mnesia_tm.erl

%%
%% %CopyrightBegin%
%%
%% Copyright Ericsson AB 1996-2022. All Rights Reserved.
%%
%% Licensed under the Apache License, Version 2.0 (the "License");
%% you may not use this file except in compliance with the License.
%% You may obtain a copy of the License at
%%
%%     http://www.apache.org/licenses/LICENSE-2.0
%%
%% Unless required by applicable law or agreed to in writing, software
%% distributed under the License is distributed on an "AS IS" BASIS,
%% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
%% See the License for the specific language governing permissions and
%% limitations under the License.
%%
%% %CopyrightEnd%
%%

%%
-module(mnesia_tm).

-export([
	 start/0,
	 init/1,
	 non_transaction/5,
	 transaction/6,
	 commit_participant/6,
	 dirty/2,
	 display_info/2,
	 do_update_op/3,
	 get_info/1,
	 get_transactions/0,
	 info/1,
	 mnesia_down/1,
	 prepare_checkpoint/2,
	 prepare_checkpoint/1, % Internal
	 prepare_snmp/3,
	 do_snmp/2,
	 put_activity_id/1,
	 put_activity_id/2,
	 block_tab/1,
	 unblock_tab/1,
	 fixtable/3,
	 new_cr_format/1
	]).

%% sys callback functions
-export([system_continue/3,
	 system_terminate/4,
	 system_code_change/4
	]).

-include("mnesia.hrl").

-import(mnesia_lib, [set/2]).
-import(mnesia_lib, [fatal/2, verbose/2, dbg_out/2]).

-record(state, {coordinators = gb_trees:empty(), participants = gb_trees:empty(), supervisor,
		blocked_tabs = [], dirty_queue = [], fixed_tabs = []}).
%% Format on coordinators is [{Tid, EtsTabList} .....

-record(prep, {protocol = sym_trans,
	       %% async_dirty | sync_dirty | sym_trans | sync_sym_trans | asym_trans | sync_asym_trans
	       records = [],
	       prev_tab = [], % initiate to a non valid table name
	       prev_types,
	       prev_snmp,
	       types,
	       majority = [],
               sync = false
	      }).

-record(participant, {tid, pid, commit, disc_nodes = [],
		      ram_nodes = [], protocol = sym_trans}).

start() ->
    mnesia_monitor:start_proc(?MODULE, ?MODULE, init, [self()]).

init(Parent) ->
    register(?MODULE, self()),
    process_flag(trap_exit, true),
    process_flag(message_queue_data, off_heap),

    %% Initialize the schema
    IgnoreFallback = mnesia_monitor:get_env(ignore_fallback_at_startup),
    mnesia_bup:tm_fallback_start(IgnoreFallback),
    mnesia_schema:init(IgnoreFallback),

    %% Handshake and initialize transaction recovery
    mnesia_recover:init(),
    Early = mnesia_monitor:init(),
    AllOthers = mnesia_lib:uniq(Early ++ mnesia_lib:all_nodes()) -- [node()],
    set(original_nodes, AllOthers),
    mnesia_recover:connect_nodes(AllOthers),

    %% Recover transactions, may wait for decision
    case mnesia_monitor:use_dir() of
	true ->
	    P = mnesia_dumper:opt_dump_log(startup), % previous log
	    L = mnesia_dumper:opt_dump_log(startup), % latest log
	    Msg = "Initial dump of log during startup: ~p~n",
	    mnesia_lib:verbose(Msg, [[P, L]]),
	    mnesia_log:init();
	false ->
	    ignore
    end,

    mnesia_schema:purge_tmp_files(),
    mnesia_recover:next_garb(),
    mnesia_recover:next_check_overload(),

    ?eval_debug_fun({?MODULE, init},  [{nodes, AllOthers}]),

    case val(debug) of
	Debug when Debug /= debug, Debug /= trace ->
	    ignore;
	_ ->
	    mnesia_subscr:subscribe(whereis(mnesia_event), {table, schema})
    end,
    proc_lib:init_ack(Parent, {ok, self()}),
    doit_loop(#state{supervisor = Parent}).

%% Local function in order to avoid external function call
val(Var) ->
    case ?catch_val_and_stack(Var) of
	{'EXIT', Stacktrace} -> mnesia_lib:other_val(Var, Stacktrace);
	Value -> Value
    end.

reply({From,Ref}, R) ->
    From ! {?MODULE, Ref, R};
reply(From, R) ->
    From ! {?MODULE, node(), R}.

reply(From, R, State) ->
    reply(From, R),
    doit_loop(State).

req(R) ->
    case whereis(?MODULE) of
	undefined ->
	    {error, {node_not_running, node()}};
	Pid ->
	    Ref = make_ref(),
	    Pid ! {{self(), Ref}, R},
	    rec(Pid, Ref)
    end.

rec() ->
    rec(whereis(?MODULE)).

rec(Pid) when is_pid(Pid) ->
    receive
	{?MODULE, _, Reply} ->
	    Reply;

	{'EXIT', Pid, _} ->
	    {error, {node_not_running, node()}}
    end;
rec(undefined) ->
    {error, {node_not_running, node()}}.

rec(Pid, Ref) ->
    receive
	{?MODULE, Ref, Reply} ->
	    Reply;
	{'EXIT', Pid, _} ->
	    {error, {node_not_running, node()}}
    end.

tmlink({From, Ref}) when is_reference(Ref) ->
    link(From);
tmlink(From) ->
    link(From).
tmpid({Pid, _Ref}) when is_pid(Pid) ->
    Pid;
tmpid(Pid) ->
    Pid.

%% Returns a list of participant transaction Tid's
mnesia_down(Node) ->
    %% Synchronously call needed in order to avoid
    %% race with mnesia_tm's coordinator processes
    %% that may restart and acquire new locks.
    %% mnesia_monitor takes care of the sync
    case whereis(?MODULE) of
	undefined ->
	    mnesia_monitor:mnesia_down(?MODULE, Node);
	Pid ->
	    Pid ! {mnesia_down, Node},
	    ok
    end.

prepare_checkpoint(Nodes, Cp) ->
    rpc:multicall(Nodes, ?MODULE, prepare_checkpoint, [Cp]).

prepare_checkpoint(Cp) ->
    req({prepare_checkpoint,Cp}).

block_tab(Tab) ->
    req({block_tab, Tab}).

unblock_tab(Tab) ->
    req({unblock_tab, Tab}).

doit_loop(#state{coordinators=Coordinators,participants=Participants,supervisor=Sup}=State) ->
    receive
	{_From, {async_dirty, Tid, Commit, Tab}} ->
	    case lists:member(Tab, State#state.blocked_tabs) of
		false ->
		    do_async_dirty(Tid, new_cr_format(Commit), Tab),
		    doit_loop(State);
		true ->
		    Item = {async_dirty, Tid, new_cr_format(Commit), Tab},
		    State2 = State#state{dirty_queue = [Item | State#state.dirty_queue]},
		    doit_loop(State2)
	    end;

	{From, {sync_dirty, Tid, Commit, Tab}} ->
	    case lists:member(Tab, State#state.blocked_tabs) of
		false ->
		    do_sync_dirty(From, Tid, new_cr_format(Commit), Tab),
		    doit_loop(State);
		true ->
		    Item = {sync_dirty, From, Tid, new_cr_format(Commit), Tab},
		    State2 = State#state{dirty_queue = [Item | State#state.dirty_queue]},
		    doit_loop(State2)
	    end;

	{From, start_outer} -> %% Create and associate ets_tab with Tid
	    try ?ets_new_table(mnesia_trans_store, [bag, public]) of
		Etab ->
		    tmlink(From),
		    C = mnesia_recover:incr_trans_tid_serial(),
		    ?ets_insert(Etab, {nodes, node()}),
		    Tid = #tid{pid = tmpid(From), counter = C},
		    A2 = gb_trees:insert(Tid,[Etab],Coordinators),
		    S2 = State#state{coordinators = A2},
		    reply(From, {new_tid, Tid, Etab}, S2)
	    catch error:Reason -> %% system limit
		    Msg = "Cannot create an ets table for the "
			"local transaction store",
		    reply(From, {error, {system_limit, Msg, Reason}}, State)
	    end;

	{From, {ask_commit, Protocol, Tid, Commit0, DiscNs, RamNs}} ->
	    ?eval_debug_fun({?MODULE, doit_ask_commit},
			    [{tid, Tid}, {prot, Protocol}]),
	    mnesia_checkpoint:tm_enter_pending(Tid, DiscNs, RamNs),
	    Commit = new_cr_format(Commit0),
	    Pid =
                if
                    node(Tid#tid.pid) =:= node() ->
                        error({internal_error, local_node});
                    Protocol =:= asym_trans orelse Protocol =:= sync_asym_trans ->
			Args = [Protocol, tmpid(From), Tid, Commit, DiscNs, RamNs],
			spawn_link(?MODULE, commit_participant, Args);
                    true -> %% *_sym_trans
			reply(From, {vote_yes, Tid}),
			nopid
		end,
	    P = #participant{tid = Tid,
			     pid = Pid,
			     commit = Commit,
			     disc_nodes = DiscNs,
			     ram_nodes = RamNs,
			     protocol = Protocol},
	    State2 = State#state{participants = gb_trees:insert(Tid,P,Participants)},
	    doit_loop(State2);

	{Tid, do_commit} ->
	    case gb_trees:lookup(Tid, Participants) of
		none ->
		    verbose("Tried to commit a non participant transaction ~p~n",[Tid]),
		    doit_loop(State);
		{value, P} ->
		    ?eval_debug_fun({?MODULE,do_commit,pre},[{tid,Tid},{participant,P}]),
		    case P#participant.pid of
			nopid ->
			    Commit = P#participant.commit,
			    Member = lists:member(node(), P#participant.disc_nodes),
			    if Member == false ->
				    ignore;
			       P#participant.protocol == sym_trans ->
				    mnesia_log:log(Commit);
			       P#participant.protocol == sync_sym_trans ->
				    mnesia_log:slog(Commit)
			    end,
			    mnesia_recover:note_decision(Tid, committed),
			    do_commit(Tid, Commit),
			    if
				P#participant.protocol == sync_sym_trans ->
				    Tid#tid.pid ! {?MODULE, node(), {committed, Tid}};
				true ->
				    ignore
			    end,
			    mnesia_locker:release_tid(Tid),
			    transaction_terminated(Tid),
			    ?eval_debug_fun({?MODULE,do_commit,post},[{tid,Tid},{pid,nopid}]),
			    doit_loop(State#state{participants=
						  gb_trees:delete(Tid,Participants)});
			Pid when is_pid(Pid) ->
			    Pid ! {Tid, committed},
			    ?eval_debug_fun({?MODULE, do_commit, post}, [{tid, Tid}, {pid, Pid}]),
			    doit_loop(State)
		    end
	    end;

	{Tid, simple_commit} ->
	    mnesia_recover:note_decision(Tid, committed),
	    mnesia_locker:release_tid(Tid),
	    transaction_terminated(Tid),
	    doit_loop(State);

	{Tid, {do_abort, Reason}} ->
	    ?eval_debug_fun({?MODULE, do_abort, pre}, [{tid, Tid}]),
	    case gb_trees:lookup(Tid, Participants) of
		none ->
		    verbose("Tried to abort a non participant transaction ~p: ~tp~n",
			    [Tid, Reason]),
		    mnesia_locker:release_tid(Tid),
		    doit_loop(State);
		{value, P} ->
		    case P#participant.pid of
			nopid ->
			    Commit = P#participant.commit,
			    mnesia_recover:note_decision(Tid, aborted),
			    do_abort(Tid, Commit),
			    if
				P#participant.protocol == sync_sym_trans ->
				    Tid#tid.pid ! {?MODULE, node(), {aborted, Tid}};
				true ->
				    ignore
			    end,
			    transaction_terminated(Tid),
			    mnesia_locker:release_tid(Tid),
			    ?eval_debug_fun({?MODULE, do_abort, post}, [{tid, Tid}, {pid, nopid}]),
			    doit_loop(State#state{participants=
						  gb_trees:delete(Tid,Participants)});
			Pid when is_pid(Pid) ->
			    Pid ! {Tid, {do_abort, Reason}},
			    ?eval_debug_fun({?MODULE, do_abort, post},
					    [{tid, Tid}, {pid, Pid}]),
			    doit_loop(State)
		    end
	    end;

