diff options
Diffstat (limited to 'lib/mnesia/src')
-rw-r--r-- | lib/mnesia/src/Makefile | 4 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia.appup.src | 42 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia.erl | 11 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia.hrl | 3 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_bup.erl | 4 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_controller.erl | 40 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_dumper.erl | 25 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_event.erl | 6 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_frag.erl | 12 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_lib.erl | 18 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_loader.erl | 248 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_locker.erl | 69 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_log.erl | 5 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_monitor.erl | 90 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_recover.erl | 6 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_schema.erl | 650 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_text.erl | 5 | ||||
-rw-r--r-- | lib/mnesia/src/mnesia_tm.erl | 66 |
18 files changed, 776 insertions, 528 deletions
diff --git a/lib/mnesia/src/Makefile b/lib/mnesia/src/Makefile index e032f563fa..1c8ec54605 100644 --- a/lib/mnesia/src/Makefile +++ b/lib/mnesia/src/Makefile @@ -1,7 +1,7 @@ # # %CopyrightBegin% # -# Copyright Ericsson AB 1996-2009. All Rights Reserved. +# Copyright Ericsson AB 1996-2011. All Rights Reserved. # # The contents of this file are subject to the Erlang Public License, # Version 1.1, (the "License"); you may not use this file except in @@ -113,6 +113,8 @@ clean: docs: +$(TARGET_FILES): $(HRL_FILES) + # ---------------------------------------------------- # Special Build Targets # ---------------------------------------------------- diff --git a/lib/mnesia/src/mnesia.appup.src b/lib/mnesia/src/mnesia.appup.src index 7bad6c4ea6..e0954ad206 100644 --- a/lib/mnesia/src/mnesia.appup.src +++ b/lib/mnesia/src/mnesia.appup.src @@ -1,33 +1,17 @@ %% -*- erlang -*- -{"%VSN%", +{"%VSN%", [ - {"4.4.17",[ - {update, mnesia_controller, soft, soft_purge, soft_purge, []} - ]}, - {"4.4.16",[ - {update, mnesia_controller, soft, soft_purge, soft_purge, []}, - {update, mnesia_frag, soft, soft_purge, soft_purge, []}, - {update, mnesia_schema, soft, soft_purge, soft_purge, []} - ]}, - {"4.4.15",[ - {update, mnesia_controller, soft, soft_purge, soft_purge, []}, - {update, mnesia_frag, soft, soft_purge, soft_purge, []}, - {update, mnesia, soft, soft_purge, soft_purge, []}, - {update, mnesia_dumper, soft, soft_purge, soft_purge, []} - ]} + {"4.5", [{restart_application, mnesia}]}, + {"4.4.19", [{restart_application, mnesia}]}, + {"4.4.18", [{restart_application, mnesia}]}, + {"4.4.17", [{restart_application, mnesia}]}, + {"4.4.16", [{restart_application, mnesia}]} ], - {"4.4.17",[ - {update, mnesia_controller, soft, soft_purge, soft_purge, []} - ]}, - {"4.4.16",[ - {update, mnesia_controller, soft, soft_purge, soft_purge, []}, - {update, mnesia_frag, soft, soft_purge, soft_purge, []}, - {update, mnesia_schema, soft, soft_purge, soft_purge, []} - ]}, - {"4.4.15",[ - {update, mnesia_controller, soft, soft_purge, soft_purge, []}, - {update, mnesia_frag, soft, soft_purge, soft_purge, []}, - {update, mnesia, soft, soft_purge, soft_purge, []}, - {update, mnesia_dumper, soft, soft_purge, soft_purge, []} - ]} + [ + {"4.5", [{restart_application, mnesia}]}, + {"4.4.19", [{restart_application, mnesia}]}, + {"4.4.18", [{restart_application, mnesia}]}, + {"4.4.17", [{restart_application, mnesia}]}, + {"4.4.16", [{restart_application, mnesia}]} + ] }. diff --git a/lib/mnesia/src/mnesia.erl b/lib/mnesia/src/mnesia.erl index 025b32f506..980a9c6213 100644 --- a/lib/mnesia/src/mnesia.erl +++ b/lib/mnesia/src/mnesia.erl @@ -39,6 +39,7 @@ %% Access within an activity - Lock acquisition lock/2, lock/4, + lock_table/2, read_lock_table/1, write_lock_table/1, @@ -92,7 +93,7 @@ add_table_copy/3, del_table_copy/2, move_table_copy/3, add_table_index/2, del_table_index/2, transform_table/3, transform_table/4, - change_table_copy_type/3, + change_table_copy_type/3, change_table_majority/2, read_table_property/2, write_table_property/2, delete_table_property/2, change_table_frag/2, clear_table/1, clear_table/4, @@ -415,6 +416,9 @@ lock(LockItem, LockKind) -> abort(no_transaction) end. +lock_table(Tab, LockKind) -> + lock({table, Tab}, LockKind). + lock(Tid, Ts, LockItem, LockKind) -> case element(1, Tid) of tid -> @@ -467,6 +471,8 @@ lock_table(Tid, Ts, Tab, LockKind) when is_atom(Tab) -> mnesia_locker:rlock_table(Tid, Store, Tab); write -> mnesia_locker:wlock_table(Tid, Store, Tab); + load -> + mnesia_locker:load_lock_table(Tid, Store, Tab); sticky_write -> mnesia_locker:sticky_wlock_table(Tid, Store, Tab); none -> @@ -2455,6 +2461,9 @@ change_table_access_mode(T, Access) -> change_table_load_order(T, O) -> mnesia_schema:change_table_load_order(T, O). +change_table_majority(T, M) -> + mnesia_schema:change_table_majority(T, M). + set_master_nodes(Nodes) when is_list(Nodes) -> UseDir = system_info(use_dir), IsRunning = system_info(is_running), diff --git a/lib/mnesia/src/mnesia.hrl b/lib/mnesia/src/mnesia.hrl index d488d9364a..2375b72d59 100644 --- a/lib/mnesia/src/mnesia.hrl +++ b/lib/mnesia/src/mnesia.hrl @@ -1,7 +1,7 @@ %% %% %CopyrightBegin% %% -%% Copyright Ericsson AB 1996-2009. All Rights Reserved. +%% Copyright Ericsson AB 1996-2011. All Rights Reserved. %% %% The contents of this file are subject to the Erlang Public License, %% Version 1.1, (the "License"); you may not use this file except in @@ -62,6 +62,7 @@ disc_only_copies = [], % [Node] load_order = 0, % Integer access_mode = read_write, % read_write | read_only + majority = false, % true | false index = [], % [Integer] snmp = [], % Snmp Ustruct local_content = false, % true | false diff --git a/lib/mnesia/src/mnesia_bup.erl b/lib/mnesia/src/mnesia_bup.erl index 47dcdad7ac..14414537b9 100644 --- a/lib/mnesia/src/mnesia_bup.erl +++ b/lib/mnesia/src/mnesia_bup.erl @@ -372,7 +372,9 @@ mk_str() -> lists:concat([node()] ++ Now ++ ".TMP"). make_initial_backup(Ns, Opaque, Mod) -> - Schema = [{schema, schema, mnesia_schema:get_initial_schema(disc_copies, Ns)}], + Orig = mnesia_schema:get_initial_schema(disc_copies, Ns), + Modded = proplists:delete(storage_properties, proplists:delete(majority, Orig)), + Schema = [{schema, schema, Modded}], O2 = do_apply(Mod, open_write, [Opaque], Opaque), O3 = do_apply(Mod, write, [O2, [mnesia_log:backup_log_header()]], O2), O4 = do_apply(Mod, write, [O3, Schema], O3), diff --git a/lib/mnesia/src/mnesia_controller.erl b/lib/mnesia/src/mnesia_controller.erl index 0254769758..6a561394d5 100644 --- a/lib/mnesia/src/mnesia_controller.erl +++ b/lib/mnesia/src/mnesia_controller.erl @@ -57,7 +57,8 @@ release_schema_commit_lock/0, create_table/1, get_disc_copy/1, - get_cstructs/0, + get_remote_cstructs/0, % new function + get_cstructs/0, % old function sync_and_block_table_whereabouts/4, sync_del_table_copy_whereabouts/2, block_table/1, @@ -72,6 +73,7 @@ add_active_replica/4, update/1, change_table_access_mode/1, + change_table_majority/1, del_active_replica/2, wait_for_tables/2, get_network_copy/2, @@ -277,9 +279,18 @@ rec_tabs([], _, _, Init) -> unlink(Init), ok. -get_cstructs() -> +%% New function that does exactly what get_cstructs() used to do. +%% When this function is called, we know that the calling node knows +%% how to convert cstructs on the receiving end (should they differ). +get_remote_cstructs() -> call(get_cstructs). +%% Old function kept for backwards compatibility; converts cstructs before sending. +get_cstructs() -> + {cstructs, Cstructs, Running} = call(get_cstructs), + Node = node(group_leader()), + {cstructs, mnesia_schema:normalize_cs(Cstructs, Node), Running}. + update(Fun) -> call({update,Fun}). @@ -690,7 +701,8 @@ handle_call({update_where_to_write, [add, Tab, AddNode], _From}, _Dummy, State) case lists:member(AddNode, Current) and (State#state.schema_is_merged == true) of true -> - mnesia_lib:add_lsort({Tab, where_to_write}, AddNode); + mnesia_lib:add_lsort({Tab, where_to_write}, AddNode), + update_where_to_wlock(Tab); false -> ignore end, @@ -1690,6 +1702,8 @@ add_active_replica(Tab, Node, Storage, AccessMode) -> set(Var, mark_blocked_tab(Blocked, Del)), mnesia_lib:del({Tab, where_to_write}, Node) end, + + update_where_to_wlock(Tab), add({Tab, active_replicas}, Node). del_active_replica(Tab, Node) -> @@ -1699,7 +1713,8 @@ del_active_replica(Tab, Node) -> New = lists:sort(Del), set(Var, mark_blocked_tab(Blocked, New)), % where_to_commit mnesia_lib:del({Tab, active_replicas}, Node), - mnesia_lib:del({Tab, where_to_write}, Node). + mnesia_lib:del({Tab, where_to_write}, Node), + update_where_to_wlock(Tab). change_table_access_mode(Cs) -> W = fun() -> @@ -1708,7 +1723,22 @@ change_table_access_mode(Cs) -> val({Tab, active_replicas})) end, update(W). - + +change_table_majority(Cs) -> + W = fun() -> + Tab = Cs#cstruct.name, + set({Tab, majority}, Cs#cstruct.majority), + update_where_to_wlock(Tab) + end, + update(W). + +update_where_to_wlock(Tab) -> + WNodes = val({Tab, where_to_write}), + Majority = case catch val({Tab, majority}) of + true -> true; + _ -> false + end, + set({Tab, where_to_wlock}, {WNodes, Majority}). %% node To now has tab loaded, but this must be undone %% This code is rpc:call'ed from the tab_copier process diff --git a/lib/mnesia/src/mnesia_dumper.erl b/lib/mnesia/src/mnesia_dumper.erl index 644133cf5d..f8d7664156 100644 --- a/lib/mnesia/src/mnesia_dumper.erl +++ b/lib/mnesia/src/mnesia_dumper.erl @@ -1,7 +1,7 @@ %% %% %CopyrightBegin% %% -%% Copyright Ericsson AB 1996-2010. All Rights Reserved. +%% Copyright Ericsson AB 1996-2011. All Rights Reserved. %% %% The contents of this file are subject to the Erlang Public License, %% Version 1.1, (the "License"); you may not use this file except in @@ -214,7 +214,12 @@ insert_rec(Rec, InPlace, InitBy, LogV) when is_record(Rec, commit) -> {Tid, committed} -> do_insert_rec(Tid, Rec, InPlace, InitBy, LogV); {Tid, aborted} -> - mnesia_schema:undo_prepare_commit(Tid, Rec) + case InitBy of + startup -> + mnesia_schema:undo_prepare_commit(Tid, Rec); + _ -> + ok + end end; insert_rec(H, _InPlace, _InitBy, _LogV) when is_record(H, log_header) -> CurrentVersion = mnesia_log:version(), @@ -359,7 +364,7 @@ dets_insert(Op,Tab,Key,Val) -> ok = dets:delete_object(Tab, Val); clear_table -> dets_cleared(Tab), - ok = dets:match_delete(Tab, '_') + ok = dets:delete_all_objects(Tab) end. dets_updated(Tab,Key) -> @@ -871,7 +876,11 @@ insert_op(Tid, _, {op, add_index, Pos, TabDef}, InPlace, InitBy) -> startup -> ignore; _ -> - mnesia_index:init_indecies(Tab, Storage, [Pos]) + case val({Tab,where_to_read}) of + nowhere -> ignore; + _ -> + mnesia_index:init_indecies(Tab, Storage, [Pos]) + end end; insert_op(Tid, _, {op, del_index, Pos, TabDef}, InPlace, InitBy) -> @@ -896,6 +905,14 @@ insert_op(Tid, _, {op, change_table_access_mode,TabDef, _OldAccess, _Access}, In end, insert_cstruct(Tid, Cs, true, InPlace, InitBy); +insert_op(Tid, _, {op, change_table_majority,TabDef, _OldAccess, _Access}, InPlace, InitBy) -> + Cs = mnesia_schema:list2cs(TabDef), + case InitBy of + startup -> ignore; + _ -> mnesia_controller:change_table_majority(Cs) + end, + insert_cstruct(Tid, Cs, true, InPlace, InitBy); + insert_op(Tid, _, {op, change_table_load_order, TabDef, _OldLevel, _Level}, InPlace, InitBy) -> Cs = mnesia_schema:list2cs(TabDef), insert_cstruct(Tid, Cs, true, InPlace, InitBy); diff --git a/lib/mnesia/src/mnesia_event.erl b/lib/mnesia/src/mnesia_event.erl index ec6b99ecaa..5a060a28ff 100644 --- a/lib/mnesia/src/mnesia_event.erl +++ b/lib/mnesia/src/mnesia_event.erl @@ -121,7 +121,7 @@ handle_system_event({mnesia_up, Node}, State) -> {ok, State#state{nodes = Nodes}}; handle_system_event({mnesia_down, Node}, State) -> - case mnesia:system_info(fallback_activated) of + case mnesia:system_info(fallback_activated) andalso Node =/= node() of true -> case mnesia_monitor:get_env(fallback_error_function) of {mnesia, lkill} -> @@ -129,8 +129,8 @@ handle_system_event({mnesia_down, Node}, State) -> "must be restarted. Forcing shutdown " "after mnesia_down from ~p...~n", report_fatal(Msg, [Node], nocore, State#state.dumped_core), - mnesia:lkill(), - exit(fatal); + catch exit(whereis(mnesia_monitor), fatal), + {ok, State}; {UserMod, UserFunc} -> Msg = "Warning: A fallback is installed and Mnesia got mnesia_down " "from ~p. ~n", diff --git a/lib/mnesia/src/mnesia_frag.erl b/lib/mnesia/src/mnesia_frag.erl index 9e77fe0b9f..4a1616e054 100644 --- a/lib/mnesia/src/mnesia_frag.erl +++ b/lib/mnesia/src/mnesia_frag.erl @@ -758,7 +758,7 @@ make_activate(Tab, Props) -> [] -> Cs2 = Cs#cstruct{frag_properties = Props}, [Cs3] = expand_cstruct(Cs2, activate), - TabDef = mnesia_schema:cs2list(Cs3), + TabDef = mnesia_schema:vsn_cs2list(Cs3), Op = {op, change_table_frag, activate, TabDef}, [[Op]]; BadProps -> @@ -783,7 +783,7 @@ make_deactivate(Tab) -> mnesia:abort({combine_error, Tab, "Too many fragments"}); true -> Cs2 = Cs#cstruct{frag_properties = []}, - TabDef = mnesia_schema:cs2list(Cs2), + TabDef = mnesia_schema:vsn_cs2list(Cs2), Op = {op, change_table_frag, deactivate, TabDef}, [[Op]] end. @@ -850,7 +850,7 @@ make_add_frag(Tab, SortedNs) -> SplitOps = split(Tab, FH2, FromIndecies, FragNames, []), Cs2 = replace_frag_hash(Cs, FH2), - TabDef = mnesia_schema:cs2list(Cs2), + TabDef = mnesia_schema:vsn_cs2list(Cs2), BaseOp = {op, change_table_frag, {add_frag, SortedNs}, TabDef}, [BaseOp, NewOp | SplitOps]. @@ -962,7 +962,7 @@ make_del_frag(Tab) -> LastFrag = element(N, FragNames), [LastOp] = mnesia_schema:make_delete_table(LastFrag, single_frag), Cs2 = replace_frag_hash(Cs, FH2), - TabDef = mnesia_schema:cs2list(Cs2), + TabDef = mnesia_schema:vsn_cs2list(Cs2), BaseOp = {op, change_table_frag, del_frag, TabDef}, [BaseOp, LastOp | MergeOps]; _ -> @@ -1075,7 +1075,7 @@ make_add_node(Tab, Node) when is_atom(Node) -> Props = Cs#cstruct.frag_properties, Props2 = lists:keyreplace(node_pool, 1, Props, {node_pool, Pool2}), Cs2 = Cs#cstruct{frag_properties = Props2}, - TabDef = mnesia_schema:cs2list(Cs2), + TabDef = mnesia_schema:vsn_cs2list(Cs2), Op = {op, change_table_frag, {add_node, Node}, TabDef}, [Op]; true -> @@ -1104,7 +1104,7 @@ make_del_node(Tab, Node) when is_atom(Node) -> Pool2 = Pool -- [Node], Props = lists:keyreplace(node_pool, 1, Cs#cstruct.frag_properties, {node_pool, Pool2}), Cs2 = Cs#cstruct{frag_properties = Props}, - TabDef = mnesia_schema:cs2list(Cs2), + TabDef = mnesia_schema:vsn_cs2list(Cs2), Op = {op, change_table_frag, {del_node, Node}, TabDef}, [Op]; false -> diff --git a/lib/mnesia/src/mnesia_lib.erl b/lib/mnesia/src/mnesia_lib.erl index 36bcfe8de9..ae6631646c 100644 --- a/lib/mnesia/src/mnesia_lib.erl +++ b/lib/mnesia/src/mnesia_lib.erl @@ -96,6 +96,8 @@ exists/1, fatal/2, get_node_number/0, + have_majority/2, + have_majority/3, fix_error/1, important/2, incr_counter/1, @@ -411,7 +413,7 @@ pr_other(Var, Other) -> [self(), process_info(self(), registered_name), Var, Other, Why]), case Other of - {badarg, [{ets, lookup_element, _}|_]} -> + {badarg, [{ets, lookup_element, _, _}|_]} -> exit(Why); _ -> erlang:error(Why) @@ -660,6 +662,14 @@ proc_info(_) -> false. get_node_number() -> {node(), self()}. +have_majority(Tab, HaveNodes) -> + have_majority(Tab, val({Tab, all_nodes}), HaveNodes). + +have_majority(_Tab, AllNodes, HaveNodes) -> + Missing = AllNodes -- HaveNodes, + Present = AllNodes -- Missing, + length(Present) > length(Missing). + read_log_files() -> [{F, catch file:read_file(F)} || F <- mnesia_log:log_files()]. @@ -1131,12 +1141,18 @@ db_erase(ram_copies, Tab, Key) -> ?ets_delete(Tab, Key), ok; db_erase(disc_copies, Tab, Key) -> ?ets_delete(Tab, Key), ok; db_erase(disc_only_copies, Tab, Key) -> dets:delete(Tab, Key). +db_match_erase(Tab, '_') -> + db_delete_all(val({Tab, storage_type}),Tab); db_match_erase(Tab, Pat) -> db_match_erase(val({Tab, storage_type}), Tab, Pat). db_match_erase(ram_copies, Tab, Pat) -> ?ets_match_delete(Tab, Pat), ok; db_match_erase(disc_copies, Tab, Pat) -> ?ets_match_delete(Tab, Pat), ok; db_match_erase(disc_only_copies, Tab, Pat) -> dets:match_delete(Tab, Pat). +db_delete_all(ram_copies, Tab) -> ets:delete_all_objects(Tab); +db_delete_all(disc_copies, Tab) -> ets:delete_all_objects(Tab); +db_delete_all(disc_only_copies, Tab) -> dets:delete_all_objects(Tab). + db_first(Tab) -> db_first(val({Tab, storage_type}), Tab). db_first(ram_copies, Tab) -> ?ets_first(Tab); diff --git a/lib/mnesia/src/mnesia_loader.erl b/lib/mnesia/src/mnesia_loader.erl index 3de329503e..c4b22814a8 100644 --- a/lib/mnesia/src/mnesia_loader.erl +++ b/lib/mnesia/src/mnesia_loader.erl @@ -1,7 +1,7 @@ %% %% %CopyrightBegin% %% -%% Copyright Ericsson AB 1998-2010. All Rights Reserved. +%% Copyright Ericsson AB 1998-2011. All Rights Reserved. %% %% The contents of this file are subject to the Erlang Public License, %% Version 1.1, (the "License"); you may not use this file except in @@ -27,7 +27,6 @@ net_load_table/4, send_table/3]). --export([old_node_init_table/6]). %% Spawned old node protocol conversion hack -export([spawned_receiver/8]). %% Spawned lock taking process -import(mnesia_lib, [set/2, fatal/2, verbose/2, dbg_out/2]). @@ -36,7 +35,7 @@ val(Var) -> case ?catch_val(Var) of - {'EXIT', Reason} -> mnesia_lib:other_val(Var, Reason); + {'EXIT', Reason} -> mnesia_lib:other_val(Var, Reason); Value -> Value end. @@ -51,7 +50,7 @@ disc_load_table(Tab, Reason) -> ?eval_debug_fun({?MODULE, do_get_disc_copy}, [{tab, Tab}, {reason, Reason}, - {storage, Storage}, + {storage, Storage}, {type, Type}]), do_get_disc_copy2(Tab, Reason, Storage, Type). @@ -63,19 +62,19 @@ do_get_disc_copy2(Tab, Reason, Storage, Type) when Storage == disc_copies -> %% NOW we create the actual table Repair = mnesia_monitor:get_env(auto_repair), Args = [{keypos, 2}, public, named_table, Type], - case Reason of + case Reason of {dumper, _} -> %% Resources allready allocated ignore; _ -> mnesia_monitor:mktab(Tab, Args), - Count = mnesia_log:dcd2ets(Tab, Repair), + Count = mnesia_log:dcd2ets(Tab, Repair), case ets:info(Tab, size) of X when X < Count * 4 -> - ok = mnesia_log:ets2dcd(Tab); + ok = mnesia_log:ets2dcd(Tab); _ -> ignore end - end, + end, mnesia_index:init_index(Tab, Storage), snmpify(Tab, Storage), set({Tab, load_node}, node()), @@ -84,7 +83,7 @@ do_get_disc_copy2(Tab, Reason, Storage, Type) when Storage == disc_copies -> do_get_disc_copy2(Tab, Reason, Storage, Type) when Storage == ram_copies -> Args = [{keypos, 2}, public, named_table, Type], - case Reason of + case Reason of {dumper, _} -> %% Resources allready allocated ignore; _ -> @@ -94,12 +93,12 @@ do_get_disc_copy2(Tab, Reason, Storage, Type) when Storage == ram_copies -> Repair = mnesia_monitor:get_env(auto_repair), case mnesia_monitor:use_dir() of true -> - case mnesia_lib:exists(Fname) of + case mnesia_lib:exists(Fname) of true -> mnesia_log:dcd2ets(Tab, Repair); false -> case mnesia_lib:exists(Datname) of true -> - mnesia_lib:dets_to_ets(Tab, Tab, Datname, + mnesia_lib:dets_to_ets(Tab, Tab, Datname, Type, Repair, no); false -> false @@ -154,11 +153,11 @@ do_get_disc_copy2(Tab, Reason, Storage, Type) when Storage == disc_only_copies - %% Disable rehashing of table %% Release read lock on table %% Send table to receiver in chunks -%% +%% %% Grab read lock on table %% Block dirty updates %% Update wherabouts -%% +%% %% Cancel the update subscription %% Process the subscription events %% Optionally dump to disc @@ -166,7 +165,7 @@ do_get_disc_copy2(Tab, Reason, Storage, Type) when Storage == disc_only_copies - %% Release read lock on table %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% --define(MAX_TRANSFER_SIZE, 7500). +-define(MAX_TRANSFER_SIZE, 7500). -define(MAX_RAM_FILE_SIZE, 1000000). -define(MAX_RAM_TRANSFERS, (?MAX_RAM_FILE_SIZE div ?MAX_TRANSFER_SIZE) + 1). -define(MAX_NOPACKETS, 20). @@ -187,14 +186,14 @@ try_net_load_table(Tab, Reason, Ns, Cs) -> do_get_network_copy(Tab, _Reason, _Ns, unknown, _Cs) -> verbose("Local table copy of ~p has recently been deleted, ignored.