp, fsp_len, flags)
ENV *env;
MPOOLFILE *mfp;
void *argp;
- u_int32_t *countp;
+ u_int32_t fsp_len[];
u_int32_t flags;
{
DB_MPOOL *dbmp;
DB_MPOOL_FSTAT **tfsp, *tstruct;
char *name, *tname;
- size_t nlen;
+ size_t nlen, tlen;
- if (*countp == 0)
- return (0);
+ /* We walked through more files than argp was allocated for. */
+ if (fsp_len[0] == 0)
+ return DB_BUFFER_SMALL;
dbmp = env->mp_handle;
tfsp = *(DB_MPOOL_FSTAT ***)argp;
if (*tfsp == NULL) {
- /* Add 1 to count because we need to skip over the NULL. */
- tstruct = (DB_MPOOL_FSTAT *)(tfsp + *countp + 1);
- tname = (char *)(tstruct + *countp);
+ /*
+ * Add 1 to count because to skip over the NULL end marker.
+ * Align it further for DB_MPOOL_STAT's most restrictive field
+ * because uintmax_t might require stricter alignment than
+ * pointers; e.g., IP32 LL64 SPARC. [#23150]
+ */
+ tstruct = (DB_MPOOL_FSTAT *)&tfsp[fsp_len[0] + 1];
+ tstruct = ALIGNP_INC(tstruct, sizeof(uintmax_t));
+ tname = (char *)&tstruct[fsp_len[0]];
*tfsp = tstruct;
} else {
tstruct = *tfsp + 1;
@@ -317,6 +352,15 @@
name = __memp_fns(dbmp, mfp);
nlen = strlen(name) + 1;
+
+ /* The space required for file names is larger than argp was allocated for. */
+ tlen = sizeof(DB_MPOOL_FSTAT *) + sizeof(DB_MPOOL_FSTAT) + nlen;
+ if (fsp_len[1] < tlen)
+ return DB_BUFFER_SMALL;
+ else
+ /* Count down the number of bytes left in argp. */
+ fsp_len[1] -= tlen;
+
memcpy(tname, name, nlen);
memcpy(tstruct, &mfp->stat, sizeof(mfp->stat));
tstruct->file_name = tname;
@@ -325,7 +369,9 @@
tstruct->st_pagesize = mfp->pagesize;
*(DB_MPOOL_FSTAT ***)argp = tfsp;
- (*countp)--;
+
+ /* Count down the number of entries left in argp. */
+ fsp_len[0]--;
if (LF_ISSET(DB_STAT_CLEAR))
memset(&mfp->stat, 0, sizeof(mfp->stat));
diff -r -u db-5.3.21_orig/src/mp/mp_sync.c db-5.3.21/src/mp/mp_sync.c
--- db-5.3.21_orig/src/mp/mp_sync.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_sync.c 2015-05-19 15:08:05.000000000 +0800
@@ -57,11 +57,13 @@
if ((t_ret = func(env,
mfp, arg, countp, flags)) != 0 && ret == 0)
ret = t_ret;
- if (ret != 0 && !LF_ISSET(DB_STAT_MEMP_NOERROR))
+ if (ret != 0 &&
+ (!LF_ISSET(DB_STAT_MEMP_NOERROR) || ret == DB_BUFFER_SMALL))
break;
}
MUTEX_UNLOCK(env, hp->mtx_hash);
- if (ret != 0 && !LF_ISSET(DB_STAT_MEMP_NOERROR))
+ if (ret != 0 &&
+ (!LF_ISSET(DB_STAT_MEMP_NOERROR) || ret == DB_BUFFER_SMALL))
break;
}
return (ret);

