1
0
mirror of https://github.com/samba-team/samba.git synced 2025-01-08 21:18:16 +03:00
samba-mirror/source3/smbd/smbXsrv_open.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1507 lines
38 KiB
C
Raw Normal View History

/*
Unix SMB/CIFS implementation.
Copyright (C) Stefan Metzmacher 2012
Copyright (C) Michael Adam 2012
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 3 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, see <http://www.gnu.org/licenses/>.
*/
#include "smbXsrv_open.h"
#include "includes.h"
#include "system/filesys.h"
#include "lib/util/server_id.h"
#include "smbd/smbd.h"
#include "smbd/globals.h"
#include "dbwrap/dbwrap.h"
#include "dbwrap/dbwrap_rbt.h"
#include "dbwrap/dbwrap_open.h"
#include "../libcli/security/security.h"
#include "messages.h"
#include "lib/util/util_tdb.h"
#include "librpc/gen_ndr/ndr_smbXsrv.h"
#include "serverid.h"
#include "source3/include/util_tdb.h"
#include "lib/util/idtree_random.h"
#include "lib/util/time_basic.h"
struct smbXsrv_open_table {
struct {
struct idr_context *idr;
struct db_context *replay_cache_db_ctx;
uint32_t lowest_id;
uint32_t highest_id;
uint32_t max_opens;
uint32_t num_opens;
} local;
struct {
struct db_context *db_ctx;
} global;
};
static struct db_context *smbXsrv_open_global_db_ctx = NULL;
NTSTATUS smbXsrv_open_global_init(void)
{
char *global_path = NULL;
struct db_context *db_ctx = NULL;
if (smbXsrv_open_global_db_ctx != NULL) {
return NT_STATUS_OK;
}
global_path = lock_path(talloc_tos(), "smbXsrv_open_global.tdb");
if (global_path == NULL) {
return NT_STATUS_NO_MEMORY;
}
db_ctx = db_open(NULL, global_path,
SMBD_VOLATILE_TDB_HASH_SIZE,
SMBD_VOLATILE_TDB_FLAGS,
O_RDWR | O_CREAT, 0600,
DBWRAP_LOCK_ORDER_1,
DBWRAP_FLAG_NONE);
TALLOC_FREE(global_path);
if (db_ctx == NULL) {
NTSTATUS status = map_nt_error_from_unix_common(errno);
return status;
}
smbXsrv_open_global_db_ctx = db_ctx;
return NT_STATUS_OK;
}
/*
* NOTE:
* We need to store the keys in big endian so that dbwrap_rbt's memcmp
* has the same result as integer comparison between the uint32_t
* values.
*
* TODO: implement string based key
*/
struct smbXsrv_open_global_key_buf { uint8_t buf[sizeof(uint32_t)]; };
static TDB_DATA smbXsrv_open_global_id_to_key(
uint32_t id, struct smbXsrv_open_global_key_buf *key_buf)
{
RSIVAL(key_buf->buf, 0, id);
return (TDB_DATA) {
.dptr = key_buf->buf,
.dsize = sizeof(key_buf->buf),
};
}
static NTSTATUS smbXsrv_open_table_init(struct smbXsrv_connection *conn,
uint32_t lowest_id,
uint32_t highest_id,
uint32_t max_opens)
{
struct smbXsrv_client *client = conn->client;
struct smbXsrv_open_table *table;
NTSTATUS status;
uint64_t max_range;
if (lowest_id > highest_id) {
return NT_STATUS_INTERNAL_ERROR;
}
max_range = highest_id;
max_range -= lowest_id;
max_range += 1;
if (max_opens > max_range) {
return NT_STATUS_INTERNAL_ERROR;
}
table = talloc_zero(client, struct smbXsrv_open_table);
if (table == NULL) {
return NT_STATUS_NO_MEMORY;
}
table->local.idr = idr_init(table);
if (table->local.idr == NULL) {
TALLOC_FREE(table);
return NT_STATUS_NO_MEMORY;
}
table->local.replay_cache_db_ctx = db_open_rbt(table);
if (table->local.replay_cache_db_ctx == NULL) {
TALLOC_FREE(table);
return NT_STATUS_NO_MEMORY;
}
table->local.lowest_id = lowest_id;
table->local.highest_id = highest_id;
table->local.max_opens = max_opens;
status = smbXsrv_open_global_init();
if (!NT_STATUS_IS_OK(status)) {
TALLOC_FREE(table);
return status;
}
table->global.db_ctx = smbXsrv_open_global_db_ctx;
client->open_table = table;
return NT_STATUS_OK;
}
static NTSTATUS smbXsrv_open_local_lookup(struct smbXsrv_open_table *table,
uint32_t open_local_id,
uint32_t open_global_id,
NTTIME now,
struct smbXsrv_open **_open)
{
struct smbXsrv_open *op = NULL;
*_open = NULL;
if (open_local_id == 0) {
return NT_STATUS_FILE_CLOSED;
}
if (table == NULL) {
/* this might happen before the end of negprot */
return NT_STATUS_FILE_CLOSED;
}
if (table->local.idr == NULL) {
return NT_STATUS_INTERNAL_ERROR;
}
op = idr_find(table->local.idr, open_local_id);
if (op == NULL) {
return NT_STATUS_FILE_CLOSED;
}
if (open_global_id == 0) {
/* make the global check a no-op for SMB1 */
open_global_id = op->global->open_global_id;
}
if (op->global->open_global_id != open_global_id) {
return NT_STATUS_FILE_CLOSED;
}
if (now != 0) {
op->idle_time = now;
}
*_open = op;
return NT_STATUS_OK;
}
static NTSTATUS smbXsrv_open_global_parse_record(
TALLOC_CTX *mem_ctx,
TDB_DATA key,
TDB_DATA val,
struct smbXsrv_open_global0 **global)
{
DATA_BLOB blob = data_blob_const(val.dptr, val.dsize);
struct smbXsrv_open_globalB global_blob;
enum ndr_err_code ndr_err;
NTSTATUS status;
TALLOC_CTX *frame = talloc_stackframe();
ndr_err = ndr_pull_struct_blob(&blob, frame, &global_blob,
(ndr_pull_flags_fn_t)ndr_pull_smbXsrv_open_globalB);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DEBUG(1,("Invalid record in smbXsrv_open_global.tdb:"
"key '%s' ndr_pull_struct_blob - %s\n",
tdb_data_dbg(key),
ndr_errstr(ndr_err)));
status = ndr_map_error2ntstatus(ndr_err);
goto done;
}
DBG_DEBUG("\n");
if (CHECK_DEBUGLVL(10)) {
NDR_PRINT_DEBUG(smbXsrv_open_globalB, &global_blob);
}
if (global_blob.version != SMBXSRV_VERSION_0) {
status = NT_STATUS_INTERNAL_DB_CORRUPTION;
DEBUG(1,("Invalid record in smbXsrv_open_global.tdb:"
"key '%s' unsupported version - %d - %s\n",
tdb_data_dbg(key),
(int)global_blob.version,
nt_errstr(status)));
goto done;
}
if (global_blob.info.info0 == NULL) {
status = NT_STATUS_INTERNAL_DB_CORRUPTION;
DEBUG(1,("Invalid record in smbXsrv_open_global.tdb:"
"key '%s' info0 NULL pointer - %s\n",
tdb_data_dbg(key),
nt_errstr(status)));
goto done;
}
*global = talloc_move(mem_ctx, &global_blob.info.info0);
status = NT_STATUS_OK;
done:
talloc_free(frame);
return status;
}
static NTSTATUS smbXsrv_open_global_verify_record(
TDB_DATA key,
TDB_DATA val,
TALLOC_CTX *mem_ctx,
struct smbXsrv_open_global0 **_global0)
{
struct smbXsrv_open_global0 *global0 = NULL;
struct server_id_buf buf;
NTSTATUS status;
if (val.dsize == 0) {
return NT_STATUS_NOT_FOUND;
}
status = smbXsrv_open_global_parse_record(mem_ctx, key, val, &global0);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("smbXsrv_open_global_parse_record for %s failed: "
"%s\n",
tdb_data_dbg(key),
nt_errstr(status));
return status;
}
*_global0 = global0;
if (server_id_is_disconnected(&global0->server_id)) {
return NT_STATUS_OK;
}
if (serverid_exists(&global0->server_id)) {
return NT_STATUS_OK;
}
DBG_WARNING("smbd %s did not clean up record %s\n",
server_id_str_buf(global0->server_id, &buf),
tdb_data_dbg(key));
return NT_STATUS_FATAL_APP_EXIT;
}
static NTSTATUS smbXsrv_open_global_store(
struct db_record *rec,
TDB_DATA key,
TDB_DATA oldval,
struct smbXsrv_open_global0 *global)
{
struct smbXsrv_open_globalB global_blob;
DATA_BLOB blob = data_blob_null;
TDB_DATA val = { .dptr = NULL, };
NTSTATUS status;
enum ndr_err_code ndr_err;
/*
* TODO: if we use other versions than '0'
* we would add glue code here, that would be able to
* store the information in the old format.
