1
0
mirror of https://github.com/samba-team/samba.git synced 2025-01-11 05:18:09 +03:00
samba-mirror/source4/rpc_server/samr/dcesrv_samr.c

3475 lines
91 KiB
C
Raw Normal View History

/*
Unix SMB/CIFS implementation.
endpoint server for the samr pipe
Copyright (C) Andrew Tridgell 2004
Copyright (C) Volker Lendecke 2004
Copyright (C) Andrew Bartlett <abartlet@samba.org> 2004-2005
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 "librpc/gen_ndr/ndr_samr.h"
#include "rpc_server/dcerpc_server.h"
#include "rpc_server/common/common.h"
#include "rpc_server/samr/dcesrv_samr.h"
#include "system/time.h"
#include "lib/ldb/include/ldb.h"
#include "ads.h"
/*
This is a bad temporary hack until we have at least some kind of schema
support
*/
static char *ldb_hexstr(TALLOC_CTX *mem_ctx, uint32_t val)
{
return talloc_asprintf(mem_ctx, "0x%.8x", val);
}
/*
samr_Connect
create a connection to the SAM database
*/
static NTSTATUS samr_Connect(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_Connect *r)
{
struct samr_connect_state *c_state;
struct dcesrv_handle *handle;
ZERO_STRUCTP(r->out.connect_handle);
c_state = talloc(dce_call->conn, struct samr_connect_state);
if (!c_state) {
return NT_STATUS_NO_MEMORY;
}
/* make sure the sam database is accessible */
c_state->sam_ctx = samdb_connect(c_state);
if (c_state->sam_ctx == NULL) {
talloc_free(c_state);
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return NT_STATUS_INVALID_SYSTEM_SERVICE;
}
handle = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_CONNECT);
if (!handle) {
talloc_free(c_state);
return NT_STATUS_NO_MEMORY;
}
handle->data = talloc_steal(handle, c_state);
c_state->access_mask = r->in.access_mask;
*r->out.connect_handle = handle->wire_handle;
return NT_STATUS_OK;
}
/*
samr_Close
*/
static NTSTATUS samr_Close(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_Close *r)
{
struct dcesrv_handle *h;
*r->out.handle = *r->in.handle;
DCESRV_PULL_HANDLE(h, r->in.handle, DCESRV_HANDLE_ANY);
talloc_free(h);
ZERO_STRUCTP(r->out.handle);
return NT_STATUS_OK;
}
/*
samr_SetSecurity
*/
static NTSTATUS samr_SetSecurity(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetSecurity *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_QuerySecurity
*/
static NTSTATUS samr_QuerySecurity(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QuerySecurity *r)
{
struct dcesrv_handle *h;
struct sec_desc_buf *sd;
r->out.sdbuf = NULL;
DCESRV_PULL_HANDLE(h, r->in.handle, DCESRV_HANDLE_ANY);
sd = talloc(mem_ctx, struct sec_desc_buf);
if (sd == NULL) {
return NT_STATUS_NO_MEMORY;
}
sd->sd = samdb_default_security_descriptor(mem_ctx);
r->out.sdbuf = sd;
return NT_STATUS_OK;
}
/*
samr_Shutdown
we refuse this operation completely. If a admin wants to shutdown samr
in Samba then they should use the samba admin tools to disable the samr pipe
*/
static NTSTATUS samr_Shutdown(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_Shutdown *r)
{
return NT_STATUS_ACCESS_DENIED;
}
/*
samr_LookupDomain
this maps from a domain name to a SID
*/
static NTSTATUS samr_LookupDomain(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_LookupDomain *r)
{
struct samr_connect_state *c_state;
struct dcesrv_handle *h;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
struct dom_sid *sid;
const char * const dom_attrs[] = { "objectSid", NULL};
const char * const ref_attrs[] = { "ncName", NULL};
struct ldb_message **dom_msgs;
struct ldb_message **ref_msgs;
int ret;
r->out.sid = NULL;
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_PULL_HANDLE(h, r->in.connect_handle, SAMR_HANDLE_CONNECT);
c_state = h->data;
if (r->in.domain_name->string == NULL) {
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return NT_STATUS_INVALID_PARAMETER;
}
if (strcasecmp(r->in.domain_name->string, "BUILTIN") == 0) {
ret = gendb_search(c_state->sam_ctx,
mem_ctx, NULL, &dom_msgs, dom_attrs,
"(objectClass=builtinDomain)");
} else {
ret = gendb_search(c_state->sam_ctx,
mem_ctx, NULL, &ref_msgs, ref_attrs,
"(&(&(nETBIOSName=%s)(objectclass=crossRef))(ncName=*))",
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
r->in.domain_name->string);
if (ret != 1) {
return NT_STATUS_NO_SUCH_DOMAIN;
}
ret = gendb_search_dn(c_state->sam_ctx, mem_ctx,
samdb_result_dn(mem_ctx,
ref_msgs[0], "ncName", NULL),
&dom_msgs, dom_attrs);
}
if (ret != 1) {
return NT_STATUS_NO_SUCH_DOMAIN;
}
sid = samdb_result_dom_sid(mem_ctx, dom_msgs[0],
"objectSid");
if (sid == NULL) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
return NT_STATUS_NO_SUCH_DOMAIN;
}
r->out.sid = sid;
return NT_STATUS_OK;
}
/*
samr_EnumDomains
list the domains in the SAM
*/
static NTSTATUS samr_EnumDomains(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_EnumDomains *r)
{
struct samr_connect_state *c_state;
struct dcesrv_handle *h;
struct samr_SamArray *array;
int count, i, start_i;
const char * const dom_attrs[] = { "cn", NULL};
const char * const ref_attrs[] = { "nETBIOSName", NULL};
struct ldb_message **dom_msgs;
struct ldb_message **ref_msgs;
*r->out.resume_handle = 0;
r->out.sam = NULL;
r->out.num_entries = 0;
DCESRV_PULL_HANDLE(h, r->in.connect_handle, SAMR_HANDLE_CONNECT);
c_state = h->data;
count = gendb_search(c_state->sam_ctx,
mem_ctx, NULL, &dom_msgs, dom_attrs,
"(objectClass=domain)");
if (count == -1) {
DEBUG(0,("samdb: no domains found in EnumDomains\n"));
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
*r->out.resume_handle = count;
start_i = *r->in.resume_handle;
if (start_i >= count) {
/* search past end of list is not an error for this call */
return NT_STATUS_OK;
}
array = talloc(mem_ctx, struct samr_SamArray);
if (array == NULL) {
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return NT_STATUS_NO_MEMORY;
}
array->count = 0;
array->entries = NULL;
array->entries = talloc_array(mem_ctx, struct samr_SamEntry, count - start_i);
if (array->entries == NULL) {
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return NT_STATUS_NO_MEMORY;
}
for (i=0;i<count-start_i;i++) {
int ret;
array->entries[i].idx = start_i + i;
/* try and find the domain */
ret = gendb_search(c_state->sam_ctx, mem_ctx, NULL,
&ref_msgs, ref_attrs,
"(&(objectClass=crossRef)(ncName=%s))",
ldb_dn_linearize(mem_ctx, dom_msgs[i]->dn));
if (ret == 1) {
array->entries[i].name.string = samdb_result_string(ref_msgs[0], "nETBIOSName", NULL);
} else {
array->entries[i].name.string = samdb_result_string(dom_msgs[i], "cn", NULL);
}
}
r->out.sam = array;
r->out.num_entries = i;
array->count = r->out.num_entries;
return NT_STATUS_OK;
}
/*
samr_OpenDomain
*/
static NTSTATUS samr_OpenDomain(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_OpenDomain *r)
{
struct dcesrv_handle *h_conn, *h_domain;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *domain_name;
struct samr_connect_state *c_state;
struct samr_domain_state *d_state;
const char * const dom_attrs[] = { "cn", NULL};
const char * const ref_attrs[] = { "nETBIOSName", NULL};
struct ldb_message **dom_msgs;
struct ldb_message **ref_msgs;
int ret;
ZERO_STRUCTP(r->out.domain_handle);
DCESRV_PULL_HANDLE(h_conn, r->in.connect_handle, SAMR_HANDLE_CONNECT);
c_state = h_conn->data;
if (r->in.sid == NULL) {
return NT_STATUS_INVALID_PARAMETER;
}
ret = gendb_search(c_state->sam_ctx,
mem_ctx, NULL, &dom_msgs, dom_attrs,
"(&(objectSid=%s)(&(objectclass=domain)))",
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
ldap_encode_ndr_dom_sid(mem_ctx, r->in.sid));
if (ret != 1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
} else {
ret = gendb_search(c_state->sam_ctx,
mem_ctx, NULL, &ref_msgs, ref_attrs,
"(&(&(nETBIOSName=*)(objectclass=crossRef))(ncName=%s))",
ldb_dn_linearize(mem_ctx, dom_msgs[0]->dn));
if (ret == 0) {
domain_name = ldb_msg_find_string(dom_msgs[0], "cn", NULL);
if (domain_name == NULL) {
return NT_STATUS_NO_SUCH_DOMAIN;
}
} else if (ret == 1) {
domain_name = ldb_msg_find_string(ref_msgs[0], "nETBIOSName", NULL);
if (domain_name == NULL) {
return NT_STATUS_NO_SUCH_DOMAIN;
}
} else {
return NT_STATUS_NO_SUCH_DOMAIN;
}
}
d_state = talloc(c_state, struct samr_domain_state);
if (!d_state) {
return NT_STATUS_NO_MEMORY;
}
d_state->connect_state = talloc_reference(d_state, c_state);
d_state->sam_ctx = c_state->sam_ctx;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
d_state->domain_sid = dom_sid_dup(d_state, r->in.sid);
d_state->domain_name = talloc_strdup(d_state, domain_name);
d_state->domain_dn = ldb_dn_copy(d_state, dom_msgs[0]->dn);
if (!d_state->domain_sid || !d_state->domain_name || !d_state->domain_dn) {
talloc_free(d_state);
return NT_STATUS_NO_MEMORY;
}
d_state->access_mask = r->in.access_mask;
h_domain = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_DOMAIN);
if (!h_domain) {
talloc_free(d_state);
return NT_STATUS_NO_MEMORY;
}
h_domain->data = talloc_steal(h_domain, d_state);
*r->out.domain_handle = h_domain->wire_handle;
return NT_STATUS_OK;
}
/*
return DomInfo1
*/
static NTSTATUS samr_info_DomInfo1(struct samr_domain_state *state,
TALLOC_CTX *mem_ctx,
struct samr_DomInfo1 *info)
{
const char * const attrs[] = { "minPwdLength", "pwdHistoryLength",
"pwdProperties", "maxPwdAge",
"minPwdAge", NULL };
int ret;
struct ldb_message **res;
ret = gendb_search_dn(state->sam_ctx, mem_ctx,
state->domain_dn , &res, attrs);
if (ret != 1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
info->min_password_length =
samdb_result_uint(res[0], "minPwdLength", 0);
info->password_history_length =
samdb_result_uint(res[0], "pwdHistoryLength", 0);
info->password_properties =
samdb_result_uint(res[0], "pwdProperties", 0);
info->max_password_age =
samdb_result_int64(res[0], "maxPwdAge", 0);
info->min_password_age =
samdb_result_int64(res[0], "minPwdAge", 0);
return NT_STATUS_OK;
}
/*
return DomInfo2
*/
static NTSTATUS samr_info_DomInfo2(struct samr_domain_state *state, TALLOC_CTX *mem_ctx,
struct samr_DomInfo2 *info)
{
const char * const dom_attrs[] = { "comment", NULL };
int ret;
struct ldb_message **dom_msgs;
const char *domain_name;
ret = gendb_search_dn(state->sam_ctx, mem_ctx,
state->domain_dn, &dom_msgs, dom_attrs);
if (ret != 1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
domain_name = state->domain_name;
/* where is this supposed to come from? is it settable? */
info->force_logoff_time = 0x8000000000000000LL;
info->comment.string = samdb_result_string(dom_msgs[0], "comment", NULL);
info->domain_name.string = domain_name;
info->primary.string = lp_netbios_name();
info->sequence_num = 0;
info->role = ROLE_DOMAIN_PDC;
info->num_users = samdb_search_count(state->sam_ctx, mem_ctx, NULL, "(objectClass=user)");
info->num_groups = samdb_search_count(state->sam_ctx, mem_ctx, NULL,
"(&(objectClass=group)(sAMAccountType=%u))",
ATYPE_GLOBAL_GROUP);
info->num_aliases = samdb_search_count(state->sam_ctx, mem_ctx, NULL,
"(&(objectClass=group)(sAMAccountType=%u))",
ATYPE_LOCAL_GROUP);
return NT_STATUS_OK;
}
/*
return DomInfo3
*/
static NTSTATUS samr_info_DomInfo3(struct samr_domain_state *state,
TALLOC_CTX *mem_ctx,
struct samr_DomInfo3 *info)
{
/* where is this supposed to come from? is it settable? */
info->force_logoff_time = 0x8000000000000000LL;
return NT_STATUS_OK;
}
/*
samr_QueryDomainInfo
*/