718
db-5.3.21-mutex_leak.patch
View File

@ -1,718 +0,0 @@
diff -U 5 -r db-5.3.21.old/src/dbinc_auto/int_def.in db-5.3.21/src/dbinc_auto/int_def.in
--- db-5.3.21.old/src/dbinc_auto/int_def.in 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/dbinc_auto/int_def.in 2016-10-25 22:40:58.000000000 +0800
@@ -1371,10 +1371,11 @@
#define __memp_failchk __memp_failchk@DB_VERSION_UNIQUE_NAME@
#define __memp_bhwrite __memp_bhwrite@DB_VERSION_UNIQUE_NAME@
#define __memp_pgread __memp_pgread@DB_VERSION_UNIQUE_NAME@
#define __memp_pg __memp_pg@DB_VERSION_UNIQUE_NAME@
#define __memp_bhfree __memp_bhfree@DB_VERSION_UNIQUE_NAME@
+#define __memp_bh_clear_dirty __memp_bh_clear_dirty@DB_VERSION_UNIQUE_NAME@
#define __memp_fget_pp __memp_fget_pp@DB_VERSION_UNIQUE_NAME@
#define __memp_fget __memp_fget@DB_VERSION_UNIQUE_NAME@
#define __memp_fcreate_pp __memp_fcreate_pp@DB_VERSION_UNIQUE_NAME@
#define __memp_fcreate __memp_fcreate@DB_VERSION_UNIQUE_NAME@
#define __memp_set_clear_len __memp_set_clear_len@DB_VERSION_UNIQUE_NAME@
@@ -1395,10 +1396,11 @@
#define __memp_fopen __memp_fopen@DB_VERSION_UNIQUE_NAME@
#define __memp_fclose_pp __memp_fclose_pp@DB_VERSION_UNIQUE_NAME@
#define __memp_fclose __memp_fclose@DB_VERSION_UNIQUE_NAME@
#define __memp_mf_discard __memp_mf_discard@DB_VERSION_UNIQUE_NAME@
#define __memp_inmemlist __memp_inmemlist@DB_VERSION_UNIQUE_NAME@
+#define __memp_mf_mark_dead __memp_mf_mark_dead@DB_VERSION_UNIQUE_NAME@
#define __memp_fput_pp __memp_fput_pp@DB_VERSION_UNIQUE_NAME@
#define __memp_fput __memp_fput@DB_VERSION_UNIQUE_NAME@
#define __memp_unpin_buffers __memp_unpin_buffers@DB_VERSION_UNIQUE_NAME@
#define __memp_dirty __memp_dirty@DB_VERSION_UNIQUE_NAME@
#define __memp_shared __memp_shared@DB_VERSION_UNIQUE_NAME@
@@ -1453,10 +1455,11 @@
#define __memp_fsync_pp __memp_fsync_pp@DB_VERSION_UNIQUE_NAME@
#define __memp_fsync __memp_fsync@DB_VERSION_UNIQUE_NAME@
#define __mp_xxx_fh __mp_xxx_fh@DB_VERSION_UNIQUE_NAME@
#define __memp_sync_int __memp_sync_int@DB_VERSION_UNIQUE_NAME@
#define __memp_mf_sync __memp_mf_sync@DB_VERSION_UNIQUE_NAME@
+#define __memp_purge_dead_files __memp_purge_dead_files@DB_VERSION_UNIQUE_NAME@
#define __memp_trickle_pp __memp_trickle_pp@DB_VERSION_UNIQUE_NAME@
#define __mutex_alloc __mutex_alloc@DB_VERSION_UNIQUE_NAME@
#define __mutex_alloc_int __mutex_alloc_int@DB_VERSION_UNIQUE_NAME@
#define __mutex_free __mutex_free@DB_VERSION_UNIQUE_NAME@
#define __mutex_free_int __mutex_free_int@DB_VERSION_UNIQUE_NAME@
diff -U 5 -r db-5.3.21.old/src/dbinc_auto/mp_ext.h db-5.3.21/src/dbinc_auto/mp_ext.h
--- db-5.3.21.old/src/dbinc_auto/mp_ext.h 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/dbinc_auto/mp_ext.h 2016-10-25 22:40:58.000000000 +0800
@@ -14,10 +14,11 @@
int __memp_failchk __P((ENV *));
int __memp_bhwrite __P((DB_MPOOL *, DB_MPOOL_HASH *, MPOOLFILE *, BH *, int));
int __memp_pgread __P((DB_MPOOLFILE *, BH *, int));
int __memp_pg __P((DB_MPOOLFILE *, db_pgno_t, void *, int));
int __memp_bhfree __P((DB_MPOOL *, REGINFO *, MPOOLFILE *, DB_MPOOL_HASH *, BH *, u_int32_t));
+void __memp_bh_clear_dirty __P((ENV*, DB_MPOOL_HASH *, BH *));
int __memp_fget_pp __P((DB_MPOOLFILE *, db_pgno_t *, DB_TXN *, u_int32_t, void *));
int __memp_fget __P((DB_MPOOLFILE *, db_pgno_t *, DB_THREAD_INFO *, DB_TXN *, u_int32_t, void *));
int __memp_fcreate_pp __P((DB_ENV *, DB_MPOOLFILE **, u_int32_t));
int __memp_fcreate __P((ENV *, DB_MPOOLFILE **));
int __memp_set_clear_len __P((DB_MPOOLFILE *, u_int32_t));
@@ -38,10 +39,11 @@
int __memp_fopen __P((DB_MPOOLFILE *, MPOOLFILE *, const char *, const char **, u_int32_t, int, size_t));
int __memp_fclose_pp __P((DB_MPOOLFILE *, u_int32_t));
int __memp_fclose __P((DB_MPOOLFILE *, u_int32_t));
int __memp_mf_discard __P((DB_MPOOL *, MPOOLFILE *, int));
int __memp_inmemlist __P((ENV *, char ***, int *));
+void __memp_mf_mark_dead __P((DB_MPOOL *, MPOOLFILE *, int*));
int __memp_fput_pp __P((DB_MPOOLFILE *, void *, DB_CACHE_PRIORITY, u_int32_t));
int __memp_fput __P((DB_MPOOLFILE *, DB_THREAD_INFO *, void *, DB_CACHE_PRIORITY));
int __memp_unpin_buffers __P((ENV *, DB_THREAD_INFO *));
int __memp_dirty __P((DB_MPOOLFILE *, void *, DB_THREAD_INFO *, DB_TXN *, DB_CACHE_PRIORITY, u_int32_t));
int __memp_shared __P((DB_MPOOLFILE *, void *));
@@ -96,10 +98,11 @@
int __memp_fsync_pp __P((DB_MPOOLFILE *));
int __memp_fsync __P((DB_MPOOLFILE *));
int __mp_xxx_fh __P((DB_MPOOLFILE *, DB_FH **));
int __memp_sync_int __P((ENV *, DB_MPOOLFILE *, u_int32_t, u_int32_t, u_int32_t *, int *));
int __memp_mf_sync __P((DB_MPOOL *, MPOOLFILE *, int));
+int __memp_purge_dead_files __P((ENV *));
int __memp_trickle_pp __P((DB_ENV *, int, int *));
#if defined(__cplusplus)
}
#endif
diff -U 5 -r db-5.3.21.old/src/mp/mp_bh.c db-5.3.21/src/mp/mp_bh.c
--- db-5.3.21.old/src/mp/mp_bh.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_bh.c 2016-10-25 17:09:35.000000000 +0800
@@ -472,15 +472,12 @@
* a shared latch.
*/
if (F_ISSET(bhp, BH_DIRTY | BH_TRASH)) {
MUTEX_LOCK(env, hp->mtx_hash);
DB_ASSERT(env, !SH_CHAIN_HASNEXT(bhp, vc));
- if (ret == 0 && F_ISSET(bhp, BH_DIRTY)) {
- F_CLR(bhp, BH_DIRTY | BH_DIRTY_CREATE);
- DB_ASSERT(env, atomic_read(&hp->hash_page_dirty) > 0);
- atomic_dec(env, &hp->hash_page_dirty);
- }
+ if (ret == 0)
+ __memp_bh_clear_dirty(env, hp, bhp);
/* put the page back if necessary. */
if ((ret != 0 || BH_REFCOUNT(bhp) > 1) &&
F_ISSET(bhp, BH_TRASH)) {
ret = __memp_pg(dbmfp, bhp->pgno, bhp->buf, 1);
@@ -686,5 +683,31 @@
} else
MUTEX_UNLOCK(env, mfp->mutex);
return (ret);
}
+
+/*
+ * __memp_bh_clear_dirty --
+ * Clear the dirty flag of of a buffer. Calls on the same buffer must be
+ * serialized to get the accounting correct. This can be achieved by
+ * acquiring an exclusive lock on the buffer, a shared lock on the
+ * buffer plus an exclusive lock on the hash bucket, or some other
+ * mechanism that guarantees single-thread access to the entire region
+ * (e.g. during __memp_region_bhfree()).
+ *
+ * PUBLIC: void __memp_bh_clear_dirty __P((ENV*, DB_MPOOL_HASH *, BH *));
+ */
+void
+__memp_bh_clear_dirty(env, hp, bhp)
+ ENV *env;
+ DB_MPOOL_HASH *hp;
+ BH *bhp;
+{
+ COMPQUIET(env, env);
+ if (F_ISSET(bhp, BH_DIRTY)) {
+ F_CLR(bhp, BH_DIRTY | BH_DIRTY_CREATE);
+ DB_ASSERT(env, atomic_read(&hp->hash_page_dirty) > 0);
+ (void)atomic_dec(env, &hp->hash_page_dirty);
+ }
+}
+
diff -U 5 -r db-5.3.21.old/src/mp/mp_fget.c db-5.3.21/src/mp/mp_fget.c
--- db-5.3.21.old/src/mp/mp_fget.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_fget.c 2016-10-25 17:11:08.000000000 +0800
@@ -437,16 +437,11 @@
* complain and get out.
*/
if (flags == DB_MPOOL_FREE) {
freebuf: MUTEX_LOCK(env, hp->mtx_hash);
h_locked = 1;
- if (F_ISSET(bhp, BH_DIRTY)) {
- F_CLR(bhp, BH_DIRTY | BH_DIRTY_CREATE);
- DB_ASSERT(env,
- atomic_read(&hp->hash_page_dirty) > 0);
- atomic_dec(env, &hp->hash_page_dirty);
- }
+ __memp_bh_clear_dirty(env, hp, bhp);
/*
* If the buffer we found is already freed, we're done.
* If the ref count is not 1 then someone may be
* peeking at the buffer. We cannot free it until they
diff -U 5 -r db-5.3.21.old/src/mp/mp_fopen.c db-5.3.21/src/mp/mp_fopen.c
--- db-5.3.21.old/src/mp/mp_fopen.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_fopen.c 2016-10-25 22:31:05.000000000 +0800
@@ -12,10 +12,11 @@
#include "dbinc/log.h"
#include "dbinc/mp.h"
#include "dbinc/db_page.h"
#include "dbinc/hash.h"
+static int __memp_count_dead_mutex __P((DB_MPOOL *, u_int32_t *));
static int __memp_mpf_alloc __P((DB_MPOOL *,
DB_MPOOLFILE *, const char *, u_int32_t, u_int32_t, MPOOLFILE **));
static int __memp_mpf_find __P((ENV *,
DB_MPOOLFILE *, DB_MPOOL_HASH *, const char *, u_int32_t, MPOOLFILE **));
@@ -709,11 +710,15 @@
* We should be able to set mfp to NULL and break out of the
* loop, but I like the idea of checking all the entries.
*/
if (LF_ISSET(DB_TRUNCATE)) {
MUTEX_LOCK(env, mfp->mutex);
- mfp->deadfile = 1;
+ /*
+ * We cannot purge dead files here, because the caller
+ * is holding the mutex of the hash bucket of mfp.
+ */
+ __memp_mf_mark_dead(dbmp, mfp, NULL);
MUTEX_UNLOCK(env, mfp->mutex);
continue;
}
/*
@@ -907,14 +912,15 @@
DB_MPOOL *dbmp;
ENV *env;
MPOOLFILE *mfp;
char *rpath;
u_int32_t ref;
- int deleted, ret, t_ret;
+ int deleted, purge_dead, ret, t_ret;
env = dbmfp->env;
dbmp = env->mp_handle;
+ purge_dead = 0;
ret = 0;
/*
* Remove the DB_MPOOLFILE from the process' list.
*
@@ -1004,11 +1010,11 @@
}
DB_ASSERT(env, mfp->neutral_cnt < mfp->mpf_cnt);
if (--mfp->mpf_cnt == 0 || LF_ISSET(DB_MPOOL_DISCARD)) {
if (LF_ISSET(DB_MPOOL_DISCARD) ||
F_ISSET(mfp, MP_TEMP) || mfp->unlink_on_close) {
- mfp->deadfile = 1;
+ __memp_mf_mark_dead(dbmp, mfp, &purge_dead);
}
if (mfp->unlink_on_close) {
if ((t_ret = __db_appname(dbmp->env, DB_APP_DATA,
R_ADDR(dbmp->reginfo, mfp->path_off), NULL,
&rpath)) != 0 && ret == 0)
@@ -1037,10 +1043,12 @@
deleted = 1;
}
}
if (!deleted && !LF_ISSET(DB_MPOOL_NOLOCK))
MUTEX_UNLOCK(env, mfp->mutex);
+ if (purge_dead)
+ (void)__memp_purge_dead_files(env);
done: /* Discard the DB_MPOOLFILE structure. */
if (dbmfp->pgcookie != NULL) {
__os_free(env, dbmfp->pgcookie->data);
__os_free(env, dbmfp->pgcookie);
@@ -1091,11 +1099,11 @@
/*
* We have to release the MPOOLFILE mutex before acquiring the region
* mutex so we don't deadlock. Make sure nobody ever looks at this
* structure again.
*/
- mfp->deadfile = 1;
+ __memp_mf_mark_dead(dbmp, mfp, NULL);
/* Discard the mutex we're holding and return it too the pool. */
MUTEX_UNLOCK(env, mfp->mutex);
if ((t_ret = __mutex_free(env, &mfp->mutex)) != 0 && ret == 0)
ret = t_ret;
@@ -1216,5 +1224,106 @@
/* Make sure we don't return any garbage. */
*cntp = 0;
*namesp = NULL;
return (ret);
}
+
+/*
+ * __memp_mf_mark_dead --
+ * Mark an MPOOLFILE as dead because its contents are no longer necessary.
+ * This happens when removing, truncation, or closing an unnamed in-memory
+ * database. Return, in the purgep parameter, whether the caller should
+ * call __memp_purge_dead_files() after the lock on mfp is released. The
+ * caller must hold an exclusive lock on the mfp handle.
+ *
+ * PUBLIC: void __memp_mf_mark_dead __P((DB_MPOOL *, MPOOLFILE *, int*));
+ */
+void
+__memp_mf_mark_dead(dbmp, mfp, purgep)
+ DB_MPOOL *dbmp;
+ MPOOLFILE *mfp;
+ int *purgep;
+{
+ ENV *env;
+#ifdef HAVE_MUTEX_SUPPORT
+ REGINFO *infop;
+ DB_MUTEXREGION *mtxregion;
+ u_int32_t mutex_max, mutex_inuse, dead_mutex;
+#endif
+
+ if (purgep != NULL)
+ *purgep = 0;
+
+ env = dbmp->env;
+
+#ifdef HAVE_MUTEX_SUPPORT
+ MUTEX_REQUIRED(env, mfp->mutex);
+
+ if (MUTEX_ON(env) && mfp->deadfile == 0) {
+ infop = &env->mutex_handle->reginfo;
+ mtxregion = infop->primary;
+
+ mutex_inuse = mtxregion->stat.st_mutex_inuse;
+ if ((mutex_max = env->dbenv->mutex_max) == 0)
+ mutex_max = infop->rp->max / mtxregion->mutex_size;
+
+ /*
+ * Purging dead pages requires a full scan of the entire cache
+ * buffer, so it is a slow operation. We only want to do it
+ * when it is necessary and provides enough benefits. Below is
+ * a simple heuristic that determines when to purge all dead
+ * pages.
+ */
+ if (purgep != NULL && mutex_inuse > mutex_max - 200) {
+ /*
+ * If the mutex region is almost full and there are
+ * many mutexes held by dead files, purge dead files.
+ */
+ (void)__memp_count_dead_mutex(dbmp, &dead_mutex);
+ dead_mutex += mfp->block_cnt + 1;
+
+ if (dead_mutex > mutex_inuse / 20)
+ *purgep = 1;
+ }
+ }
+#endif
+
+ mfp->deadfile = 1;
+}
+
+/*
+ * __memp_count_dead_mutex --
+ * Estimate the number of mutexes held by dead files.
+ */
+static int
+__memp_count_dead_mutex(dbmp, dead_mutex)
+ DB_MPOOL *dbmp;
+ u_int32_t *dead_mutex;
+{
+ ENV *env;
+ DB_MPOOL_HASH *hp;
+ MPOOL *mp;
+ MPOOLFILE *mfp;
+ u_int32_t mutex_per_file;
+ int busy, i;
+
+ env = dbmp->env;
+ *dead_mutex = 0;
+ mutex_per_file = 1;
+#ifndef HAVE_ATOMICFILEREAD
+ mutex_per_file = 2;
+#endif
+ mp = dbmp->reginfo[0].primary;
+ hp = R_ADDR(dbmp->reginfo, mp->ftab);
+ for (i = 0; i < MPOOL_FILE_BUCKETS; i++, hp++) {
+ busy = MUTEX_TRYLOCK(env, hp->mtx_hash);
+ if (busy)
+ continue;
+ SH_TAILQ_FOREACH(mfp, &hp->hash_bucket, q, __mpoolfile) {
+ if (mfp->deadfile)
+ *dead_mutex += mfp->block_cnt + mutex_per_file;
+ }
+ MUTEX_UNLOCK(env, hp->mtx_hash);
+ }
+
+ return (0);
+}
diff -U 5 -r db-5.3.21.old/src/mp/mp_method.c db-5.3.21/src/mp/mp_method.c
--- db-5.3.21.old/src/mp/mp_method.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_method.c 2016-10-25 17:22:23.000000000 +0800
@@ -638,11 +638,11 @@
DB_MPOOL_HASH *hp, *nhp;
MPOOL *mp;
MPOOLFILE *mfp;
roff_t newname_off;
u_int32_t bucket;
- int locked, ret;
+ int locked, purge_dead, ret;
size_t nlen;
void *p;
#undef op_is_remove
#define op_is_remove (newname == NULL)
@@ -655,10 +655,11 @@
dbmp = NULL;
mfp = NULL;
nhp = NULL;
p = NULL;
locked = ret = 0;
+ purge_dead = 0;
if (!MPOOL_ON(env))
goto fsop;
dbmp = env->mp_handle;
@@ -747,11 +748,11 @@
* they do not get reclaimed as long as they exist. Since we
* are now deleting the database, we need to dec that count.
*/
if (mfp->no_backing_file)
mfp->mpf_cnt--;
- mfp->deadfile = 1;
+ __memp_mf_mark_dead(dbmp, mfp, &purge_dead);
MUTEX_UNLOCK(env, mfp->mutex);
} else {
/*
* Else, it's a rename. We've allocated memory for the new
* name. Swap it with the old one. If it's in memory we
@@ -806,10 +807,16 @@
if (locked == 1) {
MUTEX_UNLOCK(env, hp->mtx_hash);
if (nhp != NULL && nhp != hp)
MUTEX_UNLOCK(env, nhp->mtx_hash);
}
+ /*
+ * __memp_purge_dead_files() must be called when the hash bucket is
+ * unlocked.
+ */
+ if (purge_dead)
+ (void)__memp_purge_dead_files(env);
return (ret);
}
/*
* __memp_ftruncate __
diff -U 5 -r db-5.3.21.old/src/mp/mp_sync.c db-5.3.21/src/mp/mp_sync.c
--- db-5.3.21.old/src/mp/mp_sync.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_sync.c 2016-10-25 17:26:58.000000000 +0800
@@ -24,10 +24,11 @@
static int __bhcmp __P((const void *, const void *));
static int __memp_close_flush_files __P((ENV *, int));
static int __memp_sync_files __P((ENV *));
static int __memp_sync_file __P((ENV *,
MPOOLFILE *, void *, u_int32_t *, u_int32_t));
+static inline void __update_err_ret(int, int*);
/*
* __memp_walk_files --
* PUBLIC: int __memp_walk_files __P((ENV *, MPOOL *,
* PUBLIC: int (*) __P((ENV *, MPOOLFILE *, void *,
@@ -961,5 +962,125 @@
return (-1);
if (bhp1->track_pgno > bhp2->track_pgno)
return (1);
return (0);
}
+
+/*
+ * __memp_purge_dead_files --
+ * Remove all dead files and their buffers from the mpool. The caller
+ * cannot hold any lock on the dead MPOOLFILE handles, their buffers
+ * or their hash buckets.
+ *
+ * PUBLIC: int __memp_purge_dead_files __P((ENV *));
+ */
+int
+__memp_purge_dead_files(env)
+ ENV *env;
+{
+ BH *bhp;
+ DB_MPOOL *dbmp;
+ DB_MPOOL_HASH *hp, *hp_end;
+ REGINFO *infop;
+ MPOOL *c_mp, *mp;
+ MPOOLFILE *mfp;
+ u_int32_t i_cache;
+ int ret, t_ret, h_lock;
+
+ if (!MPOOL_ON(env))
+ return (0);
+
+ dbmp = env->mp_handle;
+ mp = dbmp->reginfo[0].primary;
+ ret = t_ret = h_lock = 0;
+
+ /*
+ * Walk each cache's list of buffers and free all buffers whose
+ * MPOOLFILE is marked as dead.
+ */
+ for (i_cache = 0; i_cache < mp->nreg; i_cache++) {
+ infop = &dbmp->reginfo[i_cache];
+ c_mp = infop->primary;
+
+ hp = R_ADDR(infop, c_mp->htab);
+ hp_end = &hp[c_mp->htab_buckets];
+ for (; hp < hp_end; hp++) {
+ /* Skip empty buckets. */
+ if (SH_TAILQ_FIRST(&hp->hash_bucket, __bh) == NULL)
+ continue;
+
+ /*
+ * Search for a dead buffer. Other places that call
+ * __memp_bhfree() acquire the buffer lock before the
+ * hash bucket lock. Even though we acquire the two
+ * locks in reverse order, we cannot deadlock here
+ * because we don't block waiting for the locks.
+ */
+ t_ret = MUTEX_TRYLOCK(env, hp->mtx_hash);
+ if (t_ret != 0) {
+ __update_err_ret(t_ret, &ret);
+ continue;
+ }
+ h_lock = 1;
+ SH_TAILQ_FOREACH(bhp, &hp->hash_bucket, hq, __bh) {
+ /* Skip buffers that are being used. */
+ if (BH_REFCOUNT(bhp) > 0)
+ continue;
+
+ mfp = R_ADDR(dbmp->reginfo, bhp->mf_offset);
+ if (!mfp->deadfile)
+ continue;
+
+ /* Found a dead buffer. Prepare to free it. */
+ t_ret = MUTEX_TRYLOCK(env, bhp->mtx_buf);
+ if (t_ret != 0) {
+ __update_err_ret(t_ret, &ret);
+ continue;
+ }
+
+ DB_ASSERT(env, (!F_ISSET(bhp, BH_EXCLUSIVE) &&
+ BH_REFCOUNT(bhp) == 0));
+ F_SET(bhp, BH_EXCLUSIVE);
+ (void)atomic_inc(env, &bhp->ref);
+
+ __memp_bh_clear_dirty(env, hp, bhp);
+
+ /*
+ * Free the buffer. The buffer and hash bucket
+ * are unlocked by __memp_bhfree.
+ */
+ if ((t_ret = __memp_bhfree(dbmp, infop, mfp,
+ hp, bhp, BH_FREE_FREEMEM)) == 0)
+ /*
+ * Decrement hp, so the next turn will
+ * search the same bucket again.
+ */
+ hp--;
+ else
+ __update_err_ret(t_ret, &ret);
+
+ /*
+ * The hash bucket is unlocked, we need to
+ * start over again.
+ */
+ h_lock = 0;
+ break;
+ }
+
+ if (h_lock) {
+ MUTEX_UNLOCK(env, hp->mtx_hash);
+ h_lock = 0;
+ }
+ }
+ }
+
+ return (ret);
+}
+
+static inline void
+__update_err_ret(t_ret, retp)
+ int t_ret;
+ int *retp;
+{
+ if (t_ret != 0 && t_ret != DB_LOCK_NOTGRANTED && *retp == 0)
+ *retp = t_ret;
+}
diff -U 5 -r db-5.3.21.old/src/mp/mp_trickle.c db-5.3.21/src/mp/mp_trickle.c
--- db-5.3.21.old/src/mp/mp_trickle.c 2012-05-12 01:57:53.000000000 +0800
+++ db-5.3.21/src/mp/mp_trickle.c 2016-10-25 17:27:57.000000000 +0800
@@ -65,10 +65,14 @@
"DB_ENV->memp_trickle: %d: percent must be between 1 and 100",
"%d"), pct);
return (EINVAL);
}
+ /* First we purge all dead files and their buffers. */
+ if ((ret = __memp_purge_dead_files(env)) != 0)
+ return (ret);
+
/*
* Loop through the caches counting total/dirty buffers.
*
* XXX
* Using hash_page_dirty is our only choice at the moment, but it's not
diff -U 5 -r db-5.3.21.old/src/mutex/mut_region.c db-5.3.21/src/mutex/mut_region.c
--- db-5.3.21.old/src/mutex/mut_region.c 2012-05-12 01:57:54.000000000 +0800
+++ db-5.3.21/src/mutex/mut_region.c 2016-10-25 17:34:22.000000000 +0800
@@ -15,11 +15,11 @@
#include "dbinc/txn.h"
static db_size_t __mutex_align_size __P((ENV *));
static int __mutex_region_init __P((ENV *, DB_MUTEXMGR *));
static size_t __mutex_region_size __P((ENV *));
-static size_t __mutex_region_max __P((ENV *));
+static size_t __mutex_region_max __P((ENV *, u_int32_t));
/*
* __mutex_open --
* Open a mutex region.
*
@@ -32,11 +32,11 @@
{
DB_ENV *dbenv;
DB_MUTEXMGR *mtxmgr;
DB_MUTEXREGION *mtxregion;
size_t size;
- u_int32_t cpu_count;
+ u_int32_t cpu_count, mutex_needed;
int ret;
#ifndef HAVE_ATOMIC_SUPPORT
u_int i;
#endif
@@ -59,23 +59,24 @@
cpu_count : cpu_count * MUTEX_SPINS_PER_PROCESSOR)) != 0)
return (ret);
}
/*
- * If the user didn't set an absolute value on the number of mutexes
- * we'll need, figure it out. We're conservative in our allocation,
- * we need mutexes for DB handles, group-commit queues and other things
- * applications allocate at run-time. The application may have kicked
- * up our count to allocate its own mutexes, add that in.
+ * Figure out the number of mutexes we'll need. We're conservative in
+ * our allocation, we need mutexes for DB handles, group-commit queues
+ * and other things applications allocate at run-time. The application
+ * may have kicked up our count to allocate its own mutexes, add that
+ * in.
*/
+ mutex_needed =
+ __lock_region_mutex_count(env) +
+ __log_region_mutex_count(env) +
+ __memp_region_mutex_count(env) +
+ __txn_region_mutex_count(env);
if (dbenv->mutex_cnt == 0 &&
F_ISSET(env, ENV_PRIVATE | ENV_THREAD) != ENV_PRIVATE)
- dbenv->mutex_cnt =
- __lock_region_mutex_count(env) +
- __log_region_mutex_count(env) +
- __memp_region_mutex_count(env) +
- __txn_region_mutex_count(env);
+ dbenv->mutex_cnt = mutex_needed;
if (dbenv->mutex_max != 0 && dbenv->mutex_cnt > dbenv->mutex_max)
dbenv->mutex_cnt = dbenv->mutex_max;
/* Create/initialize the mutex manager structure. */
@@ -88,12 +89,12 @@
mtxmgr->reginfo.id = INVALID_REGION_ID;
mtxmgr->reginfo.flags = REGION_JOIN_OK;
size = __mutex_region_size(env);
if (create_ok)
F_SET(&mtxmgr->reginfo, REGION_CREATE_OK);
- if ((ret = __env_region_attach(env,
- &mtxmgr->reginfo, size, size + __mutex_region_max(env))) != 0)
+ if ((ret = __env_region_attach(env, &mtxmgr->reginfo,
+ size, size + __mutex_region_max(env, mutex_needed))) != 0)
goto err;
/* If we created the region, initialize it. */
if (F_ISSET(&mtxmgr->reginfo, REGION_CREATE))
if ((ret = __mutex_region_init(env, mtxmgr)) != 0)
@@ -350,44 +351,62 @@
dbenv = env->dbenv;
s = sizeof(DB_MUTEXMGR) + 1024;
- /* We discard one mutex for the OOB slot. */
+ /*
+ * We discard one mutex for the OOB slot. Make sure mutex_cnt doesn't
+ * overflow.
+ */
s += __env_alloc_size(
- (dbenv->mutex_cnt + 1) *__mutex_align_size(env));
+ (dbenv->mutex_cnt + (dbenv->mutex_cnt == UINT32_MAX ? 0 : 1)) *
+ __mutex_align_size(env));
return (s);
}
/*
* __mutex_region_max --
* Return the amount of space needed to reach the maximum size.
*/
static size_t
-__mutex_region_max(env)
+__mutex_region_max(env, mutex_needed)
ENV *env;
+ u_int32_t mutex_needed;
{
DB_ENV *dbenv;
- u_int32_t max;
+ u_int32_t max, mutex_cnt;
dbenv = env->dbenv;
+ mutex_cnt = dbenv->mutex_cnt;
- if ((max = dbenv->mutex_max) == 0) {
+ /*
+ * We want to limit the region size to accommodate at most UINT32_MAX
+ * mutexes. If mutex_cnt is UINT32_MAX, no more space is allowed.
+ */
+ if ((max = dbenv->mutex_max) == 0 && mutex_cnt != UINT32_MAX)
if (F_ISSET(env, ENV_PRIVATE | ENV_THREAD) == ENV_PRIVATE)
- max = dbenv->mutex_inc + 1;
- else
+ if (dbenv->mutex_inc + 1 < UINT32_MAX - mutex_cnt)
+ max = dbenv->mutex_inc + 1 + mutex_cnt;
+ else
+ max = UINT32_MAX;
+ else {
max = __lock_region_mutex_max(env) +
__txn_region_mutex_max(env) +
__log_region_mutex_max(env) +
dbenv->mutex_inc + 100;
- } else if (max <= dbenv->mutex_cnt)
+ if (max < UINT32_MAX - mutex_needed)
+ max += mutex_needed;
+ else
+ max = UINT32_MAX;
+ }
+
+ if (max <= mutex_cnt)
return (0);
else
- max -= dbenv->mutex_cnt;
-
- return ( __env_alloc_size(max * __mutex_align_size(env)));
+ return (__env_alloc_size(
+ (max - mutex_cnt) * __mutex_align_size(env)));
}
#ifdef HAVE_MUTEX_SYSTEM_RESOURCES
/*
* __mutex_resource_return