	{From, {add_store, Tid}} -> %% new store for nested  transaction
	    try ?ets_new_table(mnesia_trans_store, [bag, public]) of
		Etab ->
		    A2 = add_coord_store(Coordinators, Tid, Etab),
		    reply(From, {new_store, Etab},
			  State#state{coordinators = A2})
	    catch error:Reason -> %% system limit
		    Msg = "Cannot create an ets table for a nested "
			"local transaction store",
		    reply(From, {error, {system_limit, Msg, Reason}}, State)
	    end;

	{From, {del_store, Tid, Current, Obsolete, PropagateStore}} ->
	    opt_propagate_store(Current, Obsolete, PropagateStore),
	    A2 = del_coord_store(Coordinators, Tid, Current, Obsolete),
	    reply(From, store_erased, State#state{coordinators = A2});

	{'EXIT', Pid, Reason} ->
	    handle_exit(Pid, Reason, State);

	{From, {restart, Tid, Store}} ->
	    A2 = restore_stores(Coordinators, Tid, Store),
	    clear_fixtable([Store]),
	    ?ets_match_delete(Store, '_'),
	    ?ets_insert(Store, {nodes, node()}),
	    reply(From, {restarted, Tid}, State#state{coordinators = A2});

	{delete_transaction, Tid} ->
	    %% used to clear transactions which are committed
	    %% in coordinator or participant processes
	    case gb_trees:is_defined(Tid, Participants) of
		false ->
		    case gb_trees:lookup(Tid, Coordinators) of
			none ->
			    verbose("** ERROR ** Tried to delete a non transaction ~p~n",
				    [Tid]),
			    doit_loop(State);
			{value, Etabs} ->
			    clear_fixtable(Etabs),
			    erase_ets_tabs(Etabs),
			    transaction_terminated(Tid),
			    doit_loop(State#state{coordinators =
						  gb_trees:delete(Tid,Coordinators)})
		    end;
		true ->
		    transaction_terminated(Tid),
		    State2 = State#state{participants=gb_trees:delete(Tid,Participants)},
		    doit_loop(State2)
	    end;

	{sync_trans_serial, Tid} ->
	    %% Do the Lamport thing here
	    mnesia_recover:sync_trans_tid_serial(Tid),
	    doit_loop(State);

	{From, info} ->
	    reply(From, {info, gb_trees:values(Participants),
			 gb_trees:to_list(Coordinators)}, State);

	{mnesia_down, N} ->
	    verbose("Got mnesia_down from ~p, reconfiguring...~n", [N]),
	    reconfigure_coordinators(N, gb_trees:to_list(Coordinators)),

	    Tids = gb_trees:keys(Participants),
	    reconfigure_participants(N, gb_trees:values(Participants)),
	    NewState = clear_fixtable(N, State),

	    mnesia_locker:mnesia_down(N, Tids),
	    mnesia_monitor:mnesia_down(?MODULE, N),
	    doit_loop(NewState);

	{From, {unblock_me, Tab}} ->
	    case lists:member(Tab, State#state.blocked_tabs) of
		false ->
		    verbose("Wrong dirty Op blocked on ~p ~tp ~p",
			    [node(), Tab, From]),
		    reply(From, unblocked),
		    doit_loop(State);
		true ->
		    Item = {Tab, unblock_me, From},
		    State2 = State#state{dirty_queue = [Item | State#state.dirty_queue]},
		    doit_loop(State2)
	    end;

	{From, {block_tab, Tab}} ->
	    State2 = State#state{blocked_tabs = [Tab | State#state.blocked_tabs]},
	    reply(From, ok, State2);

	{From, {unblock_tab, Tab}} ->
	    BlockedTabs2 = State#state.blocked_tabs -- [Tab],
	    case lists:member(Tab, BlockedTabs2) of
		false ->
		    mnesia_controller:unblock_table(Tab),
		    Queue = process_dirty_queue(Tab, State#state.dirty_queue),
		    State2 = State#state{blocked_tabs = BlockedTabs2,
					 dirty_queue = Queue},
		    reply(From, ok, State2);
		true ->
		    State2 = State#state{blocked_tabs = BlockedTabs2},
		    reply(From, ok, State2)
	    end;

	{From, {prepare_checkpoint, Cp}} ->
	    Res = mnesia_checkpoint:tm_prepare(Cp),
	    case Res of
		{ok, _Name, IgnoreNew, _Node} ->
		    prepare_pending_coordinators(gb_trees:to_list(Coordinators), IgnoreNew),
		    prepare_pending_participants(gb_trees:values(Participants), IgnoreNew);
		{error, _Reason} ->
		    ignore
	    end,
	    reply(From, Res, State);
	{From, {fixtable, [Tab,Lock,Requester]}} ->
	    case ?catch_val({Tab, storage_type}) of
		{'EXIT', _} ->
		    reply(From, error, State);
		Storage ->
		    mnesia_lib:db_fixtable(Storage,Tab,Lock),
		    NewState = manage_fixtable(Tab,Lock,Requester,State),
		    reply(From, node(), NewState)
	    end;

	{system, From, Msg} ->
	    dbg_out("~p got {system, ~p, ~tp}~n", [?MODULE, From, Msg]),
	    sys:handle_system_msg(Msg, From, Sup, ?MODULE, [], State);

	Msg ->
	    verbose("** ERROR ** ~p got unexpected message: ~tp~n", [?MODULE, Msg]),
	    doit_loop(State)
    end.

do_sync_dirty(From, Tid, Commit, _Tab) ->
    ?eval_debug_fun({?MODULE, sync_dirty, pre}, [{tid, Tid}]),
    Res = do_dirty(Tid, Commit),
    ?eval_debug_fun({?MODULE, sync_dirty, post}, [{tid, Tid}]),
    From ! {?MODULE, node(), {dirty_res, Res}}.

do_async_dirty(Tid, Commit, _Tab) ->
    ?eval_debug_fun({?MODULE, async_dirty, pre}, [{tid, Tid}]),
    do_dirty(Tid, Commit),
    ?eval_debug_fun({?MODULE, async_dirty, post}, [{tid, Tid}]).


%% Process items in fifo order
process_dirty_queue(Tab, [Item | Queue]) ->
    Queue2 = process_dirty_queue(Tab, Queue),
    case Item of
	{async_dirty, Tid, Commit, Tab} ->
	    do_async_dirty(Tid, Commit, Tab),
	    Queue2;
	{sync_dirty, From, Tid, Commit, Tab} ->
	    do_sync_dirty(From, Tid, Commit, Tab),
	    Queue2;
	{Tab, unblock_me, From} ->
	    reply(From, unblocked),
	    Queue2;
	_ ->
	    [Item | Queue2]
    end;
process_dirty_queue(_Tab, []) ->
    [].

prepare_pending_coordinators([{Tid, [Store | _Etabs]} | Coords], IgnoreNew) ->
    try ?ets_lookup(Store, pending) of
	[] ->
	    prepare_pending_coordinators(Coords, IgnoreNew);
	[Pending] ->
	    case lists:member(Tid, IgnoreNew) of
		false ->
		    mnesia_checkpoint:tm_enter_pending(Pending);
		true ->
		    ignore
	    end,
	    prepare_pending_coordinators(Coords, IgnoreNew)
    catch error:_ ->
	    prepare_pending_coordinators(Coords, IgnoreNew)
    end;
prepare_pending_coordinators([], _IgnoreNew) ->
    ok.

prepare_pending_participants([Part | Parts], IgnoreNew) ->
    Tid = Part#participant.tid,
    D = Part#participant.disc_nodes,
    R = Part#participant.ram_nodes,
    case lists:member(Tid, IgnoreNew) of
	false ->
	    mnesia_checkpoint:tm_enter_pending(Tid, D, R);
	true ->
	    ignore
    end,
    prepare_pending_participants(Parts, IgnoreNew);
prepare_pending_participants([], _IgnoreNew) ->
    ok.

handle_exit(Pid, _Reason, State) when node(Pid) /= node() ->
    %% We got exit from a remote fool
    doit_loop(State);

handle_exit(Pid, _Reason, State) when Pid == State#state.supervisor ->
    %% Our supervisor has died, time to stop
    do_stop(State);

handle_exit(Pid, Reason, State) ->
    %% Check if it is a coordinator
    case pid_search_delete(Pid, gb_trees:to_list(State#state.coordinators)) of
	{none, _} ->
	    %% Check if it is a participant
	    Ps = gb_trees:values(State#state.participants),
	    case mnesia_lib:key_search_delete(Pid,#participant.pid,Ps) of
		{none, _} ->
		    %% We got exit from a local fool
		    doit_loop(State);
		{P = #participant{}, _RestP} ->
		    fatal("Participant ~p in transaction ~p died ~tp~n",
			  [P#participant.pid, P#participant.tid, Reason]),
		    NewPs = gb_trees:delete(P#participant.tid,State#state.participants),
		    doit_loop(State#state{participants = NewPs})
	    end;

	{{Tid, Etabs}, RestC} ->
	    %% A local coordinator has died and
	    %% we must determine the outcome of the
	    %% transaction and tell mnesia_tm on the
	    %% other nodes about it and then recover
	    %% locally.
	    recover_coordinator(Tid, Etabs),
	    doit_loop(State#state{coordinators = RestC})
    end.

recover_coordinator(Tid, Etabs) ->
    verbose("Coordinator ~p in transaction ~p died.~n", [Tid#tid.pid, Tid]),

    Store = hd(Etabs),
    CheckNodes = get_elements(nodes,Store),
    TellNodes = CheckNodes -- [node()],
    try arrange(Tid, Store, async) of
	{_N, Prep} ->
	    %% Tell the participants about the outcome
	    Protocol = Prep#prep.protocol,
	    Outcome = tell_outcome(Tid, Protocol, node(), CheckNodes, TellNodes),

	    %% Recover locally
	    CR = Prep#prep.records,
	    {DiscNs, RamNs} = commit_nodes(CR, [], []),
	    case lists:keysearch(node(), #commit.node, CR) of
		{value, Local} ->
		    ?eval_debug_fun({?MODULE, recover_coordinator, pre},
				    [{tid, Tid}, {outcome, Outcome}, {prot, Protocol}]),
		    recover_coordinator(Tid, Protocol, Outcome, Local, DiscNs, RamNs),
		    ?eval_debug_fun({?MODULE, recover_coordinator, post},
				    [{tid, Tid}, {outcome, Outcome}, {prot, Protocol}]);
		false ->  %% When killed before store haven't been copied to
		    ok    %% to the new nested trans store.
	    end
    catch _:Reason:Stacktrace ->
	    dbg_out("Recovery of coordinator ~p failed: ~tp~n",
		    [Tid, {Reason, Stacktrace}]),
	    Protocol = asym_trans,
	    tell_outcome(Tid, Protocol, node(), CheckNodes, TellNodes)
    end,
    erase_ets_tabs(Etabs),
    transaction_terminated(Tid),
    mnesia_locker:release_tid(Tid).

recover_coordinator(Tid, sym_trans, committed, Local, _, _) ->
    mnesia_recover:note_decision(Tid, committed),
    do_dirty(Tid, Local);
recover_coordinator(Tid, sym_trans, aborted, _Local, _, _) ->
    mnesia_recover:note_decision(Tid, aborted);
recover_coordinator(Tid, sync_sym_trans, committed, Local, _, _) ->
    mnesia_recover:note_decision(Tid, committed),
    do_dirty(Tid, Local);
recover_coordinator(Tid, sync_sym_trans, aborted, _Local, _, _) ->
    mnesia_recover:note_decision(Tid, aborted);

recover_coordinator(Tid, Protocol, committed, Local, DiscNs, RamNs)
  when Protocol =:= asym_trans; Protocol =:= sync_asym_trans ->
    D = #decision{tid = Tid, outcome = committed,
		  disc_nodes = DiscNs, ram_nodes = RamNs},
    mnesia_recover:log_decision(D),
    do_commit(Tid, Local);
recover_coordinator(Tid, Protocol, aborted, Local, DiscNs, RamNs)
  when Protocol =:= asym_trans; Protocol =:= sync_asym_trans ->
    D = #decision{tid = Tid, outcome = aborted,
		  disc_nodes = DiscNs, ram_nodes = RamNs},
    mnesia_recover:log_decision(D),
    do_abort(Tid, Local).

restore_stores(Coords, Tid, Store) ->
    Etstabs = gb_trees:get(Tid,Coords),
    Remaining  = lists:delete(Store, Etstabs),
    erase_ets_tabs(Remaining),
    gb_trees:update(Tid,[Store],Coords).

add_coord_store(Coords, Tid, Etab) ->
    Stores = gb_trees:get(Tid, Coords),
    gb_trees:update(Tid, [Etab|Stores], Coords).

del_coord_store(Coords, Tid, Current, Obsolete) ->
    Stores = gb_trees:get(Tid, Coords),
    Rest =
    	case Stores of
    	    [Obsolete, Current | Tail] -> Tail;
    	    [Current, Obsolete | Tail] -> Tail
    	end,
    ?ets_delete_table(Obsolete),
    gb_trees:update(Tid, [Current|Rest], Coords).

erase_ets_tabs([H | T]) ->
    ?ets_delete_table(H),
    erase_ets_tabs(T);
erase_ets_tabs([]) ->
    ok.