static NTSTATUS samr_QueryDomainInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryDomainInfo *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
r->out.info = NULL;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
r->out.info = talloc(mem_ctx, union samr_DomainInfo);
if (!r->out.info) {
return NT_STATUS_NO_MEMORY;
}
ZERO_STRUCTP(r->out.info);
switch (r->in.level) {
case 1:
return samr_info_DomInfo1(d_state, mem_ctx,
&r->out.info->info1);
case 2:
return samr_info_DomInfo2(d_state, mem_ctx, &r->out.info->info2);
case 3:
return samr_info_DomInfo3(d_state, mem_ctx,
&r->out.info->info3);
}
return NT_STATUS_INVALID_INFO_CLASS;
}
/*
samr_SetDomainInfo
*/
static NTSTATUS samr_SetDomainInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetDomainInfo *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_CreateDomainGroup
*/
static NTSTATUS samr_CreateDomainGroup(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_CreateDomainGroup *r)
{
struct samr_domain_state *d_state;
struct samr_account_state *a_state;
struct dcesrv_handle *h;
const char *name;
struct ldb_message *msg;
struct dom_sid *sid;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *groupname;
struct dcesrv_handle *g_handle;
int ret;
ZERO_STRUCTP(r->out.group_handle);
*r->out.rid = 0;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
groupname = r->in.name->string;
if (groupname == NULL) {
return NT_STATUS_INVALID_PARAMETER;
}
/* check if the group already exists */
name = samdb_search_string(d_state->sam_ctx, mem_ctx, NULL,
"sAMAccountName",
"(&(sAMAccountName=%s)(objectclass=group))",
groupname);
if (name != NULL) {
return NT_STATUS_GROUP_EXISTS;
}
msg = ldb_msg_new(mem_ctx);
if (msg == NULL) {
return NT_STATUS_NO_MEMORY;
}
/* add core elements to the ldb_message for the user */
msg->dn = ldb_dn_string_compose(mem_ctx, d_state->domain_dn,
"CN=%s, CN=Users", groupname);
if (!msg->dn) {
return NT_STATUS_NO_MEMORY;
}
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg, "sAMAccountName", groupname);
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg, "objectClass", "group");
/* create the group */
ret = samdb_add(d_state->sam_ctx, mem_ctx, msg);
if (ret != 0) {
DEBUG(0,("Failed to create group record %s\n",
ldb_dn_linearize(mem_ctx, msg->dn)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
a_state = talloc(d_state, struct samr_account_state);
if (!a_state) {
return NT_STATUS_NO_MEMORY;
}
a_state->sam_ctx = d_state->sam_ctx;
a_state->access_mask = r->in.access_mask;
a_state->domain_state = talloc_reference(a_state, d_state);
a_state->account_dn = talloc_steal(a_state, msg->dn);
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
/* retrieve the sid for the group just created */
sid = samdb_search_dom_sid(d_state->sam_ctx, a_state,
msg->dn, "objectSid", "dn=%s",
ldb_dn_linearize(mem_ctx, msg->dn));
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
if (sid == NULL) {
return NT_STATUS_UNSUCCESSFUL;
}
a_state->account_name = talloc_strdup(a_state, groupname);
if (!a_state->account_name) {
return NT_STATUS_NO_MEMORY;
}
/* create the policy handle */
g_handle = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_GROUP);
if (!g_handle) {
return NT_STATUS_NO_MEMORY;
}
g_handle->data = talloc_steal(g_handle, a_state);
*r->out.group_handle = g_handle->wire_handle;
*r->out.rid = sid->sub_auths[sid->num_auths-1];
return NT_STATUS_OK;
}
/*
comparison function for sorting SamEntry array
*/
static int compare_SamEntry(struct samr_SamEntry *e1, struct samr_SamEntry *e2)
{
return e1->idx - e2->idx;
}
/*
samr_EnumDomainGroups
*/
static NTSTATUS samr_EnumDomainGroups(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_EnumDomainGroups *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
struct ldb_message **res;
int ldb_cnt, count, i, first;
struct samr_SamEntry *entries;
const char * const attrs[3] = { "objectSid", "sAMAccountName", NULL };
*r->out.resume_handle = 0;
r->out.sam = NULL;
r->out.num_entries = 0;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
/* search for all domain groups in this domain. This could possibly be
cached and resumed based on resume_key */
ldb_cnt = samdb_search_domain(d_state->sam_ctx, mem_ctx,
d_state->domain_dn, &res, attrs,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
d_state->domain_sid,
"(&(grouptype=%s)(objectclass=group))",
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_GLOBAL_GROUP));
if (ldb_cnt == -1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
if (ldb_cnt == 0 || r->in.max_size == 0) {
return NT_STATUS_OK;
}
/* convert to SamEntry format */
entries = talloc_array(mem_ctx, struct samr_SamEntry, ldb_cnt);
if (!entries) {
return NT_STATUS_NO_MEMORY;
}
count = 0;
for (i=0;i<ldb_cnt;i++) {
struct dom_sid *group_sid;
group_sid = samdb_result_dom_sid(mem_ctx, res[i],
"objectSid");
if (group_sid == NULL)
continue;
entries[count].idx =
group_sid->sub_auths[group_sid->num_auths-1];
entries[count].name.string =
samdb_result_string(res[i], "sAMAccountName", "");
count += 1;
}
/* sort the results by rid */
qsort(entries, count, sizeof(struct samr_SamEntry),
(comparison_fn_t)compare_SamEntry);
/* find the first entry to return */
for (first=0;
first<count && entries[first].idx <= *r->in.resume_handle;
first++) ;
if (first == count) {
return NT_STATUS_OK;
}
/* return the rest, limit by max_size. Note that we
use the w2k3 element size value of 54 */
r->out.num_entries = count - first;
r->out.num_entries = MIN(r->out.num_entries,
1+(r->in.max_size/SAMR_ENUM_USERS_MULTIPLIER));
r->out.sam = talloc(mem_ctx, struct samr_SamArray);
if (!r->out.sam) {
return NT_STATUS_NO_MEMORY;
}
r->out.sam->entries = entries+first;
r->out.sam->count = r->out.num_entries;
if (r->out.num_entries < count - first) {
*r->out.resume_handle = entries[first+r->out.num_entries-1].idx;
return STATUS_MORE_ENTRIES;
}
return NT_STATUS_OK;
}
/*
samr_CreateUser2
TODO: This should do some form of locking, especially around the rid allocation
*/
static NTSTATUS samr_CreateUser2(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_CreateUser2 *r)
{
struct samr_domain_state *d_state;
struct samr_account_state *a_state;
struct dcesrv_handle *h;
const char *name;
struct ldb_message *msg;
struct dom_sid *sid;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *account_name;
struct dcesrv_handle *u_handle;
int ret;
const char *container, *obj_class=NULL;
char *cn_name;
int cn_name_len;
ZERO_STRUCTP(r->out.user_handle);
*r->out.access_granted = 0;
*r->out.rid = 0;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
account_name = r->in.account_name->string;
if (account_name == NULL) {
return NT_STATUS_INVALID_PARAMETER;
}
/* check if the user already exists */
name = samdb_search_string(d_state->sam_ctx, mem_ctx, NULL,
"sAMAccountName",
"(&(sAMAccountName=%s)(objectclass=user))", account_name);
if (name != NULL) {
return NT_STATUS_USER_EXISTS;
}
msg = ldb_msg_new(mem_ctx);
if (msg == NULL) {
return NT_STATUS_NO_MEMORY;
}
cn_name = talloc_strdup(mem_ctx, account_name);
NT_STATUS_HAVE_NO_MEMORY(cn_name);
cn_name_len = strlen(cn_name);
/* This must be one of these values *only* */
if (r->in.acct_flags == ACB_NORMAL) {
container = "Users";
obj_class = "user";
} else if (r->in.acct_flags == ACB_WSTRUST) {
if (cn_name[cn_name_len - 1] != '$') {
return NT_STATUS_FOOBAR;
}
cn_name[cn_name_len - 1] = '\0';
container = "Computers";
obj_class = "computer";
} else if (r->in.acct_flags == ACB_SVRTRUST) {
if (cn_name[cn_name_len - 1] != '$') {
return NT_STATUS_FOOBAR;
}
cn_name[cn_name_len - 1] = '\0';
container = "Domain Controllers";
obj_class = "computer";
} else if (r->in.acct_flags == ACB_DOMTRUST) {
container = "Users";
obj_class = "user";
} else {
return NT_STATUS_INVALID_PARAMETER;
}
/* add core elements to the ldb_message for the user */
msg->dn = ldb_dn_build_child(mem_ctx, "CN", cn_name, ldb_dn_build_child(mem_ctx, "CN", container, d_state->domain_dn));
if (!msg->dn) {
return NT_STATUS_NO_MEMORY;
}
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg, "sAMAccountName", account_name);
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg, "objectClass", obj_class);
/* create the user */
ret = samdb_add(d_state->sam_ctx, mem_ctx, msg);
if (ret != 0) {
DEBUG(0,("Failed to create user record %s\n",
ldb_dn_linearize(mem_ctx, msg->dn)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
a_state = talloc(d_state, struct samr_account_state);
if (!a_state) {
return NT_STATUS_NO_MEMORY;
}
a_state->sam_ctx = d_state->sam_ctx;
a_state->access_mask = r->in.access_mask;
a_state->domain_state = talloc_reference(a_state, d_state);
a_state->account_dn = talloc_steal(a_state, msg->dn);
/* retrieve the sid for the user just created */
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
sid = samdb_search_dom_sid(d_state->sam_ctx, a_state,
msg->dn, "objectSid", "dn=%s", ldb_dn_linearize(mem_ctx, msg->dn));
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
if (sid == NULL) {
return NT_STATUS_UNSUCCESSFUL;
}
a_state->account_name = talloc_strdup(a_state, account_name);
if (!a_state->account_name) {
return NT_STATUS_NO_MEMORY;
}
/* create the policy handle */
u_handle = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_USER);
if (!u_handle) {
return NT_STATUS_NO_MEMORY;
}
u_handle->data = talloc_steal(u_handle, a_state);
*r->out.user_handle = u_handle->wire_handle;
*r->out.access_granted = 0xf07ff; /* TODO: fix access mask calculations */
*r->out.rid = sid->sub_auths[sid->num_auths-1];
return NT_STATUS_OK;
}
/*
samr_CreateUser
*/
static NTSTATUS samr_CreateUser(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_CreateUser *r)
{
struct samr_CreateUser2 r2;
uint32_t access_granted = 0;
/* a simple wrapper around samr_CreateUser2 works nicely */
r2.in.domain_handle = r->in.domain_handle;
r2.in.account_name = r->in.account_name;
r2.in.acct_flags = ACB_NORMAL;
r2.in.access_mask = r->in.access_mask;
r2.out.user_handle = r->out.user_handle;
r2.out.access_granted = &access_granted;
r2.out.rid = r->out.rid;
return samr_CreateUser2(dce_call, mem_ctx, &r2);
}
/*
samr_EnumDomainUsers
*/
static NTSTATUS samr_EnumDomainUsers(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_EnumDomainUsers *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
struct ldb_message **res;
int count, i, first;
struct samr_SamEntry *entries;
const char * const attrs[3] = { "objectSid", "sAMAccountName", NULL };
*r->out.resume_handle = 0;
r->out.sam = NULL;
r->out.num_entries = 0;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
/* search for all users in this domain. This could possibly be cached and
resumed based on resume_key */
count = gendb_search(d_state->sam_ctx, mem_ctx, d_state->domain_dn, &res, attrs,
"objectclass=user");
if (count == -1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
if (count == 0 || r->in.max_size == 0) {
return NT_STATUS_OK;
}
/* convert to SamEntry format */
entries = talloc_array(mem_ctx, struct samr_SamEntry, count);
if (!entries) {
return NT_STATUS_NO_MEMORY;
}
for (i=0;i<count;i++) {
entries[i].idx = samdb_result_rid_from_sid(mem_ctx, res[i], "objectSid", 0);
entries[i].name.string = samdb_result_string(res[i], "sAMAccountName", "");
}
/* sort the results by rid */