20
db-5.3.28-atomic_compare_exchange.patch
View File

@ -1,20 +0,0 @@
--- db-5.3.28/src/dbinc/atomic.h.old 2018-05-23 09:20:04.216914922 +0200
+++ db-5.3.28/src/dbinc/atomic.h 2018-05-23 09:20:49.510057897 +0200
@@ -144,7 +144,7 @@
#define atomic_inc(env, p) __atomic_inc(p)
#define atomic_dec(env, p) __atomic_dec(p)
#define atomic_compare_exchange(env, p, o, n) \
- __atomic_compare_exchange((p), (o), (n))
+ __db_atomic_compare_exchange((p), (o), (n))
static inline int __atomic_inc(db_atomic_t *p)
{
int temp;
@@ -176,7 +176,7 @@
* http://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
* which configure could be changed to use.
*/
-static inline int __atomic_compare_exchange(
+static inline int __db_atomic_compare_exchange(
db_atomic_t *p, atomic_value_t oldval, atomic_value_t newval)
{
atomic_value_t was;

135
db-5.3.28-condition-variable-ppc.patch
View File

@ -1,135 +0,0 @@
diff -up db-5.3.28/src/dbinc_auto/int_def.in.pthreads_ppc db-5.3.28/src/dbinc_auto/int_def.in
--- db-5.3.28/src/dbinc_auto/int_def.in.pthreads_ppc 2017-06-26 15:06:09.282020804 +0200
+++ db-5.3.28/src/dbinc_auto/int_def.in 2017-06-26 15:06:22.238771039 +0200
@@ -1572,6 +1572,7 @@
#define __os_strerror __os_strerror@DB_VERSION_UNIQUE_NAME@
#define __os_posix_err __os_posix_err@DB_VERSION_UNIQUE_NAME@
#define __os_fileid __os_fileid@DB_VERSION_UNIQUE_NAME@
+#define __check_lock_fn __check_lock_fn@DB_VERSION_UNIQUE_NAME@
#define __os_fdlock __os_fdlock@DB_VERSION_UNIQUE_NAME@
#define __os_fsync __os_fsync@DB_VERSION_UNIQUE_NAME@
#define __os_getenv __os_getenv@DB_VERSION_UNIQUE_NAME@
diff -up db-5.3.28/src/dbinc_auto/os_ext.h.pthreads_ppc db-5.3.28/src/dbinc_auto/os_ext.h
--- db-5.3.28/src/dbinc_auto/os_ext.h.pthreads_ppc 2017-06-26 15:05:23.792897700 +0200
+++ db-5.3.28/src/dbinc_auto/os_ext.h 2017-06-26 15:06:22.172772311 +0200
@@ -41,6 +41,7 @@ void __os_set_errno __P((int));
char *__os_strerror __P((int, char *, size_t));
int __os_posix_err __P((int));
int __os_fileid __P((ENV *, const char *, int, u_int8_t *));
+int __check_lock_fn __P((char *, pid_t));
int __os_fdlock __P((ENV *, DB_FH *, off_t, db_lockmode_t, int));
int __os_fsync __P((ENV *, DB_FH *));
int __os_getenv __P((ENV *, const char *, char **, size_t));
diff -up db-5.3.28/src/env/env_region.c.pthreads_ppc db-5.3.28/src/env/env_region.c
--- db-5.3.28/src/env/env_region.c.pthreads_ppc 2017-06-26 15:04:40.805726365 +0200
+++ db-5.3.28/src/env/env_region.c 2017-06-26 15:04:40.807726326 +0200
@@ -305,14 +305,6 @@ user_map_functions:
goto creation;
}
- /* We have an old environment but cannot rebuild it safely. */
- if (ret == DB_OLD_VERSION) {
- __db_errx(env, DB_STR("1539",
- "Build signature doesn't match environment"));
- ret = DB_VERSION_MISMATCH;
- goto err;
- }
-
if (renv->majver != DB_VERSION_MAJOR ||
renv->minver != DB_VERSION_MINOR) {
/*
@@ -337,6 +329,20 @@ user_map_functions:
goto err;
}
+ /*
+ * We have an old environment but cannot rebuild it safely.
+ * The environment might still be readable by the current process
+ * (only the libpthread timestamp changed) if this process is the one
+ * that is holding the lock on the region file as we can be sure
+ * the environment did not get rebuilt under our feet.
+ */
+ if (ret == DB_OLD_VERSION && __check_lock_fn(env->lockfhp->name, getpid()) ) {
+ __db_errx(env, DB_STR("1539",
+ "Build signature doesn't match environment"));
+ ret = DB_VERSION_MISMATCH;
+ goto err;
+ }
+
/*
* Check if the environment has had a catastrophic failure.
*
diff -up db-5.3.28/src/os/os_flock.c.pthreads_ppc db-5.3.28/src/os/os_flock.c
--- db-5.3.28/src/os/os_flock.c.pthreads_ppc 2017-06-26 15:04:40.806726346 +0200
+++ db-5.3.28/src/os/os_flock.c 2017-06-26 15:04:40.807726326 +0200
@@ -15,6 +15,70 @@ static int __os_filelocking_notsup __P((
#endif
/*
+ * __check_lock_fn --
+ * Parse /proc/locks to see if the file described by 'fn' is locked.
+ * Additionally (if 'pid' is not 0) check if the process holding
+ * the lock has the same pid value as 'pid'.
+ *
+ * Returns 0 if a lock on fn is found, 1 if it is not found and -1 on error.
+ * PUBLIC: int __check_lock_fn __P((char *, pid_t));
+ */
+
+int __check_lock_fn(fn, pid)
+ char *fn;
+ pid_t pid;
+{
+ FILE* fp;
+ char buffer[PATH_MAX];
+ char *token;
+ int i, inode;
+ struct stat st;
+ pid_t lpid = 0;
+
+ if (!fn)
+ return -1;
+
+ fp = fopen("/proc/locks", "r");
+ if (!fp)
+ return -1;
+
+ /* Get the file's inode */
+ if (stat(fn, &st)) {
+ fclose(fp);
+ return -1;
+ }
+
+ while (fgets(buffer, sizeof(buffer), fp))
+ for (token = strtok(buffer, " "), i = 0; token; token = strtok(NULL, " "), i++) {
+ /* Do not parse any other fields */
+ if (i > 5)
+ break;
+ /* Save the PID */
+ if (i == 4)
+ lpid = atoi(token);
+ /* Check the inode */
+ else if (i == 5) {
+ inode = 0;
+ sscanf(token, "%*02x:%*02x:%d", &inode);
+ /* Not the inode we are looking for */
+ if (inode != st.st_ino)
+ continue;
+ /*
+ * We have the correct file.
+ * We are either looking for a specific process or we do not care at all.
+ */
+ if (!pid || lpid == pid) {
+ fclose(fp);
+ return 0;
+ }
+ /* Not the lock we are looking for */
+ }
+ }
+ fclose(fp);
+ return 1;
+}
+
+/*
* __os_fdlock --
* Acquire/release a lock on a byte in a file.
*

