1
0
mirror of https://github.com/samba-team/samba.git synced 2024-12-27 03:21:53 +03:00
samba-mirror/source4/utils/ntlm_auth.c

987 lines
26 KiB
C
Raw Normal View History

/*
Unix SMB/CIFS implementation.
Winbind status program.
Copyright (C) Tim Potter 2000-2003
Copyright (C) Andrew Bartlett <abartlet@samba.org> 2003-2004
Copyright (C) Francesco Chemolli <kinkie@kame.usr.dsi.unimi.it> 2000
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program 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 General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include "includes.h"
#include "system/filesys.h"
#include "system/passwd.h"
#include "lib/cmdline/popt_common.h"
#include "auth/auth.h"
#include "librpc/gen_ndr/ndr_security.h"
#include "pstring.h"
#define SQUID_BUFFER_SIZE 2010
enum stdio_helper_mode {
SQUID_2_4_BASIC,
SQUID_2_5_BASIC,
SQUID_2_5_NTLMSSP,
NTLMSSP_CLIENT_1,
GSS_SPNEGO_CLIENT,
GSS_SPNEGO_SERVER,
NTLM_SERVER_1,
NUM_HELPER_MODES
};
#define NTLM_AUTH_FLAG_USER_SESSION_KEY 0x0004
#define NTLM_AUTH_FLAG_LMKEY 0x0008
typedef void (*stdio_helper_function)(enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **private2);
static void manage_squid_basic_request (enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **private2);
static void manage_gensec_request (enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **private2);
static void manage_ntlm_server_1_request (enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **private2);
static void manage_squid_request(enum stdio_helper_mode helper_mode,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
stdio_helper_function fn, void **private2);
static const struct {
enum stdio_helper_mode mode;
const char *name;
stdio_helper_function fn;
} stdio_helper_protocols[] = {
{ SQUID_2_4_BASIC, "squid-2.4-basic", manage_squid_basic_request},
{ SQUID_2_5_BASIC, "squid-2.5-basic", manage_squid_basic_request},
{ SQUID_2_5_NTLMSSP, "squid-2.5-ntlmssp", manage_gensec_request},
{ GSS_SPNEGO_CLIENT, "gss-spnego-client", manage_gensec_request},
{ GSS_SPNEGO_SERVER, "gss-spnego", manage_gensec_request},
{ NTLMSSP_CLIENT_1, "ntlmssp-client-1", manage_gensec_request},
{ NTLM_SERVER_1, "ntlm-server-1", manage_ntlm_server_1_request},
{ NUM_HELPER_MODES, NULL, NULL}
};
extern int winbindd_fd;
static const char *opt_username;
static const char *opt_domain;
static const char *opt_workstation;
static const char *opt_password;
static int opt_multiplex;
static void mux_printf(unsigned int mux_id, const char *format, ...) PRINTF_ATTRIBUTE(2, 3);
static void mux_printf(unsigned int mux_id, const char *format, ...)