qsort(entries, count, sizeof(struct samr_SamEntry),
(comparison_fn_t)compare_SamEntry);
/* find the first entry to return */
for (first=0;
first<count && entries[first].idx <= *r->in.resume_handle;
first++) ;
if (first == count) {
return NT_STATUS_OK;
}
/* return the rest, limit by max_size. Note that we
use the w2k3 element size value of 54 */
r->out.num_entries = count - first;
r->out.num_entries = MIN(r->out.num_entries,
1+(r->in.max_size/SAMR_ENUM_USERS_MULTIPLIER));
r->out.sam = talloc(mem_ctx, struct samr_SamArray);
if (!r->out.sam) {
return NT_STATUS_NO_MEMORY;
}
r->out.sam->entries = entries+first;
r->out.sam->count = r->out.num_entries;
if (r->out.num_entries < count - first) {
*r->out.resume_handle = entries[first+r->out.num_entries-1].idx;
return STATUS_MORE_ENTRIES;
}
return NT_STATUS_OK;
}
/*
samr_CreateDomAlias
*/
static NTSTATUS samr_CreateDomAlias(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_CreateDomAlias *r)
{
struct samr_domain_state *d_state;
struct samr_account_state *a_state;
struct dcesrv_handle *h;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *alias_name, *name;
struct ldb_message *msg;
struct dom_sid *sid;
struct dcesrv_handle *a_handle;
int ret;
ZERO_STRUCTP(r->out.alias_handle);
*r->out.rid = 0;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
alias_name = r->in.alias_name->string;
if (alias_name == NULL) {
return NT_STATUS_INVALID_PARAMETER;
}
/* Check if alias already exists */
name = samdb_search_string(d_state->sam_ctx, mem_ctx, NULL,
"sAMAccountName",
"(sAMAccountName=%s)(objectclass=group))",
alias_name);
if (name != NULL) {
return NT_STATUS_ALIAS_EXISTS;
}
msg = ldb_msg_new(mem_ctx);
if (msg == NULL) {
return NT_STATUS_NO_MEMORY;
}
/* add core elements to the ldb_message for the alias */
msg->dn = ldb_dn_string_compose(mem_ctx, d_state->domain_dn,
"CN=%s, CN=Users", alias_name);
if (!msg->dn) {
return NT_STATUS_NO_MEMORY;
}
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg, "sAMAccountName", alias_name);
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg, "objectClass", "group");
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg, "groupType", "0x80000004");
/* create the alias */
ret = samdb_add(d_state->sam_ctx, mem_ctx, msg);
if (ret != 0) {
DEBUG(0,("Failed to create alias record %s\n",
ldb_dn_linearize(mem_ctx, msg->dn)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
a_state = talloc(d_state, struct samr_account_state);
if (!a_state) {
return NT_STATUS_NO_MEMORY;
}
a_state->sam_ctx = d_state->sam_ctx;
a_state->access_mask = r->in.access_mask;
a_state->domain_state = talloc_reference(a_state, d_state);
a_state->account_dn = talloc_steal(a_state, msg->dn);
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
/* retrieve the sid for the alias just created */
sid = samdb_search_dom_sid(d_state->sam_ctx, a_state,
msg->dn, "objectSid", "dn=%s",
ldb_dn_linearize(mem_ctx, msg->dn));
a_state->account_name = talloc_strdup(a_state, alias_name);
if (!a_state->account_name) {
return NT_STATUS_NO_MEMORY;
}
/* create the policy handle */
a_handle = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_ALIAS);
if (a_handle == NULL)
return NT_STATUS_NO_MEMORY;
a_handle->data = talloc_steal(a_handle, a_state);
*r->out.alias_handle = a_handle->wire_handle;
*r->out.rid = sid->sub_auths[sid->num_auths-1];
return NT_STATUS_OK;
}
/*
samr_EnumDomainAliases
*/
static NTSTATUS samr_EnumDomainAliases(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_EnumDomainAliases *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
struct ldb_message **res;
int ldb_cnt, count, i, first;
struct samr_SamEntry *entries;
const char * const attrs[3] = { "objectSid", "sAMAccountName", NULL };
*r->out.resume_handle = 0;
r->out.sam = NULL;
r->out.num_entries = 0;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
/* search for all domain groups in this domain. This could possibly be
cached and resumed based on resume_key */
ldb_cnt = samdb_search_domain(d_state->sam_ctx, mem_ctx,
d_state->domain_dn,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
&res, attrs,
d_state->domain_sid,
"(&(|(grouptype=%s)(grouptype=%s)))"
"(objectclass=group))",
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_BUILTIN_LOCAL_GROUP),
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_DOMAIN_LOCAL_GROUP));
if (ldb_cnt == -1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
if (ldb_cnt == 0) {
return NT_STATUS_OK;
}
/* convert to SamEntry format */
entries = talloc_array(mem_ctx, struct samr_SamEntry, ldb_cnt);
if (!entries) {
return NT_STATUS_NO_MEMORY;
}
count = 0;
for (i=0;i<ldb_cnt;i++) {
struct dom_sid *alias_sid;
alias_sid = samdb_result_dom_sid(mem_ctx, res[i],
"objectSid");
if (alias_sid == NULL)
continue;
entries[count].idx =
alias_sid->sub_auths[alias_sid->num_auths-1];
entries[count].name.string =
samdb_result_string(res[i], "sAMAccountName", "");
count += 1;
}
/* sort the results by rid */
qsort(entries, count, sizeof(struct samr_SamEntry),
(comparison_fn_t)compare_SamEntry);
/* find the first entry to return */
for (first=0;
first<count && entries[first].idx <= *r->in.resume_handle;
first++) ;
if (first == count) {
return NT_STATUS_OK;
}
r->out.num_entries = count - first;
r->out.num_entries = MIN(r->out.num_entries, 1000);
r->out.sam = talloc(mem_ctx, struct samr_SamArray);
if (!r->out.sam) {
return NT_STATUS_NO_MEMORY;
}
r->out.sam->entries = entries+first;
r->out.sam->count = r->out.num_entries;
if (r->out.num_entries < count - first) {
*r->out.resume_handle =
entries[first+r->out.num_entries-1].idx;
return STATUS_MORE_ENTRIES;
}
return NT_STATUS_OK;
}
/*
samr_GetAliasMembership
*/
static NTSTATUS samr_GetAliasMembership(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetAliasMembership *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
struct ldb_message **res;
int i, count = 0;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
if (r->in.sids->num_sids > 0) {
const char *filter;
const char * const attrs[2] = { "objectSid", NULL };
filter = talloc_asprintf(mem_ctx,
"(&(|(grouptype=%s)(grouptype=%s))"
"(objectclass=group)(|",
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_BUILTIN_LOCAL_GROUP),
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_DOMAIN_LOCAL_GROUP));
if (filter == NULL)
return NT_STATUS_NO_MEMORY;
for (i=0; i<r->in.sids->num_sids; i++) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *memberdn;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
memberdn =
samdb_search_string(d_state->sam_ctx,
mem_ctx, NULL, "dn",
"(objectSid=%s)",
ldap_encode_ndr_dom_sid(mem_ctx,
r->in.sids->sids[i].sid));
if (memberdn == NULL)
continue;
filter = talloc_asprintf(mem_ctx, "%s(member=%s)",
filter, memberdn);
if (filter == NULL)
return NT_STATUS_NO_MEMORY;
}
count = samdb_search_domain(d_state->sam_ctx, mem_ctx,
d_state->domain_dn, &res, attrs,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
d_state->domain_sid, "%s))", filter);
if (count < 0)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
r->out.rids->count = 0;
r->out.rids->ids = talloc_array(mem_ctx, uint32_t, count);
if (r->out.rids->ids == NULL)
return NT_STATUS_NO_MEMORY;
for (i=0; i<count; i++) {
struct dom_sid *alias_sid;
alias_sid = samdb_result_dom_sid(mem_ctx, res[i], "objectSid");
if (alias_sid == NULL) {
DEBUG(0, ("Could not find objectSid\n"));
continue;
}
r->out.rids->ids[r->out.rids->count] =
alias_sid->sub_auths[alias_sid->num_auths-1];
r->out.rids->count += 1;
}
return NT_STATUS_OK;
}
/*
samr_LookupNames
*/
static NTSTATUS samr_LookupNames(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_LookupNames *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
int i;
NTSTATUS status = NT_STATUS_OK;
const char * const attrs[] = { "sAMAccountType", "objectSid", NULL };
int count;
ZERO_STRUCT(r->out.rids);
ZERO_STRUCT(r->out.types);
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
if (r->in.num_names == 0) {
return NT_STATUS_OK;
}
r->out.rids.ids = talloc_array(mem_ctx, uint32_t, r->in.num_names);
r->out.types.ids = talloc_array(mem_ctx, uint32_t, r->in.num_names);
if (!r->out.rids.ids || !r->out.types.ids) {
return NT_STATUS_NO_MEMORY;
}
r->out.rids.count = r->in.num_names;
r->out.types.count = r->in.num_names;
for (i=0;i<r->in.num_names;i++) {
struct ldb_message **res;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
struct dom_sid *sid;
uint32_t atype, rtype;
r->out.rids.ids[i] = 0;
r->out.types.ids[i] = SID_NAME_UNKNOWN;
count = gendb_search(d_state->sam_ctx, mem_ctx, d_state->domain_dn, &res, attrs,
"sAMAccountName=%s", r->in.names[i].string);
if (count != 1) {
status = STATUS_SOME_UNMAPPED;
continue;
}
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
sid = samdb_result_dom_sid(mem_ctx, res[0], "objectSid");
if (sid == NULL) {
status = STATUS_SOME_UNMAPPED;
continue;
}
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
atype = samdb_result_uint(res[0], "sAMAccountType", 0);
if (atype == 0) {
status = STATUS_SOME_UNMAPPED;
continue;
}
rtype = samdb_atype_map(atype);
if (rtype == SID_NAME_UNKNOWN) {
status = STATUS_SOME_UNMAPPED;
continue;
}
r->out.rids.ids[i] = sid->sub_auths[sid->num_auths-1];
r->out.types.ids[i] = rtype;
}
return status;
}
/*
samr_LookupRids
*/
static NTSTATUS samr_LookupRids(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_LookupRids *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
int i, total;
NTSTATUS status = NT_STATUS_OK;
struct lsa_String *names;
uint32_t *ids;
ZERO_STRUCT(r->out.names);
ZERO_STRUCT(r->out.types);
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
if (r->in.num_rids == 0)
return NT_STATUS_OK;
names = talloc_array(mem_ctx, struct lsa_String, r->in.num_rids);
ids = talloc_array(mem_ctx, uint32_t, r->in.num_rids);
if ((names == NULL) || (ids == NULL))
return NT_STATUS_NO_MEMORY;
total = 0;
for (i=0; i<r->in.num_rids; i++) {
struct ldb_message **res;
int count;
const char * const attrs[] = { "sAMAccountType",
"sAMAccountName", NULL };
uint32_t atype;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
struct dom_sid *sid;
ids[i] = SID_NAME_UNKNOWN;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
sid = dom_sid_add_rid(mem_ctx, d_state->domain_sid, r->in.rids[i]);
if (sid == NULL) {
names[i].string = NULL;
status = STATUS_SOME_UNMAPPED;
continue;
}
count = gendb_search(d_state->sam_ctx, mem_ctx,
d_state->domain_dn, &res, attrs,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
"(objectSid=%s)",
ldap_encode_ndr_dom_sid(mem_ctx, sid));
if (count != 1) {
names[i].string = NULL;
status = STATUS_SOME_UNMAPPED;
continue;
}
names[i].string = samdb_result_string(res[0], "sAMAccountName",
NULL);
atype = samdb_result_uint(res[0], "sAMAccountType", 0);
if (atype == 0) {
status = STATUS_SOME_UNMAPPED;
continue;
}
ids[i] = samdb_atype_map(atype);
if (ids[i] == SID_NAME_UNKNOWN) {
status = STATUS_SOME_UNMAPPED;
continue;
}
}
r->out.names.names = names;
r->out.names.count = r->in.num_rids;
r->out.types.ids = ids;
r->out.types.count = r->in.num_rids;
return status;
}
/*
samr_OpenGroup
*/
static NTSTATUS samr_OpenGroup(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_OpenGroup *r)
{
struct samr_domain_state *d_state;