%% Clear one transactions all fixtables
clear_fixtable([Store|_]) ->
    Fixed = get_elements(fixtable, Store),
    lists:foreach(fun({Tab,Node}) ->
			  rpc:cast(Node, ?MODULE, fixtable, [Tab,false,self()])
		  end, Fixed).

%% Clear all fixtable Node have done
clear_fixtable(Node, State=#state{fixed_tabs = FT0}) ->
    case mnesia_lib:key_search_delete(Node, 1, FT0) of
	{none, _Ft} ->
	    State;
	{{Node,Tabs},FT} ->
	    lists:foreach(
	      fun(Tab) ->
		      case ?catch_val({Tab, storage_type}) of
			  {'EXIT', _} ->
			      ignore;
			  Storage ->
			      mnesia_lib:db_fixtable(Storage,Tab,false)
		      end
	      end, Tabs),
	    State#state{fixed_tabs=FT}
    end.

manage_fixtable(Tab,true,Requester,State=#state{fixed_tabs = FT0}) ->
    Node = node(Requester),
    case mnesia_lib:key_search_delete(Node, 1, FT0) of
	{none, FT}->
	    State#state{fixed_tabs=[{Node, [Tab]}|FT]};
	{{Node,Tabs},FT} ->
	    State#state{fixed_tabs=[{Node, [Tab|Tabs]}|FT]}
    end;
manage_fixtable(Tab,false,Requester,State = #state{fixed_tabs = FT0}) ->
    Node = node(Requester),
    case mnesia_lib:key_search_delete(Node, 1, FT0) of
	{none,_FT} -> State; % Hmm? Safeguard
	{{Node, Tabs0},FT} ->
	    case lists:delete(Tab, Tabs0) of
		[] -> State#state{fixed_tabs=FT};
		Tabs -> State#state{fixed_tabs=[{Node,Tabs}|FT]}
	    end
    end.

%% Deletes a pid from a list of participants
%% or from a gb_trees of coordinators
%% {none, All} or {Tr, Rest}
pid_search_delete(Pid, Trs) ->
    pid_search_delete(Pid, Trs, none, []).
pid_search_delete(Pid, [Tr = {Tid, _Ts} | Trs], _Val, Ack) when Tid#tid.pid == Pid ->
    pid_search_delete(Pid, Trs, Tr, Ack);
pid_search_delete(Pid, [Tr | Trs], Val, Ack) ->
    pid_search_delete(Pid, Trs, Val, [Tr | Ack]);

pid_search_delete(_Pid, [], Val, Ack) ->
    {Val, gb_trees:from_orddict(lists:reverse(Ack))}.

transaction_terminated(Tid)  ->
    mnesia_checkpoint:tm_exit_pending(Tid),
    Pid = Tid#tid.pid,
    if
	node(Pid) == node() ->
	    unlink(Pid);
	true ->  %% Do the Lamport thing here
	    mnesia_recover:sync_trans_tid_serial(Tid)
    end.

%% If there are an surrounding transaction, we inherit it's context
non_transaction(OldState={_,_,Trans}, Fun, Args, ActivityKind, Mod)
  when Trans /= non_transaction ->
    Kind = case ActivityKind of
	       sync_dirty -> sync;
	       _ -> async
	   end,
    case transaction(OldState, Fun, Args, infinity, Mod, Kind) of
	{atomic, Res} -> Res;
	{aborted,Res} -> exit(Res)
    end;
non_transaction(OldState, Fun, Args, ActivityKind, Mod) ->
    Id = {ActivityKind, self()},
    NewState = {Mod, Id, non_transaction},
    put(mnesia_activity_state, NewState),
    try apply(Fun, Args) of
	{'EXIT', Reason} -> exit(Reason);
	{aborted, Reason} -> mnesia:abort(Reason);
	Res -> Res
    catch
        throw:Throw     -> throw(Throw);
        error:Reason:ST -> exit({Reason, ST});
        exit:Reason     -> exit(Reason)
    after
	case OldState of
	    undefined -> erase(mnesia_activity_state);
	    _ -> put(mnesia_activity_state, OldState)
	end
    end.

transaction(OldTidTs, Fun, Args, Retries, Mod, Type) ->
    Factor = 1,
    case OldTidTs of
	undefined -> % Outer
	    execute_outer(Mod, Fun, Args, Factor, Retries, Type);
	{_, _, non_transaction} -> % Transaction inside ?sync_dirty
	    Res = execute_outer(Mod, Fun, Args, Factor, Retries, Type),
	    put(mnesia_activity_state, OldTidTs),
	    Res;
	{OldMod, Tid, Ts} ->  % Nested
	    execute_inner(Mod, Tid, OldMod, Ts, Fun, Args, Factor, Retries, Type);
	_ -> % Bad nesting
	    {aborted, nested_transaction}
    end.

execute_outer(Mod, Fun, Args, Factor, Retries, Type) ->
    case req(start_outer) of
	{error, Reason} ->
	    {aborted, Reason};
	{new_tid, Tid, Store} ->
	    Ts = #tidstore{store = Store},
	    NewTidTs = {Mod, Tid, Ts},
	    put(mnesia_activity_state, NewTidTs),
	    execute_transaction(Fun, Args, Factor, Retries, Type)
    end.

execute_inner(Mod, Tid, OldMod, Ts, Fun, Args, Factor, Retries, Type) ->
    case req({add_store, Tid}) of
	{error, Reason} ->
	    {aborted, Reason};
	{new_store, Ets} ->
	    copy_ets(Ts#tidstore.store, Ets),
	    Up = [{OldMod,Ts#tidstore.store} | Ts#tidstore.up_stores],
	    NewTs = Ts#tidstore{level = 1 + Ts#tidstore.level,
				store = Ets,
				up_stores = Up},
	    NewTidTs = {Mod, Tid, NewTs},
	    put(mnesia_activity_state, NewTidTs),
	    execute_transaction(Fun, Args, Factor, Retries, Type)
    end.

copy_ets(From, To) ->
    do_copy_ets(?ets_first(From), From, To).
do_copy_ets('$end_of_table', _,_) ->
    ok;
do_copy_ets(K, From, To) ->
    Objs = ?ets_lookup(From, K),
    insert_objs(Objs, To),
    do_copy_ets(?ets_next(From, K), From, To).

insert_objs([H|T], Tab) ->
    ?ets_insert(Tab, H),
    insert_objs(T, Tab);
insert_objs([], _Tab) ->
    ok.

execute_transaction(Fun, Args, Factor, Retries, Type) ->
    try apply_fun(Fun, Args, Type) of
	{atomic, Value} ->
	    mnesia_lib:incr_counter(trans_commits),
	    erase(mnesia_activity_state),
	    %% no need to clear locks, already done by commit ...
	    %% Flush any un processed mnesia_down messages we might have
	    flush_downs(),
	    ?SAFE(unlink(whereis(?MODULE))),
	    {atomic, Value};
	{do_abort, Reason} ->
	    check_exit(Fun, Args, Factor, Retries, {aborted, Reason}, Type);
	{nested_atomic, Value} ->
	    mnesia_lib:incr_counter(trans_commits),
	    {atomic, Value}
    catch throw:Value ->  %% User called throw
	    Reason = {aborted, {throw, Value}},
	    return_abort(Fun, Args, Reason);
	  error:Reason:ST ->
	    check_exit(Fun, Args, Factor, Retries, {Reason,ST}, Type);
	  _:Reason ->
	    check_exit(Fun, Args, Factor, Retries, Reason, Type)
    end.

apply_fun(Fun, Args, Type) ->
    Result = apply(Fun, Args),
    case t_commit(Type) of
	do_commit ->
            {atomic, Result};
        do_commit_nested ->
            {nested_atomic, Result};
	{do_abort, {aborted, Reason}} ->
	    {do_abort, Reason};
	{do_abort, _} = Abort ->
	    Abort
    end.

check_exit(Fun, Args, Factor, Retries, Reason, Type) ->
    case Reason of
	{aborted, C = #cyclic{}} ->
	    maybe_restart(Fun, Args, Factor, Retries, Type, C);
	{aborted, {node_not_running, N}} ->
	    maybe_restart(Fun, Args, Factor, Retries, Type, {node_not_running, N});
	{aborted, {bad_commit, N}} ->
	    maybe_restart(Fun, Args, Factor, Retries, Type, {bad_commit, N});
	_ ->
	    return_abort(Fun, Args, Reason)
    end.

maybe_restart(Fun, Args, Factor, Retries, Type, Why) ->
    {Mod, Tid, Ts} = get(mnesia_activity_state),
    case try_again(Retries) of
	yes when Ts#tidstore.level == 1 ->
	    restart(Mod, Tid, Ts, Fun, Args, Factor, Retries, Type, Why);
	yes ->
	    return_abort(Fun, Args, Why);
	no ->
	    return_abort(Fun, Args, {aborted, nomore})
    end.

try_again(infinity) -> yes;
try_again(X) when is_number(X) , X > 1 -> yes;
try_again(_) -> no.

%% We can only restart toplevel transactions.
%% If a deadlock situation occurs in a nested transaction
%% The whole thing including all nested transactions need to be
%% restarted. The stack is thus popped by a consecutive series of
%% exit({aborted, #cyclic{}}) calls

restart(Mod, Tid, Ts, Fun, Args, Factor0, Retries0, Type, Why) ->
    mnesia_lib:incr_counter(trans_restarts),
    Retries = decr(Retries0),
    case Why of
	{bad_commit, _N} ->
	    return_abort(Fun, Args, Why),
	    Factor = 1,
	    SleepTime = mnesia_lib:random_time(Factor, Tid#tid.counter),
	    dbg_out("Restarting transaction ~w: in ~wms ~w~n", [Tid, SleepTime, Why]),
	    timer:sleep(SleepTime),
	    execute_outer(Mod, Fun, Args, Factor, Retries, Type);
	{node_not_running, _N} ->   %% Avoids hanging in receive_release_tid_ack
	    return_abort(Fun, Args, Why),
	    Factor = 1,
	    SleepTime = mnesia_lib:random_time(Factor, Tid#tid.counter),
	    dbg_out("Restarting transaction ~w: in ~wms ~w~n", [Tid, SleepTime, Why]),
	    timer:sleep(SleepTime),
	    execute_outer(Mod, Fun, Args, Factor, Retries, Type);
	_ ->
	    SleepTime = mnesia_lib:random_time(Factor0, Tid#tid.counter),
	    dbg_out("Restarting transaction ~w: in ~wms ~w~n", [Tid, SleepTime, Why]),

	    if
		Factor0 /= 10 ->
		    ignore;
		true ->
		    %% Our serial may be much larger than other nodes ditto
		    AllNodes = val({current, db_nodes}),
		    verbose("Sync serial ~p~n", [Tid]),
		    rpc:abcast(AllNodes, ?MODULE, {sync_trans_serial, Tid})
	    end,
	    intercept_friends(Tid, Ts),
	    Store = Ts#tidstore.store,
	    Nodes = get_elements(nodes,Store),
	    ?MODULE ! {self(), {restart, Tid, Store}},
	    mnesia_locker:send_release_tid(Nodes, Tid),
	    timer:sleep(SleepTime),
	    mnesia_locker:receive_release_tid_acc(Nodes, Tid),
	    case get_restarted(Tid) of
		{restarted, Tid} ->
		    execute_transaction(Fun, Args, Factor0 + 1,
					Retries, Type);
		{error, Reason} ->
		    mnesia:abort(Reason)
	    end
    end.

get_restarted(Tid) ->
    case Res = rec() of
	{restarted, Tid} ->
	    Res;
	{error,_} ->
	    Res;
	_ -> %% We could get a couple of aborts to many.
	    get_restarted(Tid)
    end.

decr(infinity) -> infinity;
decr(X) when is_integer(X), X > 1 -> X - 1;
decr(_X) -> 0.

