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.
>Initialise user_rid value in WINBIND_USERINFO structure returned by
>the rpc version of query_user(). This fixes a caching bug found by
>Gavrie Philipson from disksite.
(This used to be commit 77bde1fa33)
only this file, and not any others. It includes the function prototypes.
(Forgot to commit with earlier patch)
Andrew Bartlett
(This used to be commit 3ec3861445)
The global winbind file descriptor can cause havoc in some situations -
particulary when it becomes 0, 1 or 2. This patch (based on some very nice
work by Hannes Schmidt <mail@schmidt-net.via.t-online.de>) starts to recitfy
the problem by ensuring that the close-on-exec flag is set, and that we move
above 3 in the file descriptor table.
I've also decided that the PAM module can close it's pipe handle on every
request - this isn't performance-critical code.
The next step is to do the same for nss_winbind. (But things like getent()
might get in our way there).
This also cleans up some function prototypes, puts them in just one place.
Andrew Bartlett
(This used to be commit 442eb39657)
to extend the ADS_STATUS system to include NTSTATUS, and to provide a better
general infrustructure for his sam_ads work.
I've also added some extra failure mode DEBUG()s to parts of the code.
NOTE: The ADS_ERR_OK() macro is rather sensitive to braketing issues - without
the final set of brakets, the test is essentially inverted - causing some
intersting 'error = success' messages...
Andrew Bartlett
(This used to be commit 5b9a7ab901)
changed cli_nt_setup_creds() to call cli_net_auth_2 or cli_net_auth_3 based on a switch.
pass also the negociation flags all the way.
all the places calling cli_nt_setup_creds() are still using cli_net_aut2(), it's just for future use and for rpcclient.
in the future we will be able to call auth_2 or auth_3 as we want.
J.F.
(This used to be commit 4d38caca40)
actually work. Also, the idea of 'loopback winbind' isn't that bad an idea
anyway (potential PDC/BDC applications).
Given all that, remove it...
Andrew Bartlett
(This used to be commit fc0d6e53fc)
Tridge suggested a generic caching mechanism for Samba to avoid the
proliferation of little cache files hanging around limpet like in the
locks directory. Someone should probably implement this at some
stage.
(This used to be commit dad31483b3)
setups.
- split up the ads structure into logical pieces. This makes it much
easier to keep things like the authentication realm and the server
realm separate (they can be different).
- allow ads callers to specify that no sasl bind should be performed
(used by "net ads info" for example)
- fix an error with handing ADS_ERROR_SYSTEM() when errno is 0
- completely rewrote the code for finding the LDAP server. Now try DNS
methods first, and try all DNS servers returned from the SRV DNS
query, sorted by closeness to our interfaces (using the same sort code
as we use in replies from WINS servers). This allows us to cope with
ADS DCs that are down, and ensures we don't pick one that is on the
other side of the country unless absolutely necessary.
- recognise dnsRecords as binary when displaying them
- cope with the realm not being configured in smb.conf (work it out
from the LDAP server)
- look at the trustDirection when looking up trusted domains and don't
include trusts that trust our domains but we don't trust
theirs.
- use LDAP to query the alternate (netbios) name for a realm, and make
sure that both and long and short forms of the name are accepted by
winbindd. Use the short form by default for listing users/groups.
- rescan the list of trusted domains every 5 minutes in case new trust
relationships are added while winbindd is running
- include transient trust relationships (ie. C trusts B, B trusts A,
so C trusts A) in winbindd.
- don't do a gratuituous node status lookup when finding an ADS DC (we
don't need it and it could fail)
- remove unused sid_to_distinguished_name function
- make sure we find the allternate name of our primary domain when
operating with a netbiosless ADS DC (using LDAP to do the lookup)
- fixed the rpc trusted domain enumeration to support up to approx
2000 trusted domains (the old limit was 3)
- use the IP for the remote_machine (%m) macro when the client doesn't
supply us with a name via a netbios session request (eg. port 445)
- if the client uses SPNEGO then use the machine name from the SPNEGO
auth packet for remote_machine (%m) macro
- add new 'net ads workgroup' command to find the netbios workgroup
name for a realm
(This used to be commit e358d7b24c)
This updates the 'winbind' authentication module and winbind's 'PAM' (actually
netlogon) code to allow smbd to cache connections to the DC.
This is particulary relevent when we need mutex locks already - there is no
parallelism to be gained anyway.
The winbind code authenticates the user, and if successful, passes back the
'info3' struct describing the user. smbd then interprets that in exactly the
same way as an 'ntdomain' logon.
Also, add parinoia to winbind about null termination.
Andrew Bartlett
(This used to be commit 167f122b67)
a getgr*() function that lists groups without numerating all the
group members. Instead of definiing a new nss method (which might
cause problems) I added an environment variable WINBIND_GETGRLST
that tells winbind not to fill in the group members in a gergrent()
request. This can speed up group listing by a factor of 20 or more
(on my test system with 50000 groups it reduces the time from an hour
to 2 minutes)
(This used to be commit e3f73256d3)
membership from an ADS server. We now use a 'member' query on the
group and do a separate call to convert the resulting distinguished
name to a name, rid etc. This is *much* faster for very large numbers
of groups (on a quantum test system with 10000 groups it drops the
time from an hour to about 35 seconds).
strangely enough, this actually *increases* the amount of ldap
traffic, its just that the MS LDAP server answers these queries much
faster.
(This used to be commit 5538048e4f)