{
va_list ap;
if (opt_multiplex) {
x_fprintf(x_stdout, "%d ", mux_id);
}
va_start(ap, format);
x_vfprintf(x_stdout, format, ap);
va_end(ap);
}
/* Copy of parse_domain_user from winbindd_util.c. Parse a string of the
form DOMAIN/user into a domain and a user */
static BOOL parse_ntlm_auth_domain_user(const char *domuser, fstring domain,
fstring user)
{
char *p = strchr(domuser,*lp_winbind_separator());
if (!p) {
return False;
}
fstrcpy(user, p+1);
fstrcpy(domain, domuser);
domain[PTR_DIFF(p, domuser)] = 0;
return True;
}
/* Authenticate a user with a plaintext password */
static BOOL check_plaintext_auth(const char *user, const char *pass,
BOOL stdout_diagnostics)
{
return (strcmp(pass, opt_password) == 0);
}
/* authenticate a user with an encrypted username/password */
static NTSTATUS local_pw_check_specified(const char *username,
const char *domain,
const char *workstation,
const DATA_BLOB *challenge,
const DATA_BLOB *lm_response,
const DATA_BLOB *nt_response,
uint32_t flags,
DATA_BLOB *lm_session_key,
DATA_BLOB *user_session_key,
char **error_string,
char **unix_name)
{
NTSTATUS nt_status;
struct samr_Password lm_pw, nt_pw;
struct samr_Password *lm_pwd, *nt_pwd;
TALLOC_CTX *mem_ctx = talloc_init("local_pw_check_specified");
if (!mem_ctx) {
nt_status = NT_STATUS_NO_MEMORY;
} else {
E_md4hash(opt_password, nt_pw.hash);
if (E_deshash(opt_password, lm_pw.hash)) {
lm_pwd = &lm_pw;
} else {
lm_pwd = NULL;
}
nt_pwd = &nt_pw;
nt_status = ntlm_password_check(mem_ctx,
MSV1_0_ALLOW_SERVER_TRUST_ACCOUNT |
MSV1_0_ALLOW_WORKSTATION_TRUST_ACCOUNT,
challenge,
lm_response,
nt_response,
username,
username,
domain,
lm_pwd, nt_pwd, user_session_key, lm_session_key);
if (NT_STATUS_IS_OK(nt_status)) {
if (unix_name) {
asprintf(unix_name,
"%s%c%s", domain,
*lp_winbind_separator(),
username);
}
} else {
DEBUG(3, ("Login for user [%s]\\[%s]@[%s] failed due to [%s]\n",
domain, username, workstation,
nt_errstr(nt_status)));
}
talloc_free(mem_ctx);
}
if (error_string) {
*error_string = strdup(nt_errstr(nt_status));
}
return nt_status;
}
static void manage_squid_basic_request(enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **private2)
{
char *user, *pass;
user=buf;
pass=memchr(buf,' ',length);
if (!pass) {
DEBUG(2, ("Password not found. Denying access\n"));
mux_printf(mux_id, "ERR\n");
return;
}
*pass='\0';
pass++;
if (stdio_helper_mode == SQUID_2_5_BASIC) {
rfc1738_unescape(user);
rfc1738_unescape(pass);
}
if (check_plaintext_auth(user, pass, False)) {
mux_printf(mux_id, "OK\n");
} else {
mux_printf(mux_id, "ERR\n");
}
}
/* This is a bit hairy, but the basic idea is to do a password callback
to the calling application. The callback comes from within gensec */
static void manage_gensec_get_pw_request(enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **password)
{
DATA_BLOB in;
if (strlen(buf) < 2) {
DEBUG(1, ("query [%s] invalid", buf));
mux_printf(mux_id, "BH\n");
return;
}
if (strlen(buf) > 3) {
in = base64_decode_data_blob(NULL, buf + 3);
} else {
in = data_blob(NULL, 0);
}
if (strncmp(buf, "PW ", 3) == 0) {
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
*password = talloc_strndup(*private /* hopefully the right gensec context, useful to use for talloc */,
(const char *)in.data, in.length);
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
if (*password == NULL) {
DEBUG(1, ("Out of memory\n"));
mux_printf(mux_id, "BH\n");
data_blob_free(&in);
return;
}
mux_printf(mux_id, "OK\n");
data_blob_free(&in);
return;
}
DEBUG(1, ("Asked for (and expected) a password\n"));
mux_printf(mux_id, "BH\n");
data_blob_free(&in);
}
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
/**
* Callback for password credentails. This is not async, and when
* GENSEC and the credentails code is made async, it will look rather
* different.