struct samr_account_state *a_state;
struct dcesrv_handle *h;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *groupname;
struct dom_sid *sid;
struct ldb_message **msgs;
struct dcesrv_handle *g_handle;
const char * const attrs[2] = { "sAMAccountName", NULL };
int ret;
ZERO_STRUCTP(r->out.group_handle);
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
/* form the group SID */
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
sid = dom_sid_add_rid(mem_ctx, d_state->domain_sid, r->in.rid);
if (!sid) {
return NT_STATUS_NO_MEMORY;
}
/* search for the group record */
ret = gendb_search(d_state->sam_ctx,
mem_ctx, d_state->domain_dn, &msgs, attrs,
"(&(objectSid=%s)(objectclass=group)"
"(grouptype=%s))",
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
ldap_encode_ndr_dom_sid(mem_ctx, sid),
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_GLOBAL_GROUP));
if (ret == 0) {
return NT_STATUS_NO_SUCH_GROUP;
}
if (ret != 1) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
DEBUG(0,("Found %d records matching sid %s\n",
ret, dom_sid_string(mem_ctx, sid)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
groupname = samdb_result_string(msgs[0], "sAMAccountName", NULL);
if (groupname == NULL) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
DEBUG(0,("sAMAccountName field missing for sid %s\n",
dom_sid_string(mem_ctx, sid)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
a_state = talloc(d_state, struct samr_account_state);
if (!a_state) {
return NT_STATUS_NO_MEMORY;
}
a_state->sam_ctx = d_state->sam_ctx;
a_state->access_mask = r->in.access_mask;
a_state->domain_state = talloc_reference(a_state, d_state);
a_state->account_dn = talloc_steal(a_state, msgs[0]->dn);
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
a_state->account_sid = talloc_steal(a_state, sid);
a_state->account_name = talloc_strdup(a_state, groupname);
if (!a_state->account_name) {
return NT_STATUS_NO_MEMORY;
}
/* create the policy handle */
g_handle = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_GROUP);
if (!g_handle) {
return NT_STATUS_NO_MEMORY;
}
g_handle->data = talloc_steal(g_handle, a_state);
*r->out.group_handle = g_handle->wire_handle;
return NT_STATUS_OK;
}
/* these query macros make samr_Query[User|Group]Info a bit easier to read */
#define QUERY_STRING(msg, field, attr) \
r->out.info->field = samdb_result_string(msg, attr, "");
#define QUERY_UINT(msg, field, attr) \
r->out.info->field = samdb_result_uint(msg, attr, 0);
#define QUERY_RID(msg, field, attr) \
r->out.info->field = samdb_result_rid_from_sid(mem_ctx, msg, attr, 0);
#define QUERY_NTTIME(msg, field, attr) \
r->out.info->field = samdb_result_nttime(msg, attr, 0);
#define QUERY_APASSC(msg, field, attr) \
r->out.info->field = samdb_result_allow_password_change(a_state->sam_ctx, mem_ctx, \
a_state->domain_state->domain_dn, msg, attr);
#define QUERY_FPASSC(msg, field, attr) \
r->out.info->field = samdb_result_force_password_change(a_state->sam_ctx, mem_ctx, \
a_state->domain_state->domain_dn, msg, attr);
#define QUERY_LHOURS(msg, field, attr) \
r->out.info->field = samdb_result_logon_hours(mem_ctx, msg, attr);
#define QUERY_AFLAGS(msg, field, attr) \
r->out.info->field = samdb_result_acct_flags(msg, attr);
/* these are used to make the Set[User|Group]Info code easier to follow */
#define SET_STRING(mod, field, attr) do { \
if (r->in.info->field == NULL) return NT_STATUS_INVALID_PARAMETER; \
if (samdb_msg_add_string(a_state->sam_ctx, mem_ctx, mod, attr, r->in.info->field) != 0) { \
return NT_STATUS_NO_MEMORY; \
} \
} while (0)
#define SET_UINT(mod, field, attr) do { \
if (samdb_msg_add_uint(a_state->sam_ctx, mem_ctx, mod, attr, r->in.info->field) != 0) { \
return NT_STATUS_NO_MEMORY; \
} \
} while (0)
#define SET_AFLAGS(msg, field, attr) do { \
if (samdb_msg_add_acct_flags(a_state->sam_ctx, mem_ctx, msg, attr, r->in.info->field) != 0) { \
return NT_STATUS_NO_MEMORY; \
} \
} while (0)
#define SET_LHOURS(msg, field, attr) do { \
if (samdb_msg_add_logon_hours(a_state->sam_ctx, mem_ctx, msg, attr, &r->in.info->field) != 0) { \
return NT_STATUS_NO_MEMORY; \
} \
} while (0)
/*
samr_QueryGroupInfo
*/
static NTSTATUS samr_QueryGroupInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryGroupInfo *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct ldb_message *msg, **res;
const char * const attrs[4] = { "sAMAccountName", "description",
"numMembers", NULL };
int ret;
r->out.info = NULL;
DCESRV_PULL_HANDLE(h, r->in.group_handle, SAMR_HANDLE_GROUP);
a_state = h->data;
/* pull all the group attributes */
ret = gendb_search_dn(a_state->sam_ctx, mem_ctx,
a_state->account_dn, &res, attrs);
if (ret != 1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
msg = res[0];
/* allocate the info structure */
r->out.info = talloc(mem_ctx, union samr_GroupInfo);
if (r->out.info == NULL) {
return NT_STATUS_NO_MEMORY;
}
ZERO_STRUCTP(r->out.info);
/* Fill in the level */
switch (r->in.level) {
case GROUPINFOALL:
QUERY_STRING(msg, all.name.string, "sAMAccountName");
r->out.info->all.attributes = 7; /* Do like w2k3 */
QUERY_UINT (msg, all.num_members, "numMembers")
QUERY_STRING(msg, all.description.string, "description");
break;
case GROUPINFONAME:
QUERY_STRING(msg, name.string, "sAMAccountName");
break;
case GROUPINFOX:
r->out.info->unknown.unknown = 7;
break;
case GROUPINFODESCRIPTION:
QUERY_STRING(msg, description.string, "description");
break;
default:
r->out.info = NULL;
return NT_STATUS_INVALID_INFO_CLASS;
}
return NT_STATUS_OK;
}
/*
samr_SetGroupInfo
*/
static NTSTATUS samr_SetGroupInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetGroupInfo *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct ldb_message *msg;
int ret;
DCESRV_PULL_HANDLE(h, r->in.group_handle, SAMR_HANDLE_GROUP);
a_state = h->data;
msg = ldb_msg_new(mem_ctx);
if (msg == NULL) {
return NT_STATUS_NO_MEMORY;
}
msg->dn = ldb_dn_copy(mem_ctx, a_state->account_dn);
if (!msg->dn) {
return NT_STATUS_NO_MEMORY;
}
switch (r->in.level) {
case GROUPINFODESCRIPTION:
SET_STRING(msg, description.string, "description");
break;
case GROUPINFONAME:
/* On W2k3 this does not change the name, it changes the
* sAMAccountName attribute */
SET_STRING(msg, name.string, "sAMAccountName");
break;
case GROUPINFOX:
/* This does not do anything obviously visible in W2k3 LDAP */
break;
default:
return NT_STATUS_INVALID_INFO_CLASS;
}
/* modify the samdb record */
ret = samdb_replace(a_state->sam_ctx, mem_ctx, msg);
if (ret != 0) {
/* we really need samdb.c to return NTSTATUS */
return NT_STATUS_UNSUCCESSFUL;
}
return NT_STATUS_OK;
}
/*
samr_AddGroupMember
*/
static NTSTATUS samr_AddGroupMember(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_AddGroupMember *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct samr_domain_state *d_state;
struct ldb_message *mod;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
struct dom_sid *membersid;
const char *memberdn;
struct ldb_message **msgs;
const char * const attrs[2] = { "dn", NULL };
int ret;
DCESRV_PULL_HANDLE(h, r->in.group_handle, SAMR_HANDLE_GROUP);
a_state = h->data;
d_state = a_state->domain_state;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
membersid = dom_sid_add_rid(mem_ctx, d_state->domain_sid, r->in.rid);
if (membersid == NULL)
return NT_STATUS_NO_MEMORY;
/* In native mode, AD can also nest domain groups. Not sure yet
* whether this is also available via RPC. */
ret = gendb_search(d_state->sam_ctx, mem_ctx, d_state->domain_dn,
&msgs, attrs, "(&(objectSid=%s)(objectclass=user))",
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
ldap_encode_ndr_dom_sid(mem_ctx, membersid));
if (ret == 0)
return NT_STATUS_NO_SUCH_USER;
if (ret > 1)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
memberdn = samdb_result_string(msgs[0], "dn", NULL);
if (memberdn == NULL)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
mod = ldb_msg_new(mem_ctx);
if (mod == NULL) {
return NT_STATUS_NO_MEMORY;
}
mod->dn = talloc_reference(mem_ctx, a_state->account_dn);
if (samdb_msg_add_addval(d_state->sam_ctx, mem_ctx, mod, "member",
memberdn) != 0)
return NT_STATUS_UNSUCCESSFUL;
if (samdb_modify(a_state->sam_ctx, mem_ctx, mod) != 0)
return NT_STATUS_UNSUCCESSFUL;
return NT_STATUS_OK;
}
/*
samr_DeleteDomainGroup
*/
static NTSTATUS samr_DeleteDomainGroup(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_DeleteDomainGroup *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
int ret;
*r->out.group_handle = *r->in.group_handle;
DCESRV_PULL_HANDLE(h, r->in.group_handle, SAMR_HANDLE_GROUP);
a_state = h->data;
ret = samdb_delete(a_state->sam_ctx, mem_ctx, a_state->account_dn);
if (ret != 0) {
return NT_STATUS_UNSUCCESSFUL;
}
ZERO_STRUCTP(r->out.group_handle);
return NT_STATUS_OK;
}
/*
samr_DeleteGroupMember
*/
static NTSTATUS samr_DeleteGroupMember(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_DeleteGroupMember *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct samr_domain_state *d_state;
struct ldb_message *mod;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
struct dom_sid *membersid;
const char *memberdn;
struct ldb_message **msgs;
const char * const attrs[2] = { "dn", NULL };
int ret;
DCESRV_PULL_HANDLE(h, r->in.group_handle, SAMR_HANDLE_GROUP);
a_state = h->data;
d_state = a_state->domain_state;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
membersid = dom_sid_add_rid(mem_ctx, d_state->domain_sid, r->in.rid);
if (membersid == NULL)
return NT_STATUS_NO_MEMORY;
/* In native mode, AD can also nest domain groups. Not sure yet
* whether this is also available via RPC. */
ret = gendb_search(d_state->sam_ctx, mem_ctx, d_state->domain_dn,
&msgs, attrs, "(&(objectSid=%s)(objectclass=user))",
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
ldap_encode_ndr_dom_sid(mem_ctx, membersid));
if (ret == 0)
return NT_STATUS_NO_SUCH_USER;
if (ret > 1)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
memberdn = samdb_result_string(msgs[0], "dn", NULL);
if (memberdn == NULL)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
mod = ldb_msg_new(mem_ctx);
if (mod == NULL) {
return NT_STATUS_NO_MEMORY;
}
mod->dn = talloc_reference(mem_ctx, a_state->account_dn);
if (samdb_msg_add_delval(d_state->sam_ctx, mem_ctx, mod, "member",
memberdn) != 0)
return NT_STATUS_UNSUCCESSFUL;
if (samdb_modify(a_state->sam_ctx, mem_ctx, mod) != 0)
return NT_STATUS_UNSUCCESSFUL;
return NT_STATUS_OK;
}
/*
samr_QueryGroupMember
*/
static NTSTATUS samr_QueryGroupMember(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryGroupMember *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct ldb_message **res;
struct ldb_message_element *el;
struct samr_RidTypeArray *array;
const char * const attrs[2] = { "member", NULL };
int ret;
DCESRV_PULL_HANDLE(h, r->in.group_handle, SAMR_HANDLE_GROUP);
a_state = h->data;
/* pull the member attribute */
ret = gendb_search_dn(a_state->sam_ctx, mem_ctx,
a_state->account_dn, &res, attrs);
if (ret != 1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
array = talloc(mem_ctx, struct samr_RidTypeArray);
if (array == NULL)
return NT_STATUS_NO_MEMORY;
ZERO_STRUCTP(array);
el = ldb_msg_find_element(res[0], "member");
if (el != NULL) {
int i;
