protocol design and packet details
*/
#include "includes.h"
-#include "lib/events/events.h"
+#include "lib/tevent/tevent.h"
#include "lib/tdb/include/tdb.h"
#include "lib/util/dlinklist.h"
#include "system/network.h"
return ctdb_db;
}
-
/*
a varient of input packet that can be used in lock requeue
*/
int msglen, len;
if (ctdb->methods == NULL) {
- DEBUG(DEBUG_ERR,(__location__ " Failed to send error. Transport is DOWN\n"));
+ DEBUG(DEBUG_INFO,(__location__ " Failed to send error. Transport is DOWN\n"));
return;
}
}
-/*
- send a redirect reply
-*/
+/**
+ * send a redirect reply
+ *
+ * The logic behind this function is this:
+ *
+ * A client wants to grab a record and sends a CTDB_REQ_CALL packet
+ * to its local ctdb (ctdb_request_call). If the node is not itself
+ * the record's DMASTER, it first redirects the packet to the
+ * record's LMASTER. The LMASTER then redirects the call packet to
+ * the current DMASTER. But there is a race: The record may have
+ * been migrated off the DMASTER while the redirected packet is
+ * on the wire (or in the local queue). So in case the record has
+ * migrated off the new destinaton of the call packet, instead of
+ * going back to the LMASTER to get the new DMASTER, we try to
+ * reduce rountrips by fist chasing the record a couple of times
+ * before giving up the direct chase and finally going back to the
+ * LMASTER (again). Note that this works because auf this: When
+ * a record is migrated off a node, then the new DMASTER is stored
+ * in the record's copy on the former DMASTER.
+ *
+ * The maxiumum number of attempts for direct chase to make before
+ * going back to the LMASTER is configurable by the tunable
+ * "MaxRedirectCount".
+ */
static void ctdb_call_send_redirect(struct ctdb_context *ctdb,
TDB_DATA key,
struct ctdb_req_call *c,
}
if (ctdb->methods == NULL) {
- ctdb_fatal(ctdb, "ctdb_send_dmaster_reply cant update dmaster sicne transport is down");
+ ctdb_fatal(ctdb, "ctdb_send_dmaster_reply cant update dmaster since transport is down");
return;
}
tmp_ctx = talloc_new(ctdb);
/* send the CTDB_REPLY_DMASTER */
- len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize;
+ len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize + sizeof(uint32_t);
r = ctdb_transport_allocate(ctdb, tmp_ctx, CTDB_REPLY_DMASTER, len,
struct ctdb_reply_dmaster);
CTDB_NO_MEMORY_FATAL(ctdb, r);
r->db_id = ctdb_db->db_id;
memcpy(&r->data[0], key.dptr, key.dsize);
memcpy(&r->data[key.dsize], data.dptr, data.dsize);
+ memcpy(&r->data[key.dsize+data.dsize], &header->flags, sizeof(uint32_t));
ctdb_queue_packet(ctdb, &r->hdr);
return;
}
+ if (data->dsize != 0) {
+ header->flags |= CTDB_REC_FLAG_MIGRATED_WITH_DATA;
+ }
+
if (lmaster == ctdb->pnn) {
ctdb_send_dmaster_reply(ctdb_db, header, *key, *data,
c->hdr.srcnode, c->hdr.reqid);
return;
}
- len = offsetof(struct ctdb_req_dmaster, data) + key->dsize + data->dsize;
+ len = offsetof(struct ctdb_req_dmaster, data) + key->dsize + data->dsize
+ + sizeof(uint32_t);
r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REQ_DMASTER, len,
struct ctdb_req_dmaster);
CTDB_NO_MEMORY_FATAL(ctdb, r);
r->datalen = data->dsize;
memcpy(&r->data[0], key->dptr, key->dsize);
memcpy(&r->data[key->dsize], data->dptr, data->dsize);
+ memcpy(&r->data[key->dsize + data->dsize], &header->flags, sizeof(uint32_t));
header->dmaster = c->hdr.srcnode;
if (ctdb_ltdb_store(ctdb_db, *key, header, *data) != 0) {
talloc_free(r);
}
-static void ctdb_hold_back_key(struct ctdb_db_context *db, TDB_DATA key)
-{
- size_t num_keys;
- uint8_t **tmp;
-
- DEBUG(DEBUG_INFO, ("Holding back key %08x\n", ctdb_hash(&key)));
-
- num_keys = talloc_array_length(db->holdback_keys);
- tmp = talloc_realloc(db, db->holdback_keys, uint8_t *, num_keys+1);
- if (tmp == NULL) {
- DEBUG(DEBUG_ERR, ("talloc_realloc failed\n"));
- return;
- }
- db->holdback_keys = tmp;
-
- db->holdback_keys[num_keys] = (uint8_t *)talloc_memdup(
- db->holdback_keys, key.dptr, key.dsize);
- if (db->holdback_keys[num_keys] == NULL) {
- DEBUG(DEBUG_ERR, ("talloc_memdup failed\n"));
- db->holdback_keys = talloc_realloc(db, db->holdback_keys,
- uint8_t *, num_keys);
- }
-}
-
/*
called when a CTDB_REPLY_DMASTER packet comes in, or when the lmaster
gets a CTDB_REQUEST_DMASTER for itself. We become the dmaster.