*/
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
static const char *get_password(struct cli_credentials *credentials)
{
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
char *password = NULL;
/* Ask for a password */
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
mux_printf((unsigned int)credentials->priv_data, "PW\n");
credentials->priv_data = NULL;
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
manage_squid_request(NUM_HELPER_MODES /* bogus */, manage_gensec_get_pw_request, (void **)&password);
return password;
}
static void manage_gensec_request(enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **private2)
{
DATA_BLOB in;
DATA_BLOB out = data_blob(NULL, 0);
char *out_base64 = NULL;
const char *reply_arg = NULL;
struct gensec_ntlm_state {
struct gensec_security *gensec_state;
const char *set_password;
};
struct gensec_ntlm_state *state;
NTSTATUS nt_status;
BOOL first = False;
const char *reply_code;
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
struct cli_credentials *creds;
TALLOC_CTX *mem_ctx;
if (*private) {
state = *private;
} else {
state = talloc_zero(NULL, struct gensec_ntlm_state);
if (!state) {
mux_printf(mux_id, "BH No Memory\n");
exit(1);
}
*private = state;
if (opt_password) {
state->set_password = opt_password;
}
}
if (strlen(buf) < 2) {
DEBUG(1, ("query [%s] invalid", buf));
mux_printf(mux_id, "BH\n");
return;
}
if (strlen(buf) > 3) {
in = base64_decode_data_blob(NULL, buf + 3);
} else {
in = data_blob(NULL, 0);
}
if (strncmp(buf, "YR", 2) == 0) {
if (state->gensec_state) {
talloc_free(state->gensec_state);
state->gensec_state = NULL;
}
} else if ( (strncmp(buf, "OK", 2) == 0)) {
/* do nothing */
data_blob_free(&in);
return;
} else if ( (strncmp(buf, "TT ", 3) != 0) &&
(strncmp(buf, "KK ", 3) != 0) &&
(strncmp(buf, "AF ", 3) != 0) &&
(strncmp(buf, "NA ", 3) != 0) &&
(strncmp(buf, "UG", 2) != 0) &&
(strncmp(buf, "PW ", 3) != 0)) {
DEBUG(1, ("SPNEGO request [%s] invalid\n", buf));
mux_printf(mux_id, "BH\n");
data_blob_free(&in);
return;
}
/* setup gensec */
if (!(state->gensec_state)) {
switch (stdio_helper_mode) {
case GSS_SPNEGO_CLIENT:
case NTLMSSP_CLIENT_1:
/* setup the client side */
nt_status = gensec_client_start(NULL, &state->gensec_state, NULL);
if (!NT_STATUS_IS_OK(nt_status)) {
exit(1);
}
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
break;
case GSS_SPNEGO_SERVER:
case SQUID_2_5_NTLMSSP:
if (!NT_STATUS_IS_OK(gensec_server_start(NULL, &state->gensec_state, NULL))) {
exit(1);
}
break;
default:
abort();
}
creds = cli_credentials_init(state->gensec_state);
cli_credentials_set_conf(creds);
if (opt_username) {
cli_credentials_set_username(creds, opt_username, CRED_SPECIFIED);
}
if (opt_domain) {
cli_credentials_set_domain(creds, opt_domain, CRED_SPECIFIED);
}
if (state->set_password) {
cli_credentials_set_password(creds, state->set_password, CRED_SPECIFIED);
} else {
cli_credentials_set_password_callback(creds, get_password);
creds->priv_data = (void*)mux_id;
}
if (opt_workstation) {
cli_credentials_set_workstation(creds, opt_workstation, CRED_SPECIFIED);
}
switch (stdio_helper_mode) {
case GSS_SPNEGO_SERVER:
case SQUID_2_5_NTLMSSP:
cli_credentials_set_machine_account(creds);
break;
default:
break;
}
gensec_set_credentials(state->gensec_state, creds);
switch (stdio_helper_mode) {
case GSS_SPNEGO_CLIENT:
case GSS_SPNEGO_SERVER:
nt_status = gensec_start_mech_by_oid(state->gensec_state, GENSEC_OID_SPNEGO);
if (!in.length) {
first = True;
}
break;
case NTLMSSP_CLIENT_1:
if (!in.length) {
first = True;