array->count = el->num_values;
array->rids = talloc_array(mem_ctx, uint32_t,
el->num_values);
if (array->rids == NULL)
return NT_STATUS_NO_MEMORY;
array->types = talloc_array(mem_ctx, uint32_t,
el->num_values);
if (array->types == NULL)
return NT_STATUS_NO_MEMORY;
for (i=0; i<el->num_values; i++) {
struct ldb_message **res2;
const char * const attrs2[2] = { "objectSid", NULL };
ret = gendb_search_dn(a_state->sam_ctx, mem_ctx,
ldb_dn_explode(mem_ctx, el->values[i].data),
&res2, attrs2);
if (ret != 1)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
array->rids[i] =
samdb_result_rid_from_sid(mem_ctx, res2[0],
"objectSid", 0);
if (array->rids[i] == 0)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
array->types[i] = 7; /* RID type of some kind, not sure what the value means. */
}
}
r->out.rids = array;
return NT_STATUS_OK;
}
/*
samr_SetMemberAttributesOfGroup
*/
static NTSTATUS samr_SetMemberAttributesOfGroup(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetMemberAttributesOfGroup *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_OpenAlias
*/
static NTSTATUS samr_OpenAlias(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_OpenAlias *r)
{
struct samr_domain_state *d_state;
struct samr_account_state *a_state;
struct dcesrv_handle *h;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *alias_name;
struct dom_sid *sid;
struct ldb_message **msgs;
struct dcesrv_handle *g_handle;
const char * const attrs[2] = { "sAMAccountName", NULL };
int ret;
ZERO_STRUCTP(r->out.alias_handle);
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
/* form the alias SID */
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
sid = dom_sid_add_rid(mem_ctx, d_state->domain_sid, r->in.rid);
if (sid == NULL)
return NT_STATUS_NO_MEMORY;
/* search for the group record */
ret = gendb_search(d_state->sam_ctx,
mem_ctx, d_state->domain_dn, &msgs, attrs,
"(&(objectSid=%s)(objectclass=group)"
"(|(grouptype=%s)(grouptype=%s)))",
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
ldap_encode_ndr_dom_sid(mem_ctx, sid),
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_BUILTIN_LOCAL_GROUP),
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_DOMAIN_LOCAL_GROUP));
if (ret == 0) {
return NT_STATUS_NO_SUCH_ALIAS;
}
if (ret != 1) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
DEBUG(0,("Found %d records matching sid %s\n",
ret, dom_sid_string(mem_ctx, sid)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
alias_name = samdb_result_string(msgs[0], "sAMAccountName", NULL);
if (alias_name == NULL) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
DEBUG(0,("sAMAccountName field missing for sid %s\n",
dom_sid_string(mem_ctx, sid)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
a_state = talloc(d_state, struct samr_account_state);
if (!a_state) {
return NT_STATUS_NO_MEMORY;
}
a_state->sam_ctx = d_state->sam_ctx;
a_state->access_mask = r->in.access_mask;
a_state->domain_state = talloc_reference(a_state, d_state);
a_state->account_dn = talloc_steal(a_state, msgs[0]->dn);
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
a_state->account_sid = talloc_steal(a_state, sid);
a_state->account_name = talloc_strdup(a_state, alias_name);
if (!a_state->account_name) {
return NT_STATUS_NO_MEMORY;
}
/* create the policy handle */
g_handle = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_ALIAS);
if (!g_handle) {
return NT_STATUS_NO_MEMORY;
}
g_handle->data = talloc_steal(g_handle, a_state);
*r->out.alias_handle = g_handle->wire_handle;
return NT_STATUS_OK;
}
/*
samr_QueryAliasInfo
*/
static NTSTATUS samr_QueryAliasInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryAliasInfo *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct ldb_message *msg, **res;
const char * const attrs[4] = { "sAMAccountName", "description",
"numMembers", NULL };
int ret;
r->out.info = NULL;
DCESRV_PULL_HANDLE(h, r->in.alias_handle, SAMR_HANDLE_ALIAS);
a_state = h->data;
/* pull all the alias attributes */
ret = gendb_search_dn(a_state->sam_ctx, mem_ctx,
a_state->account_dn ,&res, attrs);
if (ret != 1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
msg = res[0];
/* allocate the info structure */
r->out.info = talloc(mem_ctx, union samr_AliasInfo);
if (r->out.info == NULL) {
return NT_STATUS_NO_MEMORY;
}
ZERO_STRUCTP(r->out.info);
switch(r->in.level) {
case ALIASINFOALL:
QUERY_STRING(msg, all.name.string, "sAMAccountName");
QUERY_UINT (msg, all.num_members, "numMembers");
QUERY_STRING(msg, all.description.string, "description");
break;
case ALIASINFONAME:
QUERY_STRING(msg, name.string, "sAMAccountName");
break;
case ALIASINFODESCRIPTION:
QUERY_STRING(msg, description.string, "description");
break;
default:
r->out.info = NULL;
return NT_STATUS_INVALID_INFO_CLASS;
}
return NT_STATUS_OK;
}
/*
samr_SetAliasInfo
*/
static NTSTATUS samr_SetAliasInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetAliasInfo *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct ldb_message *msg;
int ret;
DCESRV_PULL_HANDLE(h, r->in.alias_handle, SAMR_HANDLE_ALIAS);
a_state = h->data;
msg = ldb_msg_new(mem_ctx);
if (msg == NULL) {
return NT_STATUS_NO_MEMORY;
}
msg->dn = ldb_dn_copy(mem_ctx, a_state->account_dn);
if (!msg->dn) {
return NT_STATUS_NO_MEMORY;
}
switch (r->in.level) {
case ALIASINFODESCRIPTION:
SET_STRING(msg, description.string, "description");
break;
case ALIASINFONAME:
/* On W2k3 this does not change the name, it changes the
* sAMAccountName attribute */
SET_STRING(msg, name.string, "sAMAccountName");
break;
default:
return NT_STATUS_INVALID_INFO_CLASS;
}
/* modify the samdb record */
ret = samdb_replace(a_state->sam_ctx, mem_ctx, msg);
if (ret != 0) {
/* we really need samdb.c to return NTSTATUS */
return NT_STATUS_UNSUCCESSFUL;
}
return NT_STATUS_OK;
}
/*
samr_DeleteDomAlias
*/
static NTSTATUS samr_DeleteDomAlias(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_DeleteDomAlias *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
int ret;
*r->out.alias_handle = *r->in.alias_handle;
DCESRV_PULL_HANDLE(h, r->in.alias_handle, SAMR_HANDLE_ALIAS);
a_state = h->data;
ret = samdb_delete(a_state->sam_ctx, mem_ctx, a_state->account_dn);
if (ret != 0) {
return NT_STATUS_UNSUCCESSFUL;
}
ZERO_STRUCTP(r->out.alias_handle);
return NT_STATUS_OK;
}
/*
samr_AddAliasMember
*/
static NTSTATUS samr_AddAliasMember(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_AddAliasMember *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct samr_domain_state *d_state;
struct ldb_message *mod;
struct ldb_message **msgs;
const char * const attrs[2] = { "dn", NULL };
struct ldb_dn *memberdn = NULL;
int ret;
DCESRV_PULL_HANDLE(h, r->in.alias_handle, SAMR_HANDLE_ALIAS);
a_state = h->data;
d_state = a_state->domain_state;
ret = gendb_search(d_state->sam_ctx, mem_ctx, NULL,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
&msgs, attrs, "(objectsid=%s)",
ldap_encode_ndr_dom_sid(mem_ctx, r->in.sid));
if (ret == 1) {
memberdn = ldb_dn_explode(mem_ctx, ldb_msg_find_string(msgs[0], "dn", NULL));
} else if (ret > 1) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
DEBUG(0,("Found %d records matching sid %s\n",
ret, dom_sid_string(mem_ctx, r->in.sid)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
} else if (ret == 0) {
struct ldb_message *msg;
struct ldb_dn *basedn;
const char *sidstr;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
sidstr = dom_sid_string(mem_ctx, r->in.sid);
NT_STATUS_HAVE_NO_MEMORY(sidstr);
/* We might have to create a ForeignSecurityPrincipal, but
* only if it's not our own domain */
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
if (dom_sid_in_domain(d_state->domain_sid, r->in.sid))
return NT_STATUS_OBJECT_NAME_NOT_FOUND;
msg = ldb_msg_new(mem_ctx);
if (msg == NULL) {
return NT_STATUS_NO_MEMORY;
}
/* TODO: Hmmm. This feels wrong. How do I find the base dn to
* put the ForeignSecurityPrincipals? d_state->domain_dn does
* not work, this is wrong for the Builtin domain, there's no
* cn=For...,cn=Builtin,dc={BASEDN}. -- vl
*/
basedn = samdb_search_dn(d_state->sam_ctx, mem_ctx, NULL,
"(&(objectClass=container)(cn=ForeignSecurityPrincipals))");
if (basedn == NULL) {
DEBUG(0, ("Failed to find DN for "
"ForeignSecurityPrincipal container\n"));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
/* add core elements to the ldb_message for the alias */
msg->dn = ldb_dn_build_child(mem_ctx, "CN", sidstr, basedn);
if (msg->dn == NULL)
return NT_STATUS_NO_MEMORY;
memberdn = msg->dn;
samdb_msg_add_string(d_state->sam_ctx, mem_ctx, msg,
"objectClass",
"foreignSecurityPrincipal");
/* create the alias */
ret = samdb_add(d_state->sam_ctx, mem_ctx, msg);
if (ret != 0) {
DEBUG(0,("Failed to create foreignSecurityPrincipal "
"record %s\n", ldb_dn_linearize(mem_ctx, msg->dn)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
} else {
DEBUG(0, ("samdb_search returned %d\n", ret));
}
if (memberdn == NULL) {
DEBUG(0, ("Could not find memberdn\n"));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
mod = ldb_msg_new(mem_ctx);
if (mod == NULL) {
return NT_STATUS_NO_MEMORY;
}
mod->dn = talloc_reference(mem_ctx, a_state->account_dn);
if (samdb_msg_add_addval(d_state->sam_ctx, mem_ctx, mod, "member",
ldb_dn_linearize(mem_ctx, memberdn)) != 0)
return NT_STATUS_UNSUCCESSFUL;
if (samdb_modify(a_state->sam_ctx, mem_ctx, mod) != 0)
return NT_STATUS_UNSUCCESSFUL;
return NT_STATUS_OK;
}
/*
samr_DeleteAliasMember
*/
static NTSTATUS samr_DeleteAliasMember(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_DeleteAliasMember *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct samr_domain_state *d_state;
struct ldb_message *mod;
const char *memberdn;
DCESRV_PULL_HANDLE(h, r->in.alias_handle, SAMR_HANDLE_ALIAS);
a_state = h->data;
d_state = a_state->domain_state;
memberdn = samdb_search_string(d_state->sam_ctx, mem_ctx, NULL,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
"dn", "(objectSid=%s)",
ldap_encode_ndr_dom_sid(mem_ctx, r->in.sid));
if (memberdn == NULL)
return NT_STATUS_OBJECT_NAME_NOT_FOUND;
mod = ldb_msg_new(mem_ctx);
if (mod == NULL) {
return NT_STATUS_NO_MEMORY;
}
mod->dn = talloc_reference(mem_ctx, a_state->account_dn);
if (samdb_msg_add_delval(d_state->sam_ctx, mem_ctx, mod, "member",
memberdn) != 0)
return NT_STATUS_UNSUCCESSFUL;
if (samdb_modify(a_state->sam_ctx, mem_ctx, mod) != 0)
return NT_STATUS_UNSUCCESSFUL;
return NT_STATUS_OK;
}
/*
samr_GetMembersInAlias
*/
static NTSTATUS samr_GetMembersInAlias(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetMembersInAlias *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct samr_domain_state *d_state;
struct ldb_message **msgs;
struct lsa_SidPtr *sids;
struct ldb_message_element *el;
const char * const attrs[2] = { "member", NULL};
int ret;
DCESRV_PULL_HANDLE(h, r->in.alias_handle, SAMR_HANDLE_ALIAS);
a_state = h->data;
d_state = a_state->domain_state;
ret = gendb_search_dn(d_state->sam_ctx, mem_ctx,
a_state->account_dn, &msgs, attrs);
if (ret != 1)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
r->out.sids->num_sids = 0;
r->out.sids->sids = NULL;
el = ldb_msg_find_element(msgs[0], "member");
if (el != NULL) {
int i;
sids = talloc_array(mem_ctx, struct lsa_SidPtr,
el->num_values);