return_abort(Fun, Args, Reason)  ->
    {_Mod, Tid, Ts} = get(mnesia_activity_state),
    dbg_out("Transaction ~p calling ~tp with ~tp failed: ~n ~tp~n",
	    [Tid, Fun, Args, Reason]),
    OldStore = Ts#tidstore.store,
    Nodes = get_elements(nodes, OldStore),
    intercept_friends(Tid, Ts),
    ?SAFE(mnesia_lib:incr_counter(trans_failures)),
    Level = Ts#tidstore.level,
    if
	Level == 1 ->
	    mnesia_locker:async_release_tid(Nodes, Tid),
	    ?SAFE(?MODULE ! {delete_transaction, Tid}),
	    erase(mnesia_activity_state),
	    flush_downs(),
	    ?SAFE(unlink(whereis(?MODULE))),
	    {aborted, mnesia_lib:fix_error(Reason)};
	true ->
	    %% Nested transaction
	    [{OldMod,NewStore} | Tail] = Ts#tidstore.up_stores,
	    req({del_store, Tid, NewStore, OldStore, true}),
	    Ts2 = Ts#tidstore{store = NewStore,
			      up_stores = Tail,
			      level = Level - 1},
	    NewTidTs = {OldMod, Tid, Ts2},
	    put(mnesia_activity_state, NewTidTs),
	    case Reason of
		#cyclic{} ->
		    exit({aborted, Reason});
		{node_not_running, _N} ->
		    exit({aborted, Reason});
		{bad_commit, _N}->
		    exit({aborted, Reason});
		_ ->
		    {aborted, mnesia_lib:fix_error(Reason)}
	    end
    end.

flush_downs() ->
    receive
	{?MODULE, _, _} -> flush_downs(); % Votes
	{mnesia_down, _} -> flush_downs()
    after 0 -> flushed
    end.


put_activity_id(MTT) ->
    put_activity_id(MTT, undefined).
put_activity_id(undefined,_) ->
    erase_activity_id();
put_activity_id({Mod, Tid = #tid{}, Ts = #tidstore{}},Fun) ->
    flush_downs(),
    Store = Ts#tidstore.store,
    if
	is_function(Fun) ->
	    ?ets_insert(Store, {friends, {stop,Fun}});
	true ->
	    ?ets_insert(Store, {friends, self()})
    end,
    NewTidTs = {Mod, Tid, Ts},
    put(mnesia_activity_state, NewTidTs);
put_activity_id(SimpleState,_) ->
    put(mnesia_activity_state, SimpleState).

erase_activity_id() ->
    flush_downs(),
    erase(mnesia_activity_state).

get_elements(Type,Store) ->
    try ?ets_lookup(Store, Type) of
	[] -> [];
	[{_,Val}] -> [Val];
	Vals -> [Val|| {_,Val} <- Vals]
    catch error:_ -> []
    end.

opt_propagate_store(_Current, _Obsolete, false) ->
    ok;
opt_propagate_store(Current, Obsolete, true) ->
    propagate_store(Current, nodes, get_elements(nodes,Obsolete)),
    propagate_store(Current, fixtable, get_elements(fixtable,Obsolete)),
    propagate_store(Current, friends, get_elements(friends, Obsolete)).

propagate_store(Store, Var, [Val | Vals]) ->
    ?ets_insert(Store, {Var, Val}),
    propagate_store(Store, Var, Vals);
propagate_store(_Store, _Var, []) ->
    ok.

%% Tell all processes that are cooperating with the current transaction
intercept_friends(_Tid, Ts) ->
    Friends = get_elements(friends,Ts#tidstore.store),
    intercept_best_friend(Friends, false).

intercept_best_friend([],_) ->    ok;
intercept_best_friend([{stop,Fun} | R],Ignore) ->
    ?CATCH(Fun()),
    intercept_best_friend(R,Ignore);
intercept_best_friend([Pid | R],false) ->
    Pid ! {activity_ended, undefined, self()},
    wait_for_best_friend(Pid, 0),
    intercept_best_friend(R,true);
intercept_best_friend([_|R],true) ->
    intercept_best_friend(R,true).

wait_for_best_friend(Pid, Timeout) ->
    receive
	{'EXIT', Pid, _} -> ok;
	{activity_ended, _, Pid} -> ok
    after Timeout ->
	    case erlang:is_process_alive(Pid) of
		true -> wait_for_best_friend(Pid, 1000);
		false -> ok
	    end
    end.

dirty(Protocol, Item) ->
    {{Tab, Key}, _Val, _Op} = Item,
    Tid = {dirty, self()},
    Prep = prepare_items(Tid, Tab, Key, [Item], #prep{protocol= Protocol}),
    CR =  Prep#prep.records,
    case Protocol of
	async_dirty ->
	    %% Send commit records to the other involved nodes,
	    %% but do only wait for one node to complete.
	    %% Preferably, the local node if possible.

	    ReadNode = val({Tab, where_to_read}),
	    {WaitFor, FirstRes} = async_send_dirty(Tid, CR, Tab, ReadNode),
	    rec_dirty(WaitFor, FirstRes);

	sync_dirty ->
	    %% Send commit records to the other involved nodes,
	    %% and wait for all nodes to complete
	    {WaitFor, FirstRes} = sync_send_dirty(Tid, CR, Tab, []),
	    rec_dirty(WaitFor, FirstRes);
	_ ->
	    mnesia:abort({bad_activity, Protocol})
    end.

%% This is the commit function, The first thing it does,
%% is to find out which nodes that have been participating
%% in this particular transaction, all of the mnesia_locker:lock*
%% functions insert the names of the nodes where it acquires locks
%% into the local shadow Store
%% This function exacutes in the context of the user process
t_commit(Type) ->
    {_Mod, Tid, Ts} = get(mnesia_activity_state),
    Store = Ts#tidstore.store,
    if
	Ts#tidstore.level == 1 ->
	    intercept_friends(Tid, Ts),
	    %% N is number of updates
	    case arrange(Tid, Store, Type) of
		{N, Prep} when N > 0 ->
		    multi_commit(Prep#prep.protocol,
				 majority_attr(Prep),
				 Tid, Prep#prep.records, Store);
		{0, Prep} ->
		    multi_commit(read_only,
				 majority_attr(Prep),
				 Tid, Prep#prep.records, Store)
	    end;
	true ->
	    %% nested commit
	    Level = Ts#tidstore.level,
	    [{OldMod,Obsolete} | Tail] = Ts#tidstore.up_stores,
	    req({del_store, Tid, Store, Obsolete, false}),
	    NewTs = Ts#tidstore{store = Store,
				up_stores = Tail,
				level = Level - 1},
	    NewTidTs = {OldMod, Tid, NewTs},
	    put(mnesia_activity_state, NewTidTs),
	    do_commit_nested
    end.

majority_attr(#prep{majority = M}) ->
    M.


%% This function arranges for all objects we shall write in S to be
%% in a list of {Node, CommitRecord}
%% Important function for the performance of mnesia.

arrange(Tid, Store, Type) ->
    %% The local node is always included
    Nodes = get_elements(nodes,Store),
    Recs = prep_recs(Nodes, []),
    Key = ?ets_first(Store),
    N = 0,
    Prep =
	case Type of
	    async -> #prep{protocol = sym_trans, records = Recs};
	    sync -> #prep{protocol = sync_sym_trans, records = Recs}
	end,
    {New, Prepared} = do_arrange(Tid, Store, Key, Prep, N),
    {New, Prepared#prep{records = reverse(Prepared#prep.records)}}.

reverse([]) ->
    [];
reverse([H=#commit{ram_copies=Ram, disc_copies=DC,
		   disc_only_copies=DOC, ext=Ext}
	 |R]) ->
    [
     H#commit{
       ram_copies       = lists:reverse(Ram),
       disc_copies      = lists:reverse(DC),
       disc_only_copies = lists:reverse(DOC),
       ext              = [{Type, lists:reverse(E)} || {Type,E} <- Ext]
      }
     | reverse(R)].

prep_recs([N | Nodes], Recs) ->
    prep_recs(Nodes, [#commit{decision = presume_commit, node = N} | Recs]);
prep_recs([], Recs) ->
    Recs.

%% storage_types is a list of {Node, Storage} tuples
%% where each tuple represents an active replica
do_arrange(Tid, Store, {Tab, Key}, Prep, N) ->
    Oid = {Tab, Key},
    Items = ?ets_lookup(Store, Oid), %% Store is a bag
    P2 = prepare_items(Tid, Tab, Key, Items, Prep),
    do_arrange(Tid, Store, ?ets_next(Store, Oid), P2, N + 1);
do_arrange(Tid, Store, SchemaKey, Prep, N) when SchemaKey == op ->
    Items = ?ets_lookup(Store, SchemaKey), %% Store is a bag
    P2 = prepare_schema_items(Tid, Items, Prep),
    do_arrange(Tid, Store, ?ets_next(Store, SchemaKey), P2, N + 1);
do_arrange(Tid, Store, RestoreKey, Prep, N) when RestoreKey == restore_op ->
    [{restore_op, R}] = ?ets_lookup(Store, RestoreKey),
    Fun = fun({Tab, Key}, CommitRecs, _RecName, Where, Snmp) ->
		  Item = [{{Tab, Key}, {Tab, Key}, delete}],
		  do_prepare_items(Tid, Tab, Key, Where, Snmp, Item, CommitRecs);
	     (BupRec, CommitRecs, RecName, Where, Snmp) ->
		  Tab = element(1, BupRec),
		  Key = element(2, BupRec),
		  Item =
		      if
			  Tab == RecName ->
			      [{{Tab, Key}, BupRec, write}];
			  true ->
			      BupRec2 = setelement(1, BupRec, RecName),
			      [{{Tab, Key}, BupRec2, write}]
		      end,
		  do_prepare_items(Tid, Tab, Key, Where, Snmp, Item, CommitRecs)
	  end,
    Recs2 = mnesia_schema:arrange_restore(R, Fun, Prep#prep.records),
    P2 = Prep#prep{protocol = asym_trans, records = Recs2},
    do_arrange(Tid, Store, ?ets_next(Store, RestoreKey), P2, N + 1);
do_arrange(_Tid, _Store, '$end_of_table', Prep, N) ->
    case Prep of
        #prep{sync=true, protocol=asym_trans} ->
            {N, Prep#prep{protocol=sync_asym_trans}};
        _ ->
            {N, Prep}
    end;
do_arrange(Tid, Store, sticky, Prep, N) ->
    P2 = Prep#prep{sync=true},
    do_arrange(Tid, Store, ?ets_next(Store, sticky), P2, N);
do_arrange(Tid, Store, IgnoredKey, Prep, N) -> %% locks, nodes ... local atoms...
    do_arrange(Tid, Store, ?ets_next(Store, IgnoredKey), Prep, N).

%% Returns a prep record  with all items in reverse order
prepare_schema_items(Tid, Items, Prep) ->
    Types = [{N, schema_ops} || N <- val({current, db_nodes})],
    Recs = prepare_nodes(Tid, Types, Items, Prep#prep.records, schema),
    Prep#prep{protocol = asym_trans, records = Recs}.