must be called with the chainlock held. This function releases the chainlock
*/
-static void ctdb_become_dmaster(struct ctdb_db_context *ctdb_db,
+static void ctdb_become_dmaster(struct ctdb_db_context *ctdb_db,
struct ctdb_req_header *hdr,
TDB_DATA key, TDB_DATA data,
- uint64_t rsn)
+ uint64_t rsn, uint32_t record_flags)
{
struct ctdb_call_state *state;
struct ctdb_context *ctdb = ctdb_db->ctdb;
struct ctdb_ltdb_header header;
+ int ret;
- DEBUG(DEBUG_INFO,("pnn %u dmaster response %08x\n", ctdb->pnn, ctdb_hash(&key)));
+ DEBUG(DEBUG_DEBUG,("pnn %u dmaster response %08x\n", ctdb->pnn, ctdb_hash(&key)));
ZERO_STRUCT(header);
header.rsn = rsn + 1;
header.dmaster = ctdb->pnn;
+ header.flags = record_flags;
+
+ state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_call_state);
+
+ if (state) {
+ if (state->call->flags & CTDB_CALL_FLAG_VACUUM_MIGRATION) {
+ /*
+ * We temporarily add the VACUUM_MIGRATED flag to
+ * the record flags, so that ctdb_ltdb_store can
+ * decide whether the record should be stored or
+ * deleted.
+ */
+ header.flags |= CTDB_REC_FLAG_VACUUM_MIGRATED;
+ }
+ }
if (ctdb_ltdb_store(ctdb_db, key, &header, data) != 0) {
ctdb_fatal(ctdb, "ctdb_reply_dmaster store failed\n");
- ctdb_ltdb_unlock(ctdb_db, key);
+
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
return;
}
- state = ctdb_reqid_find(ctdb, hdr->reqid, struct ctdb_call_state);
if (state == NULL) {
DEBUG(DEBUG_ERR,("pnn %u Invalid reqid %u in ctdb_become_dmaster from node %u\n",
ctdb->pnn, hdr->reqid, hdr->srcnode));
- ctdb_ltdb_unlock(ctdb_db, key);
+
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
+ return;
+ }
+
+ if (key.dsize != state->call->key.dsize || memcmp(key.dptr, state->call->key.dptr, key.dsize)) {
+ DEBUG(DEBUG_ERR, ("Got bogus DMASTER packet reqid:%u from node %u. Key does not match key held in matching idr.\n", hdr->reqid, hdr->srcnode));
+
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
return;
}
if (hdr->reqid != state->reqid) {
/* we found a record but it was the wrong one */
DEBUG(DEBUG_ERR, ("Dropped orphan in ctdb_become_dmaster with reqid:%u\n from node %u", hdr->reqid, hdr->srcnode));
- ctdb_ltdb_unlock(ctdb_db, key);
+
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
return;
}
- ctdb_hold_back_key(ctdb_db, key);
+ ctdb_call_local(ctdb_db, state->call, &header, state, &data, true);
- ctdb_call_local(ctdb_db, state->call, &header, state, &data, ctdb->pnn);
-
- ctdb_ltdb_unlock(ctdb_db, state->call->key);
+ ret = ctdb_ltdb_unlock(ctdb_db, state->call->key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
state->state = CTDB_CALL_DONE;
if (state->async.fn) {
TDB_DATA key, data, data2;
struct ctdb_ltdb_header header;
struct ctdb_db_context *ctdb_db;
+ uint32_t record_flags = 0;
+ size_t len;
int ret;
key.dptr = c->data;
key.dsize = c->keylen;
data.dptr = c->data + c->keylen;
data.dsize = c->datalen;
+ len = offsetof(struct ctdb_req_dmaster, data) + key.dsize + data.dsize
+ + sizeof(uint32_t);
+ if (len <= c->hdr.length) {
+ record_flags = *(uint32_t *)&c->data[c->keylen + c->datalen];
+ }
ctdb_db = find_ctdb_db(ctdb, c->db_id);
if (!ctdb_db) {
ctdb_fatal(ctdb, "ctdb_req_dmaster to non-lmaster");
}
- DEBUG(DEBUG_INFO,("pnn %u dmaster request on %08x for %u from %u\n",
+ DEBUG(DEBUG_DEBUG,("pnn %u dmaster request on %08x for %u from %u\n",
ctdb->pnn, ctdb_hash(&key), c->dmaster, c->hdr.srcnode));
/* its a protocol error if the sending node is not the current dmaster */
(unsigned long long)c->rsn, (unsigned long long)header.rsn, c->hdr.reqid,
(key.dsize >= 4)?(*(uint32_t *)key.dptr):0));
if (header.rsn != 0 || header.dmaster != ctdb->pnn) {
- ctdb_fatal(ctdb, "ctdb_req_dmaster from non-master");
+ DEBUG(DEBUG_ERR,("ctdb_req_dmaster from non-master. Force a recovery.\n"));
+
+ ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
+ ctdb_ltdb_unlock(ctdb_db, key);
return;
}
}
/* use the rsn from the sending node */
header.rsn = c->rsn;
+ /* store the record flags from the sending node */
+ header.flags = record_flags;
+
/* check if the new dmaster is the lmaster, in which case we
skip the dmaster reply */
if (c->dmaster == ctdb->pnn) {
- ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn);
+ ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn, record_flags);
} else {
ctdb_send_dmaster_reply(ctdb_db, &header, key, data, c->dmaster, hdr->reqid);
- ctdb_ltdb_unlock(ctdb_db, key);
- }
-}
-
-/*
- * Did we just pull the dmaster of the record for a client fetch_lock,
- * so should we hold it back a while and thus give our client the
- * chance to do its own tdb_lock?