759
db-5.3.28-condition_variable.patch
View File

@ -1,759 +0,0 @@
diff -up db-5.3.28/dist/android/android_config.in.pthreads db-5.3.28/dist/android/android_config.in
--- db-5.3.28/dist/android/android_config.in.pthreads 2013-09-09 17:35:02.000000000 +0200
+++ db-5.3.28/dist/android/android_config.in 2017-06-13 11:15:15.323215161 +0200
@@ -123,6 +123,9 @@
/* Define to 1 if allocated filesystem blocks are not zeroed. */
/* #undef HAVE_FILESYSTEM_NOTZERO */
+/* Define to 1 if you have the `flock' function. */
+#undef HAVE_FLOCK
+
/* Define to 1 if you have the `fopen' function. */
#define HAVE_FOPEN 1
diff -up db-5.3.28/dist/config.hin.pthreads db-5.3.28/dist/config.hin
--- db-5.3.28/dist/config.hin.pthreads 2013-09-09 17:35:02.000000000 +0200
+++ db-5.3.28/dist/config.hin 2017-06-13 11:15:15.000000000 +0200
@@ -92,6 +92,9 @@
/* Define to 1 if you have the <dlfcn.h> header file. */
#undef HAVE_DLFCN_H
+/* Define to 1 if you have the `dl_iterate_phdr' function. */
+#undef HAVE_DL_ITERATE_PHDR
+
/* Define to 1 to use dtrace for performance monitoring. */
#undef HAVE_DTRACE
@@ -125,6 +128,9 @@
/* Define to 1 if allocated filesystem blocks are not zeroed. */
#undef HAVE_FILESYSTEM_NOTZERO
+/* Define to 1 if you have the `flock' function. */
+#undef HAVE_FLOCK
+
/* Define to 1 if you have the `fopen' function. */
#undef HAVE_FOPEN
@@ -360,6 +366,9 @@
/* Define to 1 if you have the `pstat_getdynamic' function. */
#undef HAVE_PSTAT_GETDYNAMIC
+/* Define to 1 if you have dl_iterate_phdr and use pthread-based mutexes. */
+#undef HAVE_PTHREADS_TIMESTAMP
+
/* Define to 1 if it is OK to initialize an already initialized
pthread_cond_t. */
#undef HAVE_PTHREAD_COND_REINIT_OKAY
@@ -620,6 +629,11 @@
/* Define to 1 to mask harmless uninitialized memory read/writes. */
#undef UMRW
+/* Enable large inode numbers on Mac OS X 10.5. */
+#ifndef _DARWIN_USE_64_BIT_INODE
+# define _DARWIN_USE_64_BIT_INODE 1
+#endif
+
/* Number of bits in a file offset, on hosts where this is settable. */
#undef _FILE_OFFSET_BITS
diff -up db-5.3.28/dist/configure.ac.pthreads db-5.3.28/dist/configure.ac
--- db-5.3.28/dist/configure.ac.pthreads 2017-06-13 11:15:15.309215434 +0200
+++ db-5.3.28/dist/configure.ac 2017-06-13 11:15:15.323215161 +0200
@@ -698,7 +698,7 @@ AC_REPLACE_FUNCS(\
# Check for system functions we optionally use.
AC_CHECK_FUNCS(\
_fstati64 backtrace backtrace_symbols directio fchmod fclose\
- fcntl fdatasync fgetc fgets fopen fwrite getgid\
+ fcntl fdatasync fgetc fgets flock fopen fwrite getgid\
getrusage getuid hstrerror mprotect pstat_getdynamic\
pthread_self pthread_yield random sched_yield select setgid setuid\
sigaction snprintf stat sysconf vsnprintf yield)
@@ -1042,6 +1042,34 @@ if test "$db_cv_localization" = "yes"; t
[Define to 1 if you have localization function to support globalization.])
fi
+# Check for dl_iterate_phdr; do the test explicitly instead of using
+# AC_CHECK_FUNCS because <netdb.h> isn't a standard include file.
+AC_CACHE_CHECK([for dl_iterate_phdr], db_cv_dl_iterate_phdr, [
+AC_TRY_LINK([
+#include <sys/types.h>
+#include <netdb.h>], [
+ dl_iterate_phdr(0, 0);
+], [db_cv_dl_iterate_phdr=yes], [db_cv_dl_iterate_phdr=no])])
+if test "$db_cv_dl_iterate_phdr" = "yes"; then
+ AC_DEFINE(HAVE_DL_ITERATE_PHDR)
+ AH_TEMPLATE(HAVE_DL_ITERATE_PHDR,
+ [Define to 1 if you have the `dl_iterate_phdr' function.])
+fi
+
+# If we are using pthread mutex or condition variables, and dl_iterate_phdr() is
+# available, then we try to detect when libpthread is updated -- which can
+# render existing environment invalid. DB_ENV->open() tries to rebuild such
+# environments when they are idle.
+case "$db_cv_mutex" in
+ *pthreads*)
+ if test "$db_cv_dl_iterate_phdr" = "yes" ; then
+ AC_DEFINE(HAVE_PTHREADS_TIMESTAMP)
+ AH_TEMPLATE(HAVE_PTHREADS_TIMESTAMP,
+ [Define to 1 if you have dl_iterate_phdr and use pthread-based mutexes.])
+ fi
+ ;;
+esac
+
# We need to add the additional object files into the Makefile with the correct
# suffix. We can't use $LTLIBOBJS itself, because that variable has $U encoded
# in it for automake, and that's not what we want. See SR #7227 for additional
diff -up db-5.3.28/src/db/db_meta.c.pthreads db-5.3.28/src/db/db_meta.c
--- db-5.3.28/src/db/db_meta.c.pthreads 2013-09-09 17:35:07.000000000 +0200
+++ db-5.3.28/src/db/db_meta.c 2017-06-13 11:15:15.323215161 +0200
@@ -1330,8 +1330,9 @@ __db_haslock(env, locker, dbmfp, pgno, m
}
/*
* __db_has_pagelock --
- * Determine if this locker holds a particular page lock.
- * Returns 0 if lock is held, non-zero otherwise.
+ * Determine if this locker holds a particular page lock, and return an
+ * error if it is missing a page lock that it should have.
+ * Otherwise (TDS with the page locked, or DS or CDS) return 0.
*
* PUBLIC: #ifdef DIAGNOSTIC
* PUBLIC: int __db_has_pagelock __P((ENV *, DB_LOCKER *,
@@ -1348,6 +1349,9 @@ __db_has_pagelock(env, locker, dbmfp, pa
{
int ret;
+ if (!FLD_ISSET(env->open_flags, DB_INIT_TXN))
+ return (0);
+
switch (pagep->type) {
case P_OVERFLOW:
case P_INVALID:
diff -up db-5.3.28/src/dbinc_auto/int_def.in.pthreads db-5.3.28/src/dbinc_auto/int_def.in
--- db-5.3.28/src/dbinc_auto/int_def.in.pthreads 2017-06-13 11:15:15.317215278 +0200
+++ db-5.3.28/src/dbinc_auto/int_def.in 2017-06-13 11:15:15.324215141 +0200
@@ -1545,6 +1545,7 @@
#if defined(HAVE_REPLICATION_THREADS)
#define __os_freeaddrinfo __os_freeaddrinfo@DB_VERSION_UNIQUE_NAME@
#endif
+#define __os_pthreads_timestamp __os_pthreads_timestamp@DB_VERSION_UNIQUE_NAME@
#define __os_umalloc __os_umalloc@DB_VERSION_UNIQUE_NAME@
#define __os_urealloc __os_urealloc@DB_VERSION_UNIQUE_NAME@
#define __os_ufree __os_ufree@DB_VERSION_UNIQUE_NAME@
diff -up db-5.3.28/src/dbinc_auto/os_ext.h.pthreads db-5.3.28/src/dbinc_auto/os_ext.h
--- db-5.3.28/src/dbinc_auto/os_ext.h.pthreads 2013-09-09 17:35:08.000000000 +0200
+++ db-5.3.28/src/dbinc_auto/os_ext.h 2017-06-13 11:15:15.324215141 +0200
@@ -14,6 +14,7 @@ int __os_getaddrinfo __P((ENV *, const c
#if defined(HAVE_REPLICATION_THREADS)
void __os_freeaddrinfo __P((ENV *, ADDRINFO *));
#endif
+time_t __os_pthreads_timestamp __P((ENV *));
int __os_umalloc __P((ENV *, size_t, void *));
int __os_urealloc __P((ENV *, size_t, void *));
void __os_ufree __P((ENV *, void *));
@@ -40,7 +41,7 @@ void __os_set_errno __P((int));
char *__os_strerror __P((int, char *, size_t));
int __os_posix_err __P((int));
int __os_fileid __P((ENV *, const char *, int, u_int8_t *));
-int __os_fdlock __P((ENV *, DB_FH *, off_t, int, int));
+int __os_fdlock __P((ENV *, DB_FH *, off_t, db_lockmode_t, int));
int __os_fsync __P((ENV *, DB_FH *));
int __os_getenv __P((ENV *, const char *, char **, size_t));
int __os_openhandle __P((ENV *, const char *, int, int, DB_FH **));
diff -up db-5.3.28/src/dbinc/region.h.pthreads db-5.3.28/src/dbinc/region.h
--- db-5.3.28/src/dbinc/region.h.pthreads 2013-09-09 17:35:08.000000000 +0200
+++ db-5.3.28/src/dbinc/region.h 2017-06-13 11:15:15.324215141 +0200
@@ -178,6 +178,16 @@ typedef struct __db_reg_env { /* SHARED
u_int32_t envid; /* Unique environment ID. */
u_int32_t signature; /* Structure signatures. */
+#if defined(HAVE_PTHREADS_TIMESTAMP) && defined(HAVE_MUTEX_PTHREADS)
+ /*
+ * Updates to glibc/libpthread can change its pthreads implementation
+ * and invalidate on-disk environments, even without changing the
+ * version number. If using POSIX mutexes and a change in this
+ * timestamp is detecting when opening an environment with DB_CREATE,
+ * __env_attach recreates any existing on-disk environment.
+ */
+ time_t pthreads_timestamp;
+#endif
time_t timestamp; /* Creation time. */
diff -up db-5.3.28/src/env/env_open.c.pthreads db-5.3.28/src/env/env_open.c
--- db-5.3.28/src/env/env_open.c.pthreads 2013-09-09 17:35:08.000000000 +0200
+++ db-5.3.28/src/env/env_open.c 2017-06-13 11:15:15.324215141 +0200
@@ -1031,11 +1031,11 @@ __env_attach_regions(dbenv, flags, orig_
goto err;
/*
- * __env_attach will return the saved init_flags field, which contains
+ * __env_attach has returned the saved init_flags field, which contains
* the DB_INIT_* flags used when the environment was created.
*
- * We may be joining an environment -- reset our flags to match the
- * ones in the environment.
+ * We may be joining an existing environment -- reset our flags to match
+ * the ones in the environment.
*/
if (FLD_ISSET(init_flags, DB_INITENV_CDB))
LF_SET(DB_INIT_CDB);
diff -up db-5.3.28/src/env/env_region.c.pthreads db-5.3.28/src/env/env_region.c
--- db-5.3.28/src/env/env_region.c.pthreads 2013-09-09 17:35:08.000000000 +0200
+++ db-5.3.28/src/env/env_region.c 2017-06-13 11:16:02.159299222 +0200
@@ -18,13 +18,49 @@ static int __env_des_get __P((ENV *, RE
static int __env_faultmem __P((ENV *, void *, size_t, int));
static int __env_sys_attach __P((ENV *, REGINFO *, REGION *));
static int __env_sys_detach __P((ENV *, REGINFO *, int));
+static int __env_check_recreate __P((ENV *, REGENV *, u_int32_t));
static void __env_des_destroy __P((ENV *, REGION *));
static void __env_remove_file __P((ENV *));
+
+/*
+ * If the system supports flock()-like file locking, then the primary region
+ * file __db.001 is exclusively locked during creation, and is read-locked while
+ * the environment is open. Most Unix-like systems have flock(), with the
+ * notable exception of Solaris.
+ * Note: fcntl cannot be used for this locking because of the unfortunate
+ * definition of its interaction with close(2). A process's fcntl locks are
+ * released whenever it closes any file descriptor for that file. So, if an
+ * environment is opened more than once, closing one of the DB_ENV handles would
+ * release the read lock that protects the other handle.
+ */
+#ifdef HAVE_FLOCK
+#define ENV_PRIMARY_LOCK(env, lockmode, async) \
+ ((env)->lockfhp == NULL ? 0 : \
+ __os_fdlock((env), (env)->lockfhp, -1, lockmode, async))
+#define ENV_PRIMARY_UNLOCK(env) \
+ ((env)->lockfhp == NULL ? 0 : \
+ __os_fdlock((env), (env)->lockfhp, -1, DB_LOCK_NG, 0))
+#else
+#define ENV_PRIMARY_LOCK(env, lockmode, async) (0)
+#define ENV_PRIMARY_UNLOCK(env) (0)
+#endif
+
/*
* __env_attach
* Join/create the environment
*
+ * Safely detecting and managing multiple processes' environment handles:
+ * BDB uses a shared or exclusive fcntl()-style lock on the first byte
+ * of the primary region file (__db.001) to detect whether other processes
+ * have the environment open, and to single-thread attempts to create the
+ * environment. If the open includes DB_CREATE, an exclusive lock is
+ * obtained during the open call. After the creation is finished, and
+ * anytime during a non-DB_CREATE env open, the process holds a shared
+ * lock.
+ * - single-thread creation of the environment
+ * - detect whether any other processes are currently attached to it.
+ *
* PUBLIC: int __env_attach __P((ENV *, u_int32_t *, int, int));
*/
int
@@ -122,7 +158,11 @@ loop: renv = NULL;
if ((ret = __os_open(
env, infop->name, 0, DB_OSO_REGION, 0, &env->lockfhp)) != 0)
goto err;
-
+ /* Wait to get shared access to the primary region. */
+ if ((ret = ENV_PRIMARY_LOCK(env, DB_LOCK_READ, 0)) != 0) {
+ __db_err(env, ret, "__env_attach: existing: shared lock error");
+ goto err;
+ }
/*
* !!!
* The region may be in system memory not backed by the filesystem
@@ -218,11 +258,10 @@ loop: renv = NULL;
segid = ref.segid;
}
-#ifndef HAVE_MUTEX_FCNTL
+#if !defined(HAVE_FCNTL) && !defined(HAVE_PTHREADS_TIMESTAMP)
/*
- * If we're not doing fcntl locking, we can close the file handle. We
- * no longer need it and the less contact between the buffer cache and
- * the VM, the better.
+ * Without fcntl-like support, we no longer need the file handle. Close
+ * it to limit the interaction between the buffer cache and the VM.
*/
(void)__os_closehandle(env, env->lockfhp);
env->lockfhp = NULL;
@@ -233,6 +272,8 @@ loop: renv = NULL;
tregion.size = (roff_t)size;
tregion.max = (roff_t)max;
tregion.segid = segid;
+ /* Attach to the existing primary region. */
+ /* The leaking db.001 gets open inside of here, in __os_attach(). */
if ((ret = __env_sys_attach(env, infop, &tregion)) != 0)
goto err;
@@ -245,21 +286,48 @@ user_map_functions:
infop->primary = infop->addr;
infop->head = (u_int8_t *)infop->addr + sizeof(REGENV);
renv = infop->primary;
+ ret = __env_check_recreate(env, renv, signature);
+
+ if (create_ok &&
+ ret == DB_OLD_VERSION &&
+ ENV_PRIMARY_LOCK(env, DB_LOCK_WRITE, 1) == 0) {
+ if (FLD_ISSET(dbenv->verbose, DB_VERB_RECOVERY))
+ __db_msg(env, "Recreating idle environment");
+ F_SET(infop, REGION_CREATE_OK);
+
+ /*
+ * Detach from the environment region; we need to unmap it (and
+ * close any file handle) so that we don't leak memory or files.
+ */
+ DB_ASSERT(env, infop->rp == NULL);
+ infop->rp = &tregion;
+ (void)__env_sys_detach(env, infop, 0);
+ goto creation;
+ }
+
+ /* We have an old environment but cannot rebuild it safely. */
+ if (ret == DB_OLD_VERSION) {
+ __db_errx(env, DB_STR("1539",
+ "Build signature doesn't match environment"));
+ ret = DB_VERSION_MISMATCH;
+ goto err;
+ }
- /*
- * Make sure the region matches our build. Special case a region
- * that's all nul bytes, just treat it like any other corruption.
- */
if (renv->majver != DB_VERSION_MAJOR ||
renv->minver != DB_VERSION_MINOR) {
- if (renv->majver != 0 || renv->minver != 0) {
+ /*
+ * Special case a region that's all nul bytes, just treat it
+ * like any other corruption.
+ */
+ if (renv->majver == 0 && renv->minver == 0)
+ ret = EINVAL;
+ else {
__db_errx(env, DB_STR_A("1538",
- "Program version %d.%d doesn't match environment version %d.%d",
+ "Program version %d.%d doesn't match in-use environment version %d.%d",
"%d %d %d %d"), DB_VERSION_MAJOR, DB_VERSION_MINOR,
renv->majver, renv->minver);
ret = DB_VERSION_MISMATCH;
- } else
- ret = EINVAL;
+ }
goto err;
}
if (renv->signature != signature) {
@@ -289,6 +357,18 @@ user_map_functions:
}
if (renv->magic != DB_REGION_MAGIC)
goto retry;
+ /*
+ * A bad magic number means that the env is new and not yet available:
+ * wait a while and try again. If the magic number says recovery is in
+ * process, remember the env creation time to record that recovery was
+ * the reason that the open failed.
+ */
+ if (renv->magic != DB_REGION_MAGIC) {
+ __db_msg(env, "attach sees bad region magic 0x%lx",
+ (u_long)renv->magic);
+ goto retry;
+ }
+
/*
* Get a reference to the underlying REGION information for this
@@ -346,6 +426,12 @@ user_map_functions:
return (0);
creation:
+ /* Should this wait for the lock (passing 0 instead of 1)? */
+ if ((ret = ENV_PRIMARY_LOCK(env, DB_LOCK_WRITE, 1)) != 0) {
+ __db_err(env, ret, "__env_attach: creation could not lock %s",
+ env->lockfhp->name);
+ goto err;
+ }
/* Create the environment region. */
F_SET(infop, REGION_CREATE);
@@ -437,7 +523,14 @@ creation:
renv->minver = (u_int32_t)minver;
renv->patchver = (u_int32_t)patchver;
renv->signature = signature;
-
+#ifdef HAVE_PTHREADS_TIMESTAMP
+ renv->pthreads_timestamp = __os_pthreads_timestamp(env);
+ {
+ char *s = getenv("TS_ADJUST");
+ if (s != NULL)
+ renv->pthreads_timestamp -= atoi(s);
+ }
+#endif
(void)time(&renv->timestamp);
__os_unique_id(env, &renv->envid);
@@ -513,16 +606,24 @@ find_err: __db_errx(env, DB_STR_A("1544"
}
}
-#ifndef HAVE_MUTEX_FCNTL
- /*
- * If we're not doing fcntl locking, we can close the file handle. We
- * no longer need it and the less contact between the buffer cache and
- * the VM, the better.
- */
+#ifdef HAVE_FCNTL
+ if ((ret = ENV_PRIMARY_UNLOCK(env)) != 0) {
+ __db_err(env, ret, "__env_attach: release exclusive lock");
+ goto err;
+ }
+ if ((ret = ENV_PRIMARY_LOCK(env, DB_LOCK_READ, 0)) != 0) {
+ __db_err(env, ret, "__env_attach: new: acquire shared lock");
+ goto err;
+ }
+#else
+ /*
+ * We no longer need the primary region file's handle and the less
+ * contact between the buffer cache and the VM, the better.
+ */
if (env->lockfhp != NULL) {
(void)__os_closehandle(env, env->lockfhp);
env->lockfhp = NULL;
- }
+ }
#endif
/* Everything looks good, we're done. */
@@ -562,9 +663,9 @@ retry: /* Close any open file handle. */
/* If we had a temporary error, wait awhile and try again. */
if (ret == 0) {
if (!retry_ok || ++retry_cnt > 3) {
+ ret = EAGAIN;
__db_errx(env, DB_STR("1546",
"unable to join the environment"));
- ret = EAGAIN;
} else {
__os_yield(env, retry_cnt * 3, 0);
goto loop;
@@ -575,6 +676,59 @@ retry: /* Close any open file handle. */
}
/*
+ * __env_check_recreate --
+ * Determine whether an existing on-disk environment should be recreated
+ * because it is not compatible with this compiled BDB library.
+ *
+ * Returns:
+ * 0 -
+ * The env was generated by this library. No recreation needed.
+ * DB_OLD_VERSION -
+ * It was created by an earlier BDB version, or by an earlier
+ * version of libpthreads (on certain Linux systems). The caller
+ * will try to recreate it with the currently configured settings.
+ * DB_VERSION_MISMATCH -
+ * It was created by a newer version of BDB. Do not attempt to
+ * fix it, something is probably wrong with the application setup.
+ */
+static int
+__env_check_recreate(env, renv, signature)
+ ENV *env;
+ REGENV *renv;
+ u_int32_t signature;
+{
+#ifdef HAVE_PTHREADS_TIMESTAMP
+ time_t pthreads_time;
+ char envtime[CTIME_BUFLEN], libtime[CTIME_BUFLEN];
+#endif
+
+ /* First, bail out if the env is too new for this code to handle. */
+ if (renv->majver > DB_VERSION_MAJOR ||
+ (renv->majver == DB_VERSION_MAJOR &&
+ renv->minver > DB_VERSION_MINOR))
+ return (DB_VERSION_MISMATCH);
+
+#ifdef HAVE_PTHREADS_TIMESTAMP
+ pthreads_time = __os_pthreads_timestamp(env);
+ if (pthreads_time != renv->pthreads_timestamp) {
+ if (FLD_ISSET(env->dbenv->verbose, DB_VERB_RECOVERY))
+ __db_msg(env,
+ "Pthreads timestamp changed: env %.24s current %.24s",
+ __os_ctime(&renv->pthreads_timestamp, envtime),
+ __os_ctime(&pthreads_time, libtime));
+ return (DB_OLD_VERSION);
+ }
+#endif
+ if (renv->signature != signature || renv->majver != DB_VERSION_MAJOR ||
+ renv->minver != DB_VERSION_MINOR) {
+ if (FLD_ISSET(env->dbenv->verbose, DB_VERB_RECOVERY))
+ __db_msg(env, "Signature or version changed");
+ return (DB_OLD_VERSION);
+ }
+ return (0);
+}
+
+/*
* __env_turn_on --
* Turn on the created environment.
*
diff -up db-5.3.28/src/env/env_register.c.pthreads db-5.3.28/src/env/env_register.c
--- db-5.3.28/src/env/env_register.c.pthreads 2013-09-09 17:35:08.000000000 +0200
+++ db-5.3.28/src/env/env_register.c 2017-06-13 11:15:15.325215121 +0200
@@ -19,9 +19,9 @@
#define PID_LEN (25) /* PID entry length */
#define REGISTRY_LOCK(env, pos, nowait) \
- __os_fdlock(env, (env)->dbenv->registry, (off_t)(pos), 1, nowait)
+ __os_fdlock(env, (env)->dbenv->registry, (off_t)(pos), DB_LOCK_WRITE, nowait)
#define REGISTRY_UNLOCK(env, pos) \
- __os_fdlock(env, (env)->dbenv->registry, (off_t)(pos), 0, 0)
+ __os_fdlock(env, (env)->dbenv->registry, (off_t)(pos), DB_LOCK_NG, 0)
#define REGISTRY_EXCL_LOCK(env, nowait) \
REGISTRY_LOCK(env, 1, nowait)
#define REGISTRY_EXCL_UNLOCK(env) \
diff -up db-5.3.28/src/env/env_stat.c.pthreads db-5.3.28/src/env/env_stat.c
--- db-5.3.28/src/env/env_stat.c.pthreads 2013-09-09 17:35:08.000000000 +0200
+++ db-5.3.28/src/env/env_stat.c 2017-06-13 11:15:15.325215121 +0200
@@ -177,6 +177,10 @@ __env_print_stats(env, flags)
STAT_LONG("Txn version", DB_TXNVERSION);
__db_msg(env,
"%.24s\tCreation time", __os_ctime(&renv->timestamp, time_buf));
+#if defined(HAVE_PTHREADS_TIMESTAMP) && defined(HAVE_MUTEX_PTHREADS)
+ __db_msg(env,
+ "%.24s\tlibpthread timestamp", __os_ctime(&renv->pthreads_timestamp, time_buf));
+#endif
STAT_HEX("Environment ID", renv->envid);
__mutex_print_debug_single(env,
"Primary region allocation and reference count mutex",
diff -up db-5.3.28/src/os/os_addrinfo.c.pthreads db-5.3.28/src/os/os_addrinfo.c
--- db-5.3.28/src/os/os_addrinfo.c.pthreads 2013-09-09 17:35:09.000000000 +0200
+++ db-5.3.28/src/os/os_addrinfo.c 2017-06-13 11:15:15.325215121 +0200
@@ -10,6 +10,10 @@
#include "db_int.h"
+#ifdef HAVE_PTHREADS_TIMESTAMP
+#include <link.h>
+#endif
+
/*
* __os_getaddrinfo and __os_freeaddrinfo wrap the getaddrinfo and freeaddrinfo
* calls, as well as the associated platform dependent error handling, mapping
@@ -177,3 +181,48 @@ __os_freeaddrinfo(env, ai)
}
#endif
}
+
+#ifdef HAVE_PTHREADS_TIMESTAMP
+/*
+ * callback_find_pthreads --
+ * dl_iterate_phdr() calls this once for each loaded library.
+ *
+ * Returns:
+ * 0 - the library does not appear to be libpthreads.
+ * 1 - the library *does* seem to be libpthreads. Its modification time is
+ * stored into into the last argument's location.
+ */
+static int
+callback_find_pthreads(struct dl_phdr_info *info, size_t size, void *data)
+{
+ struct stat stbuf;
+
+ /* Stop (return non-zero) when libc is found. */
+ if (strstr(info->dlpi_name, "libpthread") != NULL &&
+ stat(info->dlpi_name, &stbuf) == 0) {
+ *(time_t *)data = stbuf.st_mtime;
+ return (1);
+ }
+ COMPQUIET(size, 0);
+ COMPQUIET(data, NULL);
+ return (0);
+}
+
+/*
+ * __os_pthreads_timestamp --
+ *
+ * PUBLIC: time_t __os_pthreads_timestamp __P((ENV *));
+ */
+time_t
+__os_pthreads_timestamp(env)
+ ENV *env;
+{
+ time_t timestamp;
+
+ timestamp = 0;
+ dl_iterate_phdr(callback_find_pthreads, &timestamp);
+
+ COMPQUIET(env, 0);
+ return (timestamp);
+}
+#endif
diff -up db-5.3.28/src/os/os_flock.c.pthreads db-5.3.28/src/os/os_flock.c
--- db-5.3.28/src/os/os_flock.c.pthreads 2013-09-09 17:35:09.000000000 +0200
+++ db-5.3.28/src/os/os_flock.c 2017-06-13 11:15:15.325215121 +0200
@@ -10,41 +10,98 @@
#include "db_int.h"
+#if !defined(HAVE_FCNTL) || !defined(HAVE_FLOCK)
+static int __os_filelocking_notsup __P((ENV *));
+#endif
+
/*
* __os_fdlock --
* Acquire/release a lock on a byte in a file.
*
- * PUBLIC: int __os_fdlock __P((ENV *, DB_FH *, off_t, int, int));
+ * The lock modes supported here are:
+ * DB_LOCK_NG - release the lock
+ * DB_LOCK_READ - get shared access
+ * DB_LOCK_WRITE - get exclusive access
+ *
+ * Use fcntl()-like semantics most of the time (DB_REGISTER support). Fcntl
+ * supports range locking, but has the additional broken semantics that
+ * closing any of the file's descriptors releases any locks, even if its
+ * other file descriptors remain open. Thanks SYSV & POSIX.
+ * However, if the offset is negative (which is allowed, because POSIX
+ * off_t a signed integer) then use flock() instead. It has only whole-
+ * file locks, but they persist until explicitly unlocked or the process
+ * exits.
+ * PUBLIC: int __os_fdlock __P((ENV *, DB_FH *, off_t, db_lockmode_t, int));
*/
int
-__os_fdlock(env, fhp, offset, acquire, nowait)
+__os_fdlock(env, fhp, offset, lockmode, nowait)
ENV *env;
DB_FH *fhp;
- int acquire, nowait;
off_t offset;
+ db_lockmode_t lockmode;
+ int nowait;
{
#ifdef HAVE_FCNTL
DB_ENV *dbenv;
struct flock fl;
int ret, t_ret;
+ static char *mode_string[DB_LOCK_WRITE + 1] = {
+ "unlock",
+ "read",
+ "write"
+ };
+ short mode_fcntl[DB_LOCK_WRITE + 1] = {
+ F_UNLCK,
+ F_RDLCK,
+ F_WRLCK
+ };
+#ifdef HAVE_FLOCK
+ short mode_flock[DB_LOCK_WRITE + 1] = {
+ LOCK_UN,
+ LOCK_SH,
+ LOCK_EX
+ };
+#endif
dbenv = env == NULL ? NULL : env->dbenv;
DB_ASSERT(env, F_ISSET(fhp, DB_FH_OPENED) && fhp->fd != -1);
+ DB_ASSERT(env, lockmode <= DB_LOCK_WRITE);
- if (dbenv != NULL && FLD_ISSET(dbenv->verbose, DB_VERB_FILEOPS_ALL))
- __db_msg(env, DB_STR_A("0138",
- "fileops: flock %s %s offset %lu", "%s %s %lu"), fhp->name,
- acquire ? DB_STR_P("acquire"): DB_STR_P("release"),
- (u_long)offset);
-
- fl.l_start = offset;
- fl.l_len = 1;
- fl.l_type = acquire ? F_WRLCK : F_UNLCK;
- fl.l_whence = SEEK_SET;
-
- RETRY_CHK_EINTR_ONLY(
- (fcntl(fhp->fd, nowait ? F_SETLK : F_SETLKW, &fl)), ret);
+ if (dbenv != NULL && FLD_ISSET(dbenv->verbose, DB_VERB_FILEOPS_ALL)) {
+ if (offset < 0)
+ __db_msg(env, DB_STR_A("####",
+ "fileops: flock %s %s %s", "%s %s %s"),
+ fhp->name, mode_string[lockmode],
+ nowait ? "nowait" : "");
+ else
+ __db_msg(env, DB_STR_A("0020",
+ "fileops: fcntls %s %s offset %lu", "%s %s %lu"),
+ fhp->name, mode_string[lockmode], (u_long)offset);
+ }
+
+ if (offset < 0) {
+#ifdef HAVE_FLOCK
+ RETRY_CHK_EINTR_ONLY(flock(fhp->fd,
+ mode_flock[lockmode] | (nowait ? LOCK_NB : 0)), ret);
+#else
+ ret = __os_filelocking_notsup(env);
+#endif
+ } else {
+ fl.l_start = offset;
+ fl.l_len = 1;
+ fl.l_whence = SEEK_SET;
+ fl.l_type = mode_fcntl[lockmode];
+ RETRY_CHK_EINTR_ONLY(
+ fcntl(fhp->fd, nowait ? F_SETLK : F_SETLKW, &fl), ret);
+ }
+
+ if (offset < 0 && dbenv != NULL &&
+ FLD_ISSET(dbenv->verbose, DB_VERB_FILEOPS_ALL))
+ __db_msg(env, DB_STR_A("####",
+ "fileops: flock %s %s %s returns %s", "%s %s %s"),
+ fhp->name, mode_string[lockmode],
+ nowait ? "nowait" : "", db_strerror(ret));
if (ret == 0)
return (0);
@@ -53,12 +110,29 @@ __os_fdlock(env, fhp, offset, acquire, n
__db_syserr(env, ret, DB_STR("0139", "fcntl"));
return (t_ret);
#else
+ ret = __os_filelocking_notsup(env);
COMPQUIET(fhp, NULL);
- COMPQUIET(acquire, 0);
+ COMPQUIET(lockmode, 0);
COMPQUIET(nowait, 0);
COMPQUIET(offset, 0);
+ return (ret)
+#endif
+}
+
+
+#if !defined(HAVE_FCNTL) || !defined(HAVE_FLOCK)
+/*
+ * __os_filelocking_notsup --
+ * Generate an error message if fcntl() or flock() is requested on a
+ * platform that does not support it.
+ *
+ */
+static int
+__os_filelocking_notsup(env)
+ ENV *env;
+{
__db_syserr(env, DB_OPNOTSUP, DB_STR("0140",
"advisory file locking unavailable"));
return (DB_OPNOTSUP);
-#endif
}
+#endif
diff -up db-5.3.28/src/os/os_map.c.pthreads db-5.3.28/src/os/os_map.c
--- db-5.3.28/src/os/os_map.c.pthreads 2013-09-09 17:35:09.000000000 +0200
+++ db-5.3.28/src/os/os_map.c 2017-06-13 11:15:15.325215121 +0200
@@ -32,7 +32,7 @@ static int __no_system_mem __P((ENV *));
/*
* __os_attach --
- * Create/join a shared memory region.
+ * Create/join a 'shared' region of Berkeley DB memory.
*
* PUBLIC: int __os_attach __P((ENV *, REGINFO *, REGION *));
*/
@@ -50,6 +50,7 @@ __os_attach(env, infop, rp)
* so there must be a valid handle.
*/
DB_ASSERT(env, env != NULL && env->dbenv != NULL);
+ DB_ASSERT(env, infop->fhp == NULL);
dbenv = env->dbenv;
if (DB_GLOBAL(j_region_map) != NULL) {