%% Returns a prep record with all items in reverse order
prepare_items(Tid, Tab, Key, Items, Prep) when Prep#prep.prev_tab == Tab ->
    Types = Prep#prep.prev_types,
    Snmp = Prep#prep.prev_snmp,
    Recs = Prep#prep.records,
    Recs2 = do_prepare_items(Tid, Tab, Key, Types, Snmp, Items, Recs),
    Prep#prep{records = Recs2};

prepare_items(Tid, Tab, Key, Items, Prep) ->
    Types = val({Tab, where_to_commit}),
    case Types of
	[] -> mnesia:abort({no_exists, Tab});
	{blocked, _} ->
	    unblocked = req({unblock_me, Tab}),
	    prepare_items(Tid, Tab, Key, Items, Prep);
	_ ->
	    Majority = needs_majority(Tab, Prep),
	    Snmp = val({Tab, snmp}),
	    Recs2 = do_prepare_items(Tid, Tab, Key, Types,
				     Snmp, Items, Prep#prep.records),
	    Prep2 = Prep#prep{records = Recs2, prev_tab = Tab,
			      majority = Majority,
			      prev_types = Types, prev_snmp = Snmp},
	    check_prep(Prep2, Types)
    end.

do_prepare_items(Tid, Tab, Key, Types, Snmp, Items, Recs) ->
    Recs2 = prepare_snmp(Tid, Tab, Key, Types, Snmp, Items, Recs), % May exit
    prepare_nodes(Tid, Types, Items, Recs2, normal).


needs_majority(Tab, #prep{majority = M}) ->
    case lists:keymember(Tab, 1, M) of
	true ->
	    M;
	false ->
	    case ?catch_val({Tab, majority}) of
		{'EXIT', _} ->
		    M;
		false ->
		    M;
		true ->
		    CopyHolders = val({Tab, all_nodes}),
		    [{Tab, CopyHolders} | M]
	    end
    end.

have_majority([], _) ->
    ok;
have_majority([{Tab, AllNodes} | Rest], Nodes) ->
    case mnesia_lib:have_majority(Tab, AllNodes, Nodes) of
	true ->
	    have_majority(Rest, Nodes);
	false ->
	    {error, Tab}
    end.

prepare_snmp(Tab, Key, Items) ->
    case val({Tab, snmp}) of
	[] ->
	    [];
	Ustruct when Key /= '_' ->
	    {_Oid, _Val, Op} = hd(Items),
	    %% Still making snmp oid (not used) because we want to catch errors here
	    %% And also it keeps backwards comp. with old nodes.
	    SnmpOid = mnesia_snmp_hook:key_to_oid(Tab, Key, Ustruct), % May exit
	    [{Op, Tab, Key, SnmpOid}];
	_ ->
	    [{clear_table, Tab}]
    end.

prepare_snmp(_Tid, _Tab, _Key, _Types, [], _Items, Recs) ->
    Recs;

prepare_snmp(Tid, Tab, Key, Types, Us, Items, Recs) ->
    if Key /= '_' ->
	    {_Oid, _Val, Op} = hd(Items),
	    SnmpOid = mnesia_snmp_hook:key_to_oid(Tab, Key, Us), % May exit
	    prepare_nodes(Tid, Types, [{Op, Tab, Key, SnmpOid}], Recs, snmp);
       Key == '_' ->
	    prepare_nodes(Tid, Types, [{clear_table, Tab}], Recs, snmp)
    end.

check_prep(#prep{majority = [], types = Types} = Prep, Types) ->
    Prep;
check_prep(#prep{majority = M, types = undefined} = Prep, Types) ->
    Protocol = if M == [] ->
		       Prep#prep.protocol;
		  true ->
		       asym_trans
	       end,
    Prep#prep{protocol = Protocol, types = Types};
check_prep(Prep, _Types) ->
    Prep#prep{protocol = asym_trans}.

%% Returns a list of commit records
prepare_nodes(Tid, [{Node, Storage} | Rest], Items, C, Kind) ->
    {Rec, C2} = pick_node(Tid, Node, C, []),
    Rec2 = prepare_node(Node, Storage, Items, Rec, Kind),
    [Rec2 | prepare_nodes(Tid, Rest, Items, C2, Kind)];
prepare_nodes(_Tid, [], _Items, CommitRecords, _Kind) ->
    CommitRecords.

pick_node(Tid, Node, [Rec | Rest], Done) ->
    if
	Rec#commit.node == Node ->
	    {Rec, Done ++ Rest};
	true ->
	    pick_node(Tid, Node, Rest, [Rec | Done])
    end;
pick_node({dirty,_}, Node, [], Done) ->
    {#commit{decision = presume_commit, node = Node}, Done};
pick_node(_Tid, Node, [], _Done) ->
    mnesia:abort({bad_commit, {missing_lock, Node}}).

prepare_node(Node, Storage, [Item | Items], #commit{ext=Ext0}=Rec, Kind) when Kind == snmp ->
    Rec2 = case lists:keytake(snmp, 1, Ext0) of
	       false ->
		   Rec#commit{ext = [{snmp,[Item]}|Ext0]};
	       {_, {snmp,Snmp},Ext} ->
		   Rec#commit{ext = [{snmp,[Item|Snmp]}|Ext]}
	   end,
    prepare_node(Node, Storage, Items, Rec2, Kind);
prepare_node(Node, Storage, [Item | Items], Rec, Kind) when Kind /= schema ->
    Rec2 =
	case Storage of
	    ram_copies ->
		Rec#commit{ram_copies = [Item | Rec#commit.ram_copies]};
	    disc_copies ->
		Rec#commit{disc_copies = [Item | Rec#commit.disc_copies]};
	    disc_only_copies ->
		Rec#commit{disc_only_copies =
			   [Item | Rec#commit.disc_only_copies]};
	    {ext, Alias, Mod} ->
		Ext0 = Rec#commit.ext,
		case lists:keytake(ext_copies, 1, Ext0) of
		    false ->
			Rec#commit{ext = [{ext_copies, [{{ext,Alias,Mod}, Item}]}|Ext0]};
		    {_,{_,EC},Ext} ->
			Rec#commit{ext = [{ext_copies, [{{ext,Alias,Mod}, Item}|EC]}|Ext]}
		end
	end,
    prepare_node(Node, Storage, Items, Rec2, Kind);
prepare_node(_Node, _Storage, Items, Rec, Kind)
  when Kind == schema, Rec#commit.schema_ops == []  ->
    Rec#commit{schema_ops = Items};
prepare_node(_Node, _Storage, [], Rec, _Kind) ->
    Rec.

%% multi_commit((Protocol, Tid, CommitRecords, Store)
%% Local work is always performed in users process
multi_commit(read_only, _Maj = [], Tid, CR, _Store) ->
    %% This featherweight commit protocol is used when no
    %% updates has been performed in the transaction.

    {DiscNs, RamNs} = commit_nodes(CR, [], []),
    Msg = {Tid, simple_commit},
    rpc:abcast(DiscNs -- [node()], ?MODULE, Msg),
    rpc:abcast(RamNs -- [node()], ?MODULE, Msg),
    mnesia_recover:note_decision(Tid, committed),
    mnesia_locker:release_tid(Tid),
    ?MODULE ! {delete_transaction, Tid},
    do_commit;

multi_commit(sym_trans, _Maj = [], Tid, CR, Store) ->
    %% This lightweight commit protocol is used when all
    %% the involved tables are replicated symmetrically.
    %% Their storage types must match on each node.
    %%
    %% 1  Ask the other involved nodes if they want to commit
    %%    All involved nodes votes yes if they are up
    %% 2a Somebody has voted no
    %%    Tell all yes voters to do_abort
    %% 2b Everybody has voted yes
    %%    Tell everybody to do_commit. I.e. that they should
    %%    prepare the commit, log the commit record and
    %%    perform the updates.
    %%
    %%    The outcome is kept 3 minutes in the transient decision table.
    %%
    %% Recovery:
    %%    If somebody dies before the coordinator has
    %%    broadcasted do_commit, the transaction is aborted.
    %%
    %%    If a participant dies, the table load algorithm
    %%    ensures that the contents of the involved tables
    %%    are picked from another node.
    %%
    %%    If the coordinator dies, each participants checks
    %%    the outcome with all the others. If all are uncertain
    %%    about the outcome, the transaction is aborted. If
    %%    somebody knows the outcome the others will follow.

    {DiscNs, RamNs} = commit_nodes(CR, [], []),
    Pending = mnesia_checkpoint:tm_enter_pending(Tid, DiscNs, RamNs),
    ?ets_insert(Store, Pending),

    {WaitFor, Local} = ask_commit(sym_trans, Tid, CR, DiscNs, RamNs),
    {Outcome, []} = rec_all(WaitFor, Tid, do_commit, []),
    ?eval_debug_fun({?MODULE, multi_commit_sym},
		    [{tid, Tid}, {outcome, Outcome}]),
    rpc:abcast(DiscNs -- [node()], ?MODULE, {Tid, Outcome}),
    rpc:abcast(RamNs -- [node()], ?MODULE, {Tid, Outcome}),
    case Outcome of
	do_commit ->
	    mnesia_recover:note_decision(Tid, committed),
	    do_dirty(Tid, Local),
	    mnesia_locker:release_tid(Tid),
	    ?MODULE ! {delete_transaction, Tid};
	{do_abort, _Reason} ->
	    mnesia_recover:note_decision(Tid, aborted)
    end,
    ?eval_debug_fun({?MODULE, multi_commit_sym, post},
		    [{tid, Tid}, {outcome, Outcome}]),
    Outcome;

multi_commit(sync_sym_trans, _Maj = [], Tid, CR, Store) ->
    %%   This protocol is the same as sym_trans except that it
    %%   uses synchronized calls to disk_log and synchronized commits
    %%   when several nodes are involved.

    {DiscNs, RamNs} = commit_nodes(CR, [], []),
    Pending = mnesia_checkpoint:tm_enter_pending(Tid, DiscNs, RamNs),
    ?ets_insert(Store, Pending),

    {WaitFor, Local} = ask_commit(sync_sym_trans, Tid, CR, DiscNs, RamNs),
    {Outcome, []} = rec_all(WaitFor, Tid, do_commit, []),
    ?eval_debug_fun({?MODULE, multi_commit_sym_sync},
		    [{tid, Tid}, {outcome, Outcome}]),
    [?ets_insert(Store, {waiting_for_commit_ack, Node}) || Node <- WaitFor],
    rpc:abcast(DiscNs -- [node()], ?MODULE, {Tid, Outcome}),
    rpc:abcast(RamNs -- [node()], ?MODULE, {Tid, Outcome}),
    case Outcome of
	do_commit ->
	    mnesia_recover:note_decision(Tid, committed),
	    mnesia_log:slog(Local),
	    do_commit(Tid, Local),
	    %% Just wait for completion result is ignore.
	    rec_all(WaitFor, Tid, ignore, []),
	    mnesia_locker:release_tid(Tid),
	    ?MODULE ! {delete_transaction, Tid};
	{do_abort, _Reason} ->
	    mnesia_recover:note_decision(Tid, aborted)
    end,
    ?eval_debug_fun({?MODULE, multi_commit_sym, post},
		    [{tid, Tid}, {outcome, Outcome}]),
    Outcome;

multi_commit(Protocol, Majority, Tid, CR, Store)
  when Protocol =:= asym_trans; Protocol =:= sync_asym_trans ->
    %% This more expensive commit protocol is used when
    %% table definitions are changed (schema transactions).
    %% It is also used when the involved tables are
    %% replicated asymetrically. If the storage type differs
    %% on at least one node this protocol is used.
    %%
    %% 1 Ask the other involved nodes if they want to commit.
    %%   All involved nodes prepares the commit, logs a presume_abort
    %%   commit record and votes yes or no depending of the
    %%   outcome of the prepare. The preparation is also performed
    %%   by the coordinator.
    %%
    %% 2a Somebody has died or voted no
    %%    Tell all yes voters to do_abort
    %% 2b Everybody has voted yes
    %%    Put a unclear marker in the log.
    %%    Tell the others to pre_commit. I.e. that they should
    %%    put a unclear marker in the log and reply
    %%    acc_pre_commit when they are done.
    %%
    %% 3a Somebody died
    %%    Tell the remaining participants to do_abort
    %% 3b Everybody has replied acc_pre_commit
    %%    Tell everybody to committed. I.e that they should
    %%    put a committed marker in the log, perform the updates
    %%    and reply done_commit when they are done. The coordinator
    %%    must wait with putting his committed marker inte the log
    %%    until the committed has been sent to all the others.
    %%    Then he performs local commit before collecting replies.
    %%
    %% 4  Everybody has either died or replied done_commit
    %%    Return to the caller.
    %%
    %% Recovery:
    %%    If the coordinator dies, the participants (and
    %%    the coordinator when he starts again) must do
    %%    the following:
    %%
    %%    If we have no unclear marker in the log we may
    %%    safely abort, since we know that nobody may have
    %%    decided to commit yet.
    %%
    %%    If we have a committed marker in the log we may
    %%    safely commit since we know that everybody else
    %%    also will come to this conclusion.
    %%
    %%    If we have a unclear marker but no committed
    %%    in the log we are uncertain about the real outcome
    %%    of the transaction and must ask the others before
    %%    we can decide what to do. If someone knows the
    %%    outcome we will do the same. If nobody knows, we
    %%    will wait for the remaining involved nodes to come
    %%    up. When all involved nodes are up and uncertain,
    %%    we decide to commit (first put a committed marker
    %%    in the log, then do the updates).

