| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| |
|
|
| |
Summary: Handle initials and streetAddress attributes when doing a sync total update.
|
| |
|
|
| |
Summary: Performance cleanup of sync code. Improve tombstone search performance.
|
| |
|
|
| |
Summary: Handle syncing add opererations that have a ntuniqueid present.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Description: MMR: Directory updates on same object
Reviewed by: nhosoi (Thanks!)
Fix Description: The problem does appear to be concurrency. I think the original intention of
the urp fixup code was that it should only be run inside the database lock, so
that the database could be restored to a consistent state before the next
operation was processed. However, this requires the database code to know when
the database is already locked, so that if e.g. a modrdn operation needs to
call an internal delete, the database should not be locked again. The flag
OP_FLAG_REPL_FIXUP is used to denote both that the operation is such an
internal operation, and that the database should not be locked again.
There are a couple of cases where these operations can be called from outside
of the database lock:
urp_fixup_rename_entry is called from multimaster_postop_modrdn and
multimaster_postop_delete, both of which are front end post op plugins, not
called from within the database lock. Same with urp_fixup_delete_entry and
urp_fixup_modify_entry. In other cases, such as urp_fixup_add_entry, and other
places where urp_fixup_rename_entry and urp_fixup_modify_entry are called, they
are called from a bepostop plugin function, which is called after the original
database operation has been processed, within the database lock. So the
solution appears to be to move the urp_* functions to the bepostop plugin
functions. One of these functions does an internal search -
urp_get_min_naming_conflict_entry - but it does not appear that search locks
the database, so there was nothing to be done to make it "reentrant".
Without this patch, I can crash the server in a matter of minutes (x86_64
rhel5) using the latest Fedora DS 1.1 code. With the patch, the server runs
for several hours (maybe longer, I had to stop the test).
Also, to really exercise the urp code, I added a rename operation between the
add and delete e.g.
add("ou=test");
rename("ou=test", "ou=test2");
delete("ou=test2");
The server still runs for several hours with no problems.
Platforms tested: RHEL5 x86_64
Flag Day: no
Doc impact: no
|
| |
|
|
| |
Summary: Corrected search scope used to find entries to sync in winsync total update protocol.
|
| |
|
|
| |
Summary: Handle syncing multi-valued street attribute to AD.
|
| |
|
|
| |
Summary: Check if passwords are already hashed before sync'ing with AD.
|
| |
|
|
| |
Summary: Trim initials attribute value when sync'ing to AD.
|
| |
|
|
|
|
|
|
| |
Summary: Use mozldap6 ldif_parse_line API
Changes:
1) Removed ldif.h from the DS tree.
2) Eliminated the 5-th arg of ldif_parse_line (errmsg) and the errmsg related
code.
|
| |
|
|
| |
Summary: Remove obsolete makefiles
|
| |
|
|
|
|
|
|
| |
Summary: Change format of DBVERSION and guardian files (Comment #6)
1) introduced new strings for DBVERSION
2) added the logic to compare the new DBVERSION strings
note: we don't store the current db version string in the replication
code any more. Instead, we get it from Berkeley DB header file db.h.
|
| |
|
|
|
|
|
|
|
| |
Bug Description: Allow building with bdb 4.4 or later
Reviewed by: nkinder (Thanks!)
Fix Description: db.m4 already had code to detect and use the correct version of db headers and libraries. There have been some minor api changes since 4.3, so not much code changes were required. Note that this merely allows the server to build and run with db4.4 or later, not to take advantage of the newer features of the API.
Platforms tested: FC7
Flag Day: no
Doc impact: no
|
| |
|
|
|
| |
Summary: replica initialization: random failure (Comment #1)
Change: if acquire_replica is called for replica initialization, log REPLICA_BUSY.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Description: integrating db43 into ds70
Reviewed by: nkinder (Thanks!)