if (sids == NULL)
return NT_STATUS_NO_MEMORY;
for (i=0; i<el->num_values; i++) {
struct ldb_message **msgs2;
const char * const attrs2[2] = { "objectSid", NULL };
ret = gendb_search_dn(a_state->sam_ctx, mem_ctx,
ldb_dn_explode(mem_ctx, el->values[i].data),
&msgs2, attrs2);
if (ret != 1)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
sids[i].sid = samdb_result_dom_sid(mem_ctx, msgs2[0],
"objectSid");
if (sids[i].sid == NULL)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
r->out.sids->num_sids = el->num_values;
r->out.sids->sids = sids;
}
return NT_STATUS_OK;
}
/*
samr_OpenUser
*/
static NTSTATUS samr_OpenUser(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_OpenUser *r)
{
struct samr_domain_state *d_state;
struct samr_account_state *a_state;
struct dcesrv_handle *h;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *account_name;
struct dom_sid *sid;
struct ldb_message **msgs;
struct dcesrv_handle *u_handle;
const char * const attrs[2] = { "sAMAccountName", NULL };
int ret;
ZERO_STRUCTP(r->out.user_handle);
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
/* form the users SID */
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
sid = dom_sid_add_rid(mem_ctx, d_state->domain_sid, r->in.rid);
if (!sid) {
return NT_STATUS_NO_MEMORY;
}
/* search for the user record */
ret = gendb_search(d_state->sam_ctx,
mem_ctx, d_state->domain_dn, &msgs, attrs,
"(&(objectSid=%s)(objectclass=user))",
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
ldap_encode_ndr_dom_sid(mem_ctx, sid));
if (ret == 0) {
return NT_STATUS_NO_SUCH_USER;
}
if (ret != 1) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
DEBUG(0,("Found %d records matching sid %s\n", ret,
dom_sid_string(mem_ctx, sid)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
account_name = samdb_result_string(msgs[0], "sAMAccountName", NULL);
if (account_name == NULL) {
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
DEBUG(0,("sAMAccountName field missing for sid %s\n",
dom_sid_string(mem_ctx, sid)));
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
a_state = talloc(mem_ctx, struct samr_account_state);
if (!a_state) {
return NT_STATUS_NO_MEMORY;
}
a_state->sam_ctx = d_state->sam_ctx;
a_state->access_mask = r->in.access_mask;
a_state->domain_state = talloc_reference(a_state, d_state);
a_state->account_dn = talloc_steal(a_state, msgs[0]->dn);
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
a_state->account_sid = talloc_steal(a_state, sid);
a_state->account_name = talloc_strdup(a_state, account_name);
if (!a_state->account_name) {
return NT_STATUS_NO_MEMORY;
}
/* create the policy handle */
u_handle = dcesrv_handle_new(dce_call->context, SAMR_HANDLE_USER);
if (!u_handle) {
return NT_STATUS_NO_MEMORY;
}
u_handle->data = talloc_steal(u_handle, a_state);
*r->out.user_handle = u_handle->wire_handle;
return NT_STATUS_OK;
}
/*
samr_DeleteUser
*/
static NTSTATUS samr_DeleteUser(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_DeleteUser *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
int ret;
*r->out.user_handle = *r->in.user_handle;
DCESRV_PULL_HANDLE(h, r->in.user_handle, SAMR_HANDLE_USER);
a_state = h->data;
ret = samdb_delete(a_state->sam_ctx, mem_ctx, a_state->account_dn);
if (ret != 0) {
return NT_STATUS_UNSUCCESSFUL;
}
ZERO_STRUCTP(r->out.user_handle);
return NT_STATUS_OK;
}
/*
samr_QueryUserInfo
*/
static NTSTATUS samr_QueryUserInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryUserInfo *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct ldb_message *msg, **res;
int ret;
r->out.info = NULL;
DCESRV_PULL_HANDLE(h, r->in.user_handle, SAMR_HANDLE_USER);
a_state = h->data;
/* pull all the user attributes */
ret = gendb_search_dn(a_state->sam_ctx, mem_ctx,
a_state->account_dn ,&res, NULL);
if (ret != 1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
msg = res[0];
/* allocate the info structure */
r->out.info = talloc(mem_ctx, union samr_UserInfo);
if (r->out.info == NULL) {
return NT_STATUS_NO_MEMORY;
}
ZERO_STRUCTP(r->out.info);
/* fill in the reply */
switch (r->in.level) {
case 1:
QUERY_STRING(msg, info1.account_name.string, "sAMAccountName");
QUERY_STRING(msg, info1.full_name.string, "displayName");
QUERY_UINT (msg, info1.primary_gid, "primaryGroupID");
QUERY_STRING(msg, info1.description.string, "description");
QUERY_STRING(msg, info1.comment.string, "comment");
break;
case 2:
QUERY_STRING(msg, info2.comment.string, "comment");
QUERY_UINT (msg, info2.country_code, "countryCode");
QUERY_UINT (msg, info2.code_page, "codePage");
break;
case 3:
QUERY_STRING(msg, info3.account_name.string, "sAMAccountName");
QUERY_STRING(msg, info3.full_name.string, "displayName");
QUERY_RID (msg, info3.rid, "objectSid");
QUERY_UINT (msg, info3.primary_gid, "primaryGroupID");
QUERY_STRING(msg, info3.home_directory.string, "homeDirectory");
QUERY_STRING(msg, info3.home_drive.string, "homeDrive");
QUERY_STRING(msg, info3.logon_script.string, "scriptPath");
QUERY_STRING(msg, info3.profile_path.string, "profilePath");
QUERY_STRING(msg, info3.workstations.string, "userWorkstations");
QUERY_NTTIME(msg, info3.last_logon, "lastLogon");
QUERY_NTTIME(msg, info3.last_logoff, "lastLogoff");
QUERY_NTTIME(msg, info3.last_password_change, "pwdLastSet");
QUERY_APASSC(msg, info3.allow_password_change, "pwdLastSet");
QUERY_FPASSC(msg, info3.force_password_change, "pwdLastSet");
QUERY_LHOURS(msg, info3.logon_hours, "logonHours");
QUERY_UINT (msg, info3.bad_password_count, "badPwdCount");
QUERY_UINT (msg, info3.logon_count, "logonCount");
QUERY_AFLAGS(msg, info3.acct_flags, "userAccountControl");
break;
case 4:
QUERY_LHOURS(msg, info4.logon_hours, "logonHours");
break;
case 5:
QUERY_STRING(msg, info5.account_name.string, "sAMAccountName");
QUERY_STRING(msg, info5.full_name.string, "displayName");
QUERY_RID (msg, info5.rid, "objectSid");
QUERY_UINT (msg, info5.primary_gid, "primaryGroupID");
QUERY_STRING(msg, info5.home_directory.string, "homeDirectory");
QUERY_STRING(msg, info5.home_drive.string, "homeDrive");
QUERY_STRING(msg, info5.logon_script.string, "scriptPath");
QUERY_STRING(msg, info5.profile_path.string, "profilePath");
QUERY_STRING(msg, info5.description.string, "description");
QUERY_STRING(msg, info5.workstations.string, "userWorkstations");
QUERY_NTTIME(msg, info5.last_logon, "lastLogon");
QUERY_NTTIME(msg, info5.last_logoff, "lastLogoff");
QUERY_LHOURS(msg, info5.logon_hours, "logonHours");
QUERY_UINT (msg, info5.bad_password_count, "badPwdCount");
QUERY_UINT (msg, info5.logon_count, "logonCount");
QUERY_NTTIME(msg, info5.last_password_change, "pwdLastSet");
QUERY_NTTIME(msg, info5.acct_expiry, "accountExpires");
QUERY_AFLAGS(msg, info5.acct_flags, "userAccountControl");
break;
case 6:
QUERY_STRING(msg, info6.account_name.string, "sAMAccountName");
QUERY_STRING(msg, info6.full_name.string, "displayName");
break;
case 7:
QUERY_STRING(msg, info7.account_name.string, "sAMAccountName");
break;
case 8:
QUERY_STRING(msg, info8.full_name.string, "displayName");
break;
case 9:
QUERY_UINT (msg, info9.primary_gid, "primaryGroupID");
break;
case 10:
QUERY_STRING(msg, info10.home_directory.string,"homeDirectory");
QUERY_STRING(msg, info10.home_drive.string, "homeDrive");
break;
case 11:
QUERY_STRING(msg, info11.logon_script.string, "scriptPath");
break;
case 12:
QUERY_STRING(msg, info12.profile_path.string, "profilePath");
break;
case 13:
QUERY_STRING(msg, info13.description.string, "description");
break;
case 14:
QUERY_STRING(msg, info14.workstations.string, "userWorkstations");
break;
case 16:
QUERY_AFLAGS(msg, info16.acct_flags, "userAccountControl");
break;
case 17:
QUERY_NTTIME(msg, info17.acct_expiry, "accountExpires");
case 20:
QUERY_STRING(msg, info20.parameters.string, "userParameters");
break;
case 21:
QUERY_NTTIME(msg, info21.last_logon, "lastLogon");
QUERY_NTTIME(msg, info21.last_logoff, "lastLogoff");
QUERY_NTTIME(msg, info21.last_password_change, "pwdLastSet");
QUERY_NTTIME(msg, info21.acct_expiry, "accountExpires");
QUERY_APASSC(msg, info21.allow_password_change,"pwdLastSet");
QUERY_FPASSC(msg, info21.force_password_change,"pwdLastSet");
QUERY_STRING(msg, info21.account_name.string, "sAMAccountName");
QUERY_STRING(msg, info21.full_name.string, "displayName");
QUERY_STRING(msg, info21.home_directory.string,"homeDirectory");
QUERY_STRING(msg, info21.home_drive.string, "homeDrive");
QUERY_STRING(msg, info21.logon_script.string, "scriptPath");
QUERY_STRING(msg, info21.profile_path.string, "profilePath");
QUERY_STRING(msg, info21.description.string, "description");
QUERY_STRING(msg, info21.workstations.string, "userWorkstations");
QUERY_STRING(msg, info21.comment.string, "comment");
QUERY_STRING(msg, info21.parameters.string, "userParameters");
QUERY_RID (msg, info21.rid, "objectSid");
QUERY_UINT (msg, info21.primary_gid, "primaryGroupID");
QUERY_AFLAGS(msg, info21.acct_flags, "userAccountControl");
r->out.info->info21.fields_present = 0x00FFFFFF;
QUERY_LHOURS(msg, info21.logon_hours, "logonHours");
QUERY_UINT (msg, info21.bad_password_count, "badPwdCount");
QUERY_UINT (msg, info21.logon_count, "logonCount");
QUERY_UINT (msg, info21.country_code, "countryCode");
QUERY_UINT (msg, info21.code_page, "codePage");
break;
default:
r->out.info = NULL;
return NT_STATUS_INVALID_INFO_CLASS;
}
return NT_STATUS_OK;
}
/*
samr_SetUserInfo
*/
static NTSTATUS samr_SetUserInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetUserInfo *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct ldb_message *msg;
int ret;
NTSTATUS status = NT_STATUS_OK;
DCESRV_PULL_HANDLE(h, r->in.user_handle, SAMR_HANDLE_USER);
a_state = h->data;
msg = ldb_msg_new(mem_ctx);
if (msg == NULL) {
return NT_STATUS_NO_MEMORY;
}
msg->dn = talloc_reference(mem_ctx, a_state->account_dn);
if (!msg->dn) {
return NT_STATUS_NO_MEMORY;
}
switch (r->in.level) {
case 2:
SET_STRING(msg, info2.comment.string, "comment");
SET_UINT (msg, info2.country_code, "countryCode");
SET_UINT (msg, info2.code_page, "codePage");
break;
case 4:
SET_LHOURS(msg, info4.logon_hours, "logonHours");
break;
case 6:
SET_STRING(msg, info6.full_name.string, "displayName");
break;
case 7:
SET_STRING(msg, info7.account_name.string, "samAccountName");
break;
case 8:
SET_STRING(msg, info8.full_name.string, "displayName");
break;
case 9:
SET_UINT(msg, info9.primary_gid, "primaryGroupID");
break;
case 10:
SET_STRING(msg, info10.home_directory.string, "homeDirectory");
SET_STRING(msg, info10.home_drive.string, "homeDrive");
break;
case 11:
SET_STRING(msg, info11.logon_script.string, "scriptPath");
break;
case 12:
SET_STRING(msg, info12.profile_path.string, "profilePath");
break;
case 13:
SET_STRING(msg, info13.description.string, "description");
break;
case 14:
SET_STRING(msg, info14.workstations.string, "userWorkstations");
break;
case 16:
SET_AFLAGS(msg, info16.acct_flags, "userAccountControl");
break;
case 20:
SET_STRING(msg, info20.parameters.string, "userParameters");
break;
case 21:
#define IFSET(bit) if (bit & r->in.info->info21.fields_present)
IFSET(SAMR_FIELD_ACCOUNT_NAME)
SET_STRING(msg, info21.account_name.string, "samAccountName");
IFSET(SAMR_FIELD_FULL_NAME)
SET_STRING(msg, info21.full_name.string, "displayName");