}
case SQUID_2_5_NTLMSSP:
nt_status = gensec_start_mech_by_oid(state->gensec_state, GENSEC_OID_NTLMSSP);
break;
default:
abort();
}
if (!NT_STATUS_IS_OK(nt_status)) {
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
DEBUG(1, ("GENSEC mech failed to start: %s\n", nt_errstr(nt_status)));
mux_printf(mux_id, "BH\n");
return;
}
}
/* update */
mem_ctx = talloc_named(NULL, 0, "manage_gensec_request internal mem_ctx");
if (strncmp(buf, "PW ", 3) == 0) {
state->set_password = talloc_strndup(state,
(const char *)in.data,
in.length);
cli_credentials_set_password(gensec_get_credentials(state->gensec_state),
state->set_password,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
CRED_SPECIFIED);
mux_printf(mux_id, "OK\n");
data_blob_free(&in);
talloc_free(mem_ctx);
return;
}
if (strncmp(buf, "UG", 2) == 0) {
int i;
char *grouplist = NULL;
struct auth_session_info *session_info;
if (!NT_STATUS_IS_OK(gensec_session_info(state->gensec_state, &session_info))) {
DEBUG(1, ("gensec_session_info failed: %s\n", nt_errstr(nt_status)));
mux_printf(mux_id, "BH %s\n", nt_errstr(nt_status));
data_blob_free(&in);
talloc_free(mem_ctx);
return;
}
/* get the string onto the context */
grouplist = talloc_strdup(mem_ctx, "");
for (i=0; i<session_info->security_token->num_sids; i++) {
struct security_token *token = session_info->security_token;
const char *sidstr = dom_sid_string(session_info,
token->sids[i]);
grouplist = talloc_asprintf_append(grouplist, "%s,", sidstr);
}
mux_printf(mux_id, "GL %s\n", grouplist);
talloc_free(session_info);
data_blob_free(&in);
talloc_free(mem_ctx);
return;
}
nt_status = gensec_update(state->gensec_state, mem_ctx, in, &out);
/* don't leak 'bad password'/'no such user' info to the network client */
nt_status = auth_nt_status_squash(nt_status);
if (out.length) {
out_base64 = base64_encode_data_blob(mem_ctx, out);
} else {
out_base64 = NULL;
}
if (NT_STATUS_EQUAL(nt_status, NT_STATUS_MORE_PROCESSING_REQUIRED)) {
reply_arg = "*";
if (first) {
reply_code = "YR";
} else if (state->gensec_state->gensec_role == GENSEC_CLIENT) {
reply_code = "KK";
} else if (state->gensec_state->gensec_role == GENSEC_SERVER) {
reply_code = "TT";
} else {
abort();
}
} else if (NT_STATUS_EQUAL(nt_status, NT_STATUS_ACCESS_DENIED)) {
reply_code = "BH";
reply_arg = nt_errstr(nt_status);
DEBUG(1, ("GENSEC login failed: %s\n", nt_errstr(nt_status)));
} else if (NT_STATUS_EQUAL(nt_status, NT_STATUS_UNSUCCESSFUL)) {
reply_code = "BH";
reply_arg = nt_errstr(nt_status);
DEBUG(1, ("GENSEC login failed: %s\n", nt_errstr(nt_status)));
} else if (!NT_STATUS_IS_OK(nt_status)) {
reply_code = "NA";
reply_arg = nt_errstr(nt_status);
DEBUG(1, ("GENSEC login failed: %s\n", nt_errstr(nt_status)));
} else if /* OK */ (state->gensec_state->gensec_role == GENSEC_SERVER) {
struct auth_session_info *session_info;
nt_status = gensec_session_info(state->gensec_state, &session_info);
if (!NT_STATUS_IS_OK(nt_status)) {
reply_code = "BH";
reply_arg = nt_errstr(nt_status);
DEBUG(1, ("GENSEC failed to retreive the session info: %s\n", nt_errstr(nt_status)));
} else {
reply_code = "AF";
reply_arg = talloc_asprintf(state->gensec_state,
"%s%s%s", session_info->server_info->domain_name,
lp_winbind_separator(), session_info->server_info->account_name);
talloc_free(session_info);
}
} else if (state->gensec_state->gensec_role == GENSEC_CLIENT) {
reply_code = "AF";
reply_arg = out_base64;
} else {
abort();
}
switch (stdio_helper_mode) {
case GSS_SPNEGO_SERVER:
mux_printf(mux_id, "%s %s %s\n", reply_code,
