1
0
mirror of https://github.com/samba-team/samba.git synced 2024-12-22 13:34:15 +03:00
samba-mirror/ctdb/client/client_db.c
Amitay Isaacs f721035d43 ctdb-client: Fix implementation of transaction start
Since g_lock checks if the process exists in case of conflicting lock,
there is no need to register srvid.

Transaction start returns a transaction handle and transaction
commit/cancel will free that handle.  Since we cannot call async code
in a talloc destructor, this avoids the use of talloc destructor for
cancelling the transaction.

If user frees the transaction handle instead of calling transaction
cancel, it will leave stale g_lock lock.  This stale g_lock lock  will
get cleaned up on next transaction attempt.

Signed-off-by: Amitay Isaacs <amitay@gmail.com>
Reviewed-by: Martin Schwenke <martin@meltin.net>
2016-07-05 10:53:14 +02:00

2157 lines
49 KiB
C

/*
CTDB client code
Copyright (C) Amitay Isaacs 2015
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 "replace.h"
#include "system/network.h"
#include "system/filesys.h"
#include <talloc.h>
#include <tevent.h>
#include <tdb.h>
#include "common/logging.h"
#include "lib/tdb_wrap/tdb_wrap.h"
#include "lib/util/tevent_unix.h"
#include "lib/util/dlinklist.h"
#include "lib/util/debug.h"
#include "protocol/protocol.h"
#include "protocol/protocol_api.h"
#include "client/client_private.h"
#include "client/client.h"
static struct ctdb_db_context *client_db_handle(
struct ctdb_client_context *client,
const char *db_name)
{
struct ctdb_db_context *db;
for (db = client->db; db != NULL; db = db->next) {
if (strcmp(db_name, db->db_name) == 0) {
return db;
}
}
return NULL;
}
struct ctdb_set_db_flags_state {
struct tevent_context *ev;
struct ctdb_client_context *client;
struct timeval timeout;
uint32_t db_id;
uint8_t db_flags;
bool readonly_done, sticky_done;
uint32_t *pnn_list;
int count;
};
static void ctdb_set_db_flags_nodemap_done(struct tevent_req *subreq);
static void ctdb_set_db_flags_readonly_done(struct tevent_req *subreq);
static void ctdb_set_db_flags_sticky_done(struct tevent_req *subreq);
static struct tevent_req *ctdb_set_db_flags_send(
TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_client_context *client,
uint32_t destnode, struct timeval timeout,
uint32_t db_id, uint8_t db_flags)
{
struct tevent_req *req, *subreq;
struct ctdb_set_db_flags_state *state;
struct ctdb_req_control request;
req = tevent_req_create(mem_ctx, &state,
struct ctdb_set_db_flags_state);
if (req == NULL) {
return NULL;
}
if (! (db_flags & (CTDB_DB_FLAGS_READONLY | CTDB_DB_FLAGS_STICKY))) {
tevent_req_done(req);
return tevent_req_post(req, ev);
}
state->ev = ev;
state->client = client;
state->timeout = timeout;
state->db_id = db_id;
state->db_flags = db_flags;
ctdb_req_control_get_nodemap(&request);
subreq = ctdb_client_control_send(state, ev, client, destnode, timeout,
&request);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, ctdb_set_db_flags_nodemap_done, req);
return req;
}
static void ctdb_set_db_flags_nodemap_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_set_db_flags_state *state = tevent_req_data(
req, struct ctdb_set_db_flags_state);
struct ctdb_req_control request;
struct ctdb_reply_control *reply;
struct ctdb_node_map *nodemap;
int ret;
bool status;
status = ctdb_client_control_recv(subreq, &ret, state, &reply);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_reply_control_get_nodemap(reply, state, &nodemap);
talloc_free(reply);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
state->count = list_of_connected_nodes(nodemap, CTDB_UNKNOWN_PNN,
state, &state->pnn_list);
talloc_free(nodemap);
if (state->count <= 0) {
tevent_req_error(req, ENOMEM);
return;
}
if (state->db_flags & CTDB_DB_FLAGS_READONLY) {
ctdb_req_control_set_db_readonly(&request, state->db_id);
subreq = ctdb_client_control_multi_send(
state, state->ev, state->client,
state->pnn_list, state->count,
state->timeout, &request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq,
ctdb_set_db_flags_readonly_done, req);
} else {
state->readonly_done = true;
}
if (state->db_flags & CTDB_DB_FLAGS_STICKY) {
ctdb_req_control_set_db_sticky(&request, state->db_id);
subreq = ctdb_client_control_multi_send(
state, state->ev, state->client,
state->pnn_list, state->count,
state->timeout, &request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_set_db_flags_sticky_done,
req);
} else {
state->sticky_done = true;
}
}
static void ctdb_set_db_flags_readonly_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_set_db_flags_state *state = tevent_req_data(
req, struct ctdb_set_db_flags_state);
int ret;
bool status;
status = ctdb_client_control_multi_recv(subreq, &ret, NULL, NULL,
NULL);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
state->readonly_done = true;
if (state->readonly_done && state->sticky_done) {
tevent_req_done(req);
}
}
static void ctdb_set_db_flags_sticky_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_set_db_flags_state *state = tevent_req_data(
req, struct ctdb_set_db_flags_state);
int ret;
bool status;
status = ctdb_client_control_multi_recv(subreq, &ret, NULL, NULL,
NULL);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
state->sticky_done = true;
if (state->readonly_done && state->sticky_done) {
tevent_req_done(req);
}
}
static bool ctdb_set_db_flags_recv(struct tevent_req *req, int *perr)
{
int err;
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return false;
}
return true;
}
struct ctdb_attach_state {
struct tevent_context *ev;
struct ctdb_client_context *client;
struct timeval timeout;
uint32_t destnode;
uint8_t db_flags;