    D = #decision{tid = Tid, outcome = presume_abort},
    {D2, CR2} = commit_decision(D, CR, [], []),
    DiscNs = D2#decision.disc_nodes,
    RamNs = D2#decision.ram_nodes,
    case have_majority(Majority, DiscNs ++ RamNs) of
	ok  -> ok;
	{error, Tab} -> mnesia:abort({no_majority, Tab})
    end,
    Pending = mnesia_checkpoint:tm_enter_pending(Tid, DiscNs, RamNs),
    ?ets_insert(Store, Pending),
    {WaitFor, Local} = ask_commit(Protocol, Tid, CR2, DiscNs, RamNs),
    SchemaPrep = ?CATCH(mnesia_schema:prepare_commit(Tid, Local, {coord, WaitFor})),
    {Votes, Pids} = rec_all(WaitFor, Tid, do_commit, []),

    ?eval_debug_fun({?MODULE, multi_commit_asym_got_votes},
		    [{tid, Tid}, {votes, Votes}]),
    case Votes of
	do_commit ->
	    case SchemaPrep of
		{_Modified, C = #commit{}, DumperMode} ->
		    mnesia_log:log(C), % C is not a binary
		    ?eval_debug_fun({?MODULE, multi_commit_asym_log_commit_rec},
				    [{tid, Tid}]),

		    D3 = C#commit.decision,
		    D4 = D3#decision{outcome = unclear},
		    mnesia_recover:log_decision(D4),
		    ?eval_debug_fun({?MODULE, multi_commit_asym_log_commit_dec},
				    [{tid, Tid}]),
		    tell_participants(Pids, {Tid, pre_commit}),
		    %% Now we are uncertain and we do not know
		    %% if all participants have logged that
		    %% they are uncertain or not
		    rec_acc_pre_commit(Pids, Tid, Store, {C,Local},
				       do_commit, DumperMode, [], []);
		{'EXIT', Reason} ->
		    %% The others have logged the commit
		    %% record but they are not uncertain
		    mnesia_recover:note_decision(Tid, aborted),
		    ?eval_debug_fun({?MODULE, multi_commit_asym_prepare_exit},
				    [{tid, Tid}]),
		    tell_participants(Pids, {Tid, {do_abort, Reason}}),
		    do_abort(Tid, Local),
		    {do_abort, Reason}
	    end;

	{do_abort, Reason} ->
	    %% The others have logged the commit
	    %% record but they are not uncertain
	    mnesia_recover:note_decision(Tid, aborted),
	    ?eval_debug_fun({?MODULE, multi_commit_asym_do_abort}, [{tid, Tid}]),
	    tell_participants(Pids, {Tid, {do_abort, Reason}}),
	    do_abort(Tid, Local),
	    {do_abort, Reason}
    end.

%% Returns do_commit or {do_abort, Reason}
rec_acc_pre_commit([Pid | Tail], Tid, Store, Commit, Res, DumperMode,
		   GoodPids, AckPids) ->
    receive
	{?MODULE, _, {acc_pre_commit, Tid, Pid, true}} ->
	    rec_acc_pre_commit(Tail, Tid, Store, Commit, Res, DumperMode,
			       [Pid | GoodPids], [Pid | AckPids]);

	{?MODULE, _, {acc_pre_commit, Tid, Pid, false}} ->
	    rec_acc_pre_commit(Tail, Tid, Store, Commit, Res, DumperMode,
			       [Pid | GoodPids], AckPids);

	{?MODULE, _, {acc_pre_commit, Tid, Pid}} ->
	    %% Kept for backwards compatibility. Remove after Mnesia 4.x
	    rec_acc_pre_commit(Tail, Tid, Store, Commit, Res, DumperMode,
			       [Pid | GoodPids], [Pid | AckPids]);
	{?MODULE, _, {do_abort, Tid, Pid, _Reason}} ->
	    AbortRes = {do_abort, {bad_commit, node(Pid)}},
	    rec_acc_pre_commit(Tail, Tid, Store, Commit, AbortRes, DumperMode,
			       GoodPids, AckPids);
	{mnesia_down, Node} when Node == node(Pid) ->
	    AbortRes = {do_abort, {bad_commit, Node}},
	    ?SAFE(Pid ! {Tid, AbortRes}),  %% Tell him that he has died
	    rec_acc_pre_commit(Tail, Tid, Store, Commit, AbortRes, DumperMode,
			       GoodPids, AckPids)
    end;
rec_acc_pre_commit([], Tid, Store, {Commit,OrigC}, Res, DumperMode, GoodPids, AckPids) ->
    D = Commit#commit.decision,
    case Res of
	do_commit ->
	    %% Now everybody knows that the others
	    %% has voted yes. We also know that
	    %% everybody are uncertain.
	    prepare_sync_schema_commit(Store, AckPids),
	    tell_participants(GoodPids, {Tid, committed}),
	    D2 = D#decision{outcome = committed},
	    mnesia_recover:log_decision(D2),
            ?eval_debug_fun({?MODULE, rec_acc_pre_commit_log_commit},
			    [{tid, Tid}]),

	    %% Now we have safely logged committed
	    %% and we can recover without asking others
	    do_commit(Tid, Commit, DumperMode),
            ?eval_debug_fun({?MODULE, rec_acc_pre_commit_done_commit},
			    [{tid, Tid}]),
	    sync_schema_commit(Tid, Store, AckPids),
	    mnesia_locker:release_tid(Tid),
	    ?MODULE ! {delete_transaction, Tid};

	{do_abort, Reason} ->
	    tell_participants(GoodPids, {Tid, {do_abort, Reason}}),
	    D2 = D#decision{outcome = aborted},
	    mnesia_recover:log_decision(D2),
            ?eval_debug_fun({?MODULE, rec_acc_pre_commit_log_abort},
			    [{tid, Tid}]),
	    do_abort(Tid, OrigC),
	    ?eval_debug_fun({?MODULE, rec_acc_pre_commit_done_abort},
			    [{tid, Tid}])
    end,
    Res.

%% Note all nodes in case of mnesia_down mgt
%% sync_schema_commit is (ab)used for sync_asym_trans as well.
prepare_sync_schema_commit(_Store, []) ->
    ok;
prepare_sync_schema_commit(Store, [Pid | Pids]) ->
    ?ets_insert(Store, {waiting_for_commit_ack, node(Pid)}),
    prepare_sync_schema_commit(Store, Pids).

sync_schema_commit(_Tid, _Store, []) ->
    ok;
sync_schema_commit(Tid, Store, [Pid | Tail]) ->
    receive
	{?MODULE, _, {schema_commit, Tid, Pid}} ->
	    ?ets_match_delete(Store, {waiting_for_commit_ack, node(Pid)}),
	    sync_schema_commit(Tid, Store, Tail);

	{mnesia_down, Node} when Node == node(Pid) ->
	    ?ets_match_delete(Store, {waiting_for_commit_ack, Node}),
	    sync_schema_commit(Tid, Store, Tail)
    end.

tell_participants([Pid | Pids], Msg) ->
    Pid ! Msg,
    tell_participants(Pids, Msg);
tell_participants([], _Msg) ->
    ok.

-spec commit_participant(_, _, _, _, _, _) -> no_return().
%% Trap exit because we can get a shutdown from application manager
commit_participant(Protocol, Coord, Tid, Bin, DiscNs, RamNs) when is_binary(Bin) ->
    process_flag(trap_exit, true),
    Commit = binary_to_term(Bin),
    commit_participant(Protocol, Coord, Tid, Bin, Commit, DiscNs, RamNs);
commit_participant(Protocol, Coord, Tid, C = #commit{}, DiscNs, RamNs) ->
    process_flag(trap_exit, true),
    commit_participant(Protocol, Coord, Tid, C, C, DiscNs, RamNs).

commit_participant(Protocol, Coord, Tid, Bin, C0, DiscNs, _RamNs) ->
    ?eval_debug_fun({?MODULE, commit_participant, pre}, [{tid, Tid}]),
    try mnesia_schema:prepare_commit(Tid, C0, {part, Coord}) of
	{Modified, C = #commit{}, DumperMode} ->
	    %% If we cannot find any local unclear decision
	    %% we should presume abort at startup recovery
	    case lists:member(node(), DiscNs) of
		false ->
		    ignore;
		true ->
		    case Modified of
			false -> mnesia_log:log(Bin);
			true  -> mnesia_log:log(C)
		    end
	    end,
	    ?eval_debug_fun({?MODULE, commit_participant, vote_yes},
			    [{tid, Tid}]),
	    reply(Coord, {vote_yes, Tid, self()}),

	    receive
		{Tid, pre_commit} ->
		    D = C#commit.decision,
		    mnesia_recover:log_decision(D#decision{outcome = unclear}),
		    ?eval_debug_fun({?MODULE, commit_participant, pre_commit},
				    [{tid, Tid}]),
		    ExpectAck = C#commit.schema_ops /= []
                        orelse Protocol =:= sync_asym_trans,
		    reply(Coord, {acc_pre_commit, Tid, self(), ExpectAck}),

		    %% Now we are vulnerable for failures, since
		    %% we cannot decide without asking others
		    receive
			{Tid, committed} ->
			    mnesia_recover:log_decision(D#decision{outcome = committed}),
			    ?eval_debug_fun({?MODULE, commit_participant, log_commit},
					    [{tid, Tid}]),
			    do_commit(Tid, C, DumperMode),
			    case ExpectAck of
				false -> ignore;
				true -> reply(Coord, {schema_commit, Tid, self()})
			    end,
			    ?eval_debug_fun({?MODULE, commit_participant, do_commit},
					    [{tid, Tid}]);

			{Tid, {do_abort, _Reason}} ->
			    mnesia_recover:log_decision(D#decision{outcome = aborted}),
			    ?eval_debug_fun({?MODULE, commit_participant, log_abort},
					    [{tid, Tid}]),
			    mnesia_schema:undo_prepare_commit(Tid, C0),
			    ?eval_debug_fun({?MODULE, commit_participant, undo_prepare},
					    [{tid, Tid}]);

			{'EXIT', _MnesiaTM, Reason} ->
			    reply(Coord, {do_abort, Tid, self(), {bad_commit,Reason}}),
			    mnesia_recover:log_decision(D#decision{outcome = aborted}),
			    mnesia_schema:undo_prepare_commit(Tid, C0)
		    end;
		{Tid, {do_abort, Reason}} ->
		    reply(Coord, {do_abort, Tid, self(), Reason}),
		    mnesia_schema:undo_prepare_commit(Tid, C0),
		    ?eval_debug_fun({?MODULE, commit_participant, pre_commit_undo_prepare},
				    [{tid, Tid}]);

		{'EXIT', _, Reason} ->
		    reply(Coord, {do_abort, Tid, self(), {bad_commit,Reason}}),
		    mnesia_schema:undo_prepare_commit(Tid, C0),
		    ?eval_debug_fun({?MODULE, commit_participant, pre_commit_undo_prepare}, [{tid, Tid}])
	    end
    catch _:Reason ->
	    ?eval_debug_fun({?MODULE, commit_participant, vote_no},
			    [{tid, Tid}]),
	    reply(Coord, {vote_no, Tid, Reason}),
	    mnesia_schema:undo_prepare_commit(Tid, C0)
    end,
    mnesia_locker:release_tid(Tid),
    ?MODULE ! {delete_transaction, Tid},
    unlink(whereis(?MODULE)),
    exit(normal).

do_abort(Tid, Bin) when is_binary(Bin) ->
    %% Possible optimization:
    %% If we want we could pass around a flag
    %% that tells us whether the binary contains
    %% schema ops or not. Only if the binary
    %% contains schema ops there are meningful
    %% unpack the binary and perform
    %% mnesia_schema:undo_prepare_commit/1.
    do_abort(Tid, binary_to_term(Bin));
do_abort(Tid, Commit) ->
    mnesia_schema:undo_prepare_commit(Tid, Commit),
    Commit.

do_dirty(Tid, Commit) when Commit#commit.schema_ops == [] ->
    mnesia_log:log(Commit),
    do_commit(Tid, Commit).

%% do_commit(Tid, CommitRecord)
do_commit(Tid, Bin) when is_binary(Bin) ->
    do_commit(Tid, binary_to_term(Bin));
do_commit(Tid, C) ->
    do_commit(Tid, C, optional).

do_commit(Tid, Bin, DumperMode) when is_binary(Bin) ->
    do_commit(Tid, binary_to_term(Bin), DumperMode);
do_commit(Tid, C, DumperMode) ->
    mnesia_dumper:update(Tid, C#commit.schema_ops, DumperMode),
    R  = do_snmp(Tid, proplists:get_value(snmp, C#commit.ext, [])),
    R2 = do_update(Tid, ram_copies, C#commit.ram_copies, R),
    R3 = do_update(Tid, disc_copies, C#commit.disc_copies, R2),
    R4 = do_update(Tid, disc_only_copies, C#commit.disc_only_copies, R3),
    R5 = do_update_ext(Tid, C#commit.ext, R4),
    mnesia_subscr:report_activity(Tid),
    R5.