out_base64 ? out_base64 : "*",
reply_arg ? reply_arg : "*");
break;
default:
if (out_base64) {
mux_printf(mux_id, "%s %s\n", reply_code, out_base64);
} else if (reply_arg) {
mux_printf(mux_id, "%s %s\n", reply_code, reply_arg);
} else {
mux_printf(mux_id, "%s\n", reply_code);
}
}
talloc_free(mem_ctx);
return;
}
static void manage_ntlm_server_1_request(enum stdio_helper_mode stdio_helper_mode,
char *buf, int length, void **private,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
unsigned int mux_id, void **private2)
{
char *request, *parameter;
static DATA_BLOB challenge;
static DATA_BLOB lm_response;
static DATA_BLOB nt_response;
static char *full_username;
static char *username;
static char *domain;
static char *plaintext_password;
static BOOL ntlm_server_1_user_session_key;
static BOOL ntlm_server_1_lm_session_key;
if (strequal(buf, ".")) {
if (!full_username && !username) {
mux_printf(mux_id, "Error: No username supplied!\n");
} else if (plaintext_password) {
/* handle this request as plaintext */
if (!full_username) {
if (asprintf(&full_username, "%s%c%s", domain, *lp_winbind_separator(), username) == -1) {
mux_printf(mux_id, "Error: Out of memory in asprintf!\n.\n");
return;
}
}
if (check_plaintext_auth(full_username, plaintext_password, False)) {
mux_printf(mux_id, "Authenticated: Yes\n");
} else {
mux_printf(mux_id, "Authenticated: No\n");
}
} else if (!lm_response.data && !nt_response.data) {
mux_printf(mux_id, "Error: No password supplied!\n");
} else if (!challenge.data) {
mux_printf(mux_id, "Error: No lanman-challenge supplied!\n");
} else {
char *error_string = NULL;
DATA_BLOB lm_key;
DATA_BLOB user_session_key;
uint32_t flags = 0;
if (full_username && !username) {
fstring fstr_user;
fstring fstr_domain;
if (!parse_ntlm_auth_domain_user(full_username, fstr_user, fstr_domain)) {
/* username might be 'tainted', don't print into our new-line deleimianted stream */
mux_printf(mux_id, "Error: Could not parse into domain and username\n");
}
SAFE_FREE(username);
SAFE_FREE(domain);
username = smb_xstrdup(fstr_user);
domain = smb_xstrdup(fstr_domain);
}
if (!domain) {
domain = smb_xstrdup(lp_workgroup());
}
if (ntlm_server_1_lm_session_key)
flags |= NTLM_AUTH_FLAG_LMKEY;
if (ntlm_server_1_user_session_key)
flags |= NTLM_AUTH_FLAG_USER_SESSION_KEY;
if (!NT_STATUS_IS_OK(
local_pw_check_specified(username,
domain,
lp_netbios_name(),
&challenge,
&lm_response,
&nt_response,
flags,
&lm_key,
&user_session_key,
&error_string,
NULL))) {
mux_printf(mux_id, "Authenticated: No\n");
mux_printf(mux_id, "Authentication-Error: %s\n.\n", error_string);
SAFE_FREE(error_string);
} else {
static char zeros[16];
char *hex_lm_key;
char *hex_user_session_key;
mux_printf(mux_id, "Authenticated: Yes\n");
if (ntlm_server_1_lm_session_key
&& lm_key.length
&& (memcmp(zeros, lm_key.data,
lm_key.length) != 0)) {
hex_encode(lm_key.data,
lm_key.length,
&hex_lm_key);
mux_printf(mux_id, "LANMAN-Session-Key: %s\n", hex_lm_key);
SAFE_FREE(hex_lm_key);
}
if (ntlm_server_1_user_session_key
&& user_session_key.length
&& (memcmp(zeros, user_session_key.data,
user_session_key.length) != 0)) {
hex_encode(user_session_key.data,
user_session_key.length,
&hex_user_session_key);
mux_printf(mux_id, "User-Session-Key: %s\n", hex_user_session_key);
SAFE_FREE(hex_user_session_key);
}
}
}
/* clear out the state */
challenge = data_blob(NULL, 0);
nt_response = data_blob(NULL, 0);
lm_response = data_blob(NULL, 0);
SAFE_FREE(full_username);
SAFE_FREE(username);
SAFE_FREE(domain);
SAFE_FREE(plaintext_password);