uint32_t tdb_flags;
struct ctdb_db_context *db;
};
static void ctdb_attach_mutex_done(struct tevent_req *subreq);
static void ctdb_attach_dbid_done(struct tevent_req *subreq);
static void ctdb_attach_dbpath_done(struct tevent_req *subreq);
static void ctdb_attach_health_done(struct tevent_req *subreq);
static void ctdb_attach_flags_done(struct tevent_req *subreq);
struct tevent_req *ctdb_attach_send(TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_client_context *client,
struct timeval timeout,
const char *db_name, uint8_t db_flags)
{
struct tevent_req *req, *subreq;
struct ctdb_attach_state *state;
struct ctdb_req_control request;
req = tevent_req_create(mem_ctx, &state, struct ctdb_attach_state);
if (req == NULL) {
return NULL;
}
state->db = client_db_handle(client, db_name);
if (state->db != NULL) {
tevent_req_done(req);
return tevent_req_post(req, ev);
}
state->ev = ev;
state->client = client;
state->timeout = timeout;
state->destnode = ctdb_client_pnn(client);
state->db_flags = db_flags;
state->db = talloc_zero(client, struct ctdb_db_context);
if (tevent_req_nomem(state->db, req)) {
return tevent_req_post(req, ev);
}
state->db->db_name = talloc_strdup(state->db, db_name);
if (tevent_req_nomem(state->db, req)) {
return tevent_req_post(req, ev);
}
if (db_flags & CTDB_DB_FLAGS_PERSISTENT) {
state->db->persistent = true;
}
ctdb_req_control_get_tunable(&request, "TDBMutexEnabled");
subreq = ctdb_client_control_send(state, ev, client,
ctdb_client_pnn(client), timeout,
&request);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, ctdb_attach_mutex_done, req);
return req;
}
static void ctdb_attach_mutex_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_attach_state *state = tevent_req_data(
req, struct ctdb_attach_state);
struct ctdb_reply_control *reply;
struct ctdb_req_control request;
uint32_t mutex_enabled;
int ret;
bool status;
status = ctdb_client_control_recv(subreq, &ret, state, &reply);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_reply_control_get_tunable(reply, &mutex_enabled);
if (ret != 0) {
/* Treat error as mutex support not available */
mutex_enabled = 0;
}
if (state->db->persistent) {
state->tdb_flags = TDB_DEFAULT;
} else {
state->tdb_flags = (TDB_NOSYNC | TDB_INCOMPATIBLE_HASH |
TDB_CLEAR_IF_FIRST);
if (mutex_enabled == 1) {
state->tdb_flags |= TDB_MUTEX_LOCKING;
}
}
if (state->db->persistent) {
ctdb_req_control_db_attach_persistent(&request,
state->db->db_name,
state->tdb_flags);
} else {
ctdb_req_control_db_attach(&request, state->db->db_name,
state->tdb_flags);
}
subreq = ctdb_client_control_send(state, state->ev, state->client,
state->destnode, state->timeout,
&request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_attach_dbid_done, req);
}
static void ctdb_attach_dbid_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_attach_state *state = tevent_req_data(
req, struct ctdb_attach_state);
struct ctdb_req_control request;
struct ctdb_reply_control *reply;
bool status;
int ret;
status = ctdb_client_control_recv(subreq, &ret, state, &reply);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
if (state->db->persistent) {
ret = ctdb_reply_control_db_attach_persistent(
reply, &state->db->db_id);
} else {
ret = ctdb_reply_control_db_attach(reply, &state->db->db_id);
}
talloc_free(reply);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
ctdb_req_control_getdbpath(&request, state->db->db_id);
subreq = ctdb_client_control_send(state, state->ev, state->client,
state->destnode, state->timeout,
&request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_attach_dbpath_done, req);
}
static void ctdb_attach_dbpath_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_attach_state *state = tevent_req_data(
req, struct ctdb_attach_state);
struct ctdb_reply_control *reply;
struct ctdb_req_control request;
bool status;
int ret;
status = ctdb_client_control_recv(subreq, &ret, state, &reply);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_reply_control_getdbpath(reply, state->db,
&state->db->db_path);
talloc_free(reply);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
ctdb_req_control_db_get_health(&request, state->db->db_id);
subreq = ctdb_client_control_send(state, state->ev, state->client,
state->destnode, state->timeout,
&request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_attach_health_done, req);
}
static void ctdb_attach_health_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_attach_state *state = tevent_req_data(
req, struct ctdb_attach_state);
struct ctdb_reply_control *reply;
const char *reason;
bool status;
int ret;
status = ctdb_client_control_recv(subreq, &ret, state, &reply);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_reply_control_db_get_health(reply, state, &reason);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
if (reason != NULL) {
/* Database unhealthy, avoid attach */
/* FIXME: Log here */
tevent_req_error(req, EIO);
return;
}
subreq = ctdb_set_db_flags_send(state, state->ev, state->client,
state->destnode, state->timeout,
state->db->db_id, state->db_flags);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_attach_flags_done, req);
}
static void ctdb_attach_flags_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_attach_state *state = tevent_req_data(
req, struct ctdb_attach_state);
bool status;
int ret;
status = ctdb_set_db_flags_recv(subreq, &ret);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
state->db->ltdb = tdb_wrap_open(state->db, state->db->db_path, 0,
state->tdb_flags, O_RDWR, 0);
if (tevent_req_nomem(state->db->ltdb, req)) {
return;
}
DLIST_ADD(state->client->db, state->db);