- */
-
-static bool ctdb_held_back(struct ctdb_db_context *db, TDB_DATA key,
- struct ctdb_req_header *hdr)
-{
- int i;
- size_t num_keys = talloc_array_length(db->holdback_keys);
- size_t num_hdrs;
- struct ctdb_req_header **tmp;
- for (i=0; i<num_keys; i++) {
- uint8_t *hold_key = db->holdback_keys[i];
- size_t keylength = talloc_array_length(hold_key);
-
- if ((keylength == key.dsize)
- && (memcmp(hold_key, key.dptr, keylength) == 0)) {
- break;
+ ret = ctdb_ltdb_unlock(ctdb_db, key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
}
}
- if (i == num_keys) {
- return false;
- }
- DEBUG(DEBUG_DEBUG, ("holding back record %08x after migration\n",
- ctdb_hash(&key)));
-
- num_hdrs = talloc_array_length(db->held_back);
-
- tmp = talloc_realloc(db, db->held_back, struct ctdb_req_header *,
- num_hdrs + 1);
- if (tmp == NULL) {
- DEBUG(DEBUG_ERR, (__location__ "talloc_realloc failed\n"));
- return false;
- }
- db->held_back = tmp;
- db->held_back[num_hdrs] = talloc_move(db->held_back, &hdr);
- return true;
}
+
/*
called when a CTDB_REQ_CALL packet comes in
*/
struct ctdb_db_context *ctdb_db;
if (ctdb->methods == NULL) {
- DEBUG(DEBUG_ERR,(__location__ " Failed ctdb_request_call. Transport is DOWN\n"));
+ DEBUG(DEBUG_INFO,(__location__ " Failed ctdb_request_call. Transport is DOWN\n"));
return;
}
call->key.dsize = c->keylen;
call->call_data.dptr = c->data + c->keylen;
call->call_data.dsize = c->calldatalen;
+ call->reply_data.dptr = NULL;
+ call->reply_data.dsize = 0;
/* determine if we are the dmaster for this key. This also
fetches the record data (if any), thus avoiding a 2nd fetch of the data
return;
}
- /* if we are not the dmaster, then send a redirect to the
- requesting node */
- if (header.dmaster != ctdb->pnn) {
+ /* Dont do READONLY if we dont have a tracking database */
+ if ((c->flags & CTDB_WANT_READONLY) && !ctdb_db->readonly) {
+ c->flags &= ~CTDB_WANT_READONLY;
+ }
+
+ if (header.flags & CTDB_REC_RO_REVOKE_COMPLETE) {
+ header.flags &= ~(CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY|CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_REVOKE_COMPLETE);
+ if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
+ ctdb_fatal(ctdb, "Failed to write header with cleared REVOKE flag");
+ }
+ /* and clear out the tracking data */
+ if (tdb_delete(ctdb_db->rottdb, call->key) != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to clear out trackingdb record\n"));
+ }
+ }
+
+ /* if we are revoking, we must defer all other calls until the revoke
+ * had completed.
+ */
+ if (header.flags & CTDB_REC_RO_REVOKING_READONLY) {
talloc_free(data.dptr);
- ctdb_call_send_redirect(ctdb, call->key, c, &header);
- ctdb_ltdb_unlock(ctdb_db, call->key);
+ ret = ctdb_ltdb_unlock(ctdb_db, call->key);
+
+ if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
+ ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
+ }
+ talloc_free(call);
return;
}
- if (c->hopcount > ctdb->statistics.max_hop_count) {
- ctdb->statistics.max_hop_count = c->hopcount;
+ /* if we are not the dmaster and are not hosting any delegations,
+ then send a redirect to the requesting node */
+ if ((header.dmaster != ctdb->pnn)
+ && (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS)) ) {
+ talloc_free(data.dptr);
+ ctdb_call_send_redirect(ctdb, call->key, c, &header);
+
+ ret = ctdb_ltdb_unlock(ctdb_db, call->key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
+ return;
}
- if ((c->flags & CTDB_IMMEDIATE_MIGRATION)
- && (ctdb_held_back(ctdb_db, call->key, hdr))) {
+ if ( (!(c->flags & CTDB_WANT_READONLY))
+ && (header.flags & (CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY)) ) {
+ header.flags |= CTDB_REC_RO_REVOKING_READONLY;
+ if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
+ ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
+ }
+ ret = ctdb_ltdb_unlock(ctdb_db, call->key);
+
+ if (ctdb_start_revoke_ro_record(ctdb, ctdb_db, call->key, &header, data) != 0) {