11
db-5.3.28-cwd-db_config.patch
View File

@ -1,11 +0,0 @@
--- db-5.3.28/src/env/env_open.c.old 2017-06-26 10:32:11.011419981 +0200
+++ db-5.3.28/src/env/env_open.c 2017-06-26 10:32:46.893721233 +0200
@@ -473,7 +473,7 @@
env->db_mode = mode == 0 ? DB_MODE_660 : mode;
/* Read the DB_CONFIG file. */
- if ((ret = __env_read_db_config(env)) != 0)
+ if (env->db_home != NULL && (ret = __env_read_db_config(env)) != 0)
return (ret);
/*

20
db-5.3.28-lemon_hash.patch
View File

@ -1,20 +0,0 @@
--- db-5.3.28/lang/sql/sqlite/tool/lemon.c.lemon_hash 2013-09-09 17:35:07.000000000 +0200
+++ db-5.3.28/lang/sql/sqlite/tool/lemon.c 2017-02-22 13:12:08.564106051 +0100
@@ -3428,7 +3428,7 @@
int maxdtlength; /* Maximum length of any ".datatype" field. */
char *stddt; /* Standardized name for a datatype */
int i,j; /* Loop counters */
- int hash; /* For hashing the name of a type */
+ unsigned hash; /* For hashing the name of a type */
const char *name; /* Name of the parser */
/* Allocate and initialize types[] and allocate stddt[] */
@@ -3491,7 +3491,7 @@
break;
}
hash++;
- if( hash>=arraysize ) hash = 0;
+ if( hash>=(unsigned)arraysize ) hash = 0;
}
if( types[hash]==0 ){
sp->dtnum = hash + 1;

19
db-5.3.28-mmap-high-cpu-usage.patch
View File

@ -1,19 +0,0 @@
Author: Filip Januš <fjanus@redhat.com>
Date: 6 Sep 2021
Related: https://bugzilla.redhat.com/show_bug.cgi?id=1992402
Patch was created based on the discussion in the previous link
diff -ur db-5.3.28/src/os/os_map.c db_patch/src/os/os_map.c
--- db-5.3.28/src/os/os_map.c 2013-09-09 17:35:09.000000000 +0200
+++ db_patch/src/os/os_map.c 2021-09-09 07:33:12.027328265 +0200
@@ -213,7 +213,10 @@
if (rp->max < rp->size)
rp->max = rp->size;
if (ret == 0 && F_ISSET(infop, REGION_CREATE)) {
- if (F_ISSET(dbenv, DB_ENV_REGION_INIT))
+
+ rp->size = rp->max;
+
+ if (F_ISSET(dbenv, DB_ENV_REGION_INIT))
ret = __db_file_write(env, infop->fhp,
rp->size / MEGABYTE, rp->size % MEGABYTE, 0x00);
else

99
db-5.3.28-rpm-lock-check.patch
View File

@ -1,99 +0,0 @@
diff -up db-5.3.28/src/dbinc_auto/int_def.in.rpmlock db-5.3.28/src/dbinc_auto/int_def.in
--- db-5.3.28/src/dbinc_auto/int_def.in.rpmlock 2017-06-26 15:09:17.883356255 +0200
+++ db-5.3.28/src/dbinc_auto/int_def.in 2017-06-26 15:09:27.421170401 +0200
@@ -1573,6 +1573,7 @@
#define __os_posix_err __os_posix_err@DB_VERSION_UNIQUE_NAME@
#define __os_fileid __os_fileid@DB_VERSION_UNIQUE_NAME@
#define __check_lock_fn __check_lock_fn@DB_VERSION_UNIQUE_NAME@
+#define __rpm_lock_free __rpm_lock_free@DB_VERSION_UNIQUE_NAME@
#define __os_fdlock __os_fdlock@DB_VERSION_UNIQUE_NAME@
#define __os_fsync __os_fsync@DB_VERSION_UNIQUE_NAME@
#define __os_getenv __os_getenv@DB_VERSION_UNIQUE_NAME@
diff -up db-5.3.28/src/dbinc_auto/os_ext.h.rpmlock db-5.3.28/src/dbinc_auto/os_ext.h
--- db-5.3.28/src/dbinc_auto/os_ext.h.rpmlock 2017-06-26 15:09:21.940277203 +0200
+++ db-5.3.28/src/dbinc_auto/os_ext.h 2017-06-26 15:09:27.354171707 +0200
@@ -42,6 +42,7 @@ char *__os_strerror __P((int, char *, si
int __os_posix_err __P((int));
int __os_fileid __P((ENV *, const char *, int, u_int8_t *));
int __check_lock_fn __P((char *, pid_t));
+int __rpm_lock_free __P((ENV *));
int __os_fdlock __P((ENV *, DB_FH *, off_t, db_lockmode_t, int));
int __os_fsync __P((ENV *, DB_FH *));
int __os_getenv __P((ENV *, const char *, char **, size_t));
diff -up db-5.3.28/src/env/env_region.c.rpmlock db-5.3.28/src/env/env_region.c
--- db-5.3.28/src/env/env_region.c.rpmlock 2017-06-26 15:09:12.479461558 +0200
+++ db-5.3.28/src/env/env_region.c 2017-06-26 15:09:12.481461519 +0200
@@ -291,18 +291,23 @@ user_map_functions:
if (create_ok &&
ret == DB_OLD_VERSION &&
ENV_PRIMARY_LOCK(env, DB_LOCK_WRITE, 1) == 0) {
- if (FLD_ISSET(dbenv->verbose, DB_VERB_RECOVERY))
- __db_msg(env, "Recreating idle environment");
- F_SET(infop, REGION_CREATE_OK);
+ /* If the rpm transaction lock is taken we cannot safely rebuild */
+ if (!__rpm_lock_free(env))
+ ENV_PRIMARY_UNLOCK(env);
+ else {
+ if (FLD_ISSET(dbenv->verbose, DB_VERB_RECOVERY))
+ __db_msg(env, "Recreating idle environment");
+ F_SET(infop, REGION_CREATE_OK);
- /*
- * Detach from the environment region; we need to unmap it (and
- * close any file handle) so that we don't leak memory or files.
- */
- DB_ASSERT(env, infop->rp == NULL);
- infop->rp = &tregion;
- (void)__env_sys_detach(env, infop, 0);
- goto creation;
+ /*
+ * Detach from the environment region; we need to unmap it (and
+ * close any file handle) so that we don't leak memory or files.
+ */
+ DB_ASSERT(env, infop->rp == NULL);
+ infop->rp = &tregion;
+ (void)__env_sys_detach(env, infop, 0);
+ goto creation;
+ }
}
if (renv->majver != DB_VERSION_MAJOR ||
diff -up db-5.3.28/src/os/os_flock.c.rpmlock db-5.3.28/src/os/os_flock.c
--- db-5.3.28/src/os/os_flock.c.rpmlock 2017-06-26 15:09:12.480461538 +0200
+++ db-5.3.28/src/os/os_flock.c 2017-06-26 15:09:12.481461519 +0200
@@ -79,6 +79,35 @@ int __check_lock_fn(fn, pid)
}
/*
+ * __rpm_lock_free --
+ * Try to look at a lock used by rpm to see if libdb is being
+ * updated and it is safe to access its environment files.
+ * PUBLIC: int __rpm_lock_free __P((ENV *));
+ */
+
+#define RPM_PATH SHAREDSTATEDIR "/rpm"
+#define RPMLOCK_PATH RPM_PATH "/.rpm.lock"
+
+int __rpm_lock_free(env)
+ ENV *env;
+{
+ int ret;
+
+ /* No need to check the transaction lock if not in rpm */
+ if (strstr(env->db_home, RPM_PATH) == NULL)
+ return 1;
+
+ /* Assume it is safe to rebuild if the lock file does not exist */
+ if (access(RPMLOCK_PATH, F_OK) && errno == ENOENT)
+ return 1;
+
+ ret = __check_lock_fn(RPMLOCK_PATH, 0);
+ /* __check_lock_fn can return -1 on failure - return 0 (taken) instead */
+ return ret == -1 ? 0: ret;
+
+}
+
+/*
* __os_fdlock --
* Acquire/release a lock on a byte in a file.
*

