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.
Add NTLMv2 support to our client, used when so configured ('client use NTLMv2 =
yes') and only when 'client use spengo = no'. (A new option to allow the
client and server ends to chose spnego seperatly).
NTLMv2 signing doesn't yet work, and NTLMv2 is not done for NTLMSSP yet.
Also some parinoia checks in our input parsing.
Andrew Bartlett
This checking allows us to connect to Microsoft servers the use SMB signing,
within a few restrictions:
- I've not get the NTLMSSP stuff going - it appears to work, but if you break
the sig - say by writing a zero in it - it still passes...
- We don't currently verfiy the server's reply
- It works against one of my test servers, but not the other...
However, it provides an excellent basis to work from. Enable it with 'client
signing' in your smb.conf.
Doc to come (tomorrow) and this is not for 3.0, till we get it complete.
The CIFS Spec is misleading - the session key (for NTLMv1 at least) is the
standard session key, ie MD4(NT#).
Thanks to jra for the early work on this.
Andrew Bartlett
This tries to extract our server-side code out of sessetup.c, and into a more
general lib. I hope this is only a temporay resting place - I indend to
refactor it again into an auth-subsystem independent lib, using callbacks.
Move some of our our NTLMSSP #defines into a new file, and add two that I found
in the COMsource docs - we seem to have a double-up, but I've verified from
traces that the NTLMSSP_TARGET_TYPE_{DOMAIN,SERVER} is real.
This code also copes with ASCII clients - not that we will ever see any here,
but I hope to use this for HTTP, were we can get them. Win2k authenticates
fine under forced ASCII, btw.
Tested with Win2k, NTLMv2 and Samba's smbclient.
Andrew Bartlett
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.
smbclient would announce that it can send UNICODE, but would send the
plain text password in ASCII. This confused Samba HEAD somewhat. This
change has been tested against Samba HEAD of today and Samba 2.2.1a. I
do not have any other servers that do plain text passwords. Anybody?
Volker
Also, the negotiate blob has two ASCI strings encoded in the same way that the
UNICODE strings are, they are just in ASCII. The PARSER and Generator will have to deal with that.
we now do this:
- look for suported SASL mechanisms on the LDAP server
- choose GSS-SPNEGO if possible
- within GSS-SPNEGO choose KRB5 if we can do a kinit
- otherwise use NTLMSSP
This change also means that we no longer rely on having a gssapi
library to do ADS.
todo:
- add TLS/SSL support over LDAP
- change to using LDAP/SSL for password change in ADS
- Don't print an uninitialised buffer in service.c
- Change some charcnv.c functions to take smb_ucs2_t ** instead of void **
- Update NTLMv2 code to use dynamic buffers
- Update experimental SMB signing code - still more work to do
- Move sys_getgrouplist() to SAFE_FREE() and do a DEBUG() on initgroups()
failure.
Andrew Bartlett
there were 2 bugs:
1) we were sending a null challenge when we should have sent an empty
challenge
2) the password can be in unicode if unicode is negotiated. This means
our client code was wrong too :(
patches:
Andrew Bartlett
From his e-mail:
Below I attach the following patches as a result of my work
on trusted domains support:
1) srv_samr_nt.c.diff
This fixes a bug which caused to return null string as
the first entry of enumerated accounts list (no matter what
entry, it was always null string and rid) and possibly
spoiled further names, depeding on their length.
I found that while testing my 'net rpc trustdom list'
against nt servers and samba server.
2) libsmb.diff
Now, fallback to anonymous connection works correctly.
3) smbpasswd.c.diff
Just a little fix which actually allows one to create
a trusting domain account using smbpasswd
4) typos.diff
As the name suggests, it's just a few typos fix :)
bytes which follow the header, not the full packet size.
[Yes, the length field is either 17-bits, or (per the RFCs) it is a
16-bit length field preceeded by an 8-bit flags field of which only
the low-order bit may be used. If that bit is set, then add 65536 to
the 16-bit length field. (In other words, it's a 17-bit unsigned
length field.)
...unless, of course, the transport is native TCP [port 445] in which
case the length field *might* be 24-bits wide.]
Anyway, the change is a very minor one. We were including the four bytes
of the header in the length count and, as a result, sending four bytes of
garbage at the end of the SESSION REQUEST packet.
Small fix in function cli_session_request().
rebind proc (some give an extra paramter to pass a void* paramater) and
some small changes for the SMB signing code to reset things when the
signing starts, and to 'turn off' signing if the session setup failed.
Andrew Bartlett
The problem was that *all* packets were being signed, even packets before
signing was set up. (This broke the session request).
This fixes it to be an 'opt in' measure - that is, we only attempt to sign
things after we have got a valid, non-guest session setup as per the CIFS spec.
I've not tested this against an MS server, becouse my VMware is down, but
at least it doesn't break the build farm any more.
Andrew Bartlett
(const, takes unix string as arg)
Also update cli_full_connection to take NULL pointers as 'undefined' correctly,
and therefore do its own lookup etc. This what was intended, but previously
you needed to supply a 0.0.0.0 IP address.
Andrew Bartlett
Importantly:
The removal of the silly 'delete user script' behaviour when secuity=domain.
I have left the name the same - as it still does the (previously documented,
but not in smb.conf(5)) sane behaviour of deleting users on request.
When we decide what to do with the 'add user' functionality, we might
rename it.
Andrew Bartlett
this:
More code cleanup - this lot a bit more dodgy than the last:
The aim is to trim pwd_cache down to size. Its overly complex, and a
pain to deal with. With a header comment like this:
'obfusticaion is planned'
I think it deserved to die (at least partly).
This was being done to allow 'cli_establish_connection' to die - its
functionality has been replaced by cli_full_connection(), which does
not duplicate code everywhere for creating names etc.
This also removes the little 'init' fucntions for the various pipes,
becouse they were only used in one place, and even then it was dodgy.
(I've reworked smbcacls not to use anonymous connections any more, as
this will (should) fail with a 'restrict anonymous' PDC).
This allowed me to remove cli_pipe_util.c, which was calling
cli_establish_connection.
tpot: I'm not sure what direction you were going with the client stuff,
and you may well have been wanting the init functions. If thats the case,
give me a yell and I'll reimplement them against cli_full_connection.
Andrew Bartlett
This option was badly maintained, useless and confused our users and
distirbutors. (its SSL, therfore it must be good...)
No windows client uses this protocol without help from an SSL tunnel.
I can't see any reason why setting up a unix-side SSL wrapper would
be any more difficult than the > 10 config options this mess added
to samba in any case.
On the Samba client end, I think the LIBSMB_PROG hack should be
sufficient to start stunnel on the unix side. We might extend this
to take %i and %p (IP and port) if there is demand.
Andrew Bartlett