*/
global_blob = (struct smbXsrv_open_globalB) {
.version = smbXsrv_version_global_current(),
};
if (oldval.dsize >= 8) {
global_blob.seqnum = IVAL(oldval.dptr, 4);
}
global_blob.seqnum += 1;
global_blob.info.info0 = global;
ndr_err = ndr_push_struct_blob(&blob, talloc_tos(), &global_blob,
(ndr_push_flags_fn_t)ndr_push_smbXsrv_open_globalB);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
DBG_WARNING("key '%s' ndr_push - %s\n",
tdb_data_dbg(key),
ndr_map_error2string(ndr_err));
return ndr_map_error2ntstatus(ndr_err);
}
val = make_tdb_data(blob.data, blob.length);
status = dbwrap_record_store(rec, val, TDB_REPLACE);
TALLOC_FREE(blob.data);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("key '%s' store - %s\n",
tdb_data_dbg(key),
nt_errstr(status));
return status;
}
if (CHECK_DEBUGLVL(10)) {
DBG_DEBUG("key '%s' stored\n", tdb_data_dbg(key));
NDR_PRINT_DEBUG(smbXsrv_open_globalB, &global_blob);
}
return NT_STATUS_OK;
}
struct smbXsrv_open_global_allocate_state {
uint32_t id;
struct smbXsrv_open_global0 *global;
NTSTATUS status;
};
static void smbXsrv_open_global_allocate_fn(
struct db_record *rec, TDB_DATA oldval, void *private_data)
{
struct smbXsrv_open_global_allocate_state *state = private_data;
struct smbXsrv_open_global0 *global = state->global;
struct smbXsrv_open_global0 *tmp_global0 = NULL;
TDB_DATA key = dbwrap_record_get_key(rec);
state->status = smbXsrv_open_global_verify_record(
key, oldval, talloc_tos(), &tmp_global0);
if (NT_STATUS_IS_OK(state->status)) {
/*
* Found an existing record
*/
TALLOC_FREE(tmp_global0);
state->status = NT_STATUS_RETRY;
return;
}
if (NT_STATUS_EQUAL(state->status, NT_STATUS_NOT_FOUND)) {
/*
* Found an empty slot
*/
global->open_global_id = state->id;
global->open_persistent_id = state->id;
state->status = smbXsrv_open_global_store(
rec, key, (TDB_DATA) { .dsize = 0, }, state->global);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_WARNING("smbXsrv_open_global_store() for "
"id %"PRIu32" failed: %s\n",
state->id,
nt_errstr(state->status));
}
return;
}
if (NT_STATUS_EQUAL(state->status, NT_STATUS_FATAL_APP_EXIT)) {
NTSTATUS status;
TALLOC_FREE(tmp_global0);
/*
* smbd crashed
*/
status = dbwrap_record_delete(rec);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("dbwrap_record_delete() failed "
"for record %"PRIu32": %s\n",
state->id,
nt_errstr(status));
state->status = NT_STATUS_INTERNAL_DB_CORRUPTION;
return;
}
return;
}
}
static NTSTATUS smbXsrv_open_global_allocate(
struct db_context *db, struct smbXsrv_open_global0 *global)
{
struct smbXsrv_open_global_allocate_state state = {
.global = global,
};
uint32_t i;
uint32_t last_free = 0;
const uint32_t min_tries = 3;
/*
* Here we just randomly try the whole 32-bit space
*
* We use just 32-bit, because we want to reuse the
* ID for SRVSVC.
*/
for (i = 0; i < UINT32_MAX; i++) {
struct smbXsrv_open_global_key_buf key_buf;
TDB_DATA key;
NTSTATUS status;
if (i >= min_tries && last_free != 0) {
state.id = last_free;
} else {
generate_nonce_buffer(
(uint8_t *)&state.id, sizeof(state.id));
state.id = MAX(state.id, 1);
state.id = MIN(state.id, UINT32_MAX-1);
}
key = smbXsrv_open_global_id_to_key(state.id, &key_buf);
status = dbwrap_do_locked(
db, key, smbXsrv_open_global_allocate_fn, &state);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("dbwrap_do_locked() failed: %s\n",
nt_errstr(status));
return NT_STATUS_INTERNAL_DB_ERROR;
}
if (NT_STATUS_IS_OK(state.status)) {
/*
* Found an empty slot, done.
*/
DBG_DEBUG("Found slot %"PRIu32"\n", state.id);
return NT_STATUS_OK;
}
if (NT_STATUS_EQUAL(state.status, NT_STATUS_FATAL_APP_EXIT)) {
if ((i < min_tries) && (last_free == 0)) {
/*
* Remember "id" as free but also try
* others to not recycle ids too
* quickly.