~n", [Tab]), {not_loaded, storage_unknown}; -do_get_network_copy(Tab, Reason, Ns, Storage, Cs) -> +do_get_network_copy(Tab, Reason, Ns, Storage, Cs) -> [Node | Tail] = Ns, case lists:member(Node,val({current, db_nodes})) of true -> dbg_out("Getting table ~p (~p) from node ~p: ~p~n", [Tab, Storage, Node, Reason]), ?eval_debug_fun({?MODULE, do_get_network_copy}, - [{tab, Tab}, {reason, Reason}, + [{tab, Tab}, {reason, Reason}, {nodes, Ns}, {storage, Storage}]), case init_receiver(Node, Tab, Storage, Cs, Reason) of ok -> @@ -208,7 +207,7 @@ do_get_network_copy(Tab, Reason, Ns, Storage, Cs) -> restart -> try_net_load_table(Tab, Reason, Tail ++ [Node], Cs); down -> - try_net_load_table(Tab, Reason, Tail, Cs) + try_net_load_table(Tab, Reason, Tail, Cs) end; false -> try_net_load_table(Tab, Reason, Tail, Cs) @@ -223,10 +222,10 @@ do_snmpify(Tab, Us, Storage) -> Snmp = mnesia_snmp_hook:create_table(Us, Tab, Storage), set({Tab, {index, snmp}}, Snmp). -%% Start the recieiver +%% Start the recieiver init_receiver(Node, Tab, Storage, Cs, Reas={dumper,add_table_copy}) -> case start_remote_sender(Node, Tab, Storage) of - {SenderPid, TabSize, DetsData} -> + {SenderPid, TabSize, DetsData} -> start_receiver(Tab,Storage,Cs,SenderPid,TabSize,DetsData,Reas); Else -> Else @@ -234,21 +233,21 @@ init_receiver(Node, Tab, Storage, Cs, Reas={dumper,add_table_copy}) -> init_receiver(Node, Tab,Storage,Cs,Reason) -> %% Grab a schema lock to avoid deadlock between table_loader and schema_commit dumping. %% Both may grab tables-locks in different order. - Load = - fun() -> - {_,Tid,Ts} = get(mnesia_activity_state), + Load = + fun() -> + {_,Tid,Ts} = get(mnesia_activity_state), mnesia_locker:rlock(Tid, Ts#tidstore.store, {schema, Tab}), - %% Check that table still exists + %% Check that table still exists Active = val({Tab, active_replicas}), %% Check that we havn't loaded it already case val({Tab,where_to_read}) == node() of true -> ok; _ -> - %% And that sender still got a copy - %% (something might have happend while + %% And that sender still got a copy + %% (something might have happend while %% we where waiting for the lock) true = lists:member(Node, Active), - {SenderPid, TabSize, DetsData} = + {SenderPid, TabSize, DetsData} = start_remote_sender(Node,Tab,Storage), Init = table_init_fun(SenderPid), Args = [self(),Tab,Storage,Cs,SenderPid, @@ -258,18 +257,18 @@ init_receiver(Node, Tab,Storage,Cs,Reason) -> wait_on_load_complete(Pid) end end, - Res = + Res = case mnesia:transaction(Load, 20) of - {atomic, {error,Result}} when - element(1,Reason) == dumper -> + {atomic, {error,Result}} when + element(1,Reason) == dumper -> {error,Result}; - {atomic, {error,Result}} -> + {atomic, {error,Result}} -> fatal("Cannot create table ~p: ~p~n", [[Tab, Storage], Result]); {atomic, Result} -> Result; {aborted, nomore} -> restart; - {aborted, _Reas} -> - verbose("Receiver failed on ~p from ~p:~nReason: ~p~n", + {aborted, _Reas} -> + verbose("Receiver failed on ~p from ~p:~nReason: ~p~n", [Tab,Node,_Reas]), down %% either this node or sender is dying end, @@ -279,7 +278,7 @@ init_receiver(Node, Tab,Storage,Cs,Reason) -> start_remote_sender(Node,Tab,Storage) -> mnesia_controller:start_remote_sender(Node, Tab, self(), Storage), put(mnesia_table_sender_node, {Tab, Node}), - receive + receive {SenderPid, {first, TabSize}} -> {SenderPid, TabSize, false}; {SenderPid, {first, TabSize, DetsData}} -> @@ -291,22 +290,14 @@ start_remote_sender(Node,Tab,Storage) -> end. table_init_fun(SenderPid) -> - PConv = mnesia_monitor:needs_protocol_conversion(node(SenderPid)), - MeMyselfAndI = self(), fun(read) -> - Receiver = - if - PConv == true -> - MeMyselfAndI ! {actual_tabrec, self()}, - MeMyselfAndI; %% Old mnesia - PConv == false -> self() - end, + Receiver = self(), SenderPid ! {Receiver, more}, get_data(SenderPid, Receiver) end. %% Add_table_copy get's it's own locks. -start_receiver(Tab,Storage,Cs,SenderPid,TabSize,DetsData,{dumper,add_table_copy}) -> +start_receiver(Tab,Storage,Cs,SenderPid,TabSize,DetsData,{dumper,add_table_copy}) -> Init = table_init_fun(SenderPid), case do_init_table(Tab,Storage,Cs,SenderPid,TabSize,DetsData,self(), Init) of Err = {error, _} -> @@ -317,8 +308,8 @@ start_receiver(Tab,Storage,Cs,SenderPid,TabSize,DetsData,{dumper,add_table_copy} end. spawned_receiver(ReplyTo,Tab,Storage,Cs, SenderPid,TabSize,DetsData, Init) -> - process_flag(trap_exit, true), - Done = do_init_table(Tab,Storage,Cs, + process_flag(trap_exit, true), + Done = do_init_table(Tab,Storage,Cs, SenderPid,TabSize,DetsData, ReplyTo, Init), ReplyTo ! {self(),Done}, @@ -327,17 +318,17 @@ spawned_receiver(ReplyTo,Tab,Storage,Cs, SenderPid,TabSize,DetsData, Init) -> exit(normal). wait_on_load_complete(Pid) -> - receive - {Pid, Res} -> + receive + {Pid, Res} -> Res; - {'EXIT', Pid, Reason} -> + {'EXIT', Pid, Reason} -> exit(Reason); - Else -> + Else -> Pid ! Else, wait_on_load_complete(Pid) end. -do_init_table(Tab,Storage,Cs,SenderPid, +do_init_table(Tab,Storage,Cs,SenderPid, TabSize,DetsInfo,OrigTabRec,Init) -> case create_table(Tab, TabSize, Storage, Cs) of {Storage,Tab} -> @@ -345,11 +336,9 @@ do_init_table(Tab,Storage,Cs,SenderPid, Node = node(SenderPid), put(mnesia_table_receiver, {Tab, Node, SenderPid}), mnesia_tm:block_tab(Tab), - PConv = mnesia_monitor:needs_protocol_conversion(Node), - - case init_table(Tab,Storage,Init,PConv,DetsInfo,SenderPid) of - ok -> - tab_receiver(Node,Tab,Storage,Cs,PConv,OrigTabRec); + case init_table(Tab,Storage,Init,DetsInfo,SenderPid) of + ok -> + tab_receiver(Node,Tab,Storage,Cs,OrigTabRec); Reason -> Msg = "[d]ets:init table failed", verbose("~s: ~p: ~p~n", [Msg, Tab, Reason]), @@ -360,7 +349,7 @@ do_init_table(Tab,Storage,Cs,SenderPid, end. create_table(Tab, TabSize, Storage, Cs) -> - if + if Storage == disc_only_copies -> mnesia_lib:lock_table(Tab), Tmp = mnesia_lib:tab2tmp(Tab), @@ -390,54 +379,30 @@ create_table(Tab, TabSize, Storage, Cs) -> end end. -tab_receiver(Node, Tab, Storage, Cs, PConv, OrigTabRec) -> +tab_receiver(Node, Tab, Storage, Cs, OrigTabRec) -> receive - {SenderPid, {no_more, DatBin}} when PConv == false -> + {SenderPid, {no_more, DatBin}} -> finish_copy(Storage,Tab,Cs,SenderPid,DatBin,OrigTabRec); - - %% Protocol conversion hack - {SenderPid, {no_more, DatBin}} when is_pid(PConv) -> - PConv ! {SenderPid, no_more}, - receive - {old_init_table_complete, ok} -> - finish_copy(Storage, Tab, Cs, SenderPid, DatBin,OrigTabRec); - {old_init_table_complete, Reason} -> - Msg = "OLD: [d]ets:init table failed", - verbose("~s: ~p: ~p~n", [Msg, Tab, Reason]), - down(Tab, Storage) - end; - - {actual_tabrec, Pid} -> - tab_receiver(Node, Tab, Storage, Cs, Pid,OrigTabRec); - - {SenderPid, {more, [Recs]}} when is_pid(PConv) -> - PConv ! {SenderPid, {more, Recs}}, %% Forward Msg to OldNodes - tab_receiver(Node, Tab, Storage, Cs, PConv,OrigTabRec); - {'EXIT', PConv, Reason} -> %% [d]ets:init process crashed - Msg = "Receiver crashed", - verbose("~s: ~p: ~p~n", [Msg, Tab, Reason]), - down(Tab, Storage); - %% Protocol conversion hack {copier_done, Node} -> verbose("Sender of table ~p crashed on node ~p ~n", [Tab, Node]), down(Tab, Storage); - + {'EXIT', Pid, Reason} -> handle_exit(Pid, Reason), - tab_receiver(Node, Tab, Storage, Cs, PConv,OrigTabRec) + tab_receiver(Node, Tab, Storage, Cs, OrigTabRec) end. make_table_fun(Pid, TabRec) -> fun(close) -> ok; (read) -> - get_data(Pid, TabRec) + get_data(Pid, TabRec) end. get_data(Pid, TabRec) -> - receive + receive {Pid, {more_z, CompressedRecs}} when is_binary(CompressedRecs) -> Pid ! {TabRec, more}, {zlib_uncompress(CompressedRecs), make_table_fun(Pid,TabRec)}; @@ -448,7 +413,7 @@ get_data(Pid, TabRec) -> end_of_input; {copier_done, Node} -> case node(Pid) of - Node -> + Node -> {copier_done, Node}; _ -> get_data(Pid, TabRec) @@ -458,13 +423,13 @@ get_data(Pid, TabRec) -> get_data(Pid, TabRec) end. -init_table(Tab, disc_only_copies, Fun, false, DetsInfo,Sender) -> +init_table(Tab, disc_only_copies, Fun, DetsInfo,Sender) -> ErtsVer = erlang:system_info(version), case DetsInfo of - {ErtsVer, DetsData} -> + {ErtsVer, DetsData} -> Res = (catch dets:is_compatible_bchunk_format(Tab, DetsData)), case Res of - {'EXIT',{undef,[{dets,_,_}|_]}} -> + {'EXIT',{undef,[{dets,_,_,_}|_]}} -> Sender ! {self(), {old_protocol, Tab}}, dets:init_table(Tab, Fun); %% Old dets version {'EXIT', What} -> @@ -481,28 +446,19 @@ init_table(Tab, disc_only_copies, Fun, false, DetsInfo,Sender) -> _ -> dets:init_table(Tab, Fun) end; -init_table(Tab, _, Fun, false, _DetsInfo,_) -> +init_table(Tab, _, Fun, _DetsInfo,_) -> case catch ets:init_table(Tab, Fun) of true -> ok; {'EXIT', Else} -> Else - end; -init_table(Tab, Storage, Fun, true, _DetsInfo, Sender) -> %% Old Nodes - spawn_link(?MODULE, old_node_init_table, - [Tab, Storage, Fun, self(), false, Sender]), - ok. + end. -old_node_init_table(Tab, Storage, Fun, TabReceiver, DetsInfo,Sender) -> - Res = init_table(Tab, Storage, Fun, false, DetsInfo,Sender), - TabReceiver ! {old_init_table_complete, Res}, - unlink(TabReceiver), - ok. finish_copy(Storage,Tab,Cs,SenderPid,DatBin,OrigTabRec) -> TabRef = {Storage, Tab}, subscr_receiver(TabRef, Cs#cstruct.record_name), case handle_last(TabRef, Cs#cstruct.type, DatBin) of - ok -> + ok -> mnesia_index:init_index(Tab, Storage), snmpify(Tab, Storage), %% OrigTabRec must not be the spawned tab-receiver @@ -534,7 +490,7 @@ subscr_receiver(TabRef = {_, Tab}, RecName) -> ok end. -handle_event(TabRef, write, Rec) -> +handle_event(TabRef, write, Rec) -> db_put(TabRef, Rec); handle_event(TabRef, delete, {_Tab, Key}) -> db_erase(TabRef, Key); @@ -545,8 +501,8 @@ handle_event(TabRef, clear_table, {_Tab, _Key}) -> handle_last({disc_copies, Tab}, _Type, nobin) -> Ret = mnesia_log:ets2dcd(Tab), - Fname = mnesia_lib:tab2dat(Tab), - case mnesia_lib:exists(Fname) of + Fname = mnesia_lib:tab2dat(Tab), + case mnesia_lib:exists(Fname) of true -> %% Remove old .DAT files. file:delete(Fname); false -> @@ -653,31 +609,29 @@ send_table(Pid, Tab, RemoteS) -> {error, {no_exists, Tab}}; Storage -> %% Send first - TabSize = mnesia:table_info(Tab, size), - Pconvert = mnesia_monitor:needs_protocol_conversion(node(Pid)), + TabSize = mnesia:table_info(Tab, size), KeysPerTransfer = calc_nokeys(Storage, Tab), ChunkData = dets:info(Tab, bchunk_format), - UseDetsChunk = - Storage == RemoteS andalso - Storage == disc_only_copies andalso - ChunkData /= undefined andalso - Pconvert == false, - if + UseDetsChunk = + Storage == RemoteS andalso + Storage == disc_only_copies andalso + ChunkData /= undefined, + if UseDetsChunk == true -> DetsInfo = erlang:system_info(version), Pid ! {self(), {first, TabSize, {DetsInfo, ChunkData}}}; true -> Pid ! {self(), {first, TabSize}} end, - + %% Debug info put(mnesia_table_sender, {Tab, node(Pid), Pid}), {Init, Chunk} = reader_funcs(UseDetsChunk, Tab, Storage, KeysPerTransfer), - + SendIt = fun() -> prepare_copy(Pid, Tab, Storage), - send_more(Pid, 1, Chunk, Init(), Tab, Pconvert), + send_more(Pid, 1, Chunk, Init(), Tab), finish_copy(Pid, Tab, Storage, RemoteS) end, @@ -698,11 +652,11 @@ send_table(Pid, Tab, RemoteS) -> {error, Reason} end end. - + prepare_copy(Pid, Tab, Storage) -> Trans = fun() -> - mnesia:write_lock_table(Tab), + mnesia:lock_table(Tab, load), mnesia_subscr:subscribe(Pid, {table, Tab}), update_where_to_write(Tab, node(Pid)), mnesia_lib:db_fixtable(Storage, Tab, true), @@ -717,11 +671,11 @@ prepare_copy(Pid, Tab, Storage) -> update_where_to_write(Tab, Node) -> case val({Tab, access_mode}) of - read_only -> + read_only -> ignore; - read_write -> + read_write -> Current = val({current, db_nodes}), - Ns = + Ns = case lists:member(Node, Current) of true -> Current; false -> [Node | Current] @@ -729,27 +683,27 @@ update_where_to_write(Tab, Node) -> update_where_to_write(Ns, Tab, Node) end. -update_where_to_write([], _, _) -> +update_where_to_write([], _, _) -> ok; update_where_to_write([H|T], Tab, AddNode) -> - rpc:call(H, mnesia_controller, call, + rpc:call(H, mnesia_controller, call, [{update_where_to_write, [add, Tab, AddNode], self()}]), update_where_to_write(T, Tab, AddNode). -send_more(Pid, N, Chunk, DataState, Tab, OldNode) -> +send_more(Pid, N, Chunk, DataState, Tab) -> receive {NewPid, more} -> - case send_packet(N - 1, NewPid, Chunk, DataState, OldNode) of - New when is_integer(New) -> + case send_packet(N - 1, NewPid, Chunk, DataState) of + New when is_integer(New) -> New - 1; NewData -> - send_more(NewPid, ?MAX_NOPACKETS, Chunk, NewData, Tab, OldNode) + send_more(NewPid, ?MAX_NOPACKETS, Chunk, NewData, Tab) end; {_NewPid, {old_protocol, Tab}} -> Storage = val({Tab, storage_type}), - {Init, NewChunk} = + {Init, NewChunk} = reader_funcs(false, Tab, Storage, calc_nokeys(Storage, Tab)), - send_more(Pid, 1, NewChunk, Init(), Tab, OldNode); + send_more(Pid, 1, NewChunk, Init(), Tab); {copier_done, Node} when Node == node(Pid)-> verbose("Receiver of table ~p crashed on ~p (more)~n", [Tab, Node]), @@ -770,7 +724,7 @@ dets_bchunk(Tab, Chunk) -> %% Arrg case dets:bchunk(Tab, Chunk) of {Cont, Data} -> {Data, Cont}; Else -> Else - end. + end. zlib_compress(Data, Level) -> BinData = term_to_binary(Data), @@ -793,28 +747,20 @@ compression_level() -> Val -> Val end. -send_packet(N, Pid, _Chunk, '$end_of_table', OldNode) -> - case OldNode of - true -> ignore; %% Old nodes can't handle the new no_more - false -> Pid ! {self(), no_more} - end, +send_packet(N, Pid, _Chunk, '$end_of_table') -> + Pid ! {self(), no_more}, N; -send_packet(N, Pid, Chunk, {[], Cont}, OldNode) -> - send_packet(N, Pid, Chunk, Chunk(Cont), OldNode); -send_packet(N, Pid, Chunk, {Recs, Cont}, OldNode) when N < ?MAX_NOPACKETS -> - case OldNode of - true -> - Pid ! {self(), {more, [Recs]}}; %% Old need's wrapping list - false -> - case compression_level() of - 0 -> - Pid ! {self(), {more, Recs}}; - Level -> - Pid ! {self(), {more_z, zlib_compress(Recs, Level)}} - end +send_packet(N, Pid, Chunk, {[], Cont}) -> + send_packet(N, Pid, Chunk, Chunk(Cont)); +send_packet(N, Pid, Chunk, {Recs, Cont}) when N < ?MAX_NOPACKETS -> + case compression_level() of + 0 -> + Pid ! {self(), {more, Recs}}; + Level -> + Pid ! {self(), {more_z, zlib_compress(Recs, Level)}} end, - send_packet(N+1, Pid, Chunk, Chunk(Cont), OldNode); -send_packet(_N, _Pid, _Chunk, DataState, _OldNode) -> + send_packet(N+1, Pid, Chunk, Chunk(Cont)); +send_packet(_N, _Pid, _Chunk, DataState) -> DataState. finish_copy(Pid, Tab, Storage, RemoteS) -> @@ -855,5 +801,5 @@ dat2bin(_Tab, _LocalS, _RemoteS) -> handle_exit(Pid, Reason) when node(Pid) == node() -> exit(Reason); -handle_exit(_Pid, _Reason) -> %% Not from our node, this will be handled by +handle_exit(_Pid, _Reason) -> %% Not from our node, this will be handled by ignore. %% mnesia_down soon. diff --git a/lib/mnesia/src/mnesia_locker.erl b/lib/mnesia/src/mnesia_locker.erl index ca0cc79c45..0492d794f3 100644 --- a/lib/mnesia/src/mnesia_locker.erl +++ b/lib/mnesia/src/mnesia_locker.erl @@ -40,7 +40,8 @@ sticky_wlock_table/3, wlock/3, wlock_no_exist/4, - wlock_table/3 + wlock_table/3, + load_lock_table/3 ]). %% sys callback functions @@ -656,16 +657,17 @@ rwlock(Tid, Store, Oid) -> Lock = write, case need_lock(Store, Tab, Key, Lock) of yes -> - Ns = w_nodes(Tab), + {Ns, Majority} = w_nodes(Tab), + check_majority(Majority, Tab, Ns), Res = get_rwlocks_on_nodes(Ns, rwlock, Node, Store, Tid, Oid), ?ets_insert(Store, {{locks, Tab, Key}, Lock}), Res; no -> if Key == ?ALL -> - w_nodes(Tab); + element(2, w_nodes(Tab)); Tab == ?GLOBAL -> - w_nodes(Tab); + element(2, w_nodes(Tab)); true -> dirty_rpc(Node, Tab, Key, Lock) end @@ -677,12 +679,34 @@ rwlock(Tid, Store, Oid) -> %% in the local store under the key == nodes w_nodes(Tab) -> - Nodes = ?catch_val({Tab, where_to_write}), - case Nodes of - [_ | _] -> Nodes; + case ?catch_val({Tab, where_to_wlock}) of + {[_ | _], _} = Where -> Where; _ -> mnesia:abort({no_exists, Tab}) end. +%% If the table has the 'majority' flag set, we can +%% only take a write lock if we see a majority of the +%% nodes. + + +check_majority(true, Tab, HaveNs) -> + check_majority(Tab, HaveNs); +check_majority(false, _, _) -> + ok. + +check_majority(Tab, HaveNs) -> + case ?catch_val({Tab, majority}) of + true -> + case mnesia_lib:have_majority(Tab, HaveNs) of + true -> + ok; + false -> + mnesia:abort({no_majority, Tab}) + end; + _ -> + ok + end. + %% aquire a sticky wlock, a sticky lock is a lock %% which remains at this node after the termination of the %% transaction. @@ -708,12 +732,14 @@ sticky_lock(Tid, Store, {Tab, Key} = Oid, Lock) -> end. do_sticky_lock(Tid, Store, {Tab, Key} = Oid, Lock) -> + {WNodes, Majority} = w_nodes(Tab), + sticky_check_majority(Lock, Tab, Majority, WNodes), ?MODULE ! {self(), {test_set_sticky, Tid, Oid, Lock}}, N = node(), receive {?MODULE, N, granted} -> ?ets_insert(Store, {{locks, Tab, Key}, write}), - [?ets_insert(Store, {nodes, Node}) || Node <- w_nodes(Tab)], + [?ets_insert(Store, {nodes, Node}) || Node <- WNodes], granted; {?MODULE, N, {granted, Val}} -> %% for rwlocks case opt_lookup_in_client(Val, Oid, write) of @@ -721,7 +747,7 @@ do_sticky_lock(Tid, Store, {Tab, Key} = Oid, Lock) -> exit({aborted, C}); Val2 -> ?ets_insert(Store, {{locks, Tab, Key}, write}), - [?ets_insert(Store, {nodes, Node}) || Node <- w_nodes(Tab)], + [?ets_insert(Store, {nodes, Node}) || Node <- WNodes], Val2 end; {?MODULE, N, {not_granted, Reason}} -> @@ -737,6 +763,16 @@ do_sticky_lock(Tid, Store, {Tab, Key} = Oid, Lock) -> dirty_sticky_lock(Tab, Key, [N], Lock) end. +sticky_check_majority(W, Tab, true, WNodes) when W==write; W==read_write -> + case mnesia_lib:have_majority(Tab, WNodes) of + true -> + ok; + false -> + mnesia:abort({no_majority, Tab}) + end; +sticky_check_majority(_, _, _, _) -> + ok. + not_stuck(Tid, Store, Tab, _Key, Oid, _Lock, N) -> rlock(Tid, Store, {Tab, ?ALL}), %% needed? wlock(Tid, Store, Oid), %% perfect sync @@ -773,22 +809,33 @@ sticky_wlock_table(Tid, Store, Tab) -> %% local store when we have aquired the lock. %% wlock(Tid, Store, Oid) -> + wlock(Tid, Store, Oid, _CheckMajority = true). + +wlock(Tid, Store, Oid, CheckMajority) -> {Tab, Key} = Oid, case need_lock(Store, Tab, Key, write) of yes -> - Ns = w_nodes(Tab), + {Ns, Majority} = w_nodes(Tab), + if CheckMajority -> + check_majority(Majority, Tab, Ns); + true -> + ignore + end, Op = {self(), {write, Tid, Oid}}, ?ets_insert(Store, {{locks, Tab, Key}, write}), get_wlocks_on_nodes(Ns, Ns, Store, Op, Oid); no when Key /= ?ALL, Tab /= ?GLOBAL -> []; no -> - w_nodes(Tab) + element(2, w_nodes(Tab)) end. wlock_table(Tid, Store, Tab) -> wlock(Tid, Store, {Tab, ?ALL}). +load_lock_table(Tid, Store, Tab) -> + wlock(Tid, Store, {Tab, ?ALL}, _CheckMajority = false). + %% Write lock even if the table does not exist wlock_no_exist(Tid, Store, Tab, Ns) -> diff --git a/lib/mnesia/src/mnesia_log.erl b/lib/mnesia/src/mnesia_log.erl index 9e804cc4c2..94153473cb 100644 --- a/lib/mnesia/src/mnesia_log.erl +++ b/lib/mnesia/src/mnesia_log.erl @@ -1021,7 +1021,8 @@ add_recs([LogH|Rest], N) LogH#log_header.log_version >= "1.0" -> add_recs(Rest, N); add_recs([{{Tab, _Key}, _Val, clear_table} | Rest], N) -> - true = ets:match_delete(Tab, '_'), - add_recs(Rest, N+ets:info(Tab, size)); + Size = ets:info(Tab, size), + true = ets:delete_all_objects(Tab), + add_recs(Rest, N+Size); add_recs([], N) -> N. diff --git a/lib/mnesia/src/mnesia_monitor.erl b/lib/mnesia/src/mnesia_monitor.erl index b6eda9ad3a..8cb2e92c08 100644 --- a/lib/mnesia/src/mnesia_monitor.erl +++ b/lib/mnesia/src/mnesia_monitor.erl @@ -1,7 +1,7 @@ %% %% %CopyrightBegin% %% -%% Copyright Ericsson AB 1996-2010. All Rights Reserved. +%% Copyright Ericsson AB 1996-2011. All Rights Reserved. %% %% The contents of this file are subject to the Erlang Public License, %% Version 1.1, (the "License"); you may not use this file except in @@ -76,13 +76,13 @@ -include("mnesia.hrl"). --record(state, {supervisor, pending_negotiators = [], +-record(state, {supervisor, pending_negotiators = [], going_down = [], tm_started = false, early_connects = [], connecting, mq = []}). --define(current_protocol_version, {7,6}). +-define(current_protocol_version, {8,0}). --define(previous_protocol_version, {7,5}). +-define(previous_protocol_version, {7,6}). start() -> gen_server:start_link({local, ?MODULE}, ?MODULE, @@ -151,12 +151,12 @@ check_protocol([{Node, {accept, Mon, Version, Protocol}} | Tail], Protocols) -> case lists:member(Protocol, Protocols) of true -> case Protocol == protocol_version() of - true -> + true -> set({protocol, Node}, {Protocol, false}); false -> set({protocol, Node}, {Protocol, true}) end, - [node(Mon) | check_protocol(Tail, Protocols)]; + [node(Mon) | check_protocol(Tail, Protocols)]; false -> verbose("Failed to connect with ~p. ~p protocols rejected. " "expected version = ~p, expected protocol = ~p~n", @@ -179,7 +179,7 @@ check_protocol([], [Protocol | _Protocols]) -> set(protocol_version, Protocol), []. -protocol_version() -> +protocol_version() -> case ?catch_val(protocol_version) of {'EXIT', _} -> ?current_protocol_version; Version -> Version @@ -189,14 +189,14 @@ protocol_version() -> %% preferred protocols are first in the list acceptable_protocol_versions() -> [protocol_version(), ?previous_protocol_version]. - + needs_protocol_conversion(Node) -> case {?catch_val({protocol, Node}), protocol_version()} of {{'EXIT', _}, _} -> false; - {{_, Bool}, ?current_protocol_version} -> + {{_, Bool}, ?current_protocol_version} -> Bool; - {{_, Bool}, _} -> + {{_, Bool}, _} -> not Bool end. @@ -255,15 +255,15 @@ terminate_proc(Who, Reason, _State) -> %%---------------------------------------------------------------------- init([Parent]) -> process_flag(trap_exit, true), - ?ets_new_table(mnesia_gvar, [set, public, named_table]), - ?ets_new_table(mnesia_stats, [set, public, named_table]), + ?ets_new_table(mnesia_gvar, [set, public, named_table]), + ?ets_new_table(mnesia_stats, [set, public, named_table]), set(subscribers, []), set(activity_subscribers, []), mnesia_lib:verbose("~p starting: ~p~n", [?MODULE, self()]), Version = mnesia:system_info(version), set(version, Version), dbg_out("Version: ~p~n", [Version]), - + case catch process_config_args(env()) of ok -> mnesia_lib:set({'$$$_report', current_pos}, 0), @@ -283,7 +283,7 @@ init([Parent]) -> set(checkpoints, []), set(pending_checkpoints, []), set(pending_checkpoint_pids, []), - + {ok, #state{supervisor = Parent}}; {'EXIT', Reason} -> mnesia_lib:report_fatal("Bad configuration: ~p~n", [Reason]), @@ -398,9 +398,9 @@ handle_call({unsafe_close_log, Name}, _From, State) -> disk_log:close(Name), {reply, ok, State}; -handle_call({negotiate_protocol, Mon, _Version, _Protocols}, _From, State) +handle_call({negotiate_protocol, Mon, _Version, _Protocols}, _From, State) when State#state.tm_started == false -> - State2 = State#state{early_connects = [node(Mon) | State#state.early_connects]}, + State2 = State#state{early_connects = [node(Mon) | State#state.early_connects]}, {reply, {node(), {reject, self(), uninitialized, uninitialized}}, State2}; %% From remote monitor.. @@ -412,7 +412,7 @@ handle_call({negotiate_protocol, Mon, Version, Protocols}, From, State) true -> accept_protocol(Mon, MyVersion, Protocol, From, State); false -> - %% in this release we should be able to handle the previous + %% in this release we should be able to handle the previous %% protocol case hd(Protocols) of ?previous_protocol_version -> @@ -427,7 +427,7 @@ handle_call({negotiate_protocol, Mon, Version, Protocols}, From, State) end; %% Local request to negotiate with other monitors (nodes). -handle_call({negotiate_protocol, Nodes}, From, State) -> +handle_call({negotiate_protocol, Nodes}, From, State) -> case mnesia_lib:intersect(State#state.going_down, Nodes) of [] -> spawn_link(?MODULE, negotiate_protocol_impl, [Nodes, From]), @@ -461,7 +461,7 @@ accept_protocol(Mon, Version, Protocol, From, State) -> %% No need for wait link(Mon), %% link to remote Monitor case Protocol == protocol_version() of - true -> + true -> set({protocol, Node}, {Protocol, false}); false -> set({protocol, Node}, {Protocol, true}) @@ -509,7 +509,7 @@ handle_cast({disconnect, Node}, State) -> ignore; undefined -> ignore; - RemoteMon when is_pid(RemoteMon) -> + RemoteMon when is_pid(RemoteMon) -> unlink(RemoteMon) end, {noreply, State}; @@ -534,9 +534,13 @@ handle_info({'EXIT', Pid, R}, State) when Pid == State#state.supervisor -> dbg_out("~p was ~p by supervisor~n",[?MODULE, R]), {stop, R, State}; -handle_info({'EXIT', Pid, fatal}, State) when node(Pid) == node() -> +handle_info({'EXIT', Pid, fatal}, State) when node(Pid) == node() -> dbg_out("~p got FATAL ERROR from: ~p~n",[?MODULE, Pid]), - exit(State#state.supervisor, shutdown), + %% This may hang supervisor if a shutdown happens at the same time as an fatal + %% is in progress + %% exit(State#state.supervisor, shutdown), + %% It is better to kill an innocent process + catch exit(whereis(mnesia_locker), kill), {noreply, State}; handle_info(Msg = {'EXIT',Pid,_}, State) -> @@ -550,7 +554,7 @@ handle_info(Msg = {'EXIT',Pid,_}, State) -> Node /= node() -> {noreply, State#state{mq = State#state.mq ++ [{info, Msg}]}}; true -> - %% We have probably got an exit signal from + %% We have probably got an exit signal from %% disk_log or dets Hint = "Hint: check that the disk still is writable", fatal("~p got unexpected info: ~p; ~p~n", @@ -567,10 +571,10 @@ handle_info({nodeup, Node}, State) -> %% Let's check if Mnesia is running there in order %% to detect if the network has been partitioned %% due to communication failure. - + HasDown = mnesia_recover:has_mnesia_down(Node), ImRunning = mnesia_lib:is_running(), - + if %% If I'm not running the test will be made later. HasDown == true, ImRunning == yes -> @@ -589,7 +593,7 @@ handle_info({disk_log, _Node, Log, Info}, State) -> {truncated, _No} -> ok; _ -> - mnesia_lib:important("Warning Log file ~p error reason ~s~n", + mnesia_lib:important("Warning Log file ~p error reason ~s~n", [Log, disk_log:format_error(Info)]) end, {noreply, State}; @@ -681,38 +685,38 @@ env() -> send_compressed ]. -default_env(access_module) -> +default_env(access_module) -> mnesia; -default_env(auto_repair) -> +default_env(auto_repair) -> true; -default_env(backup_module) -> +default_env(backup_module) -> mnesia_backup; -default_env(debug) -> +default_env(debug) -> none; default_env(dir) -> Name = lists:concat(["Mnesia.", node()]), filename:absname(Name); -default_env(dump_log_load_regulation) -> +default_env(dump_log_load_regulation) -> false; -default_env(dump_log_time_threshold) -> +default_env(dump_log_time_threshold) -> timer:minutes(3); -default_env(dump_log_update_in_place) -> +default_env(dump_log_update_in_place) -> true; default_env(dump_log_write_threshold) -> 1000; -default_env(embedded_mnemosyne) -> +default_env(embedded_mnemosyne) -> false; -default_env(event_module) -> +default_env(event_module) -> mnesia_event; -default_env(extra_db_nodes) -> +default_env(extra_db_nodes) -> []; -default_env(ignore_fallback_at_startup) -> +default_env(ignore_fallback_at_startup) -> false; default_env(fallback_error_function) -> {mnesia, lkill}; -default_env(max_wait_for_decision) -> +default_env(max_wait_for_decision) -> infinity; -default_env(schema_location) -> +default_env(schema_location) -> opt_disc; default_env(core_dir) -> false; @@ -732,7 +736,7 @@ check_type(Env, Val) -> NewVal -> NewVal end. - + do_check_type(access_module, A) when is_atom(A) -> A; do_check_type(auto_repair, B) -> bool(B); do_check_type(backup_module, B) when is_atom(B) -> B; @@ -749,7 +753,7 @@ do_check_type(dump_log_update_in_place, B) -> bool(B); do_check_type(dump_log_write_threshold, I) when is_integer(I), I > 0 -> I; do_check_type(event_module, A) when is_atom(A) -> A; do_check_type(ignore_fallback_at_startup, B) -> bool(B); -do_check_type(fallback_error_function, {Mod, Func}) +do_check_type(fallback_error_function, {Mod, Func}) when is_atom(Mod), is_atom(Func) -> {Mod, Func}; do_check_type(embedded_mnemosyne, B) -> bool(B); do_check_type(extra_db_nodes, L) when is_list(L) -> @@ -804,8 +808,8 @@ detect_inconcistency(Nodes, Context) -> has_remote_mnesia_down(Node) -> HasDown = mnesia_recover:has_mnesia_down(Node), Master = mnesia_recover:get_master_nodes(schema), - if - HasDown == true, Master == [] -> + if + HasDown == true, Master == [] -> {true, node()}; true -> {false, node()} diff --git