694
db-5.3.28_cve-2019-2708.patch
View File

@ -1,694 +0,0 @@
--- db-18.1.32/src/btree/bt_cursor.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/btree/bt_cursor.c 2020-05-29 23:28:22.000000000 +0530
@@ -282,6 +282,8 @@
*
* Recno uses the btree bt_ovflsize value -- it's close enough.
*/
+ if (t->bt_minkey == 0)
+ return (DB_RECOVER);
cp->ovflsize = B_MINKEY_TO_OVFLSIZE(
dbp, F_ISSET(dbc, DBC_OPD) ? 2 : t->bt_minkey, dbp->pgsize);
--- db-18.1.32/src/btree/bt_verify.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/btree/bt_verify.c 2020-05-29 23:28:22.000000000 +0530
@@ -700,7 +700,11 @@
isbad = 1;
goto err;
default:
+ if (ret == 0) {
+ isbad = 1;
+ ret = DB_VERIFY_FATAL;
+ goto err;
+ }
- DB_ASSERT(env, ret != 0);
break;
}
@@ -1074,7 +1078,7 @@
DBT dbta, dbtb, dup_1, dup_2, *p1, *p2, *tmp;
ENV *env;
PAGE *child;
+ db_pgno_t cpgno, grandparent;
- db_pgno_t cpgno;
VRFY_PAGEINFO *pip;
db_indx_t i, *inp;
int adj, cmp, freedup_1, freedup_2, isbad, ret, t_ret;
@@ -1106,7 +1110,8 @@
buf1 = buf2 = NULL;
+ if (LF_ISSET(DB_NOORDERCHK))
+ return (EINVAL);
- DB_ASSERT(env, !LF_ISSET(DB_NOORDERCHK));
dupfunc = (dbp->dup_compare == NULL) ? __bam_defcmp : dbp->dup_compare;
if (TYPE(h) == P_LDUP)
@@ -1115,6 +1120,7 @@
func = __bam_defcmp;
if (dbp->bt_internal != NULL) {
bt = (BTREE *)dbp->bt_internal;
+ grandparent = bt->bt_root;
if (TYPE(h) == P_IBTREE && (bt->bt_compare != NULL ||
dupfunc != __bam_defcmp)) {
/*
@@ -974,8 +980,24 @@
*/
mpf = dbp->mpf;
child = h;
+ cpgno = pgno;
while (TYPE(child) == P_IBTREE) {
+ if (NUM_ENT(child) == 0) {
+ EPRINT((env, DB_STR_A("1088",
+ "Page %lu: internal page is empty and should not be",
+ "%lu"), (u_long)cpgno));
+ ret = DB_VERIFY_BAD;
+ goto err;
+ }
bi = GET_BINTERNAL(dbp, child, 0);
+ if (grandparent == bi->pgno) {
+ EPRINT((env, DB_STR_A("5552",
+ "Page %lu: found twice in the btree",
+ "%lu"), (u_long)grandparent));
+ ret = DB_VERIFY_FATAL;
+ goto err;
+ } else
+ grandparent = cpgno;
cpgno = bi->pgno;
if (child != h &&
(ret = __memp_fput(mpf,
@@ -1402,7 +1416,10 @@
*/
if (dup_1.data == NULL ||
dup_2.data == NULL) {
+ if (ovflok) {
+ isbad = 1;
+ goto err;
+ }
- DB_ASSERT(env, !ovflok);
if (pip != NULL)
F_SET(pip,
VRFY_INCOMPLETE);
@@ -1747,9 +1764,10 @@
(ret = __db_vrfy_ovfl_structure(dbp, vdp,
child->pgno, child->tlen,
flags | DB_ST_OVFL_LEAF)) != 0) {
+ if (ret == DB_VERIFY_BAD) {
- if (ret == DB_VERIFY_BAD)
isbad = 1;
+ break;
+ } else
- else
goto done;
}
@@ -1823,9 +1841,10 @@
stflags | DB_ST_TOPLEVEL,
NULL, NULL, NULL)) != 0) {
if (ret ==
+ DB_VERIFY_BAD) {
- DB_VERIFY_BAD)
isbad = 1;
+ break;
+ } else
- else
goto err;
}
}
@@ -1969,7 +1988,10 @@
*/
/* Otherwise, __db_vrfy_childput would be broken. */
+ if (child->refcnt < 1) {
+ isbad = 1;
+ goto err;
+ }
- DB_ASSERT(env, child->refcnt >= 1);
/*
* An overflow referenced more than twice here
@@ -1986,9 +2008,10 @@
if ((ret = __db_vrfy_ovfl_structure(dbp,
vdp, child->pgno, child->tlen,
flags)) != 0) {
+ if (ret == DB_VERIFY_BAD) {
- if (ret == DB_VERIFY_BAD)
isbad = 1;
+ break;
+ } else
- else
goto done;
}
}
@@ -2026,9 +2049,10 @@
if ((ret = __bam_vrfy_subtree(dbp, vdp, li->pgno,
i == 0 ? NULL : li, ri, flags, &child_level,
&child_nrecs, NULL)) != 0) {
+ if (ret == DB_VERIFY_BAD) {
- if (ret == DB_VERIFY_BAD)
isbad = 1;
+ break;
+ } else
- else
goto done;
}
@@ -2929,7 +2953,11 @@
db_pgno_t current, p;
int err_ret, ret;
+ if (pgset == NULL) {
+ EPRINT((dbp->env, DB_STR("5542",
+ "Error, database contains no visible pages.")));
+ return (DB_RUNRECOVERY);
+ }
- DB_ASSERT(dbp->env, pgset != NULL);
mpf = dbp->mpf;
h = NULL;
--- db-18.1.32/src/db/db_conv.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/db/db_conv.c 2020-05-29 23:28:22.000000000 +0530
@@ -493,8 +493,11 @@
db_indx_t i, *inp, len, tmp;
u_int8_t *end, *p, *pgend;
- if (pagesize == 0)
- return (0);
+ /* This function is also used to byteswap logs, so
+ * the pagesize might not be an actual page size.
+ */
+ if (!(pagesize >= 24 && pagesize <= DB_MAX_PGSIZE))
+ return (EINVAL);
if (pgin) {
M_32_SWAP(h->lsn.file);
@@ -513,26 +516,41 @@
pgend = (u_int8_t *)h + pagesize;
inp = P_INP(dbp, h);
- if ((u_int8_t *)inp >= pgend)
- goto out;
+ if ((u_int8_t *)inp > pgend)
+ return (__db_pgfmt(env, pg));
switch (TYPE(h)) {
case P_HASH_UNSORTED:
case P_HASH:
for (i = 0; i < NUM_ENT(h); i++) {
+ if ((u_int8_t*)(inp + i) >= pgend)
+ return (__db_pgfmt(env, pg));
+ if (inp[i] == 0)
+ continue;
if (pgin)
M_16_SWAP(inp[i]);
+ if (inp[i] >= pagesize)
+ return (__db_pgfmt(env, pg));
- if (P_ENTRY(dbp, h, i) >= pgend)
- continue;
+ if (P_ENTRY(dbp, h, i) >= pgend)
+ return (__db_pgfmt(env, pg));
switch (HPAGE_TYPE(dbp, h, i)) {
case H_KEYDATA:
break;
case H_DUPLICATE:
+ if (LEN_HITEM(dbp, h, pagesize, i) <
+ HKEYDATA_SIZE(0))
+ return (__db_pgfmt(env, pg));
+
len = LEN_HKEYDATA(dbp, h, pagesize, i);
p = HKEYDATA_DATA(P_ENTRY(dbp, h, i));
- for (end = p + len; p < end;) {
+
+ end = p + len;
+ if (end > pgend)
+ return (__db_pgfmt(env, pg));
+
+ while (p < end) {
if (pgin) {
P_16_SWAP(p);
memcpy(&tmp,
@@ -544,14 +562,20 @@
SWAP16(p);
}
p += tmp;
+ if (p >= end)
+ return (__db_pgfmt(env, pg));
SWAP16(p);
}
break;
case H_OFFDUP:
+ if ((inp[i] + HOFFDUP_SIZE) > pagesize)
+ return (__db_pgfmt(env, pg));
p = HOFFPAGE_PGNO(P_ENTRY(dbp, h, i));
SWAP32(p); /* pgno */
break;
case H_OFFPAGE:
+ if ((inp[i] + HOFFPAGE_SIZE) > pagesize)
+ return (__db_pgfmt(env, pg));
p = HOFFPAGE_PGNO(P_ENTRY(dbp, h, i));
SWAP32(p); /* pgno */
SWAP32(p); /* tlen */
@@ -559,7 +583,6 @@
default:
return (__db_pgfmt(env, pg));
}
-
}
/*
@@ -576,8 +599,12 @@
case P_LDUP:
case P_LRECNO:
for (i = 0; i < NUM_ENT(h); i++) {
+ if ((u_int8_t *)(inp + i) >= pgend)
+ return (__db_pgfmt(env, pg));
if (pgin)
M_16_SWAP(inp[i]);
+ if (inp[i] >= pagesize)
+ return (__db_pgfmt(env, pg));
/*
* In the case of on-page duplicates, key information
@@ -597,7 +624,7 @@
bk = GET_BKEYDATA(dbp, h, i);
if ((u_int8_t *)bk >= pgend)
- continue;
+ return (__db_pgfmt(env, pg));
switch (B_TYPE(bk->type)) {
case B_KEYDATA:
M_16_SWAP(bk->len);
@@ -605,6 +632,8 @@
case B_DUPLICATE:
case B_OVERFLOW:
bo = (BOVERFLOW *)bk;
+ if (((u_int8_t *)bo + BOVERFLOW_SIZE) > pgend)
+ return (__db_pgfmt(env, pg));
M_32_SWAP(bo->pgno);
M_32_SWAP(bo->tlen);
break;
@@ -618,12 +647,17 @@
break;
case P_IBTREE:
for (i = 0; i < NUM_ENT(h); i++) {
+ if ((u_int8_t *)(inp + i) > pgend)
+ return (__db_pgfmt(env, pg));
if (pgin)
M_16_SWAP(inp[i]);
+ if ((u_int16_t)(inp[i] +
+ BINTERNAL_SIZE(0) - 1) > pagesize)
+ break;
bi = GET_BINTERNAL(dbp, h, i);
- if ((u_int8_t *)bi >= pgend)
- continue;
+ if (((u_int8_t *)bi + BINTERNAL_SIZE(0)) > pgend)
+ return (__db_pgfmt(env, pg));
M_16_SWAP(bi->len);
M_32_SWAP(bi->pgno);
@@ -634,6 +668,10 @@
break;
case B_DUPLICATE:
case B_OVERFLOW:
+ if ((u_int16_t)(inp[i] +
+ BINTERNAL_SIZE(BOVERFLOW_SIZE) - 1) >
+ pagesize)
+ goto out;
bo = (BOVERFLOW *)bi->data;
M_32_SWAP(bo->pgno);
M_32_SWAP(bo->tlen);
@@ -648,12 +686,16 @@
break;
case P_IRECNO:
for (i = 0; i < NUM_ENT(h); i++) {
+ if ((u_int8_t *)(inp + i) >= pgend)
+ return (__db_pgfmt(env, pg));
if (pgin)
M_16_SWAP(inp[i]);
+ if (inp[i] >= pagesize)
+ return (__db_pgfmt(env, pg));
ri = GET_RINTERNAL(dbp, h, i);
- if ((u_int8_t *)ri >= pgend)
- continue;
+ if ((((u_int8_t *)ri) + RINTERNAL_SIZE) > pgend)
+ return (__db_pgfmt(env, pg));
M_32_SWAP(ri->pgno);
M_32_SWAP(ri->nrecs);
--- db-18.1.32/src/db/db_vrfy.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/db/db_vrfy.c 2020-05-29 23:28:22.000000000 +0530
@@ -381,8 +381,10 @@
vdp, name, 0, lp, rp, flags)) != 0) {
if (t_ret == DB_VERIFY_BAD)
isbad = 1;
+ else {
+ ret = t_ret;
+ goto err;
+ }
- else
- goto err;
}
/*
@@ -771,9 +773,10 @@
*/
if ((t_ret = __memp_fget(mpf, &i,
vdp->thread_info, NULL, 0, &h)) != 0) {
+ if ((dbp->type == DB_HASH ||
- if (dbp->type == DB_HASH ||
(dbp->type == DB_QUEUE &&
+ F_ISSET(dbp, DB_AM_INMEM))) &&
+ t_ret != DB_RUNRECOVERY) {
- F_ISSET(dbp, DB_AM_INMEM))) {
if ((t_ret =
__db_vrfy_getpageinfo(vdp, i, &pip)) != 0)
goto err1;
@@ -945,6 +948,8 @@
return (ret == 0 ? t_ret : ret);
}
+ if (ret == DB_PAGE_NOTFOUND && isbad == 1)
+ ret = 0;
return ((isbad == 1 && ret == 0) ? DB_VERIFY_BAD : ret);
}
@@ -1581,7 +1586,7 @@
if (pgno == PGNO_BASE_MD &&
dbtype != DB_QUEUE && meta->last_pgno != vdp->last_pgno) {
#ifdef HAVE_FTRUNCATE
+ ret = DB_VERIFY_FATAL;
- isbad = 1;
EPRINT((env, DB_STR_A("0552",
"Page %lu: last_pgno is not correct: %lu != %lu",
"%lu %lu %lu"), (u_long)pgno,
@@ -1622,7 +1627,11 @@
env = dbp->env;
pgset = vdp->pgset;
+ if (pgset == NULL) {
+ EPRINT((env, DB_STR("5543",
+ "Error, database contains no visible pages.")));
+ return (DB_RUNRECOVERY);
+ }
- DB_ASSERT(env, pgset != NULL);
if ((ret = __db_vrfy_getpageinfo(vdp, meta, &pip)) != 0)
return (ret);
@@ -2014,7 +2023,8 @@
int keyflag, ret, t_ret;
env = dbp->env;
+ if (!LF_ISSET(DB_SALVAGE))
+ return (EINVAL);
- DB_ASSERT(env, LF_ISSET(DB_SALVAGE));
/*
* !!!
@@ -2126,10 +2136,8 @@
int (*callback) __P((void *, const void *));
u_int32_t flags;
{
- ENV *env;
-
- env = dbp->env;
- DB_ASSERT(env, LF_ISSET(DB_SALVAGE));
+ if (!LF_ISSET(DB_SALVAGE))
+ return (EINVAL);
/* If we got this page in the subdb pass, we can safely skip it. */
if (__db_salvage_isdone(vdp, pgno))
@@ -2242,8 +2253,8 @@
ret = t_ret;
break;
case SALVAGE_OVERFLOW:
+ EPRINT((env, DB_STR("5544", "Invalid page type to salvage.")));
+ return (EINVAL);
- DB_ASSERT(env, 0); /* Shouldn't ever happen. */
- break;
case SALVAGE_HASH:
if ((t_ret = __ham_salvage(dbp, vdp,
pgno, h, handle, callback, flags)) != 0 && ret == 0)
@@ -2256,8 +2267,8 @@
* Shouldn't happen, but if it does, just do what the
* nice man says.
*/
+ EPRINT((env, DB_STR("5545", "Invalid page type to salvage.")));
+ return (EINVAL);
- DB_ASSERT(env, 0);
- break;
}
if ((t_ret = __memp_fput(mpf,
vdp->thread_info, h, dbp->priority)) != 0 && ret == 0)
@@ -2303,8 +2314,8 @@
ret = t_ret;
break;
default:
+ EPRINT((env, DB_STR("5546", "Invalid page type to salvage.")));
+ return (EINVAL);
- DB_ASSERT(env, 0); /* Shouldn't ever happen. */
- break;
}
if ((t_ret = __memp_fput(mpf,
vdp->thread_info, h, dbp->priority)) != 0 && ret == 0)
@@ -2361,7 +2372,10 @@
env = dbp->env;
+ if (himarkp == NULL) {
+ __db_msg(env, "Page %lu index has no end.", (u_long)pgno);
+ return (DB_VERIFY_FATAL);
+ }
- DB_ASSERT(env, himarkp != NULL);
inp = P_INP(dbp, h);
/*
@@ -2783,7 +2797,11 @@
goto err;
ovfl_bufsz = bkkey->len + 1;
}
+ if (subdbname == NULL) {
+ EPRINT((env, DB_STR("5547", "Subdatabase cannot be null.")));
+ ret = EINVAL;
+ goto err;
+ }
- DB_ASSERT(env, subdbname != NULL);
memcpy(subdbname, bkkey->data, bkkey->len);
subdbname[bkkey->len] = '\0';
}
--- db-18.1.32/src/db/db_vrfyutil.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/db/db_vrfyutil.c 2020-05-29 23:28:22.000000000 +0530
@@ -214,7 +214,8 @@
if ((ret = __db_get(pgdbp,
vdp->thread_info, vdp->txn, &key, &data, 0)) == 0) {
/* Found it. */
+ if (data.size != sizeof(VRFY_PAGEINFO))
+ return (DB_VERIFY_FATAL);
- DB_ASSERT(env, data.size == sizeof(VRFY_PAGEINFO));
pip = data.data;
LIST_INSERT_HEAD(&vdp->activepips, pip, links);
goto found;
@@ -342,7 +343,8 @@
F_SET(&data, DB_DBT_USERMEM);
if ((ret = __db_get(dbp, ip, txn, &key, &data, 0)) == 0) {
+ if (data.size != sizeof(int))
+ return (EINVAL);
- DB_ASSERT(dbp->env, data.size == sizeof(int));
} else if (ret == DB_NOTFOUND)
val = 0;
else
@@ -382,7 +384,8 @@
F_SET(&data, DB_DBT_USERMEM);
if ((ret = __db_get(dbp, ip, txn, &key, &data, 0)) == 0) {
+ if (data.size != sizeof(int))
+ return (DB_VERIFY_FATAL);
- DB_ASSERT(dbp->env, data.size == sizeof(int));
} else if (ret != DB_NOTFOUND)
return (ret);
@@ -419,7 +422,8 @@
if ((ret = __dbc_get(dbc, &key, &data, DB_NEXT)) != 0)
return (ret);
+ if (key.size != sizeof(db_pgno_t))
+ return (DB_VERIFY_FATAL);
- DB_ASSERT(dbc->env, key.size == sizeof(db_pgno_t));
*pgnop = pgno;
return (0);
@@ -566,7 +570,8 @@
if ((ret = __dbc_get(dbc, &key, &data, DB_SET)) != 0)
return (ret);
+ if (data.size != sizeof(VRFY_CHILDINFO))
+ return (DB_VERIFY_FATAL);
- DB_ASSERT(dbc->env, data.size == sizeof(VRFY_CHILDINFO));
*cipp = (VRFY_CHILDINFO *)data.data;
return (0);
@@ -594,7 +599,8 @@
if ((ret = __dbc_get(dbc, &key, &data, DB_NEXT_DUP)) != 0)
return (ret);
+ if (data.size != sizeof(VRFY_CHILDINFO))
+ return (DB_VERIFY_FATAL);
- DB_ASSERT(dbc->env, data.size == sizeof(VRFY_CHILDINFO));
*cipp = (VRFY_CHILDINFO *)data.data;
return (0);
@@ -721,7 +727,8 @@
return (ret);
while ((ret = __dbc_get(*dbcp, &key, &data, DB_NEXT)) == 0) {
+ if (data.size != sizeof(u_int32_t))
+ return (DB_VERIFY_FATAL);
- DB_ASSERT(dbp->env, data.size == sizeof(u_int32_t));
memcpy(&pgtype, data.data, sizeof(pgtype));
if (skip_overflow && pgtype == SALVAGE_OVERFLOW)
@@ -730,8 +737,9 @@
if ((ret = __dbc_del(*dbcp, 0)) != 0)
return (ret);
if (pgtype != SALVAGE_IGNORE) {
+ if (key.size != sizeof(db_pgno_t)
+ || data.size != sizeof(u_int32_t))
+ return (DB_VERIFY_FATAL);
- DB_ASSERT(dbp->env, key.size == sizeof(db_pgno_t));
- DB_ASSERT(dbp->env, data.size == sizeof(u_int32_t));
*pgnop = *(db_pgno_t *)key.data;
*pgtypep = *(u_int32_t *)data.data;
--- db-18.1.32/src/db/partition.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/db/partition.c 2020-05-29 23:28:22.000000000 +0530
@@ -461,9 +461,19 @@
} else
part->nparts = meta->nparts;
} else if (meta->nparts != 0 && part->nparts != meta->nparts) {
+ ret = EINVAL;
__db_errx(env, DB_STR("0656",
"Number of partitions does not match."));
- ret = EINVAL;
+ goto err;
+ }
+ /*
+ * There is no limit on the number of partitions, but I cannot imagine a real
+ * database having more than 10000.
+ */
+ if (meta->nparts > 10000) {
+ ret = EINVAL;
+ __db_errx(env, DB_STR_A("5553",
+ "Too many partitions %lu", "%lu"), (u_long)(meta->nparts));
goto err;
}
@@ -2106,10 +2116,13 @@
memcpy(rp->data, key->data, key->size);
B_TSET(rp->type, B_KEYDATA);
}
+vrfy: if ((t_ret = __db_verify(*pdbp, ip, (*pdbp)->fname,
+ NULL, handle, callback,
+ lp, rp, flags | DB_VERIFY_PARTITION)) != 0 && ret == 0) {
+ ret = t_ret;
+ if (ret == ENOENT)
+ break;
+ }
-vrfy: if ((t_ret = __db_verify(*pdbp, ip, (*pdbp)->fname,
- NULL, handle, callback,
- lp, rp, flags | DB_VERIFY_PARTITION)) != 0 && ret == 0)
- ret = t_ret;
}
err: if (lp != NULL)
--- db-18.1.32/src/hash/hash_page.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/hash/hash_page.c 2020-05-29 23:28:22.000000000 +0530
@@ -869,7 +869,11 @@
/* Validate that next, prev pointers are OK */
n = NUM_ENT(p);
dbp = dbc->dbp;
+ if (n % 2 != 0) {
+ __db_errx(dbp->env, DB_STR_A("5549",
+ "Odd number of entries on page: %lu", "%lu"), (u_long)(p->pgno));
+ return (DB_VERIFY_FATAL);
+ }
- DB_ASSERT(dbp->env, n%2 == 0 );
env = dbp->env;
t = dbp->h_internal;
@@ -940,7 +944,12 @@
if ((ret = __db_prpage(dbp, p, DB_PR_PAGE)) != 0)
return (ret);
#endif
+ if (res >= 0) {
+ __db_errx(env, DB_STR_A("5550",
+ "Odd number of entries on page: %lu", "%lu"),
+ (u_long)p->pgno);
+ return (DB_VERIFY_FATAL);
+ }
- DB_ASSERT(dbp->env, res < 0);
}
prev = curr;
--- db-18.1.32/src/hash/hash_verify.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/hash/hash_verify.c 2020-05-29 23:28:22.000000000 +0530
@@ -615,7 +615,7 @@
isbad = 1;
else
goto err;
+ }
- }
/*
* There may be unused hash pages corresponding to buckets
@@ -746,7 +746,7 @@
"Page %lu: impossible first page in bucket %lu", "%lu %lu"),
(u_long)pgno, (u_long)bucket));
/* Unsafe to continue. */
+ ret = DB_VERIFY_FATAL;
- isbad = 1;
goto err;
}
@@ -776,7 +776,7 @@
EPRINT((env, DB_STR_A("1116",
"Page %lu: hash page referenced twice", "%lu"),
(u_long)pgno));
+ ret = DB_VERIFY_FATAL;
- isbad = 1;
/* Unsafe to continue. */
goto err;
} else if ((ret = __db_vrfy_pgset_inc(vdp->pgset,
@@ -1307,7 +1307,11 @@
COMPQUIET(flags, 0);
ip = vdp->thread_info;
+ if (pgset == NULL) {
+ EPRINT((dbp->env, DB_STR("5548",
+ "Error, database contains no visible pages.")));
+ return (DB_VERIFY_FATAL);
+ }
- DB_ASSERT(dbp->env, pgset != NULL);
mpf = dbp->mpf;
totpgs = 0;
--- db-18.1.32/src/qam/qam_verify.c 2019-02-20 03:21:20.000000000 +0530
+++ db-18.1.40/src/qam/qam_verify.c 2020-05-29 23:28:22.000000000 +0530
@@ -465,7 +465,14 @@
/* Verify/salvage each page. */
if ((ret = __db_cursor(dbp, vdp->thread_info, NULL, &dbc, 0)) != 0)
return (ret);
-begin: for (; i <= stop; i++) {
+begin: if ((stop - i) > 100000) {
+ EPRINT((env, DB_STR_A("5551",
+"Warning, many possible extends files (%lu), will take a long time to verify",
+ "%lu"), (u_long)(stop - i)));
+ }
+ for (; i <= stop; i++) {
+ if (i == UINT32_MAX)
+ break;
/*
* If DB_SALVAGE is set, we inspect our database of completed
* pages, and skip any we've already printed in the subdb pass.