ntlm_server_1_user_session_key = False;
ntlm_server_1_lm_session_key = False;
mux_printf(mux_id, ".\n");
return;
}
request = buf;
/* Indicates a base64 encoded structure */
parameter = strstr(request, ":: ");
if (!parameter) {
parameter = strstr(request, ": ");
if (!parameter) {
DEBUG(0, ("Parameter not found!\n"));
mux_printf(mux_id, "Error: Parameter not found!\n.\n");
return;
}
parameter[0] ='\0';
parameter++;
parameter[0] ='\0';
parameter++;
} else {
parameter[0] ='\0';
parameter++;
parameter[0] ='\0';
parameter++;
parameter[0] ='\0';
parameter++;
base64_decode_inplace(parameter);
}
if (strequal(request, "LANMAN-Challenge")) {
challenge = strhex_to_data_blob(parameter);
if (challenge.length != 8) {
mux_printf(mux_id, "Error: hex decode of %s failed! (got %d bytes, expected 8)\n.\n",
parameter,
(int)challenge.length);
challenge = data_blob(NULL, 0);
}
} else if (strequal(request, "NT-Response")) {
nt_response = strhex_to_data_blob(parameter);
if (nt_response.length < 24) {
mux_printf(mux_id, "Error: hex decode of %s failed! (only got %d bytes, needed at least 24)\n.\n",
parameter,
(int)nt_response.length);
nt_response = data_blob(NULL, 0);
}
} else if (strequal(request, "LANMAN-Response")) {
lm_response = strhex_to_data_blob(parameter);
if (lm_response.length != 24) {
mux_printf(mux_id, "Error: hex decode of %s failed! (got %d bytes, expected 24)\n.\n",
parameter,
(int)lm_response.length);
lm_response = data_blob(NULL, 0);
}
} else if (strequal(request, "Password")) {
plaintext_password = smb_xstrdup(parameter);
} else if (strequal(request, "NT-Domain")) {
domain = smb_xstrdup(parameter);
} else if (strequal(request, "Username")) {
username = smb_xstrdup(parameter);
} else if (strequal(request, "Full-Username")) {
full_username = smb_xstrdup(parameter);
} else if (strequal(request, "Request-User-Session-Key")) {
ntlm_server_1_user_session_key = strequal(parameter, "Yes");
} else if (strequal(request, "Request-LanMan-Session-Key")) {
ntlm_server_1_lm_session_key = strequal(parameter, "Yes");
} else {
mux_printf(mux_id, "Error: Unknown request %s\n.\n", request);
}
}
static void manage_squid_request(enum stdio_helper_mode helper_mode,
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
stdio_helper_function fn, void **private2)
{
char buf[SQUID_BUFFER_SIZE+1];
unsigned int mux_id;
int length;
char *c;
static BOOL err;
struct mux_private {
unsigned int max_mux;
void **private_pointers;
};
static struct mux_private *mux_private;
static void *normal_private;
void **private;
/* this is not a typo - x_fgets doesn't work too well under squid */
if (fgets(buf, sizeof(buf)-1, stdin) == NULL) {
if (ferror(stdin)) {
DEBUG(1, ("fgets() failed! dying..... errno=%d (%s)\n", ferror(stdin),
strerror(ferror(stdin))));
exit(1); /* BIIG buffer */
}
exit(0);
}
c=memchr(buf,'\n',sizeof(buf)-1);
if (c) {
*c = '\0';
length = c-buf;
} else {
err = 1;
return;
}
if (err) {
DEBUG(0, ("Oversized message\n"));
x_fprintf(x_stdout, "ERR\n");
err = 0;
return;
}
DEBUG(10, ("Got '%s' from squid (length: %d).\n",buf,length));
if (buf[0] == '\0') {
DEBUG(0, ("Invalid Request (empty)\n"));
x_fprintf(x_stdout, "ERR\n");
return;
}
if (opt_multiplex) {
if (sscanf(buf, "%u ", &mux_id) != 1) {
DEBUG(0, ("Invalid Request - no multiplex id\n"));
x_fprintf(x_stdout, "ERR\n");
return;
}
if (!mux_private) {
mux_private = talloc(NULL, struct mux_private);
mux_private->max_mux = 0;
mux_private->private_pointers = NULL;
}
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
c=strchr(buf,' ');
if (!c) {