tevent_req_done(req);
}
bool ctdb_attach_recv(struct tevent_req *req, int *perr,
struct ctdb_db_context **out)
{
struct ctdb_attach_state *state = tevent_req_data(
req, struct ctdb_attach_state);
int err;
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return false;
}
if (out != NULL) {
*out = state->db;
}
return true;
}
int ctdb_attach(struct tevent_context *ev,
struct ctdb_client_context *client,
struct timeval timeout,
const char *db_name, uint8_t db_flags,
struct ctdb_db_context **out)
{
TALLOC_CTX *mem_ctx;
struct tevent_req *req;
bool status;
int ret;
mem_ctx = talloc_new(client);
if (mem_ctx == NULL) {
return ENOMEM;
}
req = ctdb_attach_send(mem_ctx, ev, client, timeout,
db_name, db_flags);
if (req == NULL) {
talloc_free(mem_ctx);
return ENOMEM;
}
tevent_req_poll(req, ev);
status = ctdb_attach_recv(req, &ret, out);
if (! status) {
talloc_free(mem_ctx);
return ret;
}
/*
ctdb_set_call(db, CTDB_NULL_FUNC, ctdb_null_func);
ctdb_set_call(db, CTDB_FETCH_FUNC, ctdb_fetch_func);
ctdb_set_call(db, CTDB_FETCH_WITH_HEADER_FUNC, ctdb_fetch_with_header_func);
*/
talloc_free(mem_ctx);
return 0;
}
int ctdb_detach(TALLOC_CTX *mem_ctx, struct tevent_context *ev,
struct ctdb_client_context *client,
struct timeval timeout, uint32_t db_id)
{
struct ctdb_db_context *db;
int ret;
ret = ctdb_ctrl_db_detach(mem_ctx, ev, client, client->pnn, timeout,
db_id);
if (ret != 0) {
return ret;
}
for (db = client->db; db != NULL; db = db->next) {
if (db->db_id == db_id) {
DLIST_REMOVE(client->db, db);
break;
}
}
return 0;
}
uint32_t ctdb_db_id(struct ctdb_db_context *db)
{
return db->db_id;
}
struct ctdb_db_traverse_state {
ctdb_rec_parser_func_t parser;
void *private_data;
bool extract_header;
int error;
};
static int ctdb_db_traverse_handler(struct tdb_context *tdb, TDB_DATA key,
TDB_DATA data, void *private_data)
{
struct ctdb_db_traverse_state *state =
(struct ctdb_db_traverse_state *)private_data;
int ret;
if (state->extract_header) {
struct ctdb_ltdb_header header;
ret = ctdb_ltdb_header_extract(&data, &header);
if (ret != 0) {
state->error = ret;
return 1;
}
ret = state->parser(0, &header, key, data, state->private_data);
} else {
ret = state->parser(0, NULL, key, data, state->private_data);
}
if (ret != 0) {
state->error = ret;
return 1;
}
return 0;
}
int ctdb_db_traverse(struct ctdb_db_context *db, bool readonly,
bool extract_header,
ctdb_rec_parser_func_t parser, void *private_data)
{
struct ctdb_db_traverse_state state;
int ret;
state.parser = parser;
state.private_data = private_data;
state.extract_header = extract_header;
state.error = 0;
if (readonly) {
ret = tdb_traverse_read(db->ltdb->tdb,
ctdb_db_traverse_handler, &state);
} else {
ret = tdb_traverse(db->ltdb->tdb,
ctdb_db_traverse_handler, &state);
}
if (ret == -1) {
return EIO;
}
return state.error;
}
static int ctdb_ltdb_fetch(struct ctdb_db_context *db, TDB_DATA key,
struct ctdb_ltdb_header *header,
TALLOC_CTX *mem_ctx, TDB_DATA *data)
{
TDB_DATA rec;
int ret;
rec = tdb_fetch(db->ltdb->tdb, key);
if (rec.dsize < sizeof(struct ctdb_ltdb_header)) {
/* No record present */
if (rec.dptr != NULL) {
free(rec.dptr);
}
if (tdb_error(db->ltdb->tdb) != TDB_ERR_NOEXIST) {
return EIO;
}
header->rsn = 0;
header->dmaster = CTDB_UNKNOWN_PNN;
header->flags = 0;
if (data != NULL) {
*data = tdb_null;
}
return 0;
}
ret = ctdb_ltdb_header_pull(rec.dptr, rec.dsize, header);
if (ret != 0) {
return ret;
}
ret = 0;
if (data != NULL) {
size_t offset = ctdb_ltdb_header_len(header);
data->dsize = rec.dsize - offset;
data->dptr = talloc_memdup(mem_ctx, rec.dptr + offset,
data->dsize);
if (data->dptr == NULL) {
ret = ENOMEM;
}
}
free(rec.dptr);
return ret;
}
/*
* Fetch a record from volatile database
*
* Steps:
* 1. Get a lock on the hash chain
* 2. If the record does not exist, migrate the record
* 3. If readonly=true and delegations do not exist, migrate the record.
* 4. If readonly=false and delegations exist, migrate the record.
* 5. If the local node is not dmaster, migrate the record.
* 6. Return record
*/
struct ctdb_fetch_lock_state {
struct tevent_context *ev;
struct ctdb_client_context *client;
struct ctdb_record_handle *h;
bool readonly;
uint32_t pnn;
};
static int ctdb_fetch_lock_check(struct tevent_req *req);
static void ctdb_fetch_lock_migrate(struct tevent_req *req);
static void ctdb_fetch_lock_migrate_done(struct tevent_req *subreq);
struct tevent_req *ctdb_fetch_lock_send(TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_client_context *client,
struct ctdb_db_context *db,
TDB_DATA key, bool readonly)
{
struct ctdb_fetch_lock_state *state;
struct tevent_req *req;
int ret;
req = tevent_req_create(mem_ctx, &state, struct ctdb_fetch_lock_state);
if (req == NULL) {
return NULL;
}
state->ev = ev;
state->client = client;
state->h = talloc_zero(db, struct ctdb_record_handle);
if (tevent_req_nomem(state->h, req)) {
return tevent_req_post(req, ev);
}
state->h->client = client;
state->h->db = db;
state->h->key.dptr = talloc_memdup(state->h, key.dptr, key.dsize);
if (tevent_req_nomem(state->h->key.dptr, req)) {
return tevent_req_post(req, ev);
}
state->h->key.dsize = key.dsize;
state->h->readonly = false;
state->readonly = readonly;
state->pnn = ctdb_client_pnn(client);
/* Check that database is not persistent */
if (db->persistent) {
tevent_req_error(req, EINVAL);
return tevent_req_post(req, ev);
}
ret = ctdb_fetch_lock_check(req);
if (ret == 0) {
tevent_req_done(req);
return tevent_req_post(req, ev);
}
if (ret != EAGAIN) {
tevent_req_error(req, ret);
return tevent_req_post(req, ev);
}
return req;
}
static int ctdb_fetch_lock_check(struct tevent_req *req)
