IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
sam account object, then pdb_update_sam_account() can be used to
update an account. This code path could lead to the methods element of
the account being used when uninitialised (leading to a segv)
Easiest fix is to always make that that when creating a sam_account
object we initialise the methods to null, so that the passdb code
knows that it needs to be filled in.
Also tidied up some of Richard's code (I don't think he uses the compiler
flags -g -Wall -Wshadow -Wstrict-prototypes -Wpointer-arith -Wcast-qual like
I do :-) :-).
Jeremy.
say exactly the same thing - in particular that we can algorithmic rid base ==
1000, and use the BASE_RID macro to avoid the use of magic numbers.
Andrew Bartlett
attributes rather than calling getpwnam() on the user.
This should help fix some of metze's performance issues - particularly on
enumerations.
There is a consequential change to the operation of 'non unix account's in LDAP
- they are no longer restricted to being 'within' the NUA range, but will
always be added to that range.
Finally, there is the doco for this and the previous LDAP SSL changes.
- Default is now for start-tls, on the ldap (not ldaps) port
- We check for 'I am currently root' in the right place now, and don't
accidentily use a cached connection.
- We don't loop on failure to be root, or some other errors.
- A bit cleaner error reporting for add/modify.
- Both the OpenLDAP and manual URI parsing tested.
Andrew Bartlett
This patch, from "Stefan (metze) Metzmacher" <metze@metzemix.de> implements an
LDAP connection cache. This removes the quite silly situation where every
single passdb operation involved a new LDAP connection.
The hope is that this will give us a decent performance boost in some usrmgr
related activities, and in the sid->name/sid->uid code.
The remaining things I think are 'todo' for pdb_ldap (in the near term) are:
- intergrate volker's next_rid patch for NUA accounts,
- add a 'trust ldap ids' option (remove Get_Pwnam() hit on enumerations).
- put the group mapping actually into ldap
- Schema fixes and do utf8 conversion
- server failover (try a second server for the rebind on fail)
- ensure we block between an 'add' and the ldap master replicating to our
local slave (mezte found this issue, kills domain joins)
Andrew Bartlett
Currently this calls back to mapping.c, but we have the framework
to get the information into LDAP and the passdb.tdb (should we? I
think so..).
This has received moderate testing with net rpc vampire and
usrmgr. I found the add_groupmem segfault in add_aliasmem as
well, but that will be another checkin.
Volker
The actual design change is relitivly small however:
It all goes back to jerry's 'BOOL store', added to many of the elements in a
SAM_ACCOUNT. This ensured that smb.conf defaults did not get 'fixed' into
ldap. This was a great win for admins, and this patch follows in the same way.
This patch extends the concept - we don't store values back into LDAP unless
they have been changed. So if we read a value, but don't update it, or we
read a value, find it's not there and use a default, we will not update
ldap with that value. This reduced clutter in our LDAP DB, and makes it
easier to change defaults later on.
Metze's particular problem was that when we 'write back' an unchanged value,
we would clear any muliple values in that feild. Now he can still have his
mulitivalued 'uid' feild, without Samba changing it for *every* other
operation.
This also applies to many other attributes, and helps to eliminate a nasty
race condition. (Time between get and set)
This patch is big, and needs more testing, but metze has tested usrmgr, and
I've fixed some pdbedit bugs, and tested domain joins, so it isn't compleatly
flawed ;-).
The same system will be introduced into the SAM code shortly, but this fixes
bugs that people were coming across in production uses of Samba 3.0/HEAD, hence
it's inclusion here.
Andrew Bartlett
Add const to some more functions, and reintroduce 'net rpc join oldstyle' as
*only* trying an old-style join.
This means that we can rely on it not prompting for a password on the build
farm.
Andrew Bartlett
better job of working with usrmgr. Previously we were blanking out entires,
and all sort of mischif.
The new patch (which I've now had a chance to test/modify) also takes care not
to expand % values (ie we go \\%L\%U -> \\server\user, we don't want to store
\\server\user back) and to correctly notice 'not set' compared to 'null string'
etc.
Andrew Bartlett
only the interface has been fully moved to NTSTATUS
not all the plugins make full use of it, but have been all converted.
My testings passed completely, however a bit of more testing is welcome
Simo.