+ ctdb_fatal(ctdb, "Failed to start record revoke");
+ }
talloc_free(data.dptr);
- ctdb_ltdb_unlock(ctdb_db, call->key);
+
+ if (ctdb_add_revoke_deferred_call(ctdb, ctdb_db, call->key, hdr, ctdb_call_input_pkt, ctdb) != 0) {
+ ctdb_fatal(ctdb, "Failed to add deferred call for revoke child");
+ }
+ talloc_free(call);
+
+ return;
+ }
+
+ /* If this is the first request for delegation. bump rsn and set
+ * the delegations flag
+ */
+ if ((c->flags & CTDB_WANT_READONLY)
+ && (c->callid == CTDB_FETCH_WITH_HEADER_FUNC)
+ && (!(header.flags & CTDB_REC_RO_HAVE_DELEGATIONS))) {
+ header.rsn += 3;
+ header.flags |= CTDB_REC_RO_HAVE_DELEGATIONS;
+ if (ctdb_ltdb_store(ctdb_db, call->key, &header, data) != 0) {
+ ctdb_fatal(ctdb, "Failed to store record with HAVE_DELEGATIONS set");
+ }
+ }
+ if ((c->flags & CTDB_WANT_READONLY)
+ && (call->call_id == CTDB_FETCH_WITH_HEADER_FUNC)) {
+ TDB_DATA tdata;
+
+ tdata = tdb_fetch(ctdb_db->rottdb, call->key);
+ if (ctdb_trackingdb_add_pnn(ctdb, &tdata, c->hdr.srcnode) != 0) {
+ ctdb_fatal(ctdb, "Failed to add node to trackingdb");
+ }
+ if (tdb_store(ctdb_db->rottdb, call->key, tdata, TDB_REPLACE) != 0) {
+ ctdb_fatal(ctdb, "Failed to store trackingdb data");
+ }
+ free(tdata.dptr);
+
+ ret = ctdb_ltdb_unlock(ctdb_db, call->key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
+
+ len = offsetof(struct ctdb_reply_call, data) + data.dsize + sizeof(struct ctdb_ltdb_header);
+ r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
+ struct ctdb_reply_call);
+ CTDB_NO_MEMORY_FATAL(ctdb, r);
+ r->hdr.destnode = c->hdr.srcnode;
+ r->hdr.reqid = c->hdr.reqid;
+ r->status = 0;
+ r->datalen = data.dsize + sizeof(struct ctdb_ltdb_header);
+ header.rsn -= 2;
+ header.flags |= CTDB_REC_RO_HAVE_READONLY;
+ header.flags &= ~CTDB_REC_RO_HAVE_DELEGATIONS;
+ memcpy(&r->data[0], &header, sizeof(struct ctdb_ltdb_header));
+
+ if (data.dsize) {
+ memcpy(&r->data[sizeof(struct ctdb_ltdb_header)], data.dptr, data.dsize);
+ }
+
+ ctdb_queue_packet(ctdb, &r->hdr);
+
+ talloc_free(r);
return;
}
- /* if this nodes has done enough consecutive calls on the same record
- then give them the record
- or if the node requested an immediate migration
- */
- if ( c->hdr.srcnode != ctdb->pnn &&
- ((header.laccessor == c->hdr.srcnode
- && header.lacount >= ctdb->tunable.max_lacount)
- || (c->flags & CTDB_IMMEDIATE_MIGRATION)) ) {
+ CTDB_UPDATE_STAT(ctdb, max_hop_count, c->hopcount);
+
+ /* Try if possible to migrate the record off to the caller node.
+ * From the clients perspective a fetch of the data is just as
+ * expensive as a migration.
+ */
+ if (c->hdr.srcnode != ctdb->pnn) {
if (ctdb_db->transaction_active) {
DEBUG(DEBUG_INFO, (__location__ " refusing migration"
" of key %s while transaction is active\n",
(char *)call->key.dptr));
} else {
- DEBUG(DEBUG_INFO,("pnn %u starting migration of %08x to %u\n",
+ DEBUG(DEBUG_DEBUG,("pnn %u starting migration of %08x to %u\n",
ctdb->pnn, ctdb_hash(&(call->key)), c->hdr.srcnode));
ctdb_call_send_dmaster(ctdb_db, c, &header, &(call->key), &data);
talloc_free(data.dptr);
- ctdb_ltdb_unlock(ctdb_db, call->key);
+
+ ret = ctdb_ltdb_unlock(ctdb_db, call->key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
return;
}
}
- ctdb_call_local(ctdb_db, call, &header, hdr, &data, c->hdr.srcnode);
+ ret = ctdb_call_local(ctdb_db, call, &header, hdr, &data, true);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_call_local failed\n"));
+ call->status = -1;
+ }
- ctdb_ltdb_unlock(ctdb_db, call->key);
+ ret = ctdb_ltdb_unlock(ctdb_db, call->key);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
+ }
len = offsetof(struct ctdb_reply_call, data) + call->reply_data.dsize;
r = ctdb_transport_allocate(ctdb, ctdb, CTDB_REPLY_CALL, len,
return;
}
+
+ /* read only delegation processing */
+ /* If we got a FETCH_WITH_HEADER we should check if this is a ro
+ * delegation since we may need to update the record header