Fix Description: I took the original diffs posted by Ulf and merged them in with our code which has changed slightly since the diffs were originally generated. I also put #if directives like the following:
#if 1000*DB_VERSION_MAJOR + 100*DB_VERSION_MINOR >= 4300
... db43 features ...
#else
... db42 features ...
#endif
so that we can use both db42 and db43.
Platforms tested: RHEL4/FC5
Flag Day: no
Doc impact: no
|
| |
|
|
|
|
|
|
|
|
| |
Summary: configure needs to support --with-fhs (Comment #6)
Changes: Added the following include next to the end of the copyright block.
+
+#ifdef HAVE_CONFIG_H
+# include <config.h>
+#endif
+
|
| |
|
|
|
| |
Summary: Cleaning up obsolete macros in the build
Changes: eliminated macro NSPR20 (Comment #9)
|
| |
|
|
|
| |
Summary: parameterizing the hardcoded paths (phase 3. installed binaries, change log, setup)
Comment #23
|
| |
|
|
|
|
| |
Description: Enable DS Core build/install on HP-UX
Fix Description: remove some bogus adminutil and nsdshttpd stuff from the plugin makefiles
Reviewed by: nhosoi (Thanks!)
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Description: ldapserver: Close potential security vulnerabilities in CGI code
Reviewed by: Nathan, Noriko, and Pete (Thanks!)
Fix Description: Clean up usage of sprintf, strcpy, fgets instead of
gets, fixed buffer usage, etc., mostly in the CGI code and other user
facing code (i.e. setup). Also, Steve Grubb told me about a GCC trick
to force it to check printf style varargs functions, to check the format
string against the argument string, for type mismatches, missing
arguments, and too many arguments.
In the CGI form argument parsing code, we needed to be more careful
about checking for bad input - good input is supposed to look like this:
name=value&name=value&.....
&name=value. I don't think the original code
was checking properly for something like name&name=value.
There was another place where we were not checking to see if a buffer
had enough room before appending a string to it.
I had to change a couple of functions to allow passing in the size of
the buffer.
Fixed some issues raised by Noriko and Nathan.
Platforms tested: RHEL4
Flag Day: no
Doc impact: no
QA impact: should be covered by regular nightly and manual testing
New Tests integrated into TET: none
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Description: memory leaks using ber_scanf when handling bad BER packets
Reviewed by: All (Thanks!)
Files: https://bugzilla.redhat.com/bugzilla/attachment.cgi?id=123783
Branch: HEAD
Fix Description:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=179135#c0
I basically did a search through our code for all calls to ber_scanf,
ber_get_stringa, and ber_get_stringal and made sure we properly free any
arguments that may have been allocated. There was a bug in the ldapsdk
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=179135 that causes
us to free uninitialized memory when trying to clean up the result of
ber_get_stringal (or ber_scanf with 'V'). I had to initialize some
variables to NULL so that we could properly clean them up, and added
some additional clean ups that were missing. Also, in repl_extop.c, we
were calling free on an array that we should have been calling
ch_array_free on. Yet another lesson in the evils of slapi_ch_free and
disabling compiler type checks in general.
Platforms tested: Fedora Core 4
Flag Day: no
Doc impact: no
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Description: If you delete an attribute from an entry on AD, the attribute
doesn't get deleted on the DS side. The replication code doesn't even notice
that the entry changed.
Reviewed by: Rich, Noriko, Pete (thanks!)
Files: see diffs
Branch: HEAD, Directory71Branch
Fix Description: The dirsync search control passes back deleted attributes with no
values. If you try to add a Slapi_Attr with no values to a Slapi_Entry, it doesn't
get added. This fix stuffs the deleted attributes into the deleted attributes list
in the Slapi_Entry and checks for them when creating the modification
operations to be performed on the local entry.