IFSET(SAMR_FIELD_DESCRIPTION)
SET_STRING(msg, info21.description.string, "description");
IFSET(SAMR_FIELD_COMMENT)
SET_STRING(msg, info21.comment.string, "comment");
IFSET(SAMR_FIELD_LOGON_SCRIPT)
SET_STRING(msg, info21.logon_script.string, "scriptPath");
IFSET(SAMR_FIELD_PROFILE_PATH)
SET_STRING(msg, info21.profile_path.string, "profilePath");
IFSET(SAMR_FIELD_WORKSTATIONS)
SET_STRING(msg, info21.workstations.string, "userWorkstations");
IFSET(SAMR_FIELD_LOGON_HOURS)
SET_LHOURS(msg, info21.logon_hours, "logonHours");
IFSET(SAMR_FIELD_ACCT_FLAGS)
SET_AFLAGS(msg, info21.acct_flags, "userAccountControl");
IFSET(SAMR_FIELD_PARAMETERS)
SET_STRING(msg, info21.parameters.string, "userParameters");
IFSET(SAMR_FIELD_COUNTRY_CODE)
SET_UINT (msg, info21.country_code, "countryCode");
IFSET(SAMR_FIELD_CODE_PAGE)
SET_UINT (msg, info21.code_page, "codePage");
/* Any reason the rest of these can't be set? */
#undef IFSET
break;
case 23:
#define IFSET(bit) if (bit & r->in.info->info23.info.fields_present)
IFSET(SAMR_FIELD_ACCOUNT_NAME)
SET_STRING(msg, info23.info.account_name.string, "samAccountName");
IFSET(SAMR_FIELD_FULL_NAME)
SET_STRING(msg, info23.info.full_name.string, "displayName");
IFSET(SAMR_FIELD_DESCRIPTION)
SET_STRING(msg, info23.info.description.string, "description");
IFSET(SAMR_FIELD_COMMENT)
SET_STRING(msg, info23.info.comment.string, "comment");
IFSET(SAMR_FIELD_LOGON_SCRIPT)
SET_STRING(msg, info23.info.logon_script.string, "scriptPath");
IFSET(SAMR_FIELD_PROFILE_PATH)
SET_STRING(msg, info23.info.profile_path.string, "profilePath");
IFSET(SAMR_FIELD_WORKSTATIONS)
SET_STRING(msg, info23.info.workstations.string, "userWorkstations");
IFSET(SAMR_FIELD_LOGON_HOURS)
SET_LHOURS(msg, info23.info.logon_hours, "logonHours");
IFSET(SAMR_FIELD_ACCT_FLAGS)
SET_AFLAGS(msg, info23.info.acct_flags, "userAccountControl");
IFSET(SAMR_FIELD_PARAMETERS)
SET_STRING(msg, info23.info.parameters.string, "userParameters");
IFSET(SAMR_FIELD_COUNTRY_CODE)
SET_UINT (msg, info23.info.country_code, "countryCode");
IFSET(SAMR_FIELD_CODE_PAGE)
SET_UINT (msg, info23.info.code_page, "codePage");
IFSET(SAMR_FIELD_PASSWORD) {
status = samr_set_password(dce_call,
a_state->sam_ctx,
a_state->account_dn,
a_state->domain_state->domain_dn,
mem_ctx, msg,
&r->in.info->info23.password);
} else IFSET(SAMR_FIELD_PASSWORD2) {
status = samr_set_password(dce_call,
a_state->sam_ctx,
a_state->account_dn,
a_state->domain_state->domain_dn,
mem_ctx, msg,
&r->in.info->info23.password);
}
#undef IFSET
break;
/* the set password levels are handled separately */
case 24:
status = samr_set_password(dce_call,
a_state->sam_ctx,
a_state->account_dn,
a_state->domain_state->domain_dn,
mem_ctx, msg,
&r->in.info->info24.password);
break;
case 25:
#define IFSET(bit) if (bit & r->in.info->info25.info.fields_present)
IFSET(SAMR_FIELD_ACCOUNT_NAME)
SET_STRING(msg, info25.info.account_name.string, "samAccountName");
IFSET(SAMR_FIELD_FULL_NAME)
SET_STRING(msg, info25.info.full_name.string, "displayName");
IFSET(SAMR_FIELD_DESCRIPTION)
SET_STRING(msg, info25.info.description.string, "description");
IFSET(SAMR_FIELD_COMMENT)
SET_STRING(msg, info25.info.comment.string, "comment");
IFSET(SAMR_FIELD_LOGON_SCRIPT)
SET_STRING(msg, info25.info.logon_script.string, "scriptPath");
IFSET(SAMR_FIELD_PROFILE_PATH)
SET_STRING(msg, info25.info.profile_path.string, "profilePath");
IFSET(SAMR_FIELD_WORKSTATIONS)
SET_STRING(msg, info25.info.workstations.string, "userWorkstations");
IFSET(SAMR_FIELD_LOGON_HOURS)
SET_LHOURS(msg, info25.info.logon_hours, "logonHours");
IFSET(SAMR_FIELD_ACCT_FLAGS)
SET_AFLAGS(msg, info25.info.acct_flags, "userAccountControl");
IFSET(SAMR_FIELD_PARAMETERS)
SET_STRING(msg, info25.info.parameters.string, "userParameters");
IFSET(SAMR_FIELD_COUNTRY_CODE)
SET_UINT (msg, info25.info.country_code, "countryCode");
IFSET(SAMR_FIELD_CODE_PAGE)
SET_UINT (msg, info25.info.code_page, "codePage");
IFSET(SAMR_FIELD_PASSWORD) {
status = samr_set_password_ex(dce_call,
a_state->sam_ctx,
a_state->account_dn,
a_state->domain_state->domain_dn,
mem_ctx, msg,
&r->in.info->info25.password);
} else IFSET(SAMR_FIELD_PASSWORD2) {
status = samr_set_password_ex(dce_call,
a_state->sam_ctx,
a_state->account_dn,
a_state->domain_state->domain_dn,
mem_ctx, msg,
&r->in.info->info25.password);
}
#undef IFSET
break;
/* the set password levels are handled separately */
case 26:
status = samr_set_password_ex(dce_call,
a_state->sam_ctx,
a_state->account_dn,
a_state->domain_state->domain_dn,
mem_ctx, msg,
&r->in.info->info26.password);
break;
default:
/* many info classes are not valid for SetUserInfo */
return NT_STATUS_INVALID_INFO_CLASS;
}
if (!NT_STATUS_IS_OK(status)) {
return status;
}
/* modify the samdb record */
ret = samdb_replace(a_state->sam_ctx, mem_ctx, msg);
if (ret != 0) {
/* we really need samdb.c to return NTSTATUS */
return NT_STATUS_UNSUCCESSFUL;
}
return NT_STATUS_OK;
}
/*
samr_GetGroupsForUser
*/
static NTSTATUS samr_GetGroupsForUser(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetGroupsForUser *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
struct samr_domain_state *d_state;
struct ldb_message **res;
const char * const attrs[2] = { "objectSid", NULL };
struct samr_RidWithAttributeArray *array;
int count;
DCESRV_PULL_HANDLE(h, r->in.user_handle, SAMR_HANDLE_USER);
a_state = h->data;
d_state = a_state->domain_state;
count = samdb_search_domain(a_state->sam_ctx, mem_ctx, NULL, &res,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
attrs, d_state->domain_sid,
"(&(member=%s)(grouptype=%s)(objectclass=group))",
ldb_dn_linearize(mem_ctx, a_state->account_dn),
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_GLOBAL_GROUP));
if (count < 0)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
array = talloc(mem_ctx, struct samr_RidWithAttributeArray);
if (array == NULL)
return NT_STATUS_NO_MEMORY;
array->count = 0;
array->rids = NULL;
if (count > 0) {
int i;
array->rids = talloc_array(mem_ctx, struct samr_RidWithAttribute,
count);
if (array->rids == NULL)
return NT_STATUS_NO_MEMORY;
for (i=0; i<count; i++) {
struct dom_sid *group_sid;
group_sid = samdb_result_dom_sid(mem_ctx, res[i],
"objectSid");
if (group_sid == NULL) {
DEBUG(0, ("Couldn't find objectSid attrib\n"));
continue;
}
array->rids[array->count].rid =
group_sid->sub_auths[group_sid->num_auths-1];
array->rids[array->count].attributes = SE_GROUP_MANDATORY | SE_GROUP_ENABLED_BY_DEFAULT | SE_GROUP_ENABLED;
array->count += 1;
}
}
r->out.rids = array;
return NT_STATUS_OK;
}
/*
samr_QueryDisplayInfo
*/
static NTSTATUS samr_QueryDisplayInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryDisplayInfo *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
struct ldb_message **res;
int ldb_cnt, count, i;
const char * const attrs[4] = { "objectSid", "sAMAccountName",
"description", NULL };
struct samr_DispEntryFull *entriesFull = NULL;
struct samr_DispEntryAscii *entriesAscii = NULL;
struct samr_DispEntryGeneral * entriesGeneral = NULL;
const char *filter;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
switch (r->in.level) {
case 1:
case 4:
filter = talloc_asprintf(mem_ctx, "(&(objectclass=user)"
"(sAMAccountType=%s))",
ldb_hexstr(mem_ctx,
ATYPE_NORMAL_ACCOUNT));
break;
case 2:
filter = talloc_asprintf(mem_ctx, "(&(objectclass=user)"
"(sAMAccountType=%s))",
ldb_hexstr(mem_ctx,
ATYPE_WORKSTATION_TRUST));
break;
case 3:
case 5:
filter = talloc_asprintf(mem_ctx, "(&(grouptype=%s)"
"(objectclass=group))",
ldb_hexstr(mem_ctx, GTYPE_SECURITY_GLOBAL_GROUP));
break;
default:
return NT_STATUS_INVALID_INFO_CLASS;
}
/* search for all requested objects in this domain. This could
possibly be cached and resumed based on resume_key */
ldb_cnt = samdb_search_domain(d_state->sam_ctx, mem_ctx,
d_state->domain_dn, &res, attrs,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
d_state->domain_sid, "%s", filter);
if (ldb_cnt == -1) {
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
if (ldb_cnt == 0 || r->in.max_entries == 0) {
return NT_STATUS_OK;
}
switch (r->in.level) {
case 1:
entriesGeneral = talloc_array(mem_ctx,
struct samr_DispEntryGeneral,
ldb_cnt);
break;
case 2:
case 3:
entriesFull = talloc_array(mem_ctx,
struct samr_DispEntryFull,
ldb_cnt);
break;
case 4:
case 5:
entriesAscii = talloc_array(mem_ctx,
struct samr_DispEntryAscii,
ldb_cnt);
break;
}
if ((entriesGeneral == NULL) && (entriesFull == NULL) &&
(entriesAscii == NULL))
return NT_STATUS_NO_MEMORY;
count = 0;
for (i=0; i<ldb_cnt; i++) {
struct dom_sid *objectsid;
objectsid = samdb_result_dom_sid(mem_ctx, res[i],
"objectSid");
if (objectsid == NULL)
continue;
switch(r->in.level) {
case 1:
entriesGeneral[count].idx = count + 1;
entriesGeneral[count].rid =
objectsid->sub_auths[objectsid->num_auths-1];
entriesGeneral[count].acct_flags =
samdb_result_acct_flags(res[i],
"userAccountControl");
entriesGeneral[count].account_name.string =
samdb_result_string(res[i],
"sAMAccountName", "");
entriesGeneral[count].full_name.string =
samdb_result_string(res[i], "displayName", "");
entriesGeneral[count].description.string =
samdb_result_string(res[i], "description", "");
break;
case 2:
case 3:
entriesFull[count].idx = count + 1;
entriesFull[count].rid =
objectsid->sub_auths[objectsid->num_auths-1];
entriesFull[count].acct_flags =
samdb_result_acct_flags(res[i],
"userAccountControl");
if (r->in.level == 3) {
/* We get a "7" here for groups */
entriesFull[count].acct_flags = 7;
}
entriesFull[count].account_name.string =
samdb_result_string(res[i], "sAMAccountName",
"");
entriesFull[count].description.string =
samdb_result_string(res[i], "description", "");
break;
case 4:
case 5:
entriesAscii[count].idx = count + 1;
entriesAscii[count].account_name.string =
samdb_result_string(res[i], "sAMAccountName",
"");
break;
}
count += 1;
}
r->out.total_size = count;
if (r->in.start_idx >= count) {
r->out.returned_size = 0;
switch(r->in.level) {
case 1:
r->out.info.info1.count = r->out.returned_size;
r->out.info.info1.entries = NULL;
break;
case 2:
r->out.info.info2.count = r->out.returned_size;
r->out.info.info2.entries = NULL;
break;
case 3:
r->out.info.info3.count = r->out.returned_size;
r->out.info.info3.entries = NULL;
break;
case 4:
r->out.info.info4.count = r->out.returned_size;
r->out.info.info4.entries = NULL;
break;
case 5:
r->out.info.info5.count = r->out.returned_size;
r->out.info.info5.entries = NULL;
break;
}
} else {
r->out.returned_size = MIN(count - r->in.start_idx,
r->in.max_entries);
switch(r->in.level) {
case 1:
r->out.info.info1.count = r->out.returned_size;
r->out.info.info1.entries =
&(entriesGeneral[r->in.start_idx]);
break;
case 2:
r->out.info.info2.count = r->out.returned_size;
r->out.info.info2.entries =
&(entriesFull[r->in.start_idx]);
break;
case 3:
r->out.info.info3.count = r->out.returned_size;
r->out.info.info3.entries =
&(entriesFull[r->in.start_idx]);
break;
case 4:
r->out.info.info4.count = r->out.returned_size;
r->out.info.info4.entries =
&(entriesAscii[r->in.start_idx]);
break;
case 5:
r->out.info.info5.count = r->out.returned_size;
r->out.info.info5.entries =
&(entriesAscii[r->in.start_idx]);
break;
}
}
return (r->out.returned_size < (count - r->in.start_idx)) ?