*/
last_free = state.id;
}
continue;
}
if (NT_STATUS_EQUAL(state.status, NT_STATUS_RETRY)) {
/*
* Normal collision, try next
*/
DBG_DEBUG("Found record for id %"PRIu32"\n",
state.id);
continue;
}
DBG_WARNING("smbXsrv_open_global_allocate_fn() failed: %s\n",
nt_errstr(state.status));
return state.status;
}
/* should not be reached */
return NT_STATUS_INTERNAL_ERROR;
}
static int smbXsrv_open_destructor(struct smbXsrv_open *op)
{
NTSTATUS status;
status = smbXsrv_open_close(op, 0);
if (!NT_STATUS_IS_OK(status)) {
DEBUG(0, ("smbXsrv_open_destructor: "
"smbXsrv_open_close() failed - %s\n",
nt_errstr(status)));
}
TALLOC_FREE(op->global);
return 0;
}
NTSTATUS smbXsrv_open_create(struct smbXsrv_connection *conn,
struct auth_session_info *session_info,
NTTIME now,
struct smbXsrv_open **_open)
{
struct smbXsrv_open_table *table = conn->client->open_table;
struct smbXsrv_open *op = NULL;
struct smbXsrv_open_global0 *global = NULL;
NTSTATUS status;
struct dom_sid *current_sid = NULL;
struct security_token *current_token = NULL;
int local_id;
if (session_info == NULL) {
return NT_STATUS_INVALID_HANDLE;
}
current_token = session_info->security_token;
if ((current_token == NULL) ||
(current_token->num_sids <= PRIMARY_USER_SID_INDEX)) {
return NT_STATUS_INVALID_HANDLE;
}
current_sid = &current_token->sids[PRIMARY_USER_SID_INDEX];
if (table->local.num_opens >= table->local.max_opens) {
return NT_STATUS_INSUFFICIENT_RESOURCES;
}
op = talloc_zero(table, struct smbXsrv_open);
if (op == NULL) {
return NT_STATUS_NO_MEMORY;
}
op->table = table;
op->status = NT_STATUS_OK; /* TODO: start with INTERNAL_ERROR */
op->idle_time = now;
global = talloc_zero(op, struct smbXsrv_open_global0);
if (global == NULL) {
TALLOC_FREE(op);
return NT_STATUS_NO_MEMORY;
}
op->global = global;
/*
* We mark every slot as invalid using 0xFF.
* Valid values are masked with 0xF.
*/
memset(global->lock_sequence_array, 0xFF,
sizeof(global->lock_sequence_array));
local_id = idr_get_new_random(
table->local.idr,
op,
table->local.lowest_id,
table->local.highest_id);
if (local_id == -1) {
TALLOC_FREE(op);
return NT_STATUS_INSUFFICIENT_RESOURCES;
}
op->local_id = local_id;
global->open_volatile_id = op->local_id;
global->server_id = messaging_server_id(conn->client->msg_ctx);
global->open_time = now;
global->open_owner = *current_sid;
if (conn->protocol >= PROTOCOL_SMB2_10) {
global->client_guid = conn->smb2.client.guid;
}
status = smbXsrv_open_global_allocate(table->global.db_ctx,
global);
if (!NT_STATUS_IS_OK(status)) {
int ret = idr_remove(table->local.idr, local_id);
SMB_ASSERT(ret == 0);
DBG_WARNING("smbXsrv_open_global_allocate() failed: %s\n",
nt_errstr(status));
TALLOC_FREE(op);
return status;
}
table->local.num_opens += 1;
talloc_set_destructor(op, smbXsrv_open_destructor);
if (CHECK_DEBUGLVL(10)) {
struct smbXsrv_openB open_blob = {
.version = SMBXSRV_VERSION_0,
.info.info0 = op,
};
DEBUG(10,("smbXsrv_open_create: global_id (0x%08x) stored\n",
op->global->open_global_id));
NDR_PRINT_DEBUG(smbXsrv_openB, &open_blob);
}
*_open = op;
return NT_STATUS_OK;
}
static NTSTATUS smbXsrv_open_set_replay_cache(struct smbXsrv_open *op)
{
struct GUID *create_guid;
struct GUID_txt_buf buf;
char *guid_string;
struct db_context *db = op->table->local.replay_cache_db_ctx;
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
struct smbXsrv_open_replay_cache rc = {
.idle_time = op->idle_time,
.local_id = op->local_id,
};
Fix gcc11 compiler issue "-Werror=maybe-uninitialized" BUG: https://bugzilla.samba.org/show_bug.cgi?id=14699 ../../source4/dsdb/common/util_links.c: In function ‘ndr_guid_compare’: ../../source4/dsdb/common/util_links.c:38:29: error: ‘v1_data’ may be used uninitialized [-Werror=maybe-uninitialized] 38 | struct ldb_val v1 = data_blob_const(v1_data, sizeof(v1_data)); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source4/../lib/util/samba_util.h:48, from ../../source4/include/includes.h:62, from ../../source4/dsdb/common/util_links.c:22: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source4/dsdb/common/util_links.c:37:17: note: ‘v1_data’ declared here 37 | uint8_t v1_data[16]; | ^~~~~~~ cc1: all warnings being treated as errors [1729/3991] Compiling source3/smbd/smbXsrv_open.c ../../libcli/auth/smbencrypt.c: In function ‘decode_wkssvc_join_password_buffer’: ../../libcli/auth/smbencrypt.c:1045:32: error: ‘_confounder’ may be used uninitialized [-Werror=maybe-uninitialized] 1045 | DATA_BLOB confounder = data_blob_const(_confounder, 8); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source4/../lib/util/samba_util.h:48, from ../../source4/include/includes.h:62, from ../../libcli/auth/smbencrypt.c:24: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../libcli/auth/smbencrypt.c:1044:17: note: ‘_confounder’ declared here 1044 | uint8_t _confounder[8]; | ^~~~~~~~~~~ cc1: all warnings being treated as errors [2624/3991] Compiling source4/torture/rpc/samr.c ../../source3/rpc_client/cli_samr.c: In function ‘dcerpc_samr_chgpasswd_user2’: ../../source3/rpc_client/cli_samr.c:158:33: error: ‘old_nt_hash’ may be used uninitialized [-Werror=maybe-uninitialized] 158 | DATA_BLOB session_key = data_blob_const(old_nt_hash, 16); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source3/../lib/util/samba_util.h:48, from ../../source3/include/includes.h:256, from ../../source3/rpc_client/cli_samr.c:24: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source3/rpc_client/cli_samr.c:152:17: note: ‘old_nt_hash’ declared here 152 | uint8_t old_nt_hash[16]; | ^~~~~~~~~~~ ../../source3/rpc_client/cli_samr.c: In function ‘dcerpc_samr_chgpasswd_user3’: ../../source3/rpc_client/cli_samr.c:365:33: error: ‘old_nt_hash’ may be used uninitialized [-Werror=maybe-uninitialized] 365 | DATA_BLOB session_key = data_blob_const(old_nt_hash, 16); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source3/../lib/util/samba_util.h:48, from ../../source3/include/includes.h:256, from ../../source3/rpc_client/cli_samr.c:24: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source3/rpc_client/cli_samr.c:358:17: note: ‘old_nt_hash’ declared here 358 | uint8_t old_nt_hash[16]; | ^~~~~~~~~~~ cc1: all warnings being treated as errors [3399/3991] Compiling source3/rpcclient/cmd_spotlight.c ../../source3/smbd/smbXsrv_open.c: In function ‘smbXsrv_open_set_replay_cache’: ../../source3/smbd/smbXsrv_open.c:936:26: error: ‘data’ may be used uninitialized [-Werror=maybe-uninitialized] 936 | DATA_BLOB blob = data_blob_const(data, ARRAY_SIZE(data)); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source3/../lib/util/samba_util.h:48, from ../../source3/include/includes.h:256, from ../../source3/smbd/smbXsrv_open.c:21: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source3/smbd/smbXsrv_open.c:935:17: note: ‘data’ declared here 935 | uint8_t data[SMBXSRV_OPEN_REPLAY_CACHE_FIXED_SIZE]; | ^~~~ cc1: all warnings being treated as errors ../../source3/rpcclient/cmd_spotlight.c: In function ‘cmd_mdssvc_fetch_properties’: ../../source3/rpcclient/cmd_spotlight.c:60:18: error: ‘share_path’ may be used uninitialized [-Werror=maybe-uninitialized] 60 | status = dcerpc_mdssvc_open(b, mem_ctx, | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 61 | &device_id, | ~~~~~~~~~~~ 62 | &unkn1, | ~~~~~~~ 63 | &unkn2, | ~~~~~~~ 64 | argv[2], | ~~~~~~~~ 65 | argv[1], | ~~~~~~~~ 66 | share_path, | ~~~~~~~~~~~ 67 | &share_handle); | ~~~~~~~~~~~~~~ In file included from ../../source3/rpcclient/cmd_spotlight.c:24: source3/../librpc/gen_ndr/ndr_mdssvc_c.h:26:10: note: by argument 8 of type ‘const char *’ to ‘dcerpc_mdssvc_open’ declared here 26 | NTSTATUS dcerpc_mdssvc_open(struct dcerpc_binding_handle *h, | ^~~~~~~~~~~~~~~~~~ ../../source3/rpcclient/cmd_spotlight.c:40:14: note: ‘share_path’ declared here 40 | char share_path[1025]; | ^~~~~~~~~~ cc1: all warnings being treated as errors ../../source4/torture/rpc/samr.c: In function ‘test_ChangePasswordUser2’: ../../source4/torture/rpc/samr.c:2266:19: error: ‘old_nt_hash’ may be used uninitialized [-Werror=maybe-uninitialized] 2266 | = data_blob_const(old_nt_hash, sizeof(old_nt_hash)); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source4/../lib/util/samba_util.h:48, from ../../source4/include/includes.h:62, from ../../source4/torture/rpc/samr.c:24: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source4/torture/rpc/samr.c:2263:17: note: ‘old_nt_hash’ declared here 2263 | uint8_t old_nt_hash[16], new_nt_hash[16]; | ^~~~~~~~~~~ ../../source4/torture/rpc/samr.c: In function ‘test_ChangePasswordUser2_ntstatus’: ../../source4/torture/rpc/samr.c:2371:19: error: ‘old_nt_hash’ may be used uninitialized [-Werror=maybe-uninitialized] 2371 | = data_blob_const(old_nt_hash, sizeof(old_nt_hash)); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source4/../lib/util/samba_util.h:48, from ../../source4/include/includes.h:62, from ../../source4/torture/rpc/samr.c:24: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source4/torture/rpc/samr.c:2368:17: note: ‘old_nt_hash’ declared here 2368 | uint8_t old_nt_hash[16], new_nt_hash[16]; | ^~~~~~~~~~~ ../../source4/torture/rpc/samr.c: In function ‘test_ChangePasswordUser3’: ../../source4/torture/rpc/samr.c:2478:38: error: ‘old_nt_hash’ may be used uninitialized [-Werror=maybe-uninitialized] 2478 | DATA_BLOB old_nt_hash_blob = data_blob_const(old_nt_hash, 16); | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../../source4/../lib/util/samba_util.h:48, from ../../source4/include/includes.h:62, from ../../source4/torture/rpc/samr.c:24: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source4/torture/rpc/samr.c:2473:17: note: ‘old_nt_hash’ declared here 2473 | uint8_t old_nt_hash[16], new_nt_hash[16]; | ^~~~~~~~~~~ ../../source4/torture/rpc/samr.c: In function ‘test_ChangePasswordRandomBytes’: ../../source4/torture/rpc/samr.c:2794:19: error: ‘old_nt_hash’ may be used uninitialized [-Werror=maybe-uninitialized] 2794 | = data_blob_const(old_nt_hash, | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2795 | sizeof(old_nt_hash)); | ~~~~~~~~~~~~~~~~~~~~ In file included from ../../source4/../lib/util/samba_util.h:48, from ../../source4/include/includes.h:62, from ../../source4/torture/rpc/samr.c:24: ../../lib/util/data_blob.h:116:20: note: by argument 1 of type ‘const void *’ to ‘data_blob_const’ declared here 116 | _PUBLIC_ DATA_BLOB data_blob_const(const void *p, size_t length); | ^~~~~~~~~~~~~~~ ../../source4/torture/rpc/samr.c:2792:17: note: ‘old_nt_hash’ declared here 2792 | uint8_t old_nt_hash[16], new_nt_hash[16]; | ^~~~~~~~~~~ cc1: all warnings being treated as errors Guenther Signed-off-by: Guenther Deschner <gd@samba.org> Reviewed-by: Andreas Schneider <asn@samba.org>
2021-05-03 22:27:43 +03:00
uint8_t data[SMBXSRV_OPEN_REPLAY_CACHE_FIXED_SIZE] = { 0 };
DATA_BLOB blob = { .data = data, .length = sizeof(data), };
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
enum ndr_err_code ndr_err;
NTSTATUS status;
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
TDB_DATA val;
if (!(op->flags & SMBXSRV_OPEN_NEED_REPLAY_CACHE)) {
return NT_STATUS_OK;
}
if (op->flags & SMBXSRV_OPEN_HAVE_REPLAY_CACHE) {
return NT_STATUS_OK;
}
create_guid = &op->global->create_guid;
guid_string = GUID_buf_string(create_guid, &buf);
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
ndr_err = ndr_push_struct_into_fixed_blob(&blob, &rc,
(ndr_push_flags_fn_t)ndr_push_smbXsrv_open_replay_cache);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
status = ndr_map_error2ntstatus(ndr_err);
return status;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
val = make_tdb_data(blob.data, blob.length);
status = dbwrap_store_bystring(db, guid_string, val, TDB_REPLACE);
if (NT_STATUS_IS_OK(status)) {
op->flags |= SMBXSRV_OPEN_HAVE_REPLAY_CACHE;
op->flags &= ~SMBXSRV_OPEN_NEED_REPLAY_CACHE;
}
return status;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