a/lib/mnesia/src/mnesia_recover.erl b/lib/mnesia/src/mnesia_recover.erl index b3eed1de6e..4750291a10 100644 --- a/lib/mnesia/src/mnesia_recover.erl +++ b/lib/mnesia/src/mnesia_recover.erl @@ -227,11 +227,13 @@ do_log_decision(D, DoTell, NodeD) -> note_outcome(D2), case mnesia_monitor:use_dir() of true -> - mnesia_log:append(latest_log, D2), if DoTell == true, Outcome /= unclear -> tell_im_certain(NodeD#decision.disc_nodes--[node()],D2), - tell_im_certain(NodeD#decision.ram_nodes--[node()], D2); + tell_im_certain(NodeD#decision.ram_nodes--[node()], D2), + mnesia_log:log(D2); + Outcome /= unclear -> + mnesia_log:log(D2); true -> ignore end; diff --git a/lib/mnesia/src/mnesia_schema.erl b/lib/mnesia/src/mnesia_schema.erl index d1d892a387..179e15197e 100644 --- a/lib/mnesia/src/mnesia_schema.erl +++ b/lib/mnesia/src/mnesia_schema.erl @@ -37,10 +37,12 @@ change_table_copy_type/3, change_table_access_mode/2, change_table_load_order/2, + change_table_majority/2, change_table_frag/2, - clear_table/1, +%% clear_table/1, %% removed since it is not a schema op anymore create_table/1, cs2list/1, + vsn_cs2list/1, del_snmp/1, del_table_copy/2, del_table_index/2, @@ -64,6 +66,7 @@ merge_schema/0, merge_schema/1, move_table/3, + normalize_cs/2, opt_create_dir/2, prepare_commit/3, purge_dir/2, @@ -99,7 +102,7 @@ ]). %% Needed outside to be able to use/set table_properties -%% from user (not supported) +%% from user (not supported) -export([schema_transaction/1, insert_schema_ops/2, do_create_table/1, @@ -117,9 +120,9 @@ %% Here comes the init function which also resides in %% this module, it is called upon by the trans server %% at startup of the system -%% +%% %% We have a meta table which looks like -%% {table, schema, +%% {table, schema, %% {type, set}, %% {disc_copies, all}, %% {arity, 2} @@ -148,14 +151,14 @@ exit_on_error(GoodRes) -> val(Var) -> case ?catch_val(Var) of - {'EXIT', Reason} -> mnesia_lib:other_val(Var, Reason); - Value -> Value + {'EXIT', Reason} -> mnesia_lib:other_val(Var, Reason); + Value -> Value end. %% This function traverses all cstructs in the schema and %% sets all values in mnesia_gvar accordingly for each table/cstruct -set_schema('$end_of_table') -> +set_schema('$end_of_table') -> []; set_schema(Tab) -> do_set_schema(Tab), @@ -178,6 +181,8 @@ do_set_schema(Tab, Cs) -> set({Tab, disc_only_copies}, Cs#cstruct.disc_only_copies), set({Tab, load_order}, Cs#cstruct.load_order), set({Tab, access_mode}, Cs#cstruct.access_mode), + set({Tab, majority}, Cs#cstruct.majority), + set({Tab, all_nodes}, mnesia_lib:cs_to_nodes(Cs)), set({Tab, snmp}, Cs#cstruct.snmp), set({Tab, user_properties}, Cs#cstruct.user_properties), [set({Tab, user_property, element(1, P)}, P) || P <- Cs#cstruct.user_properties], @@ -250,8 +255,8 @@ version() -> incr_version(Cs) -> {{Major, Minor}, _} = Cs#cstruct.version, Nodes = mnesia_lib:intersect(val({schema, disc_copies}), - mnesia_lib:cs_to_nodes(Cs)), - V = + mnesia_lib:cs_to_nodes(Cs)), + V = case Nodes -- val({Cs#cstruct.name, active_replicas}) of [] -> {Major + 1, 0}; % All replicas are active _ -> {Major, Minor + 1} % Some replicas are inactive @@ -356,7 +361,7 @@ delete_schema2() -> {error, Reason} -> {error, Reason} end. - + ensure_no_schema([H|T]) when is_atom(H) -> case rpc:call(H, ?MODULE, remote_read_schema, []) of {badrpc, Reason} -> @@ -404,7 +409,7 @@ opt_create_dir(UseDir, Dir) when UseDir == true-> check_can_write(Dir); false -> case file:make_dir(Dir) of - ok -> + ok -> verbose("Create Directory ~p~n", [Dir]), ok; {error, Reason} -> @@ -414,7 +419,7 @@ opt_create_dir(UseDir, Dir) when UseDir == true-> end; opt_create_dir(false, _) -> {error, {has_no_disc, node()}}. - + check_can_write(Dir) -> case file:read_file_info(Dir) of {ok, FI} when FI#file_info.type == directory, @@ -447,7 +452,7 @@ read_schema(Keep) -> read_schema(Keep, IgnoreFallback) -> lock_schema(), - Res = + Res = case mnesia:system_info(is_running) of yes -> {ok, ram, get_create_list(schema)}; @@ -474,7 +479,7 @@ read_disc_schema(Keep, IgnoreFallback) -> case mnesia_bup:fallback_exists() of true when IgnoreFallback == false, Running /= yes -> mnesia_bup:fallback_to_schema(); - _ -> + _ -> %% If we're running, we read the schema file even %% if fallback exists Dat = mnesia_lib:tab2dat(schema), @@ -496,7 +501,7 @@ read_disc_schema(Keep, IgnoreFallback) -> end. do_read_disc_schema(Fname, Keep) -> - T = + T = case Keep of false -> Args = [{keypos, 2}, public, set], @@ -520,7 +525,7 @@ do_read_disc_schema(Fname, Keep) -> get_initial_schema(SchemaStorage, Nodes) -> Cs = #cstruct{name = schema, record_name = schema, - attributes = [table, cstruct]}, + attributes = [table, cstruct]}, Cs2 = case SchemaStorage of ram_copies -> Cs#cstruct{ram_copies = Nodes}; @@ -529,7 +534,7 @@ get_initial_schema(SchemaStorage, Nodes) -> cs2list(Cs2). read_cstructs_from_disc() -> - %% Assumptions: + %% Assumptions: %% - local schema lock in global %% - use_dir is true %% - Mnesia is not running @@ -549,14 +554,14 @@ read_cstructs_from_disc() -> end, Cstructs = dets:traverse(Tab, Fun), dets:close(Tab), - {ok, Cstructs}; + {ok, Cstructs}; {error, Reason} -> {error, Reason} end; false -> {error, "No schema file exists"} end. - + %% We run a very special type of transactions when we %% we want to manipulate the schema. @@ -590,20 +595,20 @@ schema_transaction(Fun) -> %% This process may dump the transaction log, and should %% therefore not be run in an application process -%% +%% schema_coordinator(Client, _Fun, undefined) -> Res = {aborted, {node_not_running, node()}}, Client ! {transaction_done, Res, self()}, unlink(Client); - + schema_coordinator(Client, Fun, Controller) when is_pid(Controller) -> %% Do not trap exit in order to automatically die %% when the controller dies link(Controller), unlink(Client), - - %% Fulfull the transaction even if the client dies + + %% Fulfull the transaction even if the client dies Res = mnesia:transaction(Fun), Client ! {transaction_done, Res, self()}, unlink(Controller), % Avoids spurious exit message @@ -616,24 +621,82 @@ schema_coordinator(Client, Fun, Controller) when is_pid(Controller) -> insert_schema_ops({_Mod, _Tid, Ts}, SchemaIOps) -> do_insert_schema_ops(Ts#tidstore.store, SchemaIOps). - + do_insert_schema_ops(Store, [Head | Tail]) -> ?ets_insert(Store, Head), do_insert_schema_ops(Store, Tail); do_insert_schema_ops(_Store, []) -> ok. +api_list2cs(List) when is_list(List) -> + Name = pick(unknown, name, List, must), + Keys = check_keys(Name, List, record_info(fields, cstruct)), + check_duplicates(Name, Keys), + list2cs(List); +api_list2cs(Other) -> + mnesia:abort({badarg, Other}). + +vsn_cs2list(Cs) -> + cs2list(need_old_cstructs(), Cs). + cs2list(Cs) when is_record(Cs, cstruct) -> Tags = record_info(fields, cstruct), - rec2list(Tags, 2, Cs); + rec2list(Tags, Tags, 2, Cs); cs2list(CreateList) when is_list(CreateList) -> - CreateList. - -rec2list([Tag | Tags], Pos, Rec) -> + CreateList; +%% 4.4.19 +cs2list(Cs) when element(1, Cs) == cstruct, tuple_size(Cs) == 18 -> + Tags = [name,type,ram_copies,disc_copies,disc_only_copies, + load_order,access_mode,majority,index,snmp,local_content, + record_name,attributes,user_properties,frag_properties, + cookie,version], + rec2list(Tags, Tags, 2, Cs); +%% 4.4.18 and earlier +cs2list(Cs) when element(1, Cs) == cstruct, tuple_size(Cs) == 17 -> + Tags = [name,type,ram_copies,disc_copies,disc_only_copies, + load_order,access_mode,index,snmp,local_content, + record_name,attributes,user_properties,frag_properties, + cookie,version], + rec2list(Tags, Tags, 2, Cs). + +cs2list(false, Cs) -> + cs2list(Cs); +cs2list(ver4_4_18, Cs) -> %% Or earlier + Orig = record_info(fields, cstruct), + Tags = [name,type,ram_copies,disc_copies,disc_only_copies, + load_order,access_mode,index,snmp,local_content, + record_name,attributes,user_properties,frag_properties, + cookie,version], + rec2list(Tags, Orig, 2, Cs); +cs2list(ver4_4_19, Cs) -> + Orig = record_info(fields, cstruct), + Tags = [name,type,ram_copies,disc_copies,disc_only_copies, + load_order,access_mode,majority,index,snmp,local_content, + record_name,attributes,user_properties,frag_properties, + cookie,version], + rec2list(Tags, Orig, 2, Cs). + +rec2list([Tag | Tags], [Tag | Orig], Pos, Rec) -> Val = element(Pos, Rec), - [{Tag, Val} | rec2list(Tags, Pos + 1, Rec)]; -rec2list([], _Pos, _Rec) -> - []. + [{Tag, Val} | rec2list(Tags, Orig, Pos + 1, Rec)]; +rec2list([], _, _Pos, _Rec) -> + []; +rec2list(Tags, [_|Orig], Pos, Rec) -> + rec2list(Tags, Orig, Pos+1, Rec). + +normalize_cs(Cstructs, Node) -> + %% backward-compatibility hack; normalize before returning + case need_old_cstructs([Node]) of + false -> + Cstructs; + Version -> + %% some other format + [convert_cs(Version, Cs) || Cs <- Cstructs] + end. + +convert_cs(Version, Cs) -> + Fields = [Value || {_, Value} <- cs2list(Version, Cs)], + list_to_tuple([cstruct|Fields]). list2cs(List) when is_list(List) -> Name = pick(unknown, name, List, must), @@ -651,6 +714,7 @@ list2cs(List) when is_list(List) -> Snmp = pick(Name, snmp, List, []), LoadOrder = pick(Name, load_order, List, 0), AccessMode = pick(Name, access_mode, List, read_write), + Majority = pick(Name, majority, List, false), UserProps = pick(Name, user_properties, List, []), verify({alt, [nil, list]}, mnesia_lib:etype(UserProps), {bad_type, Name, {user_properties, UserProps}}), @@ -663,10 +727,7 @@ list2cs(List) when is_list(List) -> Frag = pick(Name, frag_properties, List, []), verify({alt, [nil, list]}, mnesia_lib:etype(Frag), - {badarg, Name, {frag_properties, Frag}}), - - Keys = check_keys(Name, List, record_info(fields, cstruct)), - check_duplicates(Name, Keys), + {badarg, Name, {frag_properties, Frag}}), #cstruct{name = Name, ram_copies = Rc, disc_copies = Dc, @@ -676,15 +737,14 @@ list2cs(List) when is_list(List) -> snmp = Snmp, load_order = LoadOrder, access_mode = AccessMode, + majority = Majority, local_content = LC, record_name = RecName, attributes = Attrs, user_properties = lists:sort(UserProps), frag_properties = lists:sort(Frag), cookie = Cookie, - version = Version}; -list2cs(Other) -> - mnesia:abort({badarg, Other}). + version = Version}. pick(Tab, Key, List, Default) -> case lists:keysearch(Key, 1, List) of @@ -703,7 +763,7 @@ attr_tab_to_pos(_Tab, Pos) when is_integer(Pos) -> Pos; attr_tab_to_pos(Tab, Attr) -> attr_to_pos(Attr, val({Tab, attributes})). - + %% Convert attribute name to integer if neccessary attr_to_pos(Pos, _Attrs) when is_integer(Pos) -> Pos; @@ -718,7 +778,7 @@ attr_to_pos(Attr, [_ | Attrs], Pos) -> attr_to_pos(Attr, Attrs, Pos + 1); attr_to_pos(Attr, _, _) -> mnesia:abort({bad_type, Attr}). - + check_keys(Tab, [{Key, _Val} | Tail], Items) -> case lists:member(Key, Items) of true -> [Key | check_keys(Tab, Tail, Items)]; @@ -754,7 +814,7 @@ verify_cstruct(Cs) when is_record(Cs, cstruct) -> {bad_type, Tab, {type, Type}}), %% Currently ordered_set is not supported for disk_only_copies. - if + if Type == ordered_set, Cs#cstruct.disc_only_copies /= [] -> mnesia:abort({bad_type, Tab, {not_supported, Type, disc_only_copies}}); true -> @@ -771,10 +831,10 @@ verify_cstruct(Cs) when is_record(Cs, cstruct) -> Arity = length(Attrs) + 1, verify(true, Arity > 2, {bad_type, Tab, {attributes, Attrs}}), - + lists:foldl(fun(Attr,_Other) when Attr == snmp -> mnesia:abort({bad_type, Tab, {attributes, [Attr]}}); - (Attr,Other) -> + (Attr,Other) -> verify(atom, mnesia_lib:etype(Attr), {bad_type, Tab, {attributes, [Attr]}}), verify(false, lists:member(Attr, Other), @@ -787,7 +847,7 @@ verify_cstruct(Cs) when is_record(Cs, cstruct) -> Index = Cs#cstruct.index, verify({alt, [nil, list]}, mnesia_lib:etype(Index), {bad_type, Tab, {index, Index}}), - + IxFun = fun(Pos) -> verify(true, fun() -> @@ -802,14 +862,23 @@ verify_cstruct(Cs) when is_record(Cs, cstruct) -> {bad_type, Tab, {index, [Pos]}}) end, lists:foreach(IxFun, Index), - + LC = Cs#cstruct.local_content, verify({alt, [true, false]}, LC, {bad_type, Tab, {local_content, LC}}), Access = Cs#cstruct.access_mode, verify({alt, [read_write, read_only]}, Access, {bad_type, Tab, {access_mode, Access}}), - + Majority = Cs#cstruct.majority, + verify({alt, [true, false]}, Majority, + {bad_type, Tab, {majority, Majority}}), + case Majority of + true -> + verify(false, LC, + {combine_error, Tab, [{local_content,true},{majority,true}]}); + false -> + ok + end, Snmp = Cs#cstruct.snmp, verify(true, mnesia_snmp_hook:check_ustruct(Snmp), {badarg, Tab, {snmp, Snmp}}), @@ -820,7 +889,7 @@ verify_cstruct(Cs) when is_record(Cs, cstruct) -> lists:foreach(CheckProp, Cs#cstruct.user_properties), case Cs#cstruct.cookie of - {{MegaSecs, Secs, MicroSecs}, _Node} + {{MegaSecs, Secs, MicroSecs}, _Node} when is_integer(MegaSecs), is_integer(Secs), is_integer(MicroSecs), is_atom(node) -> ok; @@ -856,15 +925,15 @@ verify_nodes(Cs) -> end, verify(integer, mnesia_lib:etype(LoadOrder), {bad_type, Tab, {load_order, LoadOrder}}), - + Nodes = Ram ++ Disc ++ DiscOnly, verify(list, mnesia_lib:etype(Nodes), {combine_error, Tab, [{ram_copies, []}, {disc_copies, []}, {disc_only_copies, []}]}), verify(false, has_duplicates(Nodes), {combine_error, Tab, Nodes}), - AtomCheck = fun(N) -> verify(atom, mnesia_lib:etype(N), {bad_type, Tab, N}) end, + AtomCheck = fun(N) -> verify(atom, mnesia_lib:etype(N), {bad_type, Tab, N}) end, lists:foreach(AtomCheck, Nodes). - + verify(Expected, Fun, Error) when is_function(Fun) -> do_verify(Expected, catch Fun(), Error); verify(Expected, Actual, Error) -> @@ -895,7 +964,7 @@ ensure_active(Cs, What) -> W = {Tab, What}, ensure_non_empty(W), Nodes = mnesia_lib:intersect(val({schema, disc_copies}), - mnesia_lib:cs_to_nodes(Cs)), + mnesia_lib:cs_to_nodes(Cs)), case Nodes -- val(W) of [] -> ok; @@ -922,7 +991,7 @@ ensure_non_empty({Tab, Vhat}) -> ensure_not_active(Tab = schema, Node) -> Active = val({Tab, active_replicas}), - case lists:member(Node, Active) of + case lists:member(Node, Active) of false when Active =/= [] -> ok; false -> @@ -956,7 +1025,7 @@ create_table(TabDef) -> do_multi_create_table(TabDef) -> get_tid_ts_and_lock(schema, write), ensure_writable(schema), - Cs = list2cs(TabDef), + Cs = api_list2cs(TabDef), case Cs#cstruct.frag_properties of [] -> do_create_table(Cs); @@ -985,7 +1054,7 @@ unsafe_make_create_table(Cs) -> {_Mod, Tid, Ts} = get_tid_ts_and_lock(schema, none), verify_cstruct(Cs), Tab = Cs#cstruct.name, - + %% Check that we have all disc replica nodes running DiscNodes = Cs#cstruct.disc_copies ++ Cs#cstruct.disc_only_copies, RunningNodes = val({current, db_nodes}), @@ -998,12 +1067,12 @@ unsafe_make_create_table(Cs) -> Nodes = mnesia_lib:intersect(mnesia_lib:cs_to_nodes(Cs), RunningNodes), Store = Ts#tidstore.store, mnesia_locker:wlock_no_exist(Tid, Store, Tab, Nodes), - [{op, create_table, cs2list(Cs)}]. + [{op, create_table, vsn_cs2list(Cs)}]. check_if_exists(Tab) -> TidTs = get_tid_ts_and_lock(schema, write), {_, _, Ts} = TidTs, - Store = Ts#tidstore.store, + Store = Ts#tidstore.store, ets:foldl( fun({op, create_table, [{name, T}|_]}, _Acc) when T==Tab -> true; @@ -1040,7 +1109,7 @@ make_delete_table(Tab, Mode) -> %% nodes etc. TidTs = get_tid_ts_and_lock(schema, write), {_, _, Ts} = TidTs, - Store = Ts#tidstore.store, + Store = Ts#tidstore.store, Deleted = ets:select_delete( Store, [{{op,'$1',[{name,Tab}|'_']}, [{'or', @@ -1063,9 +1132,9 @@ make_delete_table(Tab, Mode) -> [] -> [make_delete_table2(Tab)]; _Props -> - %% Check if it is a base table - mnesia_frag:lookup_frag_hash(Tab), - + %% Check if it is a base table + mnesia_frag:lookup_frag_hash(Tab), + %% Check for foreigners F = mnesia_frag:lookup_foreigners(Tab), verify([], F, {combine_error, @@ -1083,11 +1152,11 @@ make_delete_table2(Tab) -> Cs = val({Tab, cstruct}), ensure_active(Cs), ensure_writable(Tab), - {op, delete_table, cs2list(Cs)}. + {op, delete_table, vsn_cs2list(Cs)}. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% Change fragmentation of a table - + change_table_frag(Tab, Change) -> schema_transaction(fun() -> do_change_table_frag(Tab, Change) end). @@ -1098,14 +1167,10 @@ do_change_table_frag(Tab, Change) when is_atom(Tab), Tab /= schema -> ok; do_change_table_frag(Tab, _Change) -> mnesia:abort({bad_type, Tab}). - + %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% Clear a table -%% No need for a schema transaction -clear_table(Tab) -> - schema_transaction(fun() -> do_clear_table(Tab) end). - do_clear_table(schema) -> mnesia:abort({bad_type, schema}); do_clear_table(Tab) -> @@ -1116,7 +1181,7 @@ do_clear_table(Tab) -> make_clear_table(Tab) -> Cs = val({Tab, cstruct}), ensure_writable(Tab), - [{op, clear_table, cs2list(Cs)}]. + [{op, clear_table, vsn_cs2list(Cs)}]. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @@ -1136,7 +1201,7 @@ make_add_table_copy(Tab, Node, Storage) -> verify(false, lists:member(Node, Ns), {already_exists, Tab, Node}), Cs2 = new_cs(Cs, Node, Storage, add), verify_cstruct(Cs2), - + %% Check storage and if node is running IsRunning = lists:member(Node, val({current, db_nodes})), if @@ -1156,28 +1221,28 @@ make_add_table_copy(Tab, Node, Storage) -> IsRunning == false -> mnesia:abort({not_active, schema, Node}) end, - [{op, add_table_copy, Storage, Node, cs2list(Cs2)}]. + [{op, add_table_copy, Storage, Node, vsn_cs2list(Cs2)}]. del_table_copy(Tab, Node) -> schema_transaction(fun() -> do_del_table_copy(Tab, Node) end). do_del_table_copy(Tab, Node) when is_atom(Node) -> TidTs = get_tid_ts_and_lock(schema, write), -%% get_tid_ts_and_lock(Tab, write), +%% get_tid_ts_and_lock(Tab, write), insert_schema_ops(TidTs, make_del_table_copy(Tab, Node)); do_del_table_copy(Tab, Node) -> mnesia:abort({badarg, Tab, Node}). - + make_del_table_copy(Tab, Node) -> ensure_writable(schema), Cs = incr_version(val({Tab, cstruct})), Storage = mnesia_lib:schema_cs_to_storage_type(Node, Cs), - Cs2 = new_cs(Cs, Node, Storage, del), + Cs2 = new_cs(Cs, Node, Storage, del), case mnesia_lib:cs_to_nodes(Cs2) of [] when Tab == schema -> mnesia:abort({combine_error, Tab, "Last replica"}); [] -> - ensure_active(Cs), + ensure_active(Cs), dbg_out("Last replica deleted in table ~p~n", [Tab]), make_delete_table(Tab, whole_table); _ when Tab == schema -> @@ -1185,25 +1250,25 @@ make_del_table_copy(Tab, Node) -> ensure_not_active(Tab, Node), verify_cstruct(Cs2), Ops = remove_node_from_tabs(val({schema, tables}), Node), - [{op, del_table_copy, ram_copies, Node, cs2list(Cs2)} | Ops]; + [{op, del_table_copy, ram_copies, Node, vsn_cs2list(Cs2)} | Ops]; _ -> ensure_active(Cs), verify_cstruct(Cs2), - [{op, del_table_copy, Storage, Node, cs2list(Cs2)}] + [{op, del_table_copy, Storage, Node, vsn_cs2list(Cs2)}] end. remove_node_from_tabs([], _Node) -> []; remove_node_from_tabs([schema|Rest], Node) -> remove_node_from_tabs(Rest, Node); -remove_node_from_tabs([Tab|Rest], Node) -> - {Cs, IsFragModified} = +remove_node_from_tabs([Tab|Rest], Node) -> + {Cs, IsFragModified} = mnesia_frag:remove_node(Node, incr_version(val({Tab, cstruct}))), case mnesia_lib:schema_cs_to_storage_type(Node, Cs) of unknown -> case IsFragModified of true -> - [{op, change_table_frag, {del_node, Node}, cs2list(Cs)} | + [{op, change_table_frag, {del_node, Node}, vsn_cs2list(Cs)} | remove_node_from_tabs(Rest, Node)]; false -> remove_node_from_tabs(Rest, Node) @@ -1212,11 +1277,11 @@ remove_node_from_tabs([Tab|Rest], Node) -> Cs2 = new_cs(Cs, Node, Storage, del), case mnesia_lib:cs_to_nodes(Cs2) of [] -> - [{op, delete_table, cs2list(Cs)} | + [{op, delete_table, vsn_cs2list(Cs)} | remove_node_from_tabs(Rest, Node)]; _Ns -> verify_cstruct(Cs2), - [{op, del_table_copy, ram_copies, Node, cs2list(Cs2)}| + [{op, del_table_copy, ram_copies, Node, vsn_cs2list(Cs2)}| remove_node_from_tabs(Rest, Node)] end end. @@ -1232,7 +1297,7 @@ new_cs(Cs, Node, ram_copies, del) -> new_cs(Cs, Node, disc_copies, del) -> Cs#cstruct{disc_copies = lists:delete(Node , Cs#cstruct.disc_copies)}; new_cs(Cs, Node, disc_only_copies, del) -> - Cs#cstruct{disc_only_copies = + Cs#cstruct{disc_only_copies = lists:delete(Node , Cs#cstruct.disc_only_copies)}; new_cs(Cs, _Node, Storage, _Op) -> mnesia:abort({badarg, Cs#cstruct.name, Storage}). @@ -1264,13 +1329,13 @@ make_move_table(Tab, FromNode, ToNode) -> Running = val({current, db_nodes}), Storage = mnesia_lib:schema_cs_to_storage_type(FromNode, Cs), verify(true, lists:member(ToNode, Running), {not_active, schema, ToNode}), - + Cs2 = new_cs(Cs, ToNode, Storage, add), Cs3 = new_cs(Cs2, FromNode, Storage, del), verify_cstruct(Cs3), - [{op, add_table_copy, Storage, ToNode, cs2list(Cs2)}, + [{op, add_table_copy, Storage, ToNode, vsn_cs2list(Cs2)}, {op, sync_trans}, - {op, del_table_copy, Storage, FromNode, cs2list(Cs3)}]. + {op, del_table_copy, Storage, FromNode, vsn_cs2list(Cs3)}]. %% end of functions to add and delete nodes to tables %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @@ -1292,7 +1357,7 @@ make_change_table_copy_type(Tab, Node, unknown) -> make_change_table_copy_type(Tab, Node, ToS) -> ensure_writable(schema), Cs = incr_version(val({Tab, cstruct})), - FromS = mnesia_lib:storage_type_at_node(Node, Tab), + FromS = mnesia_lib:storage_type_at_node(Node, Tab), case compare_storage_type(false, FromS, ToS) of {same, _} -> @@ -1306,12 +1371,12 @@ make_change_table_copy_type(Tab, Node, ToS) -> Cs2 = new_cs(Cs, Node, FromS, del), Cs3 = new_cs(Cs2, Node, ToS, add), verify_cstruct(Cs3), - - [{op, change_table_copy_type, Node, FromS, ToS, cs2list(Cs3)}]. + + [{op, change_table_copy_type, Node, FromS, ToS, vsn_cs2list(Cs3)}]. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% change index functions .... -%% Pos is allready added by 1 in both of these functions +%% Pos is allready added by 1 in both of these functions add_table_index(Tab, Pos) -> schema_transaction(fun() -> do_add_table_index(Tab, Pos) end). @@ -1333,7 +1398,7 @@ make_add_table_index(Tab, Pos) -> Ix2 = lists:sort([Pos | Ix]), Cs2 = Cs#cstruct{index = Ix2}, verify_cstruct(Cs2), - [{op, add_index, Pos, cs2list(Cs2)}]. + [{op, add_index, Pos, vsn_cs2list(Cs2)}]. del_table_index(Tab, Pos) -> schema_transaction(fun() -> do_del_table_index(Tab, Pos) end). @@ -1354,7 +1419,7 @@ make_del_table_index(Tab, Pos) -> verify(true, lists:member(Pos, Ix), {no_exists, Tab, Pos}), Cs2 = Cs#cstruct{index = lists:delete(Pos, Ix)}, verify_cstruct(Cs2), - [{op, del_index, Pos, cs2list(Cs2)}]. + [{op, del_index, Pos, vsn_cs2list(Cs2)}]. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @@ -1377,7 +1442,7 @@ make_add_snmp(Tab, Ustruct) -> verify(true, mnesia_snmp_hook:check_ustruct(Ustruct), Error), Cs2 = Cs#cstruct{snmp = Ustruct}, verify_cstruct(Cs2), - [{op, add_snmp, Ustruct, cs2list(Cs2)}]. + [{op, add_snmp, Ustruct, vsn_cs2list(Cs2)}]. del_snmp(Tab) -> schema_transaction(fun() -> do_del_snmp(Tab) end). @@ -1395,17 +1460,17 @@ make_del_snmp(Tab) -> ensure_active(Cs), Cs2 = Cs#cstruct{snmp = []}, verify_cstruct(Cs2), - [{op, del_snmp, cs2list(Cs2)}]. + [{op, del_snmp, vsn_cs2list(Cs2)}]. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -%% +%% -transform_table(Tab, Fun, NewAttrs, NewRecName) - when is_function(Fun), is_list(NewAttrs), is_atom(NewRecName) -> +transform_table(Tab, Fun, NewAttrs, NewRecName) + when is_function(Fun), is_list(NewAttrs), is_atom(NewRecName) -> schema_transaction(fun() -> do_transform_table(Tab, Fun, NewAttrs, NewRecName) end); -transform_table(Tab, ignore, NewAttrs, NewRecName) - when is_list(NewAttrs), is_atom(NewRecName) -> +transform_table(Tab, ignore, NewAttrs, NewRecName) + when is_list(NewAttrs), is_atom(NewRecName) -> schema_transaction(fun() -> do_transform_table(Tab, ignore, NewAttrs, NewRecName) end); transform_table(Tab, Fun, NewAttrs, NewRecName) -> @@ -1424,33 +1489,33 @@ make_transform(Tab, Fun, NewAttrs, NewRecName) -> ensure_active(Cs), ensure_writable(Tab), case mnesia_lib:val({Tab, index}) of - [] -> + [] -> Cs2 = Cs#cstruct{attributes = NewAttrs, record_name = NewRecName}, verify_cstruct(Cs2), - [{op, transform, Fun, cs2list(Cs2)}]; + [{op, transform, Fun, vsn_cs2list(Cs2)}]; PosList -> DelIdx = fun(Pos, Ncs) -> Ix = Ncs#cstruct.index, Ncs1 = Ncs#cstruct{index = lists:delete(Pos, Ix)}, - Op = {op, del_index, Pos, cs2list(Ncs1)}, + Op = {op, del_index, Pos, vsn_cs2list(Ncs1)}, {Op, Ncs1} end, AddIdx = fun(Pos, Ncs) -> Ix = Ncs#cstruct.index, Ix2 = lists:sort([Pos | Ix]), Ncs1 = Ncs#cstruct{index = Ix2}, - Op = {op, add_index, Pos, cs2list(Ncs1)}, + Op = {op, add_index, Pos, vsn_cs2list(Ncs1)}, {Op, Ncs1} end, {DelOps, Cs1} = lists:mapfoldl(DelIdx, Cs, PosList), Cs2 = Cs1#cstruct{attributes = NewAttrs, record_name = NewRecName}, {AddOps, Cs3} = lists:mapfoldl(AddIdx, Cs2, PosList), verify_cstruct(Cs3), - lists:flatten([DelOps, {op, transform, Fun, cs2list(Cs2)}, AddOps]) + lists:flatten([DelOps, {op, transform, Fun, vsn_cs2list(Cs2)}, AddOps]) end. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -%% +%% change_table_access_mode(Tab, Mode) -> schema_transaction(fun() -> do_change_table_access_mode(Tab, Mode) end). @@ -1470,7 +1535,7 @@ make_change_table_access_mode(Tab, Mode) -> verify(false, OldMode == Mode, {already_exists, Tab, Mode}), Cs2 = Cs#cstruct{access_mode = Mode}, verify_cstruct(Cs2), - [{op, change_table_access_mode, cs2list(Cs2), OldMode, Mode}]. + [{op, change_table_access_mode, vsn_cs2list(Cs2), OldMode, Mode}]. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @@ -1491,7 +1556,44 @@ make_change_table_load_order(Tab, LoadOrder) -> OldLoadOrder = Cs#cstruct.load_order, Cs2 = Cs#cstruct{load_order = LoadOrder}, verify_cstruct(Cs2), - [{op, change_table_load_order, cs2list(Cs2), OldLoadOrder, LoadOrder}]. + [{op, change_table_load_order, vsn_cs2list(Cs2), OldLoadOrder, LoadOrder}]. + +%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% + +change_table_majority(Tab, Majority) when is_boolean(Majority) -> + schema_transaction(fun() -> do_change_table_majority(Tab, Majority) end). + +do_change_table_majority(schema, _Majority) -> + mnesia:abort({bad_type, schema}); +do_change_table_majority(Tab, Majority) -> + TidTs = get_tid_ts_and_lock(schema, write), + get_tid_ts_and_lock(Tab, none), + insert_schema_ops(TidTs, make_change_table_majority(Tab, Majority)). + +make_change_table_majority(Tab, Majority) -> + ensure_writable(schema), + Cs = incr_version(val({Tab, cstruct})), + ensure_active(Cs), + OldMajority = Cs#cstruct.majority, + Cs2 = Cs#cstruct{majority = Majority}, + FragOps = case lists:keyfind(base_table, 1, Cs#cstruct.frag_properties) of + {_, Tab} -> + FragNames = mnesia_frag:frag_names(Tab) -- [Tab], + lists:map( + fun(T) -> + get_tid_ts_and_lock(Tab, none), + CsT = incr_version(val({T, cstruct})), + ensure_active(CsT), + CsT2 = CsT#cstruct{majority = Majority}, + verify_cstruct(CsT2), + {op, change_table_majority, vsn_cs2list(CsT2), + OldMajority, Majority} + end, FragNames); + false -> []; + {_, _} -> mnesia:abort({bad_type, Tab}) + end, + verify_cstruct(Cs2), + [{op, change_table_majority, vsn_cs2list(Cs2), OldMajority, Majority} | FragOps]. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @@ -1532,7 +1634,7 @@ make_write_table_properties(Tab, [Prop | Props], Cs) -> MergedProps = lists:merge(DelProps, [Prop]), Cs2 = Cs#cstruct{user_properties = MergedProps}, verify_cstruct(Cs2), - [{op, write_property, cs2list(Cs2), Prop} | + [{op, write_property, vsn_cs2list(Cs2), Prop} | make_write_table_properties(Tab, Props, Cs2)]; make_write_table_properties(_Tab, [], _Cs) -> []. @@ -1547,9 +1649,9 @@ change_prop_in_existing_op(Tab, Prop, How, Store) -> false -> false end. - -update_existing_op([{op, Op, L = [{name,Tab}|_], _OldProp}|Ops], - Tab, Prop, How, Acc) when Op == write_property; + +update_existing_op([{op, Op, L = [{name,Tab}|_], _OldProp}|Ops], + Tab, Prop, How, Acc) when Op == write_property; Op == delete_property -> %% Apparently, mnesia_dumper doesn't care about OldProp here -- just L, %% so we will throw away OldProp (not that it matters...) and insert Prop. @@ -1574,7 +1676,7 @@ update_existing_op([], _, _, _, _) -> do_read_table_property(Tab, Key) -> TidTs = get_tid_ts_and_lock(schema, read), {_, _, Ts} = TidTs, - Store = Ts#tidstore.store, + Store = Ts#tidstore.store, Props = ets:foldl( fun({op, create_table, [{name, T}|Opts]}, _Acc) when T==Tab -> @@ -1638,7 +1740,7 @@ do_delete_table_property(Tab, PropKey) -> [Tab,PropKey]), %% this must be an existing table get_tid_ts_and_lock(Tab, none), - insert_schema_ops(TidTs, + insert_schema_ops(TidTs, make_delete_table_properties(Tab, [PropKey])) end. @@ -1653,24 +1755,24 @@ make_delete_table_properties(Tab, [PropKey | PropKeys], Cs) -> Props = lists:keydelete(PropKey, 1, OldProps), Cs2 = Cs#cstruct{user_properties = Props}, verify_cstruct(Cs2), - [{op, delete_property, cs2list(Cs2), PropKey} | + [{op, delete_property, vsn_cs2list(Cs2), PropKey} | make_delete_table_properties(Tab, PropKeys, Cs2)]; make_delete_table_properties(_Tab, [], _Cs) -> []. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% -%% Ensure that the transaction can be committed even +%% Ensure that the transaction can be committed even %% if the node crashes and Mnesia is restarted prepare_commit(Tid, Commit, WaitFor) -> case Commit#commit.schema_ops of [] -> {false, Commit, optional}; OrigOps -> - {Modified, Ops, DumperMode} = + {Modified, Ops