STATUS_MORE_ENTRIES : NT_STATUS_OK;
}
/*
samr_GetDisplayEnumerationIndex
*/
static NTSTATUS samr_GetDisplayEnumerationIndex(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetDisplayEnumerationIndex *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_TestPrivateFunctionsDomain
*/
static NTSTATUS samr_TestPrivateFunctionsDomain(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_TestPrivateFunctionsDomain *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_TestPrivateFunctionsUser
*/
static NTSTATUS samr_TestPrivateFunctionsUser(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_TestPrivateFunctionsUser *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_GetUserPwInfo
*/
static NTSTATUS samr_GetUserPwInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetUserPwInfo *r)
{
struct dcesrv_handle *h;
struct samr_account_state *a_state;
ZERO_STRUCT(r->out.info);
DCESRV_PULL_HANDLE(h, r->in.user_handle, SAMR_HANDLE_USER);
a_state = h->data;
r->out.info.min_password_length = samdb_search_uint(a_state->sam_ctx, mem_ctx, 0,
a_state->domain_state->domain_dn, "minPwdLength",
"dn=%s",
ldb_dn_linearize(mem_ctx, a_state->domain_state->domain_dn));
r->out.info.password_properties = samdb_search_uint(a_state->sam_ctx, mem_ctx, 0,
a_state->account_dn,
"pwdProperties", "dn=%s",
ldb_dn_linearize(mem_ctx, a_state->account_dn));
return NT_STATUS_OK;
}
/*
samr_RemoveMemberFromForeignDomain
*/
static NTSTATUS samr_RemoveMemberFromForeignDomain(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_RemoveMemberFromForeignDomain *r)
{
struct dcesrv_handle *h;
struct samr_domain_state *d_state;
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
const char *memberdn;
struct ldb_message **res;
const char * const attrs[3] = { "dn", "objectSid", NULL };
int i, count;
DCESRV_PULL_HANDLE(h, r->in.domain_handle, SAMR_HANDLE_DOMAIN);
d_state = h->data;
memberdn = samdb_search_string(d_state->sam_ctx, mem_ctx, NULL,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
"dn", "(objectSid=%s)",
ldap_encode_ndr_dom_sid(mem_ctx, r->in.sid));
if (memberdn == NULL)
return NT_STATUS_OBJECT_NAME_NOT_FOUND;
/* TODO: Does this call only remove alias members, or does it do this
* for domain groups as well? */
count = samdb_search_domain(d_state->sam_ctx, mem_ctx,
d_state->domain_dn, &res, attrs,
r7860: switch our ldb storage format to use a NDR encoded objectSid. This is quite a large change as we had lots of code that assumed that objectSid was a string in S- format. metze and simo tried to convince me to use NDR format months ago, but I didn't listen, so its fair that I have the pain of fixing all the code now :-) This builds on the ldb_register_samba_handlers() and ldif handlers code I did earlier this week. There are still three parts of this conversion I have not finished: - the ltdb index records need to use the string form of the objectSid (to keep the DNs sane). Until that it done I have disabled indexing on objectSid, which is a big performance hit, but allows us to pass all our tests while I rejig the indexing system to use a externally supplied conversion function - I haven't yet put in place the code that allows client to use the "S-xxx-yyy" form for objectSid in ldap search expressions. w2k3 supports this, presumably by looking for the "S-" prefix to determine what type of objectSid form is being used by the client. I have been working on ways to handle this, but am not happy with them yet so they aren't part of this patch - I need to change pidl to generate push functions that take a "const void *" instead of a "void*" for the data pointer. That will fix the couple of new warnings this code generates. Luckily it many places the conversion to NDR formatted records actually simplified the code, as it means we no longer need as many calls to dom_sid_parse_talloc(). In some places it got more complex, but not many. (This used to be commit d40bc2fa8ddd43560315688eebdbe98bdd02756c)
2005-06-24 04:18:20 +04:00
d_state->domain_sid,
"(&(member=%s)(objectClass=group)"
"(|(groupType=%s)(groupType=%s)))",
memberdn,
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_BUILTIN_LOCAL_GROUP),
ldb_hexstr(mem_ctx,
GTYPE_SECURITY_DOMAIN_LOCAL_GROUP));
if (count < 0)
return NT_STATUS_INTERNAL_DB_CORRUPTION;
for (i=0; i<count; i++) {
struct ldb_message *mod;
mod = ldb_msg_new(mem_ctx);
if (mod == NULL) {
return NT_STATUS_NO_MEMORY;
}
mod->dn = samdb_result_dn(mod, res[i], "dn", NULL);
if (mod->dn == NULL) {
talloc_free(mod);
continue;
}
if (samdb_msg_add_delval(d_state->sam_ctx, mem_ctx, mod,
"member", memberdn) != 0)
return NT_STATUS_NO_MEMORY;
if (samdb_modify(d_state->sam_ctx, mem_ctx, mod) != 0)
return NT_STATUS_UNSUCCESSFUL;
talloc_free(mod);
}
return NT_STATUS_OK;
}
/*
samr_QueryDomainInfo2
*/
static NTSTATUS samr_QueryDomainInfo2(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryDomainInfo2 *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_QueryUserInfo2
just an alias for samr_QueryUserInfo
*/
static NTSTATUS samr_QueryUserInfo2(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryUserInfo2 *r)
{
struct samr_QueryUserInfo r1;
NTSTATUS status;
ZERO_STRUCT(r1.out);
r1.in.user_handle = r->in.user_handle;
r1.in.level = r->in.level;
status = samr_QueryUserInfo(dce_call, mem_ctx, &r1);
r->out.info = r1.out.info;
return status;
}
/*
samr_QueryDisplayInfo2
*/
static NTSTATUS samr_QueryDisplayInfo2(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryDisplayInfo2 *r)
{
struct samr_QueryDisplayInfo q;
NTSTATUS result;
q.in.domain_handle = r->in.domain_handle;
q.in.level = r->in.level;
q.in.start_idx = r->in.start_idx;
q.in.max_entries = r->in.max_entries;
q.in.buf_size = r->in.buf_size;
ZERO_STRUCT(q.out);
result = samr_QueryDisplayInfo(dce_call, mem_ctx, &q);
r->out.total_size = q.out.total_size;
r->out.returned_size = q.out.returned_size;
r->out.info = q.out.info;
return result;
}
/*
samr_GetDisplayEnumerationIndex2
*/
static NTSTATUS samr_GetDisplayEnumerationIndex2(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetDisplayEnumerationIndex2 *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_QueryDisplayInfo3
*/
static NTSTATUS samr_QueryDisplayInfo3(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_QueryDisplayInfo3 *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_AddMultipleMembersToAlias
*/
static NTSTATUS samr_AddMultipleMembersToAlias(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_AddMultipleMembersToAlias *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_RemoveMultipleMembersFromAlias
*/
static NTSTATUS samr_RemoveMultipleMembersFromAlias(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_RemoveMultipleMembersFromAlias *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_GetDomPwInfo
this fetches the default password properties for a domain
note that w2k3 completely ignores the domain name in this call, and
always returns the information for the servers primary domain
*/
static NTSTATUS samr_GetDomPwInfo(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetDomPwInfo *r)
{
struct ldb_message **msgs;
int ret;
const char * const attrs[] = {"minPwdLength", "pwdProperties", NULL };
struct ldb_context *sam_ctx;
ZERO_STRUCT(r->out.info);
sam_ctx = samdb_connect(mem_ctx);
if (sam_ctx == NULL) {
return NT_STATUS_INVALID_SYSTEM_SERVICE;
}
/* The domain name in this call is ignored */
ret = gendb_search_dn(sam_ctx,
mem_ctx, samdb_base_dn(mem_ctx), &msgs, attrs);
if (ret <= 0) {
return NT_STATUS_NO_SUCH_DOMAIN;
}
if (ret > 1) {
talloc_free(msgs);
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
r->out.info.min_password_length = samdb_result_uint(msgs[0], "minPwdLength", 0);
r->out.info.password_properties = samdb_result_uint(msgs[0], "pwdProperties", 1);
talloc_free(msgs);
talloc_free(sam_ctx);
return NT_STATUS_OK;
}
/*
samr_Connect2
*/
static NTSTATUS samr_Connect2(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_Connect2 *r)
{
struct samr_Connect c;
c.in.system_name = NULL;
c.in.access_mask = r->in.access_mask;
c.out.connect_handle = r->out.connect_handle;
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return samr_Connect(dce_call, mem_ctx, &c);
}
/*
samr_SetUserInfo2
just an alias for samr_SetUserInfo
*/
static NTSTATUS samr_SetUserInfo2(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetUserInfo2 *r)
{
struct samr_SetUserInfo r2;
r2.in.user_handle = r->in.user_handle;
r2.in.level = r->in.level;
r2.in.info = r->in.info;
return samr_SetUserInfo(dce_call, mem_ctx, &r2);
}
/*
samr_SetBootKeyInformation
*/
static NTSTATUS samr_SetBootKeyInformation(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetBootKeyInformation *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_GetBootKeyInformation
*/
static NTSTATUS samr_GetBootKeyInformation(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_GetBootKeyInformation *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_Connect3
*/
static NTSTATUS samr_Connect3(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_Connect3 *r)
{
struct samr_Connect c;
c.in.system_name = NULL;
c.in.access_mask = r->in.access_mask;
c.out.connect_handle = r->out.connect_handle;
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return samr_Connect(dce_call, mem_ctx, &c);
}
/*
samr_Connect4
*/
static NTSTATUS samr_Connect4(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_Connect4 *r)
{
struct samr_Connect c;
c.in.system_name = NULL;
c.in.access_mask = r->in.access_mask;
c.out.connect_handle = r->out.connect_handle;
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
return samr_Connect(dce_call, mem_ctx, &c);
}
/*
samr_Connect5
*/
static NTSTATUS samr_Connect5(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_Connect5 *r)
{
struct samr_Connect c;
NTSTATUS status;
c.in.system_name = NULL;
c.in.access_mask = r->in.access_mask;
c.out.connect_handle = r->out.connect_handle;
status = samr_Connect(dce_call, mem_ctx, &c);
r->out.info->info1.unknown1 = 3;
r->out.info->info1.unknown2 = 0;
r->out.level = r->in.level;
return status;
}
/*
samr_RidToSid
*/
static NTSTATUS samr_RidToSid(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_RidToSid *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_SetDsrmPassword
*/
static NTSTATUS samr_SetDsrmPassword(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_SetDsrmPassword *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/*
samr_ValidatePassword
*/
static NTSTATUS samr_ValidatePassword(struct dcesrv_call_state *dce_call, TALLOC_CTX *mem_ctx,
struct samr_ValidatePassword *r)
{
r464: a big improvement to the API for writing server-side RPC servers. Previously the server pipe code needed to return the RPC level status (nearly always "OK") and separately set the function call return using r->out.result. All the programmers writing servers (metze, jelmer and me) were often getting this wrong, by doing things like "return NT_STATUS_NO_MEMORY" which was really quite meaningless as there is no code like that at the dcerpc level. I have now modified pidl to generate the necessary boilerplate so that just returning the status you want from the function will work. So for a NTSTATUS function you return NT_STATUS_XXX and from a WERROR function you return WERR_XXX. If you really want to generate a DCERPC level fault rather than just a return value in your function then you should use the DCESRV_FAULT() macro which will correctly generate a fault for you. As a side effect, this also adds automatic type checking of all of our server side rpc functions, which was impossible with the old API. When I changed the API I found and fixed quite a few functions with the wrong type information, so this is definately useful. I have also changed the server side template generation to generate a DCERPC "operation range error" by default when you have not yet filled in a server side function. This allows us to correctly implement functions in any order in our rpc pipe servers and give the client the right information about the fault. (This used to be commit a4df5c7cf88891a78d82c8d6d7f058d8485e73f0)
2004-05-04 10:07:52 +04:00
DCESRV_FAULT(DCERPC_FAULT_OP_RNG_ERROR);
}
/* include the generated boilerplate */
#include "librpc/gen_ndr/ndr_samr_s.c"