NTSTATUS smbXsrv_open_purge_replay_cache(struct smbXsrv_client *client,
const struct GUID *create_guid)
{
struct GUID_txt_buf buf;
NTSTATUS status;
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
if (client->open_table == NULL) {
return NT_STATUS_OK;
}
status = dbwrap_purge_bystring(
client->open_table->local.replay_cache_db_ctx,
GUID_buf_string(create_guid, &buf));
return status;
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
}
static NTSTATUS smbXsrv_open_clear_replay_cache(struct smbXsrv_open *op)
{
struct GUID *create_guid;
struct GUID_txt_buf buf;
struct db_context *db;
NTSTATUS status;
if (op->table == NULL) {
return NT_STATUS_OK;
}
db = op->table->local.replay_cache_db_ctx;
if (!(op->flags & SMBXSRV_OPEN_HAVE_REPLAY_CACHE)) {
return NT_STATUS_OK;
}
create_guid = &op->global->create_guid;
if (GUID_all_zero(create_guid)) {
return NT_STATUS_OK;
}
status = dbwrap_purge_bystring(db, GUID_buf_string(create_guid, &buf));
if (NT_STATUS_IS_OK(status)) {
op->flags &= ~SMBXSRV_OPEN_HAVE_REPLAY_CACHE;
}
return status;
}
struct smbXsrv_open_update_state {
struct smbXsrv_open_global0 *global;
NTSTATUS status;
};
static void smbXsrv_open_update_fn(
struct db_record *rec, TDB_DATA oldval, void *private_data)
{
struct smbXsrv_open_update_state *state = private_data;
TDB_DATA key = dbwrap_record_get_key(rec);
state->status = smbXsrv_open_global_store(
rec, key, oldval, state->global);
}
NTSTATUS smbXsrv_open_update(struct smbXsrv_open *op)
{
struct smbXsrv_open_update_state state = { .global = op->global, };
struct smbXsrv_open_table *table = op->table;
struct smbXsrv_open_global_key_buf key_buf;
TDB_DATA key = smbXsrv_open_global_id_to_key(
op->global->open_global_id, &key_buf);
NTSTATUS status;
status = dbwrap_do_locked(
table->global.db_ctx, key, smbXsrv_open_update_fn, &state);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("global_id (0x%08x) dbwrap_do_locked failed: %s\n",
op->global->open_global_id,
nt_errstr(status));
return NT_STATUS_INTERNAL_DB_ERROR;
}
if (!NT_STATUS_IS_OK(state.status)) {
DBG_WARNING("global_id (0x%08x) smbXsrv_open_global_store "
"failed: %s\n",
op->global->open_global_id,
nt_errstr(state.status));
return state.status;
}
status = smbXsrv_open_set_replay_cache(op);
if (!NT_STATUS_IS_OK(status)) {
DBG_ERR("smbXsrv_open_set_replay_cache failed: %s\n",
nt_errstr(status));
return status;
}
if (CHECK_DEBUGLVL(10)) {
struct smbXsrv_openB open_blob = {
.version = SMBXSRV_VERSION_0,
.info.info0 = op,
};
DEBUG(10,("smbXsrv_open_update: global_id (0x%08x) stored\n",
op->global->open_global_id));
NDR_PRINT_DEBUG(smbXsrv_openB, &open_blob);
}
return NT_STATUS_OK;
}
struct smbXsrv_open_close_state {
struct smbXsrv_open *op;
NTSTATUS status;
};
static void smbXsrv_open_close_fn(
struct db_record *rec, TDB_DATA oldval, void *private_data)
{
struct smbXsrv_open_close_state *state = private_data;
struct smbXsrv_open_global0 *global = state->op->global;
TDB_DATA key = dbwrap_record_get_key(rec);
if (global->durable) {
/*
* Durable open -- we need to update the global part
* instead of deleting it
*/
state->status = smbXsrv_open_global_store(
rec, key, oldval, global);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_WARNING("failed to store global key '%s': %s\n",
tdb_data_dbg(key),
nt_errstr(state->status));
return;
}
if (CHECK_DEBUGLVL(10)) {
struct smbXsrv_openB open_blob = {
.version = SMBXSRV_VERSION_0,
.info.info0 = state->op,
};
DBG_DEBUG("(0x%08x) stored disconnect\n",
global->open_global_id);
NDR_PRINT_DEBUG(smbXsrv_openB, &open_blob);
}
return;
}
state->status = dbwrap_record_delete(rec);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_WARNING("failed to delete global key '%s': %s\n",
tdb_data_dbg(key),
nt_errstr(state->status));
}
}
NTSTATUS smbXsrv_open_close(struct smbXsrv_open *op, NTTIME now)
{
struct smbXsrv_open_close_state state = { .op = op, };
struct smbXsrv_open_global0 *global = op->global;
struct smbXsrv_open_table *table;
NTSTATUS status;
NTSTATUS error = NT_STATUS_OK;
struct smbXsrv_open_global_key_buf key_buf;
TDB_DATA key = smbXsrv_open_global_id_to_key(
global->open_global_id, &key_buf);
int ret;
error = smbXsrv_open_clear_replay_cache(op);
if (!NT_STATUS_IS_OK(error)) {
DBG_ERR("smbXsrv_open_clear_replay_cache failed: %s\n",
nt_errstr(error));
}
if (op->table == NULL) {
return error;
}
table = op->table;
op->table = NULL;
op->status = NT_STATUS_FILE_CLOSED;
global->disconnect_time = now;
server_id_set_disconnected(&global->server_id);
status = dbwrap_do_locked(
table->global.db_ctx, key, smbXsrv_open_close_fn, &state);
if (!NT_STATUS_IS_OK(status)) {
DBG_WARNING("dbwrap_do_locked() for %s failed: %s\n",
tdb_data_dbg(key),
nt_errstr(status));
error = status;
} else if (!NT_STATUS_IS_OK(state.status)) {
DBG_WARNING("smbXsrv_open_close_fn() for %s failed: %s\n",
tdb_data_dbg(key),
nt_errstr(state.status));
error = state.status;
}
ret = idr_remove(table->local.idr, op->local_id);
SMB_ASSERT(ret == 0);
table->local.num_opens -= 1;
if (op->compat) {
op->compat->op = NULL;
file_free(NULL, op->compat);
op->compat = NULL;
}
return error;
}
NTSTATUS smb1srv_open_table_init(struct smbXsrv_connection *conn)
{
uint32_t max_opens;
/*
* Allow a range from 1..65534.
*
* With real_max_open_files possible ids,
* truncated to the SMB1 limit of 16-bit.
*
* 0 and 0xFFFF are no valid ids.
*/
max_opens = conn->client->sconn->real_max_open_files;
max_opens = MIN(max_opens, UINT16_MAX - 1);
return smbXsrv_open_table_init(conn, 1, UINT16_MAX - 1, max_opens);
}
NTSTATUS smb1srv_open_lookup(struct smbXsrv_connection *conn,
uint16_t fnum, NTTIME now,
struct smbXsrv_open **_open)
{
struct smbXsrv_open_table *table = conn->client->open_table;
uint32_t local_id = fnum;
uint32_t global_id = 0;
return smbXsrv_open_local_lookup(table, local_id, global_id, now, _open);
}
NTSTATUS smb2srv_open_table_init(struct smbXsrv_connection *conn)
{
uint32_t max_opens;
uint32_t highest_id;
/*
* Allow a range from 1..4294967294.