1
dead.package Normal file
View File

@ -0,0 +1 @@
libdb package is retired on branch c10s for CS-2269

502
lgpl-2.1.txt
View File

@ -1,502 +0,0 @@
GNU LESSER GENERAL PUBLIC LICENSE
Version 2.1, February 1999
Copyright (C) 1991, 1999 Free Software Foundation, Inc.
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
[This is the first released version of the Lesser GPL. It also counts
as the successor of the GNU Library Public License, version 2, hence
the version number 2.1.]
Preamble
The licenses for most software are designed to take away your
freedom to share and change it. By contrast, the GNU General Public
Licenses are intended to guarantee your freedom to share and change
free software--to make sure the software is free for all its users.
This license, the Lesser General Public License, applies to some
specially designated software packages--typically libraries--of the
Free Software Foundation and other authors who decide to use it. You
can use it too, but we suggest you first think carefully about whether
this license or the ordinary General Public License is the better
strategy to use in any particular case, based on the explanations below.
When we speak of free software, we are referring to freedom of use,
not price. Our General Public Licenses are designed to make sure that
you have the freedom to distribute copies of free software (and charge
for this service if you wish); that you receive source code or can get
it if you want it; that you can change the software and use pieces of
it in new free programs; and that you are informed that you can do
these things.
To protect your rights, we need to make restrictions that forbid
distributors to deny you these rights or to ask you to surrender these
rights. These restrictions translate to certain responsibilities for
you if you distribute copies of the library or if you modify it.
For example, if you distribute copies of the library, whether gratis
or for a fee, you must give the recipients all the rights that we gave
you. You must make sure that they, too, receive or can get the source
code. If you link other code with the library, you must provide
complete object files to the recipients, so that they can relink them
with the library after making changes to the library and recompiling
it. And you must show them these terms so they know their rights.
We protect your rights with a two-step method: (1) we copyright the
library, and (2) we offer you this license, which gives you legal
permission to copy, distribute and/or modify the library.
To protect each distributor, we want to make it very clear that
there is no warranty for the free library. Also, if the library is
modified by someone else and passed on, the recipients should know
that what they have is not the original version, so that the original
author's reputation will not be affected by problems that might be
introduced by others.
Finally, software patents pose a constant threat to the existence of
any free program. We wish to make sure that a company cannot
effectively restrict the users of a free program by obtaining a
restrictive license from a patent holder. Therefore, we insist that
any patent license obtained for a version of the library must be
consistent with the full freedom of use specified in this license.
Most GNU software, including some libraries, is covered by the
ordinary GNU General Public License. This license, the GNU Lesser
General Public License, applies to certain designated libraries, and
is quite different from the ordinary General Public License. We use
this license for certain libraries in order to permit linking those
libraries into non-free programs.
When a program is linked with a library, whether statically or using
a shared library, the combination of the two is legally speaking a
combined work, a derivative of the original library. The ordinary
General Public License therefore permits such linking only if the
entire combination fits its criteria of freedom. The Lesser General
Public License permits more lax criteria for linking other code with
the library.
We call this license the "Lesser" General Public License because it
does Less to protect the user's freedom than the ordinary General
Public License. It also provides other free software developers Less
of an advantage over competing non-free programs. These disadvantages
are the reason we use the ordinary General Public License for many
libraries. However, the Lesser license provides advantages in certain
special circumstances.
For example, on rare occasions, there may be a special need to
encourage the widest possible use of a certain library, so that it becomes
a de-facto standard. To achieve this, non-free programs must be
allowed to use the library. A more frequent case is that a free
library does the same job as widely used non-free libraries. In this
case, there is little to gain by limiting the free library to free
software only, so we use the Lesser General Public License.
In other cases, permission to use a particular library in non-free
programs enables a greater number of people to use a large body of
free software. For example, permission to use the GNU C Library in
non-free programs enables many more people to use the whole GNU
operating system, as well as its variant, the GNU/Linux operating
system.
Although the Lesser General Public License is Less protective of the
users' freedom, it does ensure that the user of a program that is
linked with the Library has the freedom and the wherewithal to run
that program using a modified version of the Library.
The precise terms and conditions for copying, distribution and
modification follow. Pay close attention to the difference between a
"work based on the library" and a "work that uses the library". The
former contains code derived from the library, whereas the latter must
be combined with the library in order to run.
GNU LESSER GENERAL PUBLIC LICENSE
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
0. This License Agreement applies to any software library or other
program which contains a notice placed by the copyright holder or
other authorized party saying it may be distributed under the terms of
this Lesser General Public License (also called "this License").
Each licensee is addressed as "you".
A "library" means a collection of software functions and/or data
prepared so as to be conveniently linked with application programs
(which use some of those functions and data) to form executables.
The "Library", below, refers to any such software library or work
which has been distributed under these terms. A "work based on the
Library" means either the Library or any derivative work under
copyright law: that is to say, a work containing the Library or a
portion of it, either verbatim or with modifications and/or translated
straightforwardly into another language. (Hereinafter, translation is
included without limitation in the term "modification".)
"Source code" for a work means the preferred form of the work for
making modifications to it. For a library, complete source code means
all the source code for all modules it contains, plus any associated
interface definition files, plus the scripts used to control compilation
and installation of the library.
Activities other than copying, distribution and modification are not
covered by this License; they are outside its scope. The act of
running a program using the Library is not restricted, and output from
such a program is covered only if its contents constitute a work based
on the Library (independent of the use of the Library in a tool for
writing it). Whether that is true depends on what the Library does
and what the program that uses the Library does.
1. You may copy and distribute verbatim copies of the Library's
complete source code as you receive it, in any medium, provided that
you conspicuously and appropriately publish on each copy an
appropriate copyright notice and disclaimer of warranty; keep intact
all the notices that refer to this License and to the absence of any
warranty; and distribute a copy of this License along with the
Library.
You may charge a fee for the physical act of transferring a copy,
and you may at your option offer warranty protection in exchange for a
fee.
2. You may modify your copy or copies of the Library or any portion
of it, thus forming a work based on the Library, and copy and
distribute such modifications or work under the terms of Section 1
above, provided that you also meet all of these conditions:
a) The modified work must itself be a software library.
b) You must cause the files modified to carry prominent notices
stating that you changed the files and the date of any change.
c) You must cause the whole of the work to be licensed at no
charge to all third parties under the terms of this License.
d) If a facility in the modified Library refers to a function or a
table of data to be supplied by an application program that uses
the facility, other than as an argument passed when the facility
is invoked, then you must make a good faith effort to ensure that,
in the event an application does not supply such function or
table, the facility still operates, and performs whatever part of
its purpose remains meaningful.
(For example, a function in a library to compute square roots has
a purpose that is entirely well-defined independent of the
application. Therefore, Subsection 2d requires that any
application-supplied function or table used by this function must
be optional: if the application does not supply it, the square
root function must still compute square roots.)
These requirements apply to the modified work as a whole. If
identifiable sections of that work are not derived from the Library,
and can be reasonably considered independent and separate works in
themselves, then this License, and its terms, do not apply to those
sections when you distribute them as separate works. But when you
distribute the same sections as part of a whole which is a work based
on the Library, the distribution of the whole must be on the terms of
this License, whose permissions for other licensees extend to the
entire whole, and thus to each and every part regardless of who wrote
it.
Thus, it is not the intent of this section to claim rights or contest
your rights to work written entirely by you; rather, the intent is to
exercise the right to control the distribution of derivative or
collective works based on the Library.
In addition, mere aggregation of another work not based on the Library
with the Library (or with a work based on the Library) on a volume of
a storage or distribution medium does not bring the other work under
the scope of this License.
3. You may opt to apply the terms of the ordinary GNU General Public
License instead of this License to a given copy of the Library. To do
this, you must alter all the notices that refer to this License, so
that they refer to the ordinary GNU General Public License, version 2,
instead of to this License. (If a newer version than version 2 of the
ordinary GNU General Public License has appeared, then you can specify
that version instead if you wish.) Do not make any other change in
these notices.
Once this change is made in a given copy, it is irreversible for
that copy, so the ordinary GNU General Public License applies to all
subsequent copies and derivative works made from that copy.
This option is useful when you wish to copy part of the code of
the Library into a program that is not a library.
4. You may copy and distribute the Library (or a portion or
derivative of it, under Section 2) in object code or executable form
under the terms of Sections 1 and 2 above provided that you accompany
it with the complete corresponding machine-readable source code, which
must be distributed under the terms of Sections 1 and 2 above on a
medium customarily used for software interchange.
If distribution of object code is made by offering access to copy
from a designated place, then offering equivalent access to copy the
source code from the same place satisfies the requirement to
distribute the source code, even though third parties are not
compelled to copy the source along with the object code.
5. A program that contains no derivative of any portion of the
Library, but is designed to work with the Library by being compiled or
linked with it, is called a "work that uses the Library". Such a
work, in isolation, is not a derivative work of the Library, and
therefore falls outside the scope of this License.
However, linking a "work that uses the Library" with the Library
creates an executable that is a derivative of the Library (because it
contains portions of the Library), rather than a "work that uses the
library". The executable is therefore covered by this License.
Section 6 states terms for distribution of such executables.
When a "work that uses the Library" uses material from a header file
that is part of the Library, the object code for the work may be a
derivative work of the Library even though the source code is not.
Whether this is true is especially significant if the work can be
linked without the Library, or if the work is itself a library. The
threshold for this to be true is not precisely defined by law.
If such an object file uses only numerical parameters, data
structure layouts and accessors, and small macros and small inline
functions (ten lines or less in length), then the use of the object
file is unrestricted, regardless of whether it is legally a derivative
work. (Executables containing this object code plus portions of the
Library will still fall under Section 6.)
Otherwise, if the work is a derivative of the Library, you may
distribute the object code for the work under the terms of Section 6.
Any executables containing that work also fall under Section 6,
whether or not they are linked directly with the Library itself.
6. As an exception to the Sections above, you may also combine or
link a "work that uses the Library" with the Library to produce a
work containing portions of the Library, and distribute that work
under terms of your choice, provided that the terms permit
modification of the work for the customer's own use and reverse
engineering for debugging such modifications.
You must give prominent notice with each copy of the work that the
Library is used in it and that the Library and its use are covered by
this License. You must supply a copy of this License. If the work
during execution displays copyright notices, you must include the
copyright notice for the Library among them, as well as a reference
directing the user to the copy of this License. Also, you must do one
of these things:
a) Accompany the work with the complete corresponding
machine-readable source code for the Library including whatever
changes were used in the work (which must be distributed under
Sections 1 and 2 above); and, if the work is an executable linked
with the Library, with the complete machine-readable "work that
uses the Library", as object code and/or source code, so that the
user can modify the Library and then relink to produce a modified
executable containing the modified Library. (It is understood
that the user who changes the contents of definitions files in the
Library will not necessarily be able to recompile the application
to use the modified definitions.)
b) Use a suitable shared library mechanism for linking with the
Library. A suitable mechanism is one that (1) uses at run time a
copy of the library already present on the user's computer system,
rather than copying library functions into the executable, and (2)
will operate properly with a modified version of the library, if
the user installs one, as long as the modified version is
interface-compatible with the version that the work was made with.
c) Accompany the work with a written offer, valid for at
least three years, to give the same user the materials
specified in Subsection 6a, above, for a charge no more
than the cost of performing this distribution.
d) If distribution of the work is made by offering access to copy
from a designated place, offer equivalent access to copy the above
specified materials from the same place.
e) Verify that the user has already received a copy of these
materials or that you have already sent this user a copy.
For an executable, the required form of the "work that uses the
Library" must include any data and utility programs needed for
reproducing the executable from it. However, as a special exception,
the materials to be distributed need not include anything that is
normally distributed (in either source or binary form) with the major
components (compiler, kernel, and so on) of the operating system on
which the executable runs, unless that component itself accompanies
the executable.
It may happen that this requirement contradicts the license
restrictions of other proprietary libraries that do not normally
accompany the operating system. Such a contradiction means you cannot
use both them and the Library together in an executable that you
distribute.
7. You may place library facilities that are a work based on the
Library side-by-side in a single library together with other library
facilities not covered by this License, and distribute such a combined
library, provided that the separate distribution of the work based on
the Library and of the other library facilities is otherwise
permitted, and provided that you do these two things:
a) Accompany the combined library with a copy of the same work
based on the Library, uncombined with any other library
facilities. This must be distributed under the terms of the
Sections above.
b) Give prominent notice with the combined library of the fact
that part of it is a work based on the Library, and explaining
where to find the accompanying uncombined form of the same work.
8. You may not copy, modify, sublicense, link with, or distribute
the Library except as expressly provided under this License. Any
attempt otherwise to copy, modify, sublicense, link with, or
distribute the Library is void, and will automatically terminate your
rights under this License. However, parties who have received copies,
or rights, from you under this License will not have their licenses
terminated so long as such parties remain in full compliance.
9. You are not required to accept this License, since you have not
signed it. However, nothing else grants you permission to modify or
distribute the Library or its derivative works. These actions are
prohibited by law if you do not accept this License. Therefore, by
modifying or distributing the Library (or any work based on the
Library), you indicate your acceptance of this License to do so, and
all its terms and conditions for copying, distributing or modifying
the Library or works based on it.
10. Each time you redistribute the Library (or any work based on the
Library), the recipient automatically receives a license from the
original licensor to copy, distribute, link with or modify the Library
subject to these terms and conditions. You may not impose any further
restrictions on the recipients' exercise of the rights granted herein.
You are not responsible for enforcing compliance by third parties with
this License.
11. If, as a consequence of a court judgment or allegation of patent
infringement or for any other reason (not limited to patent issues),
conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot
distribute so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you
may not distribute the Library at all. For example, if a patent
license would not permit royalty-free redistribution of the Library by
all those who receive copies directly or indirectly through you, then
the only way you could satisfy both it and this License would be to
refrain entirely from distribution of the Library.
If any portion of this section is held invalid or unenforceable under any
particular circumstance, the balance of the section is intended to apply,
and the section as a whole is intended to apply in other circumstances.
It is not the purpose of this section to induce you to infringe any
patents or other property right claims or to contest validity of any
such claims; this section has the sole purpose of protecting the
integrity of the free software distribution system which is
implemented by public license practices. Many people have made
generous contributions to the wide range of software distributed
through that system in reliance on consistent application of that
system; it is up to the author/donor to decide if he or she is willing
to distribute software through any other system and a licensee cannot
impose that choice.
This section is intended to make thoroughly clear what is believed to
be a consequence of the rest of this License.
12. If the distribution and/or use of the Library is restricted in
certain countries either by patents or by copyrighted interfaces, the
original copyright holder who places the Library under this License may add
an explicit geographical distribution limitation excluding those countries,
so that distribution is permitted only in or among countries not thus
excluded. In such case, this License incorporates the limitation as if
written in the body of this License.
13. The Free Software Foundation may publish revised and/or new
versions of the Lesser General Public License from time to time.
Such new versions will be similar in spirit to the present version,
but may differ in detail to address new problems or concerns.
Each version is given a distinguishing version number. If the Library
specifies a version number of this License which applies to it and
"any later version", you have the option of following the terms and
conditions either of that version or of any later version published by
the Free Software Foundation. If the Library does not specify a
license version number, you may choose any version ever published by
the Free Software Foundation.
14. If you wish to incorporate parts of the Library into other free
programs whose distribution conditions are incompatible with these,
write to the author to ask for permission. For software which is
copyrighted by the Free Software Foundation, write to the Free
Software Foundation; we sometimes make exceptions for this. Our
decision will be guided by the two goals of preserving the free status
of all derivatives of our free software and of promoting the sharing
and reuse of software generally.
NO WARRANTY
15. BECAUSE THE LIBRARY IS LICENSED FREE OF CHARGE, THERE IS NO
WARRANTY FOR THE LIBRARY, TO THE EXTENT PERMITTED BY APPLICABLE LAW.
EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR
OTHER PARTIES PROVIDE THE LIBRARY "AS IS" WITHOUT WARRANTY OF ANY
KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE
LIBRARY IS WITH YOU. SHOULD THE LIBRARY PROVE DEFECTIVE, YOU ASSUME
THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
16. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY
AND/OR REDISTRIBUTE THE LIBRARY AS PERMITTED ABOVE, BE LIABLE TO YOU
FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR
CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE
LIBRARY (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING
RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A
FAILURE OF THE LIBRARY TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF
SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
DAMAGES.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Libraries
If you develop a new library, and you want it to be of the greatest
possible use to the public, we recommend making it free software that
everyone can redistribute and change. You can do so by permitting
redistribution under these terms (or, alternatively, under the terms of the
ordinary General Public License).
To apply these terms, attach the following notices to the library. It is
safest to attach them to the start of each source file to most effectively
convey the exclusion of warranty; and each file should have at least the
"copyright" line and a pointer to where the full notice is found.
<one line to give the library's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Also add information on how to contact you by electronic and paper mail.
You should also get your employer (if you work as a programmer) or your
school, if any, to sign a "copyright disclaimer" for the library, if
necessary. Here is a sample; alter the names:
Yoyodyne, Inc., hereby disclaims all copyright interest in the
library `Frob' (a library for tweaking knobs) written by James Random Hacker.
<signature of Ty Coon>, 1 April 1990
Ty Coon, President of Vice
That's all there is to it!