+ */
+ if (state->c->callid == CTDB_FETCH_WITH_HEADER_FUNC) {
+ struct ctdb_db_context *ctdb_db = state->ctdb_db;
+ struct ctdb_ltdb_header *header = (struct ctdb_ltdb_header *)&c->data[0];
+ struct ctdb_ltdb_header oldheader;
+ TDB_DATA key, data, olddata;
+ int ret;
+
+ if (!(header->flags & CTDB_REC_RO_HAVE_READONLY)) {
+ goto finished_ro;
+ return;
+ }
+
+ key.dsize = state->c->keylen;
+ key.dptr = state->c->data;
+ ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
+ ctdb_call_input_pkt, ctdb, False);
+ if (ret == -2) {
+ return;
+ }
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to get lock in ctdb_reply_call\n"));
+ return;
+ }
+
+ ret = ctdb_ltdb_fetch(ctdb_db, key, &oldheader, state, &olddata);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR, ("Failed to fetch old record in ctdb_reply_call\n"));
+ ctdb_ltdb_unlock(ctdb_db, key);
+ goto finished_ro;
+ }
+
+ if (header->rsn <= oldheader.rsn) {
+ ctdb_ltdb_unlock(ctdb_db, key);
+ goto finished_ro;
+ }
+
+ data.dptr = &c->data[sizeof(struct ctdb_ltdb_header)];
+ data.dsize = sizeof(struct ctdb_ltdb_header);
+ ret = ctdb_ltdb_store(ctdb_db, key, header, data);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR, ("Failed to store new record in ctdb_reply_call\n"));
+ ctdb_ltdb_unlock(ctdb_db, key);
+ goto finished_ro;
+ }
+
+ ctdb_ltdb_unlock(ctdb_db, key);
+ }
+finished_ro:
+
state->call->reply_data.dptr = c->data;
state->call->reply_data.dsize = c->datalen;
state->call->status = c->status;
struct ctdb_reply_dmaster *c = (struct ctdb_reply_dmaster *)hdr;
struct ctdb_db_context *ctdb_db;
TDB_DATA key, data;
+ uint32_t record_flags = 0;
+ size_t len;
int ret;
ctdb_db = find_ctdb_db(ctdb, c->db_id);
key.dsize = c->keylen;
data.dptr = &c->data[key.dsize];
data.dsize = c->datalen;
+ len = offsetof(struct ctdb_reply_dmaster, data) + key.dsize + data.dsize
+ + sizeof(uint32_t);
+ if (len <= c->hdr.length) {
+ record_flags = *(uint32_t *)&c->data[c->keylen + c->datalen];
+ }
ret = ctdb_ltdb_lock_requeue(ctdb_db, key, hdr,
ctdb_call_input_pkt, ctdb, False);
return;
}
- ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn);
+ ctdb_become_dmaster(ctdb_db, hdr, key, data, c->rsn, record_flags);
}
*(state->call) = *call;
state->ctdb_db = ctdb_db;
- ret = ctdb_call_local(ctdb_db, state->call, header, state, data, ctdb->pnn);
+ ret = ctdb_call_local(ctdb_db, state->call, header, state, data, true);
event_add_timed(ctdb->ev, state, timeval_zero(), call_local_trigger, state);
struct ctdb_context *ctdb = ctdb_db->ctdb;
if (ctdb->methods == NULL) {
- DEBUG(DEBUG_ERR,(__location__ " Failed send packet. Transport is down\n"));
+ DEBUG(DEBUG_INFO,(__location__ " Failed send packet. Transport is down\n"));
return NULL;
}
struct ctdb_req_keepalive *r;
if (ctdb->methods == NULL) {
- DEBUG(DEBUG_ERR,(__location__ " Failed to send keepalive. Transport is DOWN\n"));
+ DEBUG(DEBUG_INFO,(__location__ " Failed to send keepalive. Transport is DOWN\n"));
return;
}
r->hdr.destnode = destnode;
r->hdr.reqid = 0;
- ctdb->statistics.keepalive_packets_sent++;
+ CTDB_INCREMENT_STAT(ctdb, keepalive_packets_sent);
ctdb_queue_packet(ctdb, &r->hdr);
talloc_free(r);
}
-static void ctdb_holdback_cleanup(struct event_context *ev,
- struct timed_event *te,
- struct timeval t, void *private_data)
+
+
+struct revokechild_deferred_call {
+ struct ctdb_context *ctdb;
+ struct ctdb_req_header *hdr;
+ deferred_requeue_fn fn;
+ void *ctx;
+};
+
+struct revokechild_handle {
+ struct revokechild_handle *next, *prev;
+ struct ctdb_context *ctdb;
+ struct ctdb_db_context *ctdb_db;
+ struct fd_event *fde;
+ int status;
+ int fd[2];
+ pid_t child;
+ TDB_DATA key;
+};
+
+struct revokechild_requeue_handle {
+ struct ctdb_context *ctdb;
+ struct ctdb_req_header *hdr;
+ deferred_requeue_fn fn;
+ void *ctx;
+};
+
+static void deferred_call_requeue(struct event_context *ev, struct timed_event *te,
+ struct timeval t, void *private_data)