Flag Day: no
Doc impact: no
QA impact: A regression test needs to be added
New Tests integrated into TET: none
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
RHEL4_x86_64_gcc3_OPT.OBJ
The makefiles were assuming that if the string 86 was found in uname -m
it should use x86 for the arch. However, it should only do this if
uname -m is not x86_64. Also, for RHEL/Linux, we do not have to add the
NS64TAG just before the _OPT (or _DBG) since it's already in the arch.
Other platforms continue to use the other naming convention. The
default naming convention for the mozilla components is
Linux2.6_x86_64_glibc_PTH_64_OPT.OBJ. I considered using Wan-Teh's
instructions about source builds, but that would cause many changes to
be made to our makefiles, so I just changed the way we calculate the
mozilla OBJDIR name from the regular OBJDIR name. These changes are
pretty much the same for adminutil, setuputil, adminserver, and ldapserver.
For ldapserver, I had to change nsarch (adminserver changed nsarch in a
different way). I also got rid of the 32 bit specific setup stuff. I
assume this was due to 32 bit NES admin server, so we can get rid of
this for the future, because we will have all native 64 bit apps. I also
went ahead and rolled in the gcc4 changes since they are safe for gcc3
as well.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Description: Chain BIND requests with chain on update + global pw policy
Reviewed by: Nathan, Pete (Thanks!)
Fix Description: When using global password policy, we need to chain the
BIND request back to a master so that it can update the pw policy op
attrs in the user's entry and replicate them to all other servers. The
call to config_get_pw_is_global_policy() is expensive (acquires a lock)
so we delay it as long as possible.
The reason we have to use config_get_pw_is_global_policy() is because the entry distribution plugin interface is poor - we have no way to register an init or start function to get the config, and no way to register a statechange callback to be notified of changes to the global password policy.
Platforms tested: Fedora Core 4
Flag Day: no
Doc impact: no
|
| |
|
|
| |
Checking in Ulf's proposal; should not get connid from pid if the op is internal. (plus the value is just for debug print which is disabled now)
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
the contents. I'm not sure why we haven't caught this earlier, but
I believe it has something to do with the patch to make ds build on
Fedora Core 4 with gcc4. To do that, we turn off the -fwriteable-strings
argument to gcc. I suppose with it on, it moves those strings to
some sort of writeable memory location. With it off, constant strings
are definitely in the data section. There was one place in views that
used a constant string, and a couple of places in the windows sync code.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Bug Description: Change ldapserver version to 1.0
Reviewed by: Noriko (Thanks!)
Fix Description: This also fixes some lingering build issues involving
perldap, which is no longer a separate setup package, but just gets
included into DS in a similar manner to nspr, nss, etc.
Platforms tested: RHEL4
Flag Day: no
Doc impact: no
QA impact: should be covered by regular nightly and manual testing
New Tests integrated into TET: none
|
| |
|
|
| |
duplicate values for system-only multivalued attributes. This change just skips the dscorepropagationdata attibute when receiving a change from DirSync.
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
improper use of ldap_get_next.
|
| |
|
|
| |
certain changes to not get synched.
|
| |
|
|
| |
replication. Add an error string returned to clients if an illegal attribute is configured
|
| |
|
|
| |
the acquire process has been successful thus far.
|
| |
|
|
| |
failed when it failed due to attempting to do fractional replication to another master.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
| |
Bug Description: RL16: Server core dumped while configuring replication agreements
Reviewed by: David (Thanks!)
Fix Description: The code for handling bad agreements was wrong. It was not returning the correct error code. The result of this is that the client did not receive an appropriate error code/message from the server. The reason for the core dump was that, even though the agreement was not created, the DSE entry was still there, causing the server to become very confused. Returning the correct error code causes the server to refuse to create the bad DSE entry.
Platforms tested: RHEL3
Flag Day: no
Doc impact: no
QA impact: should be covered by regular nightly and manual testing
New Tests integrated into TET: none
|
| |
|
|
| |
with 'streetaddress'
|
| |
|
|
| |
agreements
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
(checked in on behalf of dboreham)
|