DEBUG(0, ("Invalid Request - no data after multiplex id\n"));
x_fprintf(x_stdout, "ERR\n");
return;
}
c++;
if (mux_id >= mux_private->max_mux) {
unsigned int prev_max = mux_private->max_mux;
mux_private->max_mux = mux_id + 1;
mux_private->private_pointers
= talloc_realloc(mux_private,
mux_private->private_pointers,
void *, mux_private->max_mux);
memset(&mux_private->private_pointers[prev_max], '\0',
(sizeof(*mux_private->private_pointers) * (mux_private->max_mux - prev_max)));
};
private = &mux_private->private_pointers[mux_id];
} else {
c = buf;
private = &normal_private;
}
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
fn(helper_mode, c, length, private, mux_id, private2);
}
static void squid_stream(enum stdio_helper_mode stdio_mode,
stdio_helper_function fn) {
/* initialize FDescs */
x_setbuf(x_stdout, NULL);
x_setbuf(x_stderr, NULL);
while(1) {
r6028: A MAJOR update to intergrate the new credentails system fully with GENSEC, and to pull SCHANNEL into GENSEC, by making it less 'special'. GENSEC now no longer has it's own handling of 'set username' etc, instead it uses cli_credentials calls. In order to link the credentails code right though Samba, a lot of interfaces have changed to remove 'username, domain, password' arguments, and these have been replaced with a single 'struct cli_credentials'. In the session setup code, a new parameter 'workgroup' contains the client/server current workgroup, which seems unrelated to the authentication exchange (it was being filled in from the auth info). This allows in particular kerberos to only call back for passwords when it actually needs to perform the kinit. The kerberos code has been modified not to use the SPNEGO provided 'principal name' (in the mechListMIC), but to instead use the name the host was connected to as. This better matches Microsoft behaviour, is more secure and allows better use of standard kerberos functions. To achieve this, I made changes to our socket code so that the hostname (before name resolution) is now recorded on the socket. In schannel, most of the code from librpc/rpc/dcerpc_schannel.c is now in libcli/auth/schannel.c, and it looks much more like a standard GENSEC module. The actual sign/seal code moved to libcli/auth/schannel_sign.c in a previous commit. The schannel credentails structure is now merged with the rest of the credentails, as many of the values (username, workstation, domain) where already present there. This makes handling this in a generic manner much easier, as there is no longer a custom entry-point. The auth_domain module continues to be developed, but is now just as functional as auth_winbind. The changes here are consequential to the schannel changes. The only removed function at this point is the RPC-LOGIN test (simulating the load of a WinXP login), which needs much more work to clean it up (it contains copies of too much code from all over the torture suite, and I havn't been able to penetrate its 'structure'). Andrew Bartlett (This used to be commit 2301a4b38a21aa60917973451687063d83d18d66)
2005-03-24 07:14:06 +03:00
manage_squid_request(stdio_mode, fn, NULL);
}
}
/* Main program */
enum {
OPT_USERNAME = 1000,
OPT_DOMAIN,
OPT_WORKSTATION,
OPT_CHALLENGE,
OPT_RESPONSE,
OPT_LM,
OPT_NT,
OPT_PASSWORD,
OPT_LM_KEY,
OPT_USER_SESSION_KEY,
OPT_DIAGNOSTICS,
OPT_REQUIRE_MEMBERSHIP,
OPT_MULTIPLEX,
};
int main(int argc, const char **argv)
{
static const char *helper_protocol;
int opt;
poptContext pc;
/* NOTE: DO NOT change this interface without considering the implications!