+{
+ struct revokechild_requeue_handle *requeue_handle = talloc_get_type(private_data, struct revokechild_requeue_handle);
+
+ requeue_handle->fn(requeue_handle->ctx, requeue_handle->hdr);
+ talloc_free(requeue_handle);
+}
+
+static int deferred_call_destructor(struct revokechild_deferred_call *deferred_call)
{
- struct ctdb_context *ctdb = talloc_get_type(private_data,
- struct ctdb_context);
- struct ctdb_db_context *ctdb_db;
+ struct ctdb_context *ctdb = deferred_call->ctdb;
+ struct revokechild_requeue_handle *requeue_handle = talloc(ctdb, struct revokechild_requeue_handle);
+ struct ctdb_req_call *c = (struct ctdb_req_call *)deferred_call->hdr;
+
+ requeue_handle->ctdb = ctdb;
+ requeue_handle->hdr = deferred_call->hdr;
+ requeue_handle->fn = deferred_call->fn;
+ requeue_handle->ctx = deferred_call->ctx;
+ talloc_steal(requeue_handle, requeue_handle->hdr);
- DEBUG(DEBUG_INFO, ("running ctdb_holdback_cleanup\n"));
+ /* when revoking, any READONLY requests have 1 second grace to let read/write finish first */
+ event_add_timed(ctdb->ev, requeue_handle, timeval_current_ofs(c->flags & CTDB_WANT_READONLY ? 1 : 0, 0), deferred_call_requeue, requeue_handle);
- if (te != ctdb->holdback_cleanup_te) {
- ctdb_fatal(ctdb, "te != ctdb->holdback_cleanup_te");
+ return 0;
+}
+
+
+static int revokechild_destructor(struct revokechild_handle *rc)
+{
+ if (rc->fde != NULL) {
+ talloc_free(rc->fde);
+ }
+
+ if (rc->fd[0] != -1) {
+ close(rc->fd[0]);
}
+ if (rc->fd[1] != -1) {
+ close(rc->fd[1]);
+ }
+ kill(rc->child, SIGKILL);
- for (ctdb_db=ctdb->db_list; ctdb_db; ctdb_db=ctdb_db->next) {
- size_t i, num_heldback;
+ DLIST_REMOVE(rc->ctdb_db->revokechild_active, rc);
+ return 0;
+}
- talloc_free(ctdb_db->holdback_keys);
- ctdb_db->holdback_keys = NULL;
+static void revokechild_handler(struct event_context *ev, struct fd_event *fde,
+ uint16_t flags, void *private_data)
+{
+ struct revokechild_handle *rc = talloc_get_type(private_data,
+ struct revokechild_handle);
+ int ret;
+ char c;
- num_heldback = talloc_array_length(ctdb_db->held_back);
- for (i=0; i<num_heldback; i++) {
- ctdb_queue_packet(ctdb, ctdb_db->held_back[i]);
- }
- talloc_free(ctdb_db->held_back);
- ctdb_db->held_back = NULL;
- }
+ ret = read(rc->fd[0], &c, 1);
+ if (ret != 1) {
+ DEBUG(DEBUG_ERR,("Failed to read status from revokechild. errno:%d\n", errno));
+ rc->status = -1;
+ talloc_free(rc);
+ return;
+ }
+ if (c != 0) {
+ DEBUG(DEBUG_ERR,("revokechild returned failure. status:%d\n", c));
+ rc->status = -1;
+ talloc_free(rc);
+ return;
+ }
+
+ talloc_free(rc);
+}
+
+struct ctdb_revoke_state {
+ struct ctdb_db_context *ctdb_db;
+ TDB_DATA key;
+ struct ctdb_ltdb_header *header;
+ TDB_DATA data;
+ int count;
+ int status;
+ int finished;
+};
+
+static void update_record_cb(struct ctdb_client_control_state *state)
+{
+ struct ctdb_revoke_state *revoke_state;
+ int ret;
+ int32_t res;
+
+ if (state == NULL) {
+ return;
+ }
+ revoke_state = state->async.private_data;
+
+ state->async.fn = NULL;
+ ret = ctdb_control_recv(state->ctdb, state, state, NULL, &res, NULL);
+ if ((ret != 0) || (res != 0)) {
+ DEBUG(DEBUG_ERR,("Recv for revoke update record failed ret:%d res:%d\n", ret, res));
+ revoke_state->status = -1;
+ }
- ctdb->holdback_cleanup_te = event_add_timed(
- ctdb->ev, ctdb, timeval_current_ofs(
- 0, ctdb->tunable.holdback_cleanup_interval * 1000),
- ctdb_holdback_cleanup, ctdb);
+ revoke_state->count--;
+ if (revoke_state->count <= 0) {
+ revoke_state->finished = 1;
+ }
}
-void ctdb_start_holdback_cleanup(struct ctdb_context *ctdb)
+static void revoke_send_cb(struct ctdb_context *ctdb, uint32_t pnn, void *private_data)
{
- ctdb->holdback_cleanup_te = event_add_timed(
- ctdb->ev, ctdb, timeval_current_ofs(
- 0, ctdb->tunable.holdback_cleanup_interval * 1000),
- ctdb_holdback_cleanup, ctdb);
+ struct ctdb_revoke_state *revoke_state = private_data;
+ struct ctdb_client_control_state *state;