{
struct ctdb_fetch_lock_state *state = tevent_req_data(
req, struct ctdb_fetch_lock_state);
struct ctdb_record_handle *h = state->h;
struct ctdb_ltdb_header header;
TDB_DATA data = tdb_null;
int ret, err = 0;
bool do_migrate = false;
ret = tdb_chainlock(state->h->db->ltdb->tdb, state->h->key);
if (ret != 0) {
err = EIO;
goto failed;
}
data = tdb_fetch(h->db->ltdb->tdb, h->key);
if (data.dptr == NULL) {
if (tdb_error(h->db->ltdb->tdb) == TDB_ERR_NOEXIST) {
goto migrate;
} else {
err = EIO;
goto failed;
}
}
/* Got the record */
ret = ctdb_ltdb_header_pull(data.dptr, data.dsize, &header);
if (ret != 0) {
err = ret;
goto failed;
}
if (! state->readonly) {
/* Read/write access */
if (header.dmaster == state->pnn &&
header.flags & CTDB_REC_RO_HAVE_DELEGATIONS) {
goto migrate;
}
if (header.dmaster != state->pnn) {
goto migrate;
}
} else {
/* Readonly access */
if (header.dmaster != state->pnn &&
! (header.flags & (CTDB_REC_RO_HAVE_READONLY |
CTDB_REC_RO_HAVE_DELEGATIONS))) {
goto migrate;
}
}
/* We are the dmaster or readonly delegation */
h->header = header;
h->data = data;
if (header.flags & (CTDB_REC_RO_HAVE_READONLY |
CTDB_REC_RO_HAVE_DELEGATIONS)) {
h->readonly = true;
}
return 0;
migrate:
do_migrate = true;
err = EAGAIN;
failed:
if (data.dptr != NULL) {
free(data.dptr);
}
ret = tdb_chainunlock(h->db->ltdb->tdb, h->key);
if (ret != 0) {
DEBUG(DEBUG_ERR, ("tdb_chainunlock failed on %s\n",
h->db->db_name));
return EIO;
}
if (do_migrate) {
ctdb_fetch_lock_migrate(req);
}
return err;
}
static void ctdb_fetch_lock_migrate(struct tevent_req *req)
{
struct ctdb_fetch_lock_state *state = tevent_req_data(
req, struct ctdb_fetch_lock_state);
struct ctdb_req_call request;
struct tevent_req *subreq;
ZERO_STRUCT(request);
request.flags = CTDB_IMMEDIATE_MIGRATION;
if (state->readonly) {
request.flags |= CTDB_WANT_READONLY;
}
request.db_id = state->h->db->db_id;
request.callid = CTDB_NULL_FUNC;
request.key = state->h->key;
request.calldata = tdb_null;
subreq = ctdb_client_call_send(state, state->ev, state->client,
&request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_fetch_lock_migrate_done, req);
}
static void ctdb_fetch_lock_migrate_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_fetch_lock_state *state = tevent_req_data(
req, struct ctdb_fetch_lock_state);
struct ctdb_reply_call *reply;
int ret;
bool status;
status = ctdb_client_call_recv(subreq, state, &reply, &ret);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
if (reply->status != 0) {
tevent_req_error(req, EIO);
return;
}
talloc_free(reply);
ret = ctdb_fetch_lock_check(req);
if (ret != 0) {
if (ret != EAGAIN) {
tevent_req_error(req, ret);
}
return;
}
tevent_req_done(req);
}
static int ctdb_record_handle_destructor(struct ctdb_record_handle *h)
{
tdb_chainunlock(h->db->ltdb->tdb, h->key);
free(h->data.dptr);
return 0;
}
struct ctdb_record_handle *ctdb_fetch_lock_recv(struct tevent_req *req,
struct ctdb_ltdb_header *header,
TALLOC_CTX *mem_ctx,
TDB_DATA *data, int *perr)
{
struct ctdb_fetch_lock_state *state = tevent_req_data(
req, struct ctdb_fetch_lock_state);
struct ctdb_record_handle *h = state->h;
int err;
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return NULL;
}
if (header != NULL) {
*header = h->header;
}
if (data != NULL) {
size_t offset;
offset = ctdb_ltdb_header_len(&h->header);
data->dsize = h->data.dsize - offset;
data->dptr = talloc_memdup(mem_ctx, h->data.dptr + offset,
data->dsize);
if (data->dptr == NULL) {
TALLOC_FREE(state->h);
if (perr != NULL) {
*perr = ENOMEM;
}
return NULL;
}
}
talloc_set_destructor(h, ctdb_record_handle_destructor);
return h;
}
int ctdb_fetch_lock(TALLOC_CTX *mem_ctx, struct tevent_context *ev,
struct ctdb_client_context *client,
struct ctdb_db_context *db, TDB_DATA key, bool readonly,
struct ctdb_record_handle **out,
struct ctdb_ltdb_header *header, TDB_DATA *data)
{
struct tevent_req *req;
struct ctdb_record_handle *h;
int ret;
req = ctdb_fetch_lock_send(mem_ctx, ev, client, db, key, readonly);
if (req == NULL) {
return ENOMEM;
}
tevent_req_poll(req, ev);
h = ctdb_fetch_lock_recv(req, header, mem_ctx, data, &ret);
if (h == NULL) {
return ret;
}
*out = h;
return 0;
}
int ctdb_store_record(struct ctdb_record_handle *h, TDB_DATA data)
{
TDB_DATA rec;
size_t offset;
int ret;
/* Cannot modify the record if it was obtained as a readonly copy */
if (h->readonly) {
return EINVAL;
}
/* Check if the new data is same */
if (h->data.dsize == data.dsize &&
memcmp(h->data.dptr, data.dptr, data.dsize) == 0) {
/* No need to do anything */
return 0;
}
offset = ctdb_ltdb_header_len(&h->header);
rec.dsize = offset + data.dsize;
rec.dptr = talloc_size(h, rec.dsize);
if (rec.dptr == NULL) {
return ENOMEM;
}
ctdb_ltdb_header_push(&h->header, rec.dptr);
memcpy(rec.dptr + offset, data.dptr, data.dsize);
ret = tdb_store(h->db->ltdb->tdb, h->key, rec, TDB_REPLACE);
if (ret != 0) {
DEBUG(DEBUG_ERR, ("Failed to store record in DB %s\n",
h->db->db_name));
return EIO;
}
talloc_free(rec.dptr);
return 0;
}
struct ctdb_delete_record_state {
struct ctdb_record_handle *h;
};
static void ctdb_delete_record_done(struct tevent_req *subreq);
struct tevent_req *ctdb_delete_record_send(TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_record_handle *h)
{
struct tevent_req *req, *subreq;
struct ctdb_delete_record_state *state;
struct ctdb_key_data key;
struct ctdb_req_control request;
TDB_DATA rec;
int ret;
req = tevent_req_create(mem_ctx, &state,
struct ctdb_delete_record_state);
if (req == NULL) {
return NULL;
}
state->h = h;
/* Cannot delete the record if it was obtained as a readonly copy */