*
* With real_max_open_files possible ids,
* truncated to 16-bit (the same as SMB1 for now).
*
* 0 and 0xFFFFFFFF are no valid ids.
*
* The usage of conn->sconn->real_max_open_files
* is the reason that we use one open table per
* transport connection (as we still have a 1:1 mapping
* between process and transport connection).
*/
max_opens = conn->client->sconn->real_max_open_files;
max_opens = MIN(max_opens, UINT16_MAX - 1);
/*
* idtree uses "int" for local IDs. Limit the maximum ID to
* what "int" can hold.
*/
highest_id = UINT32_MAX-1;
highest_id = MIN(highest_id, INT_MAX);
return smbXsrv_open_table_init(conn, 1, highest_id, max_opens);
}
NTSTATUS smb2srv_open_lookup(struct smbXsrv_connection *conn,
uint64_t persistent_id,
uint64_t volatile_id,
NTTIME now,
struct smbXsrv_open **_open)
{
struct smbXsrv_open_table *table = conn->client->open_table;
uint32_t local_id = volatile_id & UINT32_MAX;
uint64_t local_zeros = volatile_id & 0xFFFFFFFF00000000LLU;
uint32_t global_id = persistent_id & UINT32_MAX;
uint64_t global_zeros = persistent_id & 0xFFFFFFFF00000000LLU;
NTSTATUS status;
if (local_zeros != 0) {
return NT_STATUS_FILE_CLOSED;
}
if (global_zeros != 0) {
return NT_STATUS_FILE_CLOSED;
}
if (global_id == 0) {
return NT_STATUS_FILE_CLOSED;
}
status = smbXsrv_open_local_lookup(table, local_id, global_id, now,
_open);
if (!NT_STATUS_IS_OK(status)) {
return status;
}
/*
* Clear the replay cache for this create_guid if it exists:
* This is based on the assumption that this lookup will be
* triggered by a client request using the file-id for lookup.
* Hence the client has proven that it has in fact seen the
* reply to its initial create call. So subsequent create replays
* should be treated as invalid. Hence the index for create_guid
* lookup needs to be removed.
*/
status = smbXsrv_open_clear_replay_cache(*_open);
return status;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
/*
* This checks or marks the replay cache, we have the following
* cases:
*
* 1. There is no record in the cache
* => we add the passes caller_req_guid as holder_req_guid
* together with local_id as 0.
* => We return STATUS_FWP_RESERVED in order to indicate
* that the caller holds the current reservation
*
* 2. There is a record in the cache and holder_req_guid
* is already the same as caller_req_guid and local_id is 0
* => We return STATUS_FWP_RESERVED in order to indicate
* that the caller holds the current reservation
*
* 3. There is a record in the cache with a holder_req_guid
* other than caller_req_guid (and local_id is 0):
* => We return NT_STATUS_FILE_NOT_AVAILABLE to indicate
* the original request is still pending
*
* 4. There is a record in the cache with a zero holder_req_guid
* and a valid local_id:
* => We lookup the existing open by local_id
* => We return NT_STATUS_OK together with the smbXsrv_open
*
*
* With NT_STATUS_OK the caller can continue the replay processing.
*
* With STATUS_FWP_RESERVED the caller should continue the normal
* open processing:
* - On success:
* - smbXsrv_open_update()/smbXsrv_open_set_replay_cache()
* will convert the record to a zero holder_req_guid
* with a valid local_id.
* - On failure:
* - smbXsrv_open_purge_replay_cache() should cleanup
* the reservation.
*
* All other values should be returned to the client,
* while NT_STATUS_FILE_NOT_AVAILABLE will trigger the
* retry loop on the client.
*/
NTSTATUS smb2srv_open_lookup_replay_cache(struct smbXsrv_connection *conn,
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
struct GUID caller_req_guid,
struct GUID create_guid,
const char *name,
NTTIME now,
struct smbXsrv_open **_open)
{
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
TALLOC_CTX *frame = talloc_stackframe();
NTSTATUS status;
struct smbXsrv_open_table *table = conn->client->open_table;
struct db_context *db = table->local.replay_cache_db_ctx;
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
struct GUID_txt_buf tmp_guid_buf;
struct GUID_txt_buf _create_guid_buf;
const char *create_guid_str = GUID_buf_string(&create_guid, &_create_guid_buf);
TDB_DATA create_guid_key = string_term_tdb_data(create_guid_str);
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
struct db_record *db_rec = NULL;
struct smbXsrv_open *op = NULL;
struct smbXsrv_open_replay_cache rc = {
.holder_req_guid = caller_req_guid,
.idle_time = now,
.local_id = 0,
};
enum ndr_err_code ndr_err;
DATA_BLOB blob = data_blob_null;
TDB_DATA val;
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
*_open = NULL;
db_rec = dbwrap_fetch_locked(db, frame, create_guid_key);
if (db_rec == NULL) {
TALLOC_FREE(frame);
return NT_STATUS_INTERNAL_DB_ERROR;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
val = dbwrap_record_get_value(db_rec);
if (val.dsize == 0) {
uint8_t data[SMBXSRV_OPEN_REPLAY_CACHE_FIXED_SIZE];
blob = data_blob_const(data, ARRAY_SIZE(data));
ndr_err = ndr_push_struct_into_fixed_blob(&blob, &rc,
(ndr_push_flags_fn_t)ndr_push_smbXsrv_open_replay_cache);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
status = ndr_map_error2ntstatus(ndr_err);
TALLOC_FREE(frame);
return status;
}
val = make_tdb_data(blob.data, blob.length);
status = dbwrap_record_store(db_rec, val, TDB_REPLACE);
if (!NT_STATUS_IS_OK(status)) {
TALLOC_FREE(frame);
return status;
}
/*
* We're the new holder
*/
*_open = NULL;
TALLOC_FREE(frame);
return NT_STATUS_FWP_RESERVED;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
if (val.dsize != SMBXSRV_OPEN_REPLAY_CACHE_FIXED_SIZE) {
TALLOC_FREE(frame);
return NT_STATUS_INTERNAL_DB_CORRUPTION;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
blob = data_blob_const(val.dptr, val.dsize);
ndr_err = ndr_pull_struct_blob_all_noalloc(&blob, &rc,
(ndr_pull_flags_fn_t)ndr_pull_smbXsrv_open_replay_cache);
if (!NDR_ERR_CODE_IS_SUCCESS(ndr_err)) {
status = ndr_map_error2ntstatus(ndr_err);
TALLOC_FREE(frame);
return status;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
if (rc.local_id != 0) {
if (GUID_equal(&rc.holder_req_guid, &caller_req_guid)) {
/*
* This should not happen
*/
status = NT_STATUS_INTERNAL_ERROR;
DBG_ERR("caller %s already holds local_id %u for create %s [%s] - %s\n",
GUID_buf_string(&caller_req_guid, &tmp_guid_buf),
(unsigned)rc.local_id,
create_guid_str,
name,
nt_errstr(status));
TALLOC_FREE(frame);
return status;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
status = smbXsrv_open_local_lookup(table,
rc.local_id,
0, /* global_id */
now,
&op);
if (!NT_STATUS_IS_OK(status)) {
DBG_ERR("holder %s stale for local_id %u for create %s [%s] - %s\n",
GUID_buf_string(&rc.holder_req_guid, &tmp_guid_buf),
(unsigned)rc.local_id,
create_guid_str,
name,
nt_errstr(status));
TALLOC_FREE(frame);
return status;
}
/*
* We found an open the caller can reuse.