%% This could/should be optimized
do_update_ext(_Tid, [], OldRes) -> OldRes;
do_update_ext(Tid, Ext, OldRes) ->
    case lists:keyfind(ext_copies, 1, Ext) of
	false -> OldRes;
	{_, Ops} ->
	    Do = fun({{ext, _,_} = Storage, Op}, R) ->
			 do_update(Tid, Storage, [Op], R)
		 end,
	    lists:foldl(Do, OldRes, Ops)
    end.

%% Update the items
do_update(Tid, Storage, [Op | Ops], OldRes) ->
    try do_update_op(Tid, Storage, Op) of
	ok ->     do_update(Tid, Storage, Ops, OldRes);
	NewRes -> do_update(Tid, Storage, Ops, NewRes)
    catch _:Reason:ST ->
	    %% This may only happen when we recently have
	    %% deleted our local replica, changed storage_type
	    %% or transformed table
	    %% BUGBUG: Updates may be lost if storage_type is changed.
	    %%         Determine actual storage type and try again.
	    %% BUGBUG: Updates may be lost if table is transformed.
	    verbose("do_update in ~w failed: ~tp -> {'EXIT', ~tp}~n",
		    [Tid, Op, {Reason, ST}]),
	    do_update(Tid, Storage, Ops, OldRes)
    end;
do_update(_Tid, _Storage, [], Res) ->
    Res.

do_update_op(Tid, Storage, {{Tab, K}, Obj, write}) ->
    commit_write(?catch_val({Tab, commit_work}), Tid, Storage,
		 Tab, K, Obj, undefined),
    mnesia_lib:db_put(Storage, Tab, Obj);

do_update_op(Tid, Storage, {{Tab, K}, Val, delete}) ->
    commit_delete(?catch_val({Tab, commit_work}), Tid, Storage, Tab, K, Val, undefined),
    mnesia_lib:db_erase(Storage, Tab, K);

do_update_op(Tid, Storage, {{Tab, K}, {RecName, Incr}, update_counter}) ->
    {NewObj, OldObjs} =
        try
	    NewVal = mnesia_lib:db_update_counter(Storage, Tab, K, Incr),
	    true = is_integer(NewVal) andalso (NewVal >= 0),
	    {{RecName, K, NewVal}, [{RecName, K, NewVal - Incr}]}
	catch error:_ when Incr > 0 ->
                New = {RecName, K, Incr},
                mnesia_lib:db_put(Storage, Tab, New),
                {New, []};
	      error:_ ->
		Zero = {RecName, K, 0},
		mnesia_lib:db_put(Storage, Tab, Zero),
		{Zero, []}
        end,
    commit_update(?catch_val({Tab, commit_work}), Tid, Storage, Tab,
		  K, NewObj, OldObjs),
    element(3, NewObj);

do_update_op(Tid, Storage, {{Tab, Key}, Obj, delete_object}) ->
    commit_del_object(?catch_val({Tab, commit_work}),
		      Tid, Storage, Tab, Key, Obj),
    mnesia_lib:db_match_erase(Storage, Tab, Obj);

do_update_op(Tid, Storage, {{Tab, Key}, Obj, clear_table}) ->
    commit_clear(?catch_val({Tab, commit_work}), Tid, Storage, Tab, Key, Obj),
    mnesia_lib:db_match_erase(Storage, Tab, Obj).

commit_write([], _, _, _, _, _, _) -> ok;
commit_write([{checkpoints, CpList}|R], Tid, Storage, Tab, K, Obj, Old) ->
    mnesia_checkpoint:tm_retain(Tid, Tab, K, write, CpList),
    commit_write(R, Tid, Storage, Tab, K, Obj, Old);
commit_write([H|R], Tid, Storage, Tab, K, Obj, Old)
  when element(1, H) == subscribers ->
    mnesia_subscr:report_table_event(H, Tab, Tid, Obj, write, Old),
    commit_write(R, Tid, Storage, Tab, K, Obj, Old);
commit_write([H|R], Tid, Storage, Tab, K, Obj, Old)
  when element(1, H) == index ->
    mnesia_index:add_index(H, Storage, Tab, K, Obj, Old),
    commit_write(R, Tid, Storage, Tab, K, Obj, Old).

commit_update([], _, _, _, _, _, _) -> ok;
commit_update([{checkpoints, CpList}|R], Tid, Storage, Tab, K, Obj, _) ->
    Old = mnesia_checkpoint:tm_retain(Tid, Tab, K, write, CpList),
    commit_update(R, Tid, Storage, Tab, K, Obj, Old);
commit_update([H|R], Tid, Storage, Tab, K, Obj, Old)
  when element(1, H) == subscribers ->
    mnesia_subscr:report_table_event(H, Tab, Tid, Obj, write, Old),
    commit_update(R, Tid, Storage, Tab, K, Obj, Old);
commit_update([H|R], Tid,Storage,  Tab, K, Obj, Old)
  when element(1, H) == index ->
    mnesia_index:add_index(H, Storage, Tab, K, Obj, Old),
    commit_update(R, Tid, Storage, Tab, K, Obj, Old).

commit_delete([], _, _, _, _, _, _) ->  ok;
commit_delete([{checkpoints, CpList}|R], Tid, Storage, Tab, K, Obj, _) ->
    Old = mnesia_checkpoint:tm_retain(Tid, Tab, K, delete, CpList),
    commit_delete(R, Tid, Storage, Tab, K, Obj, Old);
commit_delete([H|R], Tid, Storage, Tab, K, Obj, Old)
  when element(1, H) == subscribers ->
    mnesia_subscr:report_table_event(H, Tab, Tid, Obj, delete, Old),
    commit_delete(R, Tid, Storage, Tab, K, Obj, Old);
commit_delete([H|R], Tid, Storage, Tab, K, Obj, Old)
  when element(1, H) == index ->
    mnesia_index:delete_index(H, Storage, Tab, K),
    commit_delete(R, Tid, Storage, Tab, K, Obj, Old).

commit_del_object([], _, _, _, _, _) -> ok;
commit_del_object([{checkpoints, CpList}|R], Tid, Storage, Tab, K, Obj) ->
    mnesia_checkpoint:tm_retain(Tid, Tab, K, delete_object, CpList),
    commit_del_object(R, Tid, Storage, Tab, K, Obj);
commit_del_object([H|R], Tid, Storage, Tab, K, Obj) when element(1, H) == subscribers ->
    mnesia_subscr:report_table_event(H, Tab, Tid, Obj, delete_object),
    commit_del_object(R, Tid, Storage, Tab, K, Obj);
commit_del_object([H|R], Tid, Storage, Tab, K, Obj) when element(1, H) == index ->
    mnesia_index:del_object_index(H, Storage, Tab, K, Obj),
    commit_del_object(R, Tid, Storage, Tab, K, Obj).

commit_clear([], _, _, _, _, _) ->  ok;
commit_clear([{checkpoints, CpList}|R], Tid, Storage, Tab, K, Obj) ->
    mnesia_checkpoint:tm_retain(Tid, Tab, K, clear_table, CpList),
    commit_clear(R, Tid, Storage, Tab, K, Obj);
commit_clear([H|R], Tid, Storage, Tab, K, Obj)
  when element(1, H) == subscribers ->
    mnesia_subscr:report_table_event(H, Tab, Tid, Obj, clear_table, undefined),
    commit_clear(R, Tid, Storage, Tab, K, Obj);
commit_clear([H|R], Tid, Storage, Tab, K, Obj)
  when element(1, H) == index ->
    mnesia_index:clear_index(H, Tab, K, Obj),
    commit_clear(R, Tid, Storage, Tab, K, Obj).

do_snmp(_, []) ->   ok;
do_snmp(Tid, [Head|Tail]) ->
    try mnesia_snmp_hook:update(Head)
    catch _:Reason:ST ->
	    %% This should only happen when we recently have
	    %% deleted our local replica or recently deattached
	    %% the snmp table
	    verbose("do_snmp in ~w failed: ~tp -> {'EXIT', ~tp}~n",
		    [Tid, Head, {Reason, ST}])
    end,
    do_snmp(Tid, Tail).

commit_nodes([C | Tail], AccD, AccR) ->
    case C of
	#commit{disc_copies=[], disc_only_copies=[], schema_ops=[], ext=Ext} ->
	    case lists:keyfind(ext_copies, 1, Ext) of
		false -> commit_nodes(Tail, AccD, [C#commit.node | AccR]);
		_ -> commit_nodes(Tail, [C#commit.node | AccD], AccR)
	    end;
	_ ->
	    commit_nodes(Tail, [C#commit.node | AccD], AccR)
    end;
commit_nodes([], AccD, AccR) ->
    {AccD, AccR}.

commit_decision(D, [C | Tail], AccD, AccR) ->
    N = C#commit.node,
    {D2, Tail2} =
	case C of
	    #commit{disc_copies=[], disc_only_copies=[], schema_ops=[], ext=Ext} ->
		case lists:keyfind(ext_copies, 1, Ext) of
		    false -> commit_decision(D, Tail, AccD, [N | AccR]);
		    _     -> commit_decision(D, Tail, [N | AccD], AccR)
		end;
	    #commit{schema_ops=[]} ->
		commit_decision(D, Tail, [N | AccD], AccR);
	    #commit{schema_ops=Ops} ->
		case ram_only_ops(N, Ops) of
		    true ->  commit_decision(D, Tail, AccD, [N | AccR]);
		    false -> commit_decision(D, Tail, [N | AccD], AccR)
		end
	end,
    {D2, [C#commit{decision = D2} | Tail2]};
commit_decision(D, [], AccD, AccR) ->
    {D#decision{disc_nodes = AccD, ram_nodes = AccR}, []}.

ram_only_ops(N, [{op, change_table_copy_type, N, _FromS, _ToS, Cs} | _Ops ]) ->
    case lists:member({name, schema}, Cs) of
	true ->
	    %% We always use disk if change type of the schema
	    false;
	false ->
	    not lists:member(N, val({schema, disc_copies}))
    end;

ram_only_ops(N, _Ops) ->
    not lists:member(N, val({schema, disc_copies})).

%% Returns {WaitFor, Res}
sync_send_dirty(Tid, [Head | Tail], Tab, WaitFor) ->
    Node = Head#commit.node,
    if
	Node == node() ->
	    {WF, _} = sync_send_dirty(Tid, Tail, Tab, WaitFor),
	    Res =  do_dirty(Tid, Head),
	    {WF, Res};
	true ->
	    {?MODULE, Node} ! {self(), {sync_dirty, Tid, Head, Tab}},
	    sync_send_dirty(Tid, Tail, Tab, [Node | WaitFor])
    end;
sync_send_dirty(_Tid, [], _Tab, WaitFor) ->
    {WaitFor, {'EXIT', {aborted, {node_not_running, WaitFor}}}}.

%% Returns {WaitFor, Res}
async_send_dirty(_Tid, _Nodes, Tab, nowhere) ->
    {[], {'EXIT', {aborted, {no_exists, Tab}}}};
async_send_dirty(Tid, Nodes, Tab, ReadNode) ->
    async_send_dirty(Tid, Nodes, Tab, ReadNode, [], ok).

async_send_dirty(Tid, [Head | Tail], Tab, ReadNode, WaitFor, Res) ->
    Node = Head#commit.node,
    if
	ReadNode == Node, Node == node() ->
	    NewRes =  do_dirty(Tid, Head),
	    async_send_dirty(Tid, Tail, Tab, ReadNode, WaitFor, NewRes);
	ReadNode == Node ->
	    {?MODULE, Node} ! {self(), {sync_dirty, Tid, Head, Tab}},
	    NewRes = {'EXIT', {aborted, {node_not_running, Node}}},
	    async_send_dirty(Tid, Tail, Tab, ReadNode, [Node | WaitFor], NewRes);
	true ->
	    {?MODULE, Node} ! {self(), {async_dirty, Tid, Head, Tab}},
	    async_send_dirty(Tid, Tail, Tab, ReadNode, WaitFor, Res)
    end;
async_send_dirty(_Tid, [], _Tab, _ReadNode, WaitFor, Res) ->
    {WaitFor, Res}.

rec_dirty([Node | Tail], Res) when Node /= node() ->
    NewRes = get_dirty_reply(Node, Res),
    rec_dirty(Tail, NewRes);
rec_dirty([], Res) ->
    Res.

get_dirty_reply(Node, Res) ->
    receive
	{?MODULE, Node, {'EXIT', Reason}} ->
	    {'EXIT', {aborted, {badarg, Reason}}};
	{?MODULE, Node, {dirty_res, ok}} ->
	    case Res of
		{'EXIT', {aborted, {node_not_running, _Node}}} ->
		    ok;
		_ ->
		    %% Prioritize bad results, but node_not_running
		    Res
	    end;
	{?MODULE, Node, {dirty_res, Reply}} ->
	    Reply;
	{mnesia_down, Node} ->
	    case get(mnesia_activity_state) of
		{_, Tid, _Ts} when element(1,Tid) == tid ->
		    %% Hmm dirty called inside a transaction, to avoid
		    %% hanging transaction we need to restart the transaction
		    mnesia:abort({node_not_running, Node});
		_ ->
		    %% It's ok to ignore mnesia_down's since we will make
		    %% the replicas consistent again when Node is started
		    Res
	    end
    after 1000 ->
	    case lists:member(Node, val({current, db_nodes})) of
		true ->
		    get_dirty_reply(Node, Res);
		false ->
		    Res
	    end
    end.