if (h->readonly) {
tevent_req_error(req, EINVAL);
return tevent_req_post(req, ev);
}
rec.dsize = ctdb_ltdb_header_len(&h->header);
rec.dptr = talloc_size(h, rec.dsize);
if (tevent_req_nomem(rec.dptr, req)) {
return tevent_req_post(req, ev);
}
ctdb_ltdb_header_push(&h->header, rec.dptr);
ret = tdb_store(h->db->ltdb->tdb, h->key, rec, TDB_REPLACE);
talloc_free(rec.dptr);
if (ret != 0) {
DEBUG(DEBUG_ERR, ("Failed to delete record in DB %s\n",
h->db->db_name));
tevent_req_error(req, EIO);
return tevent_req_post(req, ev);
}
key.db_id = h->db->db_id;
key.header = h->header;
key.key = h->key;
ctdb_req_control_schedule_for_deletion(&request, &key);
subreq = ctdb_client_control_send(state, ev, h->client,
ctdb_client_pnn(h->client),
tevent_timeval_zero(),
&request);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, ctdb_delete_record_done, req);
return req;
}
static void ctdb_delete_record_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_delete_record_state *state = tevent_req_data(
req, struct ctdb_delete_record_state);
int ret;
bool status;
status = ctdb_client_control_recv(subreq, &ret, NULL, NULL);
TALLOC_FREE(subreq);
if (! status) {
DEBUG(DEBUG_ERR,
("delete_record: %s SCHDULE_FOR_DELETION failed, "
"ret=%d\n", state->h->db->db_name, ret));
tevent_req_error(req, ret);
return;
}
tevent_req_done(req);
}
bool ctdb_delete_record_recv(struct tevent_req *req, int *perr)
{
int err;
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return false;
}
return true;
}
int ctdb_delete_record(struct ctdb_record_handle *h)
{
struct tevent_context *ev = h->ev;
TALLOC_CTX *mem_ctx;
struct tevent_req *req;
int ret;
bool status;
mem_ctx = talloc_new(NULL);
if (mem_ctx == NULL) {
return ENOMEM;
}
req = ctdb_delete_record_send(mem_ctx, ev, h);
if (req == NULL) {
talloc_free(mem_ctx);
return ENOMEM;
}
tevent_req_poll(req, ev);
status = ctdb_delete_record_recv(req, &ret);
talloc_free(mem_ctx);
if (! status) {
return ret;
}
return 0;
}
/*
* Global lock functions
*/
struct ctdb_g_lock_lock_state {
struct tevent_context *ev;
struct ctdb_client_context *client;
struct ctdb_db_context *db;
TDB_DATA key;
struct ctdb_server_id my_sid;
enum ctdb_g_lock_type lock_type;
struct ctdb_record_handle *h;
/* state for verification of active locks */
struct ctdb_g_lock_list *lock_list;
unsigned int current;
};
static void ctdb_g_lock_lock_fetched(struct tevent_req *subreq);
static void ctdb_g_lock_lock_process_locks(struct tevent_req *req);
static void ctdb_g_lock_lock_checked(struct tevent_req *subreq);
static int ctdb_g_lock_lock_update(struct tevent_req *req);
static void ctdb_g_lock_lock_retry(struct tevent_req *subreq);
static bool ctdb_g_lock_conflicts(enum ctdb_g_lock_type l1,
enum ctdb_g_lock_type l2)
{
if ((l1 == CTDB_G_LOCK_READ) && (l2 == CTDB_G_LOCK_READ)) {
return false;
}
return true;
}
struct tevent_req *ctdb_g_lock_lock_send(TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_client_context *client,
struct ctdb_db_context *db,
const char *keyname,
struct ctdb_server_id *sid,
bool readonly)
{
struct tevent_req *req, *subreq;
struct ctdb_g_lock_lock_state *state;
req = tevent_req_create(mem_ctx, &state,
struct ctdb_g_lock_lock_state);
if (req == NULL) {
return NULL;
}
state->ev = ev;
state->client = client;
state->db = db;
state->key.dptr = discard_const(keyname);
state->key.dsize = strlen(keyname) + 1;
state->my_sid = *sid;
state->lock_type = (readonly ? CTDB_G_LOCK_READ : CTDB_G_LOCK_WRITE);
subreq = ctdb_fetch_lock_send(state, ev, client, db, state->key,
false);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, ctdb_g_lock_lock_fetched, req);
return req;
}
static void ctdb_g_lock_lock_fetched(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_g_lock_lock_state *state = tevent_req_data(
req, struct ctdb_g_lock_lock_state);
TDB_DATA data;
int ret = 0;
state->h = ctdb_fetch_lock_recv(subreq, NULL, state, &data, &ret);
TALLOC_FREE(subreq);
if (state->h == NULL) {
tevent_req_error(req, ret);
return;
}
if (state->lock_list != NULL) {
TALLOC_FREE(state->lock_list);
state->current = 0;
}
ret = ctdb_g_lock_list_pull(data.dptr, data.dsize, state,
&state->lock_list);
talloc_free(data.dptr);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
ctdb_g_lock_lock_process_locks(req);
}
static void ctdb_g_lock_lock_process_locks(struct tevent_req *req)
{
struct ctdb_g_lock_lock_state *state = tevent_req_data(
req, struct ctdb_g_lock_lock_state);
struct tevent_req *subreq;
struct ctdb_g_lock *lock;
bool check_server = false;
int ret;
while (state->current < state->lock_list->num) {
lock = &state->lock_list->lock[state->current];
/* We should not ask for the same lock more than once */
if (ctdb_server_id_equal(&lock->sid, &state->my_sid)) {
tevent_req_error(req, EDEADLK);
return;
}
if (ctdb_g_lock_conflicts(lock->type, state->lock_type)) {
check_server = true;
break;
}
state->current += 1;
}
if (check_server) {
struct ctdb_req_control request;
ctdb_req_control_process_exists(&request, lock->sid.pid);
subreq = ctdb_client_control_send(state, state->ev,
state->client,
lock->sid.vnn,
tevent_timeval_zero(),
&request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_g_lock_lock_checked, req);
return;
}
/* There is no conflict, add ourself to the lock_list */
state->lock_list->lock = talloc_realloc(state->lock_list,
state->lock_list->lock,
struct ctdb_g_lock,
state->lock_list->num + 1);
if (state->lock_list->lock == NULL) {
tevent_req_error(req, ENOMEM);
return;
}
lock = &state->lock_list->lock[state->lock_list->num];
lock->type = state->lock_type;
lock->sid = state->my_sid;
state->lock_list->num += 1;