34
libdb-1.85-c99.patch
View File

@ -1,34 +0,0 @@
Do not build snprintf. Instead use the one in glibc.
This avoids an implicit function declaration of vsprintf.
snprintf is only called from db.1.85/btree/bt_open.c, and the length
checking in the glibc implementation seems harmless there.
Furthermore, NULL is not a valid integer constant, so use 0 directly
(the __bt_first function in bt_seq.c returns int).
diff -ur db-5.3.28.orig/db.1.85/PORT/linux/Makefile db-5.3.28/db.1.85/PORT/linux/Makefile
--- db-5.3.28.orig/db.1.85/PORT/linux/Makefile 2022-12-10 12:29:48.599322424 +0100
+++ db-5.3.28/db.1.85/PORT/linux/Makefile 2022-12-10 12:35:08.415288426 +0100
@@ -11,7 +11,7 @@
OBJ5= rec_close.o rec_delete.o rec_get.o rec_open.o rec_put.o rec_search.o \
rec_seq.o rec_utils.o
-MISC= snprintf.o
+MISC=
${LIBDB}: ${OBJ1} ${OBJ2} ${OBJ3} ${OBJ4} ${OBJ5} ${MISC}
rm -f $@
only in patch2:
unchanged:
--- db-5.3.28.orig/db.1.85/btree/bt_seq.c 1994-07-26 18:57:24.000000000 +0200
+++ db-5.3.28/db.1.85/btree/bt_seq.c 2024-01-05 15:53:31.859178929 +0100
@@ -358,7 +358,7 @@
* page) and return it.
*/
if ((ep = __bt_search(t, key, exactp)) == NULL)
- return (NULL);
+ return 0;
if (*exactp) {
if (F_ISSET(t, B_NODUPS)) {
*erval = *ep;

39
libdb-5.3.21-region-size-check.patch
View File

@ -1,39 +0,0 @@
diff -up db-5.3.21/src/env/env_region.c.zero-region db-5.3.21/src/env/env_region.c
--- db-5.3.21/src/env/env_region.c.zero-region 2012-05-11 19:57:53.000000000 +0200
+++ db-5.3.21/src/env/env_region.c 2017-09-06 08:50:45.310276385 +0200
@@ -1060,6 +1060,7 @@ __env_region_attach(env, infop, init, ma
REGION *rp;
int ret;
char buf[sizeof(DB_REGION_FMT) + 20];
+ struct stat st;
/*
* Find or create a REGION structure for this region. If we create
@@ -1090,6 +1091,14 @@ __env_region_attach(env, infop, init, ma
if ((ret = __env_sys_attach(env, infop, rp)) != 0)
goto err;
+ /* Check the size of the underlying file */
+ if (infop->fhp != NULL && F_ISSET(infop->fhp, DB_FH_OPENED) &&
+ infop->fhp->fd != -1 && fstat(infop->fhp->fd, &st) != -1 &&
+ st.st_size == 0) {
+ ret = DB_RUNRECOVERY;
+ goto err;
+ }
+
/*
* Fault the pages into memory. Note, do this BEFORE we initialize
* anything because we're writing pages in created regions, not just
diff -up db-5.3.21/src/os/os_map.c.zero-region db-5.3.21/src/os/os_map.c
--- db-5.3.21/src/os/os_map.c.zero-region 2012-05-11 19:57:54.000000000 +0200
+++ db-5.3.21/src/os/os_map.c 2017-09-06 08:49:39.144546552 +0200
@@ -265,6 +265,9 @@ __os_detach(env, infop, destroy)
dbenv = env->dbenv;
rp = infop->rp;
+ /* Do not touch the region information if it no longer exists */
+ if (rp == NULL)
+ return EINVAL;
/* If the user replaced the unmap call, call through their interface. */
if (DB_GLOBAL(j_region_unmap) != NULL)

150
libdb-5.3.21-trickle_cpu.patch
View File

@ -1,150 +0,0 @@
diff -up db-5.3.21/src/dbinc_auto/int_def.in.trickle db-5.3.21/src/dbinc_auto/int_def.in
--- db-5.3.21/src/dbinc_auto/int_def.in.trickle 2018-08-21 10:54:06.066757392 +0200
+++ db-5.3.21/src/dbinc_auto/int_def.in 2018-08-21 10:54:06.111756561 +0200
@@ -1458,6 +1458,7 @@
#define __memp_sync_int __memp_sync_int@DB_VERSION_UNIQUE_NAME@
#define __memp_mf_sync __memp_mf_sync@DB_VERSION_UNIQUE_NAME@
#define __memp_purge_dead_files __memp_purge_dead_files@DB_VERSION_UNIQUE_NAME@
+#define __memp_purge_dead_and_count __memp_purge_dead_and_count@DB_VERSION_UNIQUE_NAME@
#define __memp_trickle_pp __memp_trickle_pp@DB_VERSION_UNIQUE_NAME@
#define __mutex_alloc __mutex_alloc@DB_VERSION_UNIQUE_NAME@
#define __mutex_alloc_int __mutex_alloc_int@DB_VERSION_UNIQUE_NAME@
diff -up db-5.3.21/src/dbinc_auto/mp_ext.h.trickle db-5.3.21/src/dbinc_auto/mp_ext.h
--- db-5.3.21/src/dbinc_auto/mp_ext.h.trickle 2018-08-21 10:54:06.103756709 +0200
+++ db-5.3.21/src/dbinc_auto/mp_ext.h 2018-08-21 10:54:06.112756543 +0200
@@ -101,6 +101,7 @@ int __mp_xxx_fh __P((DB_MPOOLFILE *, DB_
int __memp_sync_int __P((ENV *, DB_MPOOLFILE *, u_int32_t, u_int32_t, u_int32_t *, int *));
int __memp_mf_sync __P((DB_MPOOL *, MPOOLFILE *, int));
int __memp_purge_dead_files __P((ENV *));
+int __memp_purge_dead_and_count __P((ENV *, u_int32_t *, u_int32_t *));
int __memp_trickle_pp __P((DB_ENV *, int, int *));
#if defined(__cplusplus)
diff -up db-5.3.21/src/mp/mp_sync.c.trickle db-5.3.21/src/mp/mp_sync.c
--- db-5.3.21/src/mp/mp_sync.c.trickle 2018-08-21 10:54:06.105756672 +0200
+++ db-5.3.21/src/mp/mp_sync.c 2018-09-04 09:43:57.502992291 +0200
@@ -965,17 +965,34 @@ __bhcmp(p1, p2)
return (0);
}
+
/*
* __memp_purge_dead_files --
+ * Thin wrapper over __memp_purge_dead_and_count. Does not return any
+ * information about the number of total/dirty buffers.
+ *
+ * PUBLIC: int __memp_purge_dead_files __P((ENV *));
+ */
+int
+__memp_purge_dead_files(env)
+ ENV *env;
+{
+ return __memp_purge_dead_and_count(env, NULL, NULL);
+}
+
+/*
+ * __memp_purge_dead_and_count --
* Remove all dead files and their buffers from the mpool. The caller
* cannot hold any lock on the dead MPOOLFILE handles, their buffers
* or their hash buckets.
*
- * PUBLIC: int __memp_purge_dead_files __P((ENV *));
+ * PUBLIC: int __memp_purge_dead_and_count __P((ENV *, u_int32_t *, u_int32_t *));
*/
int
-__memp_purge_dead_files(env)
+__memp_purge_dead_and_count(env, totalp, dirtyp)
ENV *env;
+ u_int32_t *totalp;
+ u_int32_t *dirtyp;
{
BH *bhp;
DB_MPOOL *dbmp;
@@ -983,7 +1000,7 @@ __memp_purge_dead_files(env)
REGINFO *infop;
MPOOL *c_mp, *mp;
MPOOLFILE *mfp;
- u_int32_t i_cache;
+ u_int32_t i_cache, dirty, total;
int ret, t_ret, h_lock;
if (!MPOOL_ON(env))
@@ -992,6 +1009,7 @@ __memp_purge_dead_files(env)
dbmp = env->mp_handle;
mp = dbmp->reginfo[0].primary;
ret = t_ret = h_lock = 0;
+ dirty = total = 0;
/*
* Walk each cache's list of buffers and free all buffers whose
@@ -1000,6 +1018,7 @@ __memp_purge_dead_files(env)
for (i_cache = 0; i_cache < mp->nreg; i_cache++) {
infop = &dbmp->reginfo[i_cache];
c_mp = infop->primary;
+ total += c_mp->pages;
hp = R_ADDR(infop, c_mp->htab);
hp_end = &hp[c_mp->htab_buckets];
@@ -1008,6 +1027,9 @@ __memp_purge_dead_files(env)
if (SH_TAILQ_FIRST(&hp->hash_bucket, __bh) == NULL)
continue;
+ /* Count dirty pages first as we do not wait on mutex locks */
+ dirty += (u_int32_t)atomic_read(&hp->hash_page_dirty);
+
/*
* Search for a dead buffer. Other places that call
* __memp_bhfree() acquire the buffer lock before the
@@ -1073,6 +1095,11 @@ __memp_purge_dead_files(env)
}
}
+ if (dirtyp != NULL)
+ *dirtyp = dirty;
+ if (totalp != NULL)
+ *totalp = total;
+
return (ret);
}
diff -up db-5.3.21/src/mp/mp_trickle.c.trickle db-5.3.21/src/mp/mp_trickle.c
--- db-5.3.21/src/mp/mp_trickle.c.trickle 2018-08-21 10:54:06.105756672 +0200
+++ db-5.3.21/src/mp/mp_trickle.c 2018-08-21 10:54:06.112756543 +0200
@@ -56,6 +56,7 @@ __memp_trickle(env, pct, nwrotep)
dbmp = env->mp_handle;
mp = dbmp->reginfo[0].primary;
+ dirty = total = 0;
if (nwrotep != NULL)
*nwrotep = 0;
@@ -67,12 +68,8 @@ __memp_trickle(env, pct, nwrotep)
return (EINVAL);
}
- /* First we purge all dead files and their buffers. */
- if ((ret = __memp_purge_dead_files(env)) != 0)
- return (ret);
-
- /*
- * Loop through the caches counting total/dirty buffers.
+ /* First we purge all dead files and their buffers and
+ * loop through the caches counting total/dirty buffers.
*
* XXX
* Using hash_page_dirty is our only choice at the moment, but it's not
@@ -80,12 +77,8 @@ __memp_trickle(env, pct, nwrotep)
* than one page size, as a free 512B buffer may not be equivalent to
* having a free 8KB buffer.
*/
- for (ret = 0, i = dirty = total = 0; i < mp->nreg; ++i) {
- c_mp = dbmp->reginfo[i].primary;
- total += c_mp->pages;
- __memp_stat_hash(&dbmp->reginfo[i], c_mp, &dtmp);
- dirty += dtmp;
- }
+ if ((ret = __memp_purge_dead_and_count(env, &total, &dirty)) != 0)
+ return (ret);
/*
* If there are sufficient clean buffers, no buffers or no dirty