%% Assume that CommitRecord is no binary
%% Return {Res, Pids}
ask_commit(Protocol, Tid, CR, DiscNs, RamNs) ->
    ask_commit(Protocol, Tid, CR, DiscNs, RamNs, [], no_local).

ask_commit(Protocol, Tid, [Head | Tail], DiscNs, RamNs, WaitFor, Local) ->
    Node = Head#commit.node,
    if
	Node == node() ->
	    ask_commit(Protocol, Tid, Tail, DiscNs, RamNs, WaitFor, Head);
	true ->
	    Msg = {ask_commit, convert_old(Protocol, Node), Tid, Head, DiscNs, RamNs},
	    {?MODULE, Node} ! {self(), Msg},
	    ask_commit(Protocol, Tid, Tail, DiscNs, RamNs, [Node | WaitFor], Local)
    end;
ask_commit(_Protocol, _Tid, [], _DiscNs, _RamNs, WaitFor, Local) ->
    {WaitFor, Local}.

convert_old(sync_asym_trans, Node) ->
    case ?catch_val({protocol, Node}) of
        {{8,3}, _} -> asym_trans;
        _ -> sync_asym_trans
    end;
convert_old(Protocol, _) ->
    Protocol.

new_cr_format(#commit{ext=[]}=Cr) -> Cr;
new_cr_format(#commit{ext=[{_,_}|_]}=Cr) -> Cr;
new_cr_format(#commit{ext=Snmp}=Cr) ->
    Cr#commit{ext=[{snmp,Snmp}]}.

rec_all([Node | Tail], Tid, Res, Pids) ->
    receive
	{?MODULE, Node, {vote_yes, Tid}} ->
	    rec_all(Tail, Tid, Res, Pids);
	{?MODULE, Node, {vote_yes, Tid, Pid}} ->
	    rec_all(Tail, Tid, Res, [Pid | Pids]);
	{?MODULE, Node, {vote_no, Tid, Reason}} ->
	    rec_all(Tail, Tid, {do_abort, Reason}, Pids);
	{?MODULE, Node, {committed, Tid}} ->
	    rec_all(Tail, Tid, Res, Pids);
	{?MODULE, Node, {aborted, Tid}} ->
	    rec_all(Tail, Tid, Res, Pids);

	{mnesia_down, Node} ->
	    %% Make sure that mnesia_tm knows it has died
	    %% it may have been restarted
	    Abort = {do_abort, {bad_commit, Node}},
	    ?SAFE({?MODULE, Node} ! {Tid, Abort}),
	    rec_all(Tail, Tid, Abort, Pids)
    end;
rec_all([], _Tid, Res, Pids) ->
    {Res, Pids}.

get_transactions() ->
    {info, Participant, Coordinator} = req(info),
    lists:map(fun({Tid, _Tabs}) ->
		      Status = tr_status(Tid,Participant),
		      {Tid#tid.counter, Tid#tid.pid, Status}
	      end,Coordinator).

tr_status(Tid,Participant) ->
    case lists:keymember(Tid, 1, Participant) of
	true -> participant;
	false  -> coordinator
    end.

get_info(Timeout) ->
    case whereis(?MODULE) of
	undefined ->
	    {timeout, Timeout};
	Pid ->
	    Pid ! {self(), info},
	    receive
		{?MODULE, _, {info, Part, Coord}} ->
		    {info, Part, Coord}
	    after Timeout ->
		    {timeout, Timeout}
	    end
    end.

display_info(Stream, {timeout, T}) ->
    io:format(Stream, "---> No info about coordinator and participant transactions, "
	      "timeout ~p <--- ~n", [T]);

display_info(Stream, {info, Part, Coord}) ->
    io:format(Stream, "---> Participant transactions <--- ~n", []),
    lists:foreach(fun(P) -> pr_participant(Stream, P) end, Part),
    io:format(Stream, "---> Coordinator transactions <---~n", []),
    lists:foreach(fun({Tid, _Tabs}) -> pr_tid(Stream, Tid) end, Coord).

pr_participant(Stream, P) ->
    Commit0 = P#participant.commit,
    Commit =
	if
	    is_binary(Commit0) -> binary_to_term(Commit0);
	    true -> Commit0
	end,
    pr_tid(Stream, P#participant.tid),
    io:format(Stream, "with participant objects ~tp~n", [Commit]).


pr_tid(Stream, Tid) ->
    io:format(Stream, "Tid: ~p (owned by ~p) ~n",
	      [Tid#tid.counter, Tid#tid.pid]).

info(Serial) ->
    io:format( "Info about transaction with serial == ~p~n", [Serial]),
    {info, Participant, Trs} = req(info),
    search_pr_participant(Serial, Participant),
    search_pr_coordinator(Serial, Trs).


search_pr_coordinator(_S, []) -> no;
search_pr_coordinator(S, [{Tid, _Ts}|Tail]) ->
    case Tid#tid.counter of
	S ->
	    io:format( "Tid is coordinator, owner == \n", []),
	    display_pid_info(Tid#tid.pid),
	    search_pr_coordinator(S, Tail);
	_ ->
	    search_pr_coordinator(S, Tail)
    end.

search_pr_participant(_S, []) ->
    false;
search_pr_participant(S, [ P | Tail]) ->
    Tid = P#participant.tid,
    Commit0 = P#participant.commit,
    if
	Tid#tid.counter == S ->
	    io:format( "Tid is participant to commit, owner == \n", []),
	    Pid = Tid#tid.pid,
	    display_pid_info(Pid),
	    io:format( "Tid wants to write objects \n",[]),
	    Commit =
		if
		    is_binary(Commit0) -> binary_to_term(Commit0);
		    true -> Commit0
		end,

	    io:format("~tp~n", [Commit]),
	    search_pr_participant(S,Tail);  %% !!!!!
	true ->
	    search_pr_participant(S, Tail)
    end.

display_pid_info(Pid) ->
    case rpc:pinfo(Pid) of
	undefined ->
	    io:format( "Dead process \n");
	Info ->
	    Call = fetch(initial_call, Info),
	    Curr = case fetch(current_function, Info) of
		       {Mod,F,Args} when is_list(Args) ->
			   {Mod,F,length(Args)};
		       Other ->
			   Other
		   end,
	    Reds  = fetch(reductions, Info),
	    LM = fetch(message_queue_len, Info),
	    pformat(io_lib:format("~p", [Pid]),
		    io_lib:format("~tp", [Call]),
		    io_lib:format("~tp", [Curr]), Reds, LM)
    end.

pformat(A1, A2, A3, A4, A5) ->
    io:format( "~-12s ~-21ts ~-21ts ~9w ~4w~n", [A1,A2,A3,A4,A5]).

fetch(Key, Info) ->
    case lists:keysearch(Key, 1, Info) of
	{value, {_, Val}} ->
	    Val;
	_ ->
	    0
    end.


%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%  reconfigure stuff comes here ......
%%%%%%%%%%%%%%%%%%%%%

reconfigure_coordinators(N, [{Tid, [Store | _]} | Coordinators]) ->
    case mnesia_recover:outcome(Tid, unknown) of
	committed ->
	    WaitingNodes = ?ets_lookup(Store, waiting_for_commit_ack),
	    case lists:keymember(N, 2, WaitingNodes) of
		false ->
		    ignore; % avoid spurious mnesia_down messages
		true ->
		    send_mnesia_down(Tid, Store, N)
	    end;
	_ ->
	    %% Tell the coordinator about the mnesia_down
	    send_mnesia_down(Tid, Store, N)
    end,
    reconfigure_coordinators(N, Coordinators);
reconfigure_coordinators(_N, []) ->
    ok.

send_mnesia_down(Tid, Store, Node) ->
    Msg = {mnesia_down, Node},
    send_to_pids([Tid#tid.pid | get_elements(friends,Store)], Msg).

send_to_pids([Pid | Pids], Msg) when is_pid(Pid) ->
    Pid ! Msg,
    send_to_pids(Pids, Msg);
send_to_pids([_ | Pids], Msg) ->
    send_to_pids(Pids, Msg);
send_to_pids([], _Msg) ->
    ok.

reconfigure_participants(N, [P | Tail]) ->
    case lists:member(N, P#participant.disc_nodes) or
	 lists:member(N, P#participant.ram_nodes) of
	false ->
	    %% Ignore, since we are not a participant
	    %% in the transaction.
	    reconfigure_participants(N, Tail);

	true ->
	    %% We are on a participant node, lets
	    %% check if the dead one was a
	    %% participant or a coordinator.
	    Tid  = P#participant.tid,
	    if
		node(Tid#tid.pid) /= N ->
		    %% Another participant node died. Ignore.
		    reconfigure_participants(N, Tail);

		true ->
		    %% The coordinator node has died and
		    %% we must determine the outcome of the
		    %% transaction and tell mnesia_tm on all
		    %% nodes (including the local node) about it
		    verbose("Coordinator ~p in transaction ~p died~n",
			    [Tid#tid.pid, Tid]),

		    Nodes = P#participant.disc_nodes ++
			    P#participant.ram_nodes,
		    AliveNodes = Nodes  -- [N],
		    Protocol =  P#participant.protocol,
		    tell_outcome(Tid, Protocol, N, AliveNodes, AliveNodes),
		    reconfigure_participants(N, Tail)
	    end
    end;
reconfigure_participants(_, []) ->
    [].

%% We need to determine the outcome of the transaction and
%% tell mnesia_tm on all involved nodes (including the local node)
%% about the outcome.
tell_outcome(Tid, Protocol, Node, CheckNodes, TellNodes) ->
    Outcome = mnesia_recover:what_happened(Tid, proto(Protocol), CheckNodes),
    case Outcome of
	aborted ->
	    rpc:abcast(TellNodes, ?MODULE, {Tid,{do_abort, {mnesia_down, Node}}});
	committed ->
	    rpc:abcast(TellNodes, ?MODULE, {Tid, do_commit})
    end,
    Outcome.

proto(sync_asym_trans) -> asym_trans;
proto(Proto) -> Proto.

do_stop(#state{coordinators = Coordinators}) ->
    Msg = {mnesia_down, node()},
    lists:foreach(fun({Tid, _}) -> Tid#tid.pid ! Msg end, gb_trees:to_list(Coordinators)),
    mnesia_checkpoint:stop(),
    mnesia_log:stop(),
    exit(shutdown).

fixtable(Tab, Lock, Me) ->
    case req({fixtable, [Tab,Lock,Me]}) of
	error ->
	    exit({no_exists, Tab});
	Else ->
	    Else
    end.

%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% System upgrade

system_continue(_Parent, _Debug, State) ->
    doit_loop(State).

-spec system_terminate(_, _, _, _) -> no_return().
system_terminate(_Reason, _Parent, _Debug, State) ->
    do_stop(State).

system_code_change(State=#state{coordinators=Cs0,participants=Ps0},_Module,_OldVsn,downgrade) ->
    case is_tuple(Cs0) of
	true ->
	    Cs = gb_trees:to_list(Cs0),
	    Ps = gb_trees:values(Ps0),
	    {ok, State#state{coordinators=Cs,participants=Ps}};
	false ->
	    {ok, State}
    end;

system_code_change(State=#state{coordinators=Cs0,participants=Ps0},_Module,_OldVsn,_Extra) ->
    case is_list(Cs0) of
	true ->
	    Cs = gb_trees:from_orddict(lists:sort(Cs0)),
	    Ps1 = [{P#participant.tid,P}|| P <- Ps0],
	    Ps = gb_trees:from_orddict(lists:sort(Ps1)),
	    {ok, State#state{coordinators=Cs,participants=Ps}};
	false ->
	    {ok, State}
    end.