ret = ctdb_g_lock_lock_update(req);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
TALLOC_FREE(state->h);
tevent_req_done(req);
}
static void ctdb_g_lock_lock_checked(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_g_lock_lock_state *state = tevent_req_data(
req, struct ctdb_g_lock_lock_state);
struct ctdb_reply_control *reply;
int ret, value;
bool status;
status = ctdb_client_control_recv(subreq, &ret, state, &reply);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_reply_control_process_exists(reply, &value);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
talloc_free(reply);
if (value == 0) {
/* server process exists, need to retry */
TALLOC_FREE(state->h);
subreq = tevent_wakeup_send(state, state->ev,
tevent_timeval_current_ofs(0,1000));
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_g_lock_lock_retry, req);
return;
}
/* server process does not exist, remove conflicting entry */
state->lock_list->lock[state->current] =
state->lock_list->lock[state->lock_list->num-1];
state->lock_list->num -= 1;
ret = ctdb_g_lock_lock_update(req);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
ctdb_g_lock_lock_process_locks(req);
}
static int ctdb_g_lock_lock_update(struct tevent_req *req)
{
struct ctdb_g_lock_lock_state *state = tevent_req_data(
req, struct ctdb_g_lock_lock_state);
TDB_DATA data;
int ret;
data.dsize = ctdb_g_lock_list_len(state->lock_list);
data.dptr = talloc_size(state, data.dsize);
if (data.dptr == NULL) {
return ENOMEM;
}
ctdb_g_lock_list_push(state->lock_list, data.dptr);
ret = ctdb_store_record(state->h, data);
talloc_free(data.dptr);
return ret;
}
static void ctdb_g_lock_lock_retry(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_g_lock_lock_state *state = tevent_req_data(
req, struct ctdb_g_lock_lock_state);
bool success;
success = tevent_wakeup_recv(subreq);
TALLOC_FREE(subreq);
if (! success) {
tevent_req_error(req, ENOMEM);
return;
}
subreq = ctdb_fetch_lock_send(state, state->ev, state->client,
state->db, state->key, false);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_g_lock_lock_fetched, req);
}
bool ctdb_g_lock_lock_recv(struct tevent_req *req, int *perr)
{
struct ctdb_g_lock_lock_state *state = tevent_req_data(
req, struct ctdb_g_lock_lock_state);
int err;
TALLOC_FREE(state->h);
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return false;
}
return true;
}
struct ctdb_g_lock_unlock_state {
struct tevent_context *ev;
struct ctdb_client_context *client;
struct ctdb_db_context *db;
TDB_DATA key;
struct ctdb_server_id my_sid;
struct ctdb_record_handle *h;
struct ctdb_g_lock_list *lock_list;
};
static void ctdb_g_lock_unlock_fetched(struct tevent_req *subreq);
static int ctdb_g_lock_unlock_update(struct tevent_req *req);
static void ctdb_g_lock_unlock_deleted(struct tevent_req *subreq);
struct tevent_req *ctdb_g_lock_unlock_send(TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_client_context *client,
struct ctdb_db_context *db,
const char *keyname,
struct ctdb_server_id sid)
{
struct tevent_req *req, *subreq;
struct ctdb_g_lock_unlock_state *state;
req = tevent_req_create(mem_ctx, &state,
struct ctdb_g_lock_unlock_state);
if (req == NULL) {
return NULL;
}
state->ev = ev;
state->client = client;
state->db = db;
state->key.dptr = discard_const(keyname);
state->key.dsize = strlen(keyname) + 1;
state->my_sid = sid;
subreq = ctdb_fetch_lock_send(state, ev, client, db, state->key,
false);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, ctdb_g_lock_unlock_fetched, req);
return req;
}
static void ctdb_g_lock_unlock_fetched(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_g_lock_unlock_state *state = tevent_req_data(
req, struct ctdb_g_lock_unlock_state);
TDB_DATA data;
int ret = 0;
state->h = ctdb_fetch_lock_recv(subreq, NULL, state, &data, &ret);
TALLOC_FREE(subreq);
if (state->h == NULL) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_g_lock_list_pull(data.dptr, data.dsize, state,
&state->lock_list);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_g_lock_unlock_update(req);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
if (state->lock_list->num == 0) {
subreq = ctdb_delete_record_send(state, state->ev, state->h);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_g_lock_unlock_deleted,
req);
return;
}
TALLOC_FREE(state->h);
tevent_req_done(req);
}
static int ctdb_g_lock_unlock_update(struct tevent_req *req)
{
struct ctdb_g_lock_unlock_state *state = tevent_req_data(
req, struct ctdb_g_lock_unlock_state);
struct ctdb_g_lock *lock;
int ret, i;
for (i=0; i<state->lock_list->num; i++) {
lock = &state->lock_list->lock[i];
if (ctdb_server_id_equal(&lock->sid, &state->my_sid)) {
break;
}
}
if (i < state->lock_list->num) {
state->lock_list->lock[i] =
state->lock_list->lock[state->lock_list->num-1];
state->lock_list->num -= 1;
}
if (state->lock_list->num != 0) {
TDB_DATA data;
data.dsize = ctdb_g_lock_list_len(state->lock_list);
data.dptr = talloc_size(state, data.dsize);
if (data.dptr == NULL) {
return ENOMEM;
}
ctdb_g_lock_list_push(state->lock_list, data.dptr);
ret = ctdb_store_record(state->h, data);
talloc_free(data.dptr);
if (ret != 0) {
return ret;
}
}
return 0;
}
static void ctdb_g_lock_unlock_deleted(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_g_lock_unlock_state *state = tevent_req_data(
req, struct ctdb_g_lock_unlock_state);
int ret;
bool status;
status = ctdb_delete_record_recv(subreq, &ret);
if (! status) {
DEBUG(DEBUG_ERR,
("g_lock_unlock %s delete record failed, ret=%d\n",
(char *)state->key.dptr, ret));
tevent_req_error(req, ret);
return;
}
TALLOC_FREE(state->h);
tevent_req_done(req);
}