*/
SMB_ASSERT(op != NULL);
*_open = op;
TALLOC_FREE(frame);
return NT_STATUS_OK;
}
if (GUID_equal(&rc.holder_req_guid, &caller_req_guid)) {
/*
* We're still the holder
*/
*_open = NULL;
TALLOC_FREE(frame);
return NT_STATUS_FWP_RESERVED;
}
smbXsrv_open: intruduce smbXsrv_open_replay_cache to support FILE_NOT_AVAILABLE Before processing an open we need to reserve the replay cache entry in order to signal that we're still in progress. If a reserved record is already present we need to return FILE_NOT_AVAILABLE in order to let the client retry again. [MS-SMB2] contains this: <152> Section 3.2.5.1: For the following error codes, Windows-based clients will retry the operation up to three times and then retry the operation every 5 seconds until the count of milliseconds specified by Open.ResilientTimeout is exceeded: - STATUS_SERVER_UNAVAILABLE - STATUS_FILE_NOT_AVAILABLE - STATUS_SHARE_UNAVAILABLE This works fine for windows clients, but current windows servers seems to return ACCESS_DENIED instead of FILE_NOT_AVAILABLE. A Windows server doesn't do any replay detection on pending opens, which wait for a HANDLE lease to be broken (because of a SHARING_VIOLATION), at all. As this is not really documented for the server part of the current [MS-SMB2], I found the key hint in "SMB 2.2: Bigger. Faster. Scalier - (Parts 1 and 2)" on page 24. There's a picture showing that a replay gets FILE_NOT_AVAILABLE as long as the original request is still in progress. See: https://www.snia.org/educational-library/smb-22-bigger-faster-scalier-parts-1-and-2-2011 A Windows client is unhappy with the current windows server behavior if it such a situation happens. There's also a very strange interaction with oplock where the replay gets SHARING_VIOLATION after 35 seconds because it conflicts with the original open. I think it's good to follow the intial design from the 2011 presentation and make the clients happy by using FILE_NOT_AVAILABLE (and differ from Windows). I'll report that to dochelp@microsoft.com in order to get this hopefully fixed in their server too). BUG: https://bugzilla.samba.org/show_bug.cgi?id=14449 Signed-off-by: Stefan Metzmacher <metze@samba.org> Reviewed-by: Jeremy Allison <jra@samba.org>
2021-03-12 17:10:46 +03:00
/*
* The original request (or a former replay) is still
* pending, ask the client to retry by sending FILE_NOT_AVAILABLE.
*/
status = NT_STATUS_FILE_NOT_AVAILABLE;
DBG_DEBUG("holder %s still pending for create %s [%s] - %s\n",
GUID_buf_string(&rc.holder_req_guid, &tmp_guid_buf),
create_guid_str,
name,
nt_errstr(status));
TALLOC_FREE(frame);
return status;
}
struct smb2srv_open_recreate_state {
struct smbXsrv_open *op;
const struct GUID *create_guid;
struct security_token *current_token;
struct server_id me;
NTSTATUS status;
};
static void smb2srv_open_recreate_fn(
struct db_record *rec, TDB_DATA oldval, void *private_data)
{
struct smb2srv_open_recreate_state *state = private_data;
TDB_DATA key = dbwrap_record_get_key(rec);
struct smbXsrv_open_global0 *global = NULL;
state->status = smbXsrv_open_global_verify_record(
key, oldval, state->op, &state->op->global);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_WARNING("smbXsrv_open_global_verify_record for %s "
"failed: %s\n",
tdb_data_dbg(key),
nt_errstr(state->status));
goto not_found;
}
global = state->op->global;
/*
* If the provided create_guid is NULL, this means that
* the reconnect request was a v1 request. In that case
* we should skip the create GUID verification, since
* it is valid to v1-reconnect a v2-opened handle.
*/
if ((state->create_guid != NULL) &&
!GUID_equal(&global->create_guid, state->create_guid)) {
struct GUID_txt_buf buf1, buf2;
DBG_NOTICE("%s != %s in %s\n",
GUID_buf_string(&global->create_guid, &buf1),
GUID_buf_string(state->create_guid, &buf2),
tdb_data_dbg(key));
goto not_found;
}
if (!security_token_is_sid(
state->current_token, &global->open_owner)) {
struct dom_sid_buf buf;
DBG_NOTICE("global owner %s not in our token in %s\n",
dom_sid_str_buf(&global->open_owner, &buf),
tdb_data_dbg(key));
goto not_found;
}
if (!global->durable) {
DBG_NOTICE("%"PRIu64"/%"PRIu64" not durable in %s\n",
global->open_persistent_id,
global->open_volatile_id,
tdb_data_dbg(key));
goto not_found;
}
global->open_volatile_id = state->op->local_id;
global->server_id = state->me;
state->status = smbXsrv_open_global_store(rec, key, oldval, global);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_WARNING("smbXsrv_open_global_store for %s failed: %s\n",
tdb_data_dbg(key),
nt_errstr(state->status));
return;
}
return;
not_found:
state->status = NT_STATUS_OBJECT_NAME_NOT_FOUND;
}
NTSTATUS smb2srv_open_recreate(struct smbXsrv_connection *conn,
struct auth_session_info *session_info,
uint64_t persistent_id,
const struct GUID *create_guid,
NTTIME now,
struct smbXsrv_open **_open)
{
struct smbXsrv_open_table *table = conn->client->open_table;
struct smb2srv_open_recreate_state state = {
.create_guid = create_guid,
.me = messaging_server_id(conn->client->msg_ctx),
};
struct smbXsrv_open_global_key_buf key_buf;
TDB_DATA key = smbXsrv_open_global_id_to_key(
persistent_id & UINT32_MAX, &key_buf);
int ret, local_id;
NTSTATUS status;
if (session_info == NULL) {
DEBUG(10, ("session_info=NULL\n"));
return NT_STATUS_INVALID_HANDLE;
}
state.current_token = session_info->security_token;
if (state.current_token == NULL) {
DEBUG(10, ("current_token=NULL\n"));
return NT_STATUS_INVALID_HANDLE;
}
if ((persistent_id & 0xFFFFFFFF00000000LLU) != 0) {
/*
* We only use 32 bit for the persistent ID
*/