+
+ state = ctdb_ctrl_updaterecord_send(ctdb, revoke_state, timeval_current_ofs(5,0), pnn, revoke_state->ctdb_db, revoke_state->key, revoke_state->header, revoke_state->data);
+ if (state == NULL) {
+ DEBUG(DEBUG_ERR,("Failure to send update record to revoke readonly delegation\n"));
+ revoke_state->status = -1;
+ return;
+ }
+ state->async.fn = update_record_cb;
+ state->async.private_data = revoke_state;
- CTDB_NO_MEMORY_FATAL(ctdb, ctdb->holdback_cleanup_te);
+ revoke_state->count++;
- DEBUG(DEBUG_NOTICE,("Holdback cleanup has been started\n"));
}
-void ctdb_stop_holdback_cleanup(struct ctdb_context *ctdb)
+static void ctdb_revoke_timeout_handler(struct event_context *ev, struct timed_event *te,
+ struct timeval yt, void *private_data)
{
- talloc_free(ctdb->holdback_cleanup_te);
- ctdb->holdback_cleanup_te = NULL;
+ struct ctdb_revoke_state *state = private_data;
+
+ DEBUG(DEBUG_ERR,("Timed out waiting for revoke to finish\n"));
+ state->finished = 1;
+ state->status = -1;
}
-int32_t ctdb_control_gotit(struct ctdb_context *ctdb, TDB_DATA indata)
+static int ctdb_revoke_all_delegations(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA tdata, TDB_DATA key, struct ctdb_ltdb_header *header, TDB_DATA data)
{
- struct ctdb_control_gotit *c =
- (struct ctdb_control_gotit *)indata.dptr;
- struct ctdb_db_context *db;
- size_t i, num_keys, num_heldback;
- TDB_DATA key;
+ struct ctdb_revoke_state *state = talloc_zero(ctdb, struct ctdb_revoke_state);
+ int status;
+
+ state->ctdb_db = ctdb_db;
+ state->key = key;
+ state->header = header;
+ state->data = data;
+
+ ctdb_trackingdb_traverse(ctdb, tdata, revoke_send_cb, state);
+
+ event_add_timed(ctdb->ev, state, timeval_current_ofs(5, 0), ctdb_revoke_timeout_handler, state);
- if (indata.dsize < sizeof(struct ctdb_control_gotit)) {
- DEBUG(DEBUG_ERR, (__location__ "Invalid data size %d\n",
- (int)indata.dsize));
+ while (state->finished == 0) {
+ event_loop_once(ctdb->ev);
+ }
+
+ status = state->status;
+
+ if (status == 0) {
+ struct ctdb_ltdb_header new_header;
+ TDB_DATA new_data;
+
+ if (ctdb_ltdb_lock(ctdb_db, key) != 0) {
+ DEBUG(DEBUG_ERR,("Failed to chainlock the database in revokechild\n"));
+ talloc_free(state);
+ return -1;
+ }
+ if (ctdb_ltdb_fetch(ctdb_db, key, &new_header, state, &new_data) != 0) {
+ ctdb_ltdb_unlock(ctdb_db, key);
+ DEBUG(DEBUG_ERR,("Failed for fetch tdb record in revokechild\n"));
+ talloc_free(state);
+ return -1;
+ }
+ header->rsn++;
+ if (new_header.rsn > header->rsn) {
+ ctdb_ltdb_unlock(ctdb_db, key);
+ DEBUG(DEBUG_ERR,("RSN too high in tdb record in revokechild\n"));
+ talloc_free(state);
+ return -1;
+ }
+ if ( (new_header.flags & (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS)) != (CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS) ) {
+ ctdb_ltdb_unlock(ctdb_db, key);
+ DEBUG(DEBUG_ERR,("Flags are wrong in tdb record in revokechild\n"));
+ talloc_free(state);
+ return -1;
+ }
+ new_header.rsn++;
+ new_header.flags |= CTDB_REC_RO_REVOKE_COMPLETE;
+ if (ctdb_ltdb_store(ctdb_db, key, &new_header, new_data) != 0) {
+ ctdb_ltdb_unlock(ctdb_db, key);
+ DEBUG(DEBUG_ERR,("Failed to write new record in revokechild\n"));
+ talloc_free(state);
+ return -1;
+ }
+ ctdb_ltdb_unlock(ctdb_db, key);
+ }
+
+ talloc_free(state);
+ return status;
+}
+
+
+int ctdb_start_revoke_ro_record(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key, struct ctdb_ltdb_header *header, TDB_DATA data)
+{
+ TDB_DATA tdata;
+ struct revokechild_handle *rc;
+ pid_t parent = getpid();
+ int ret;
+
+ header->flags &= ~(CTDB_REC_RO_REVOKING_READONLY|CTDB_REC_RO_HAVE_DELEGATIONS|CTDB_REC_RO_HAVE_READONLY);
+ header->rsn -= 1;
+
+ if ((rc = talloc_zero(ctdb_db, struct revokechild_handle)) == NULL) {
+ DEBUG(DEBUG_ERR,("Failed to allocate revokechild_handle\n"));
+ return -1;
+ }
+
+ tdata = tdb_fetch(ctdb_db->rottdb, key);
+ if (tdata.dsize > 0) {
+ uint8_t *tmp;
+
+ tmp = tdata.dptr;