bool ctdb_g_lock_unlock_recv(struct tevent_req *req, int *perr)
{
struct ctdb_g_lock_unlock_state *state = tevent_req_data(
req, struct ctdb_g_lock_unlock_state);
int err;
TALLOC_FREE(state->h);
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return false;
}
return true;
}
/*
* Persistent database functions
*/
struct ctdb_transaction_start_state {
struct tevent_context *ev;
struct ctdb_client_context *client;
struct timeval timeout;
struct ctdb_transaction_handle *h;
uint32_t destnode;
};
static void ctdb_transaction_g_lock_attached(struct tevent_req *subreq);
static void ctdb_transaction_g_lock_done(struct tevent_req *subreq);
struct tevent_req *ctdb_transaction_start_send(TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_client_context *client,
struct timeval timeout,
struct ctdb_db_context *db,
bool readonly)
{
struct ctdb_transaction_start_state *state;
struct tevent_req *req, *subreq;
struct ctdb_transaction_handle *h;
req = tevent_req_create(mem_ctx, &state,
struct ctdb_transaction_start_state);
if (req == NULL) {
return NULL;
}
if (! db->persistent) {
tevent_req_error(req, EINVAL);
return tevent_req_post(req, ev);
}
state->ev = ev;
state->client = client;
state->destnode = ctdb_client_pnn(client);
h = talloc_zero(db, struct ctdb_transaction_handle);
if (tevent_req_nomem(h, req)) {
return tevent_req_post(req, ev);
}
h->ev = ev;
h->client = client;
h->db = db;
h->readonly = readonly;
h->updated = false;
/* SRVID is unique for databases, so client can have transactions
* active for multiple databases */
h->sid = ctdb_client_get_server_id(client, db->db_id);
h->recbuf = ctdb_rec_buffer_init(h, db->db_id);
if (tevent_req_nomem(h->recbuf, req)) {
return tevent_req_post(req, ev);
}
h->lock_name = talloc_asprintf(h, "transaction_db_0x%08x", db->db_id);
if (tevent_req_nomem(h->lock_name, req)) {
return tevent_req_post(req, ev);
}
state->h = h;
subreq = ctdb_attach_send(state, ev, client, timeout, "g_lock.tdb", 0);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, ctdb_transaction_g_lock_attached, req);
return req;
}
static void ctdb_transaction_g_lock_attached(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_transaction_start_state *state = tevent_req_data(
req, struct ctdb_transaction_start_state);
bool status;
int ret;
status = ctdb_attach_recv(subreq, &ret, &state->h->db_g_lock);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
subreq = ctdb_g_lock_lock_send(state, state->ev, state->client,
state->h->db_g_lock,
state->h->lock_name,
&state->h->sid, state->h->readonly);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_transaction_g_lock_done, req);
}
static void ctdb_transaction_g_lock_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
int ret;
bool status;
status = ctdb_g_lock_lock_recv(subreq, &ret);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
tevent_req_done(req);
}
struct ctdb_transaction_handle *ctdb_transaction_start_recv(
struct tevent_req *req,
int *perr)
{
struct ctdb_transaction_start_state *state = tevent_req_data(
req, struct ctdb_transaction_start_state);
struct ctdb_transaction_handle *h = state->h;
int err;
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return NULL;
}
return h;
}
int ctdb_transaction_start(TALLOC_CTX *mem_ctx, struct tevent_context *ev,
struct ctdb_client_context *client,
struct timeval timeout,
struct ctdb_db_context *db, bool readonly,
struct ctdb_transaction_handle **out)
{
struct tevent_req *req;
struct ctdb_transaction_handle *h;
int ret;
req = ctdb_transaction_start_send(mem_ctx, ev, client, timeout, db,
readonly);
if (req == NULL) {
return ENOMEM;
}
tevent_req_poll(req, ev);
h = ctdb_transaction_start_recv(req, &ret);
if (h == NULL) {
return ret;
}
*out = h;
return 0;
}
struct ctdb_transaction_record_fetch_state {
TDB_DATA key, data;
struct ctdb_ltdb_header header;
bool found;
};
static int ctdb_transaction_record_fetch_traverse(
uint32_t reqid,
struct ctdb_ltdb_header *nullheader,
TDB_DATA key, TDB_DATA data,
void *private_data)
{
struct ctdb_transaction_record_fetch_state *state =
(struct ctdb_transaction_record_fetch_state *)private_data;
if (state->key.dsize == key.dsize &&
memcmp(state->key.dptr, key.dptr, key.dsize) == 0) {
int ret;
ret = ctdb_ltdb_header_extract(&data, &state->header);
if (ret != 0) {
DEBUG(DEBUG_ERR, ("Failed to extract header\n"));
return 1;
}
state->data = data;
state->found = true;
}
return 0;
}
static int ctdb_transaction_record_fetch(struct ctdb_transaction_handle *h,
TDB_DATA key,
struct ctdb_ltdb_header *header,
TDB_DATA *data)
{
struct ctdb_transaction_record_fetch_state state;
int ret;
state.key = key;
state.found = false;
ret = ctdb_rec_buffer_traverse(h->recbuf,
ctdb_transaction_record_fetch_traverse,
&state);
if (ret != 0) {
return ret;
}
if (state.found) {
if (header != NULL) {
*header = state.header;
}
if (data != NULL) {
*data = state.data;
}
return 0;
}
return ENOENT;
}
int ctdb_transaction_fetch_record(struct ctdb_transaction_handle *h,
TDB_DATA key,
TALLOC_CTX *mem_ctx, TDB_DATA *data)
{
TDB_DATA tmp_data;
struct ctdb_ltdb_header header;
int ret;
ret = ctdb_transaction_record_fetch(h, key, NULL, &tmp_data);
if (ret == 0) {
data->dptr = talloc_memdup(mem_ctx, tmp_data.dptr,
tmp_data.dsize);
if (data->dptr == NULL) {
return ENOMEM;
}
data->dsize = tmp_data.dsize;
return 0;
}
ret = ctdb_ltdb_fetch(h->db, key, &header, mem_ctx, data);
if (ret != 0) {
return ret;
}
ret = ctdb_rec_buffer_add(h, h->recbuf, 0, &header, key, *data);
if (ret != 0) {
return ret;
}
return 0;
}
int ctdb_transaction_store_record(struct ctdb_transaction_handle *h,
TDB_DATA key, TDB_DATA data)
{