DBG_DEBUG("persistent_id=%"PRIx64"\n", persistent_id);
return NT_STATUS_OBJECT_NAME_NOT_FOUND;
}
if (table->local.num_opens >= table->local.max_opens) {
return NT_STATUS_INSUFFICIENT_RESOURCES;
}
state.op = talloc_zero(table, struct smbXsrv_open);
if (state.op == NULL) {
return NT_STATUS_NO_MEMORY;
}
state.op->table = table;
local_id = idr_get_new_random(
table->local.idr,
state.op,
table->local.lowest_id,
table->local.highest_id);
if (local_id == -1) {
TALLOC_FREE(state.op);
return NT_STATUS_INSUFFICIENT_RESOURCES;
}
state.op->local_id = local_id;
SMB_ASSERT(state.op->local_id == local_id); /* No coercion loss */
table->local.num_opens += 1;
state.op->idle_time = now;
state.op->status = NT_STATUS_FILE_CLOSED;
status = dbwrap_do_locked(
table->global.db_ctx, key, smb2srv_open_recreate_fn, &state);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("dbwrap_do_locked() for %s failed: %s\n",
tdb_data_dbg(key),
nt_errstr(status));
goto fail;
}
if (!NT_STATUS_IS_OK(state.status)) {
status = state.status;
DBG_DEBUG("smb2srv_open_recreate_fn for %s failed: %s\n",
tdb_data_dbg(key),
nt_errstr(status));
goto fail;
}
talloc_set_destructor(state.op, smbXsrv_open_destructor);
if (CHECK_DEBUGLVL(10)) {
struct smbXsrv_openB open_blob = {
.info.info0 = state.op,
};
DBG_DEBUG("global_id (0x%08x) stored\n",
state.op->global->open_global_id);
NDR_PRINT_DEBUG(smbXsrv_openB, &open_blob);
}
*_open = state.op;
return NT_STATUS_OK;
fail:
table->local.num_opens -= 1;
ret = idr_remove(table->local.idr, state.op->local_id);
SMB_ASSERT(ret == 0);
TALLOC_FREE(state.op);
return status;
}
struct smbXsrv_open_global_traverse_state {
int (*fn)(struct db_record *rec, struct smbXsrv_open_global0 *, void *);
void *private_data;
};
static int smbXsrv_open_global_traverse_fn(struct db_record *rec, void *data)
{
struct smbXsrv_open_global_traverse_state *state =
(struct smbXsrv_open_global_traverse_state*)data;
struct smbXsrv_open_global0 *global = NULL;
TDB_DATA key = dbwrap_record_get_key(rec);
TDB_DATA val = dbwrap_record_get_value(rec);
NTSTATUS status;
int ret = -1;
status = smbXsrv_open_global_parse_record(
talloc_tos(), key, val, &global);
if (!NT_STATUS_IS_OK(status)) {
return -1;
}
ret = state->fn(rec, global, state->private_data);
talloc_free(global);
return ret;
}
NTSTATUS smbXsrv_open_global_traverse(
int (*fn)(struct db_record *rec, struct smbXsrv_open_global0 *, void *),
void *private_data)
{
NTSTATUS status;
int count = 0;
struct smbXsrv_open_global_traverse_state state = {
.fn = fn,
.private_data = private_data,
};
become_root();
status = smbXsrv_open_global_init();
if (!NT_STATUS_IS_OK(status)) {
unbecome_root();
DEBUG(0, ("Failed to initialize open_global: %s\n",
nt_errstr(status)));
return status;
}
status = dbwrap_traverse_read(smbXsrv_open_global_db_ctx,
smbXsrv_open_global_traverse_fn,
&state,
&count);
unbecome_root();
return status;
}
struct smbXsrv_open_cleanup_state {
uint32_t global_id;
NTSTATUS status;
};
static void smbXsrv_open_cleanup_fn(
struct db_record *rec, TDB_DATA oldval, void *private_data)
{
struct smbXsrv_open_cleanup_state *state = private_data;
struct smbXsrv_open_global0 *global = NULL;
TDB_DATA key = dbwrap_record_get_key(rec);
bool delete_open = false;
if (oldval.dsize == 0) {
DBG_DEBUG("[global: 0x%08x] "
"empty record in %s, skipping...\n",
state->global_id,
dbwrap_name(dbwrap_record_get_db(rec)));
state->status = NT_STATUS_OK;
return;
}
state->status = smbXsrv_open_global_parse_record(
talloc_tos(), key, oldval, &global);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_WARNING("[global: %x08x] "
"smbXsrv_open_global_parse_record() in %s "
"failed: %s, deleting record\n",
state->global_id,
dbwrap_name(dbwrap_record_get_db(rec)),
nt_errstr(state->status));
delete_open = true;
goto do_delete;
}
if (server_id_is_disconnected(&global->server_id)) {
struct timeval now = timeval_current();
struct timeval disconnect_time;
struct timeval_buf buf;
int64_t tdiff;
nttime_to_timeval(&disconnect_time, global->disconnect_time);
tdiff = usec_time_diff(&now, &disconnect_time);
delete_open = (tdiff >= 1000*global->durable_timeout_msec);
DBG_DEBUG("[global: 0x%08x] "
"disconnected at [%s] %"PRIi64"s ago with "
"timeout of %"PRIu32"s -%s reached\n",
state->global_id,
timeval_str_buf(&disconnect_time,
false,
false,
&buf),
tdiff/1000000,
global->durable_timeout_msec / 1000,
delete_open ? "" : " not");
} else if (!serverid_exists(&global->server_id)) {
struct server_id_buf idbuf;
DBG_DEBUG("[global: 0x%08x] "
"server[%s] does not exist\n",
state->global_id,
server_id_str_buf(global->server_id, &idbuf));
delete_open = true;
}
if (!delete_open) {
state->status = NT_STATUS_OK;
return;
}
do_delete:
state->status = dbwrap_record_delete(rec);
if (!NT_STATUS_IS_OK(state->status)) {
DBG_WARNING("[global: 0x%08x] "
"failed to delete record "
"from %s: %s\n",
state->global_id,
dbwrap_name(dbwrap_record_get_db(rec)),
nt_errstr(state->status));
return;
}
DBG_DEBUG("[global: 0x%08x] "
"deleted record from %s\n",
state->global_id,
dbwrap_name(dbwrap_record_get_db(rec)));
}
NTSTATUS smbXsrv_open_cleanup(uint64_t persistent_id)
{
struct smbXsrv_open_cleanup_state state = {
.global_id = persistent_id & UINT32_MAX,
};
struct smbXsrv_open_global_key_buf key_buf;
TDB_DATA key = smbXsrv_open_global_id_to_key(
state.global_id, &key_buf);
NTSTATUS status;
status = dbwrap_do_locked(
smbXsrv_open_global_db_ctx,
key,
smbXsrv_open_cleanup_fn,
&state);
if (!NT_STATUS_IS_OK(status)) {
DBG_DEBUG("[global: 0x%08x] dbwrap_do_locked failed: %s\n",
state.global_id,
nt_errstr(status));
return status;
}
return state.status;
}