This is an external interface, which other programs will use to interact
with this helper.
*/
/* We do not use single-letter command abbreviations, because they harm future
interface stability. */
struct poptOption long_options[] = {
POPT_AUTOHELP
{ "helper-protocol", 0, POPT_ARG_STRING, &helper_protocol, OPT_DOMAIN, "operate as a stdio-based helper", "helper protocol to use"},
{ "domain", 0, POPT_ARG_STRING, &opt_domain, OPT_DOMAIN, "domain name"},
{ "workstation", 0, POPT_ARG_STRING, &opt_workstation, OPT_WORKSTATION, "workstation"},
{ "username", 0, POPT_ARG_STRING, &opt_username, OPT_PASSWORD, "Username"},
{ "password", 0, POPT_ARG_STRING, &opt_password, OPT_PASSWORD, "User's plaintext password"},
{ "multiplex", 0, POPT_ARG_NONE, &opt_multiplex, OPT_MULTIPLEX, "Multiplex Mode"},
POPT_COMMON_SAMBA
POPT_COMMON_VERSION
POPT_TABLEEND
};
/* Samba client initialisation */
setup_logging(NULL, DEBUG_STDERR);
/* Parse options */
pc = poptGetContext("ntlm_auth", argc, argv, long_options, 0);
/* Parse command line options */
if (argc == 1) {
poptPrintHelp(pc, stderr, 0);
return 1;
}
pc = poptGetContext(NULL, argc, (const char **)argv, long_options,
POPT_CONTEXT_KEEP_FIRST);
while((opt = poptGetNextOpt(pc)) != -1) {
if (opt < -1) {
break;
}
}
if (opt < -1) {
fprintf(stderr, "%s: %s\n",
poptBadOption(pc, POPT_BADOPTION_NOALIAS),
poptStrerror(opt));
return 1;
}
ntlm_auth_init_subsystems;
if (opt_domain == NULL) {
opt_domain = lp_workgroup();
}
if (helper_protocol) {
int i;
for (i=0; i<NUM_HELPER_MODES; i++) {
if (strcmp(helper_protocol, stdio_helper_protocols[i].name) == 0) {
squid_stream(stdio_helper_protocols[i].mode, stdio_helper_protocols[i].fn);
exit(0);
}
}
x_fprintf(x_stderr, "unknown helper protocol [%s]\n\nValid helper protools:\n\n", helper_protocol);
for (i=0; i<NUM_HELPER_MODES; i++) {
x_fprintf(x_stderr, "%s\n", stdio_helper_protocols[i].name);
}
exit(1);
}
if (!opt_username) {
x_fprintf(x_stderr, "username must be specified!\n\n");
poptPrintHelp(pc, stderr, 0);
exit(1);
}
if (opt_workstation == NULL) {
opt_workstation = lp_netbios_name();
}
if (!opt_password) {
opt_password = getpass("password: ");
}
{
char *user;
asprintf(&user, "%s%c%s", opt_domain, *lp_winbind_separator(), opt_username);
if (!check_plaintext_auth(user, opt_password, True)) {
return 1;
}
}
/* Exit code */
poptFreeContext(pc);
return 0;
}