TALLOC_CTX *tmp_ctx;
struct ctdb_ltdb_header header;
TDB_DATA old_data;
int ret;
if (h->readonly) {
return EINVAL;
}
tmp_ctx = talloc_new(h);
if (tmp_ctx == NULL) {
return ENOMEM;
}
ret = ctdb_transaction_record_fetch(h, key, &header, &old_data);
if (ret != 0) {
ret = ctdb_ltdb_fetch(h->db, key, &header, tmp_ctx, &old_data);
if (ret != 0) {
return ret;
}
}
if (old_data.dsize == data.dsize &&
memcmp(old_data.dptr, data.dptr, data.dsize) == 0) {
talloc_free(tmp_ctx);
return 0;
}
header.dmaster = ctdb_client_pnn(h->client);
header.rsn += 1;
ret = ctdb_rec_buffer_add(h, h->recbuf, 0, &header, key, data);
talloc_free(tmp_ctx);
if (ret != 0) {
return ret;
}
h->updated = true;
return 0;
}
int ctdb_transaction_delete_record(struct ctdb_transaction_handle *h,
TDB_DATA key)
{
return ctdb_transaction_store_record(h, key, tdb_null);
}
static int ctdb_transaction_fetch_db_seqnum(struct ctdb_transaction_handle *h,
uint64_t *seqnum)
{
const char *keyname = CTDB_DB_SEQNUM_KEY;
TDB_DATA key, data;
struct ctdb_ltdb_header header;
int ret;
key.dptr = discard_const(keyname);
key.dsize = strlen(keyname) + 1;
ret = ctdb_ltdb_fetch(h->db, key, &header, h, &data);
if (ret != 0) {
DEBUG(DEBUG_ERR,
("transaction_commit: %s seqnum fetch failed, ret=%d\n",
h->db->db_name, ret));
return ret;
}
if (data.dsize == 0) {
/* initial data */
*seqnum = 0;
return 0;
}
if (data.dsize != sizeof(uint64_t)) {
talloc_free(data.dptr);
return EINVAL;
}
*seqnum = *(uint64_t *)data.dptr;
talloc_free(data.dptr);
return 0;
}
static int ctdb_transaction_store_db_seqnum(struct ctdb_transaction_handle *h,
uint64_t seqnum)
{
const char *keyname = CTDB_DB_SEQNUM_KEY;
TDB_DATA key, data;
key.dptr = discard_const(keyname);
key.dsize = strlen(keyname) + 1;
data.dptr = (uint8_t *)&seqnum;
data.dsize = sizeof(seqnum);
return ctdb_transaction_store_record(h, key, data);
}
struct ctdb_transaction_commit_state {
struct tevent_context *ev;
struct ctdb_transaction_handle *h;
uint64_t seqnum;
};
static void ctdb_transaction_commit_try(struct tevent_req *subreq);
static void ctdb_transaction_commit_done(struct tevent_req *subreq);
struct tevent_req *ctdb_transaction_commit_send(
TALLOC_CTX *mem_ctx,
struct tevent_context *ev,
struct ctdb_transaction_handle *h)
{
struct tevent_req *req, *subreq;
struct ctdb_transaction_commit_state *state;
int ret;
req = tevent_req_create(mem_ctx, &state,
struct ctdb_transaction_commit_state);
if (req == NULL) {
return NULL;
}
state->ev = ev;
state->h = h;
ret = ctdb_transaction_fetch_db_seqnum(h, &state->seqnum);
if (ret != 0) {
tevent_req_error(req, ret);
return tevent_req_post(req, ev);
}
ret = ctdb_transaction_store_db_seqnum(h, state->seqnum+1);
if (ret != 0) {
tevent_req_error(req, ret);
return tevent_req_post(req, ev);
}
subreq = ctdb_recovery_wait_send(state, ev, h->client);
if (tevent_req_nomem(subreq, req)) {
return tevent_req_post(req, ev);
}
tevent_req_set_callback(subreq, ctdb_transaction_commit_try, req);
return req;
}
static void ctdb_transaction_commit_try(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_transaction_commit_state *state = tevent_req_data(
req, struct ctdb_transaction_commit_state);
struct ctdb_req_control request;
int ret;
bool status;
status = ctdb_recovery_wait_recv(subreq, &ret);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
ctdb_req_control_trans3_commit(&request, state->h->recbuf);
subreq = ctdb_client_control_send(state, state->ev, state->h->client,
state->h->client->pnn,
tevent_timeval_zero(), &request);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_transaction_commit_done, req);
}
static void ctdb_transaction_commit_done(struct tevent_req *subreq)
{
struct tevent_req *req = tevent_req_callback_data(
subreq, struct tevent_req);
struct ctdb_transaction_commit_state *state = tevent_req_data(
req, struct ctdb_transaction_commit_state);
struct ctdb_transaction_handle *h = state->h;
struct ctdb_reply_control *reply;
uint64_t seqnum;
int ret;
bool status;
status = ctdb_client_control_recv(subreq, &ret, state, &reply);
TALLOC_FREE(subreq);
if (! status) {
tevent_req_error(req, ret);
return;
}
ret = ctdb_reply_control_trans3_commit(reply);
if (ret < 0) {
/* Control failed due to recovery */
subreq = ctdb_recovery_wait_send(state, state->ev, h->client);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_transaction_commit_try,
req);
return;
}
ret = ctdb_transaction_fetch_db_seqnum(h, &seqnum);
if (ret != 0) {
tevent_req_error(req, ret);
return;
}
if (seqnum == state->seqnum) {
/* try again */
subreq = ctdb_recovery_wait_send(state, state->ev, h->client);
if (tevent_req_nomem(subreq, req)) {
return;
}
tevent_req_set_callback(subreq, ctdb_transaction_commit_try,
req);
return;
}
if (seqnum != state->seqnum + 1) {
tevent_req_error(req, EIO);
return;
}
tevent_req_done(req);
}
bool ctdb_transaction_commit_recv(struct tevent_req *req, int *perr)
{
int err;
if (tevent_req_is_unix_error(req, &err)) {
if (perr != NULL) {
*perr = err;
}
return false;
}
return true;
}
int ctdb_transaction_commit(struct ctdb_transaction_handle *h)
{
struct tevent_req *req;
int ret;
bool status;
if (h->readonly || ! h->updated) {
talloc_free(h);
return 0;
}
req = ctdb_transaction_commit_send(h, h->ev, h);
if (req == NULL) {
talloc_free(h);
return ENOMEM;
}
tevent_req_poll(req, h->ev);
status = ctdb_transaction_commit_recv(req, &ret);
if (! status) {
talloc_free(h);
return ret;
}
talloc_free(h);
return 0;
}
int ctdb_transaction_cancel(struct ctdb_transaction_handle *h)
{
talloc_free(h);
return 0;
}
/*
* TODO:
*
* In future Samba should register SERVER_ID.
* Make that structure same as struct srvid {}.
*/