+ tdata.dptr = talloc_memdup(rc, tdata.dptr, tdata.dsize);
+ free(tmp);
+ }
+
+ rc->status = 0;
+ rc->ctdb = ctdb;
+ rc->ctdb_db = ctdb_db;
+ rc->fd[0] = -1;
+ rc->fd[1] = -1;
+
+ talloc_set_destructor(rc, revokechild_destructor);
+
+ rc->key.dsize = key.dsize;
+ rc->key.dptr = talloc_memdup(rc, key.dptr, key.dsize);
+ if (rc->key.dptr == NULL) {
+ DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
+ talloc_free(rc);
return -1;
}
- db = find_ctdb_db(ctdb, c->db_id);
- if (db == NULL) {
- DEBUG(DEBUG_ERR, ("Unknown db_id 0x%x in ctdb_reply_dmaster\n",
- (int)c->db_id));
+
+ ret = pipe(rc->fd);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,("Failed to allocate key for revokechild_handle\n"));
+ talloc_free(rc);
return -1;
}
- key.dptr = c->key;
- key.dsize = indata.dsize - offsetof(struct ctdb_control_gotit, key);
- num_keys = talloc_array_length(db->holdback_keys);
- for (i=0; i<num_keys; i++) {
- uint8_t *hold_key = db->holdback_keys[i];
- size_t keylength = talloc_array_length(hold_key);
+ rc->child = ctdb_fork(ctdb);
+ if (rc->child == (pid_t)-1) {
+ DEBUG(DEBUG_ERR,("Failed to fork child for revokechild\n"));
+ talloc_free(rc);
+ return -1;
+ }
- if ((keylength == key.dsize)
- && (memcmp(hold_key, key.dptr, keylength) == 0)) {
- break;
+ if (rc->child == 0) {
+ char c = 0;
+ close(rc->fd[0]);
+ debug_extra = talloc_asprintf(NULL, "revokechild-%s:", ctdb_db->db_name);
+
+ if (switch_from_server_to_client(ctdb, "revokechild-%s", ctdb_db->db_name) != 0) {
+ DEBUG(DEBUG_ERR,("Failed to switch from server to client for revokechild process\n"));
+ c = 1;
+ goto child_finished;
+ }
+
+ c = ctdb_revoke_all_delegations(ctdb, ctdb_db, tdata, key, header, data);
+
+child_finished:
+ write(rc->fd[1], &c, 1);
+ /* make sure we die when our parent dies */
+ while (kill(parent, 0) == 0 || errno != ESRCH) {
+ sleep(5);
}
+ _exit(0);
}
- if (i == num_keys) {
- /*
- * ctdb_holdback_cleanup has kicked in. This is okay,
- * we will just potentially have to retry.
- */
- return 0;
+
+ close(rc->fd[1]);
+ rc->fd[1] = -1;
+ set_close_on_exec(rc->fd[0]);
+
+ /* This is an active revokechild child process */
+ DLIST_ADD_END(ctdb_db->revokechild_active, rc, NULL);
+
+ rc->fde = event_add_fd(ctdb->ev, rc, rc->fd[0],
+ EVENT_FD_READ, revokechild_handler,
+ (void *)rc);
+ if (rc->fde == NULL) {
+ DEBUG(DEBUG_ERR,("Failed to set up fd event for revokechild process\n"));
+ talloc_free(rc);
+ }
+ tevent_fd_set_auto_close(rc->fde);
+
+ return 0;
+}
+
+int ctdb_add_revoke_deferred_call(struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key, struct ctdb_req_header *hdr, deferred_requeue_fn fn, void *call_context)
+{
+ struct revokechild_handle *rc;
+ struct revokechild_deferred_call *deferred_call;
+
+ for (rc = ctdb_db->revokechild_active; rc; rc = rc->next) {
+ if (rc->key.dsize == 0) {
+ continue;
+ }
+ if (rc->key.dsize != key.dsize) {
+ continue;
+ }
+ if (!memcmp(rc->key.dptr, key.dptr, key.dsize)) {
+ break;
+ }
}
- talloc_free(db->holdback_keys[i]);
- if (i < num_keys-1) {
- db->holdback_keys[i] = db->holdback_keys[num_keys-1];
+ if (rc == NULL) {
+ DEBUG(DEBUG_ERR,("Failed to add deferred call to revoke list. revoke structure not found\n"));
+ return -1;
}
- db->holdback_keys = talloc_realloc(db, db->holdback_keys, uint8_t *,
- num_keys-1);
- num_heldback = talloc_array_length(db->held_back);
- for (i=0; i<num_heldback; i++) {
- ctdb_queue_packet(ctdb, db->held_back[i]);
+ deferred_call = talloc(rc, struct revokechild_deferred_call);
+ if (deferred_call == NULL) {
+ DEBUG(DEBUG_ERR,("Failed to allocate deferred call structure for revoking record\n"));
+ return -1;
}
- talloc_free(db->held_back);
- db->held_back = NULL;
+
+ deferred_call->ctdb = ctdb;
+ deferred_call->hdr = hdr;
+ deferred_call->fn = fn;
+ deferred_call->ctx = call_context;
+
+ talloc_set_destructor(deferred_call, deferred_call_destructor);
+ talloc_steal(deferred_call, hdr);
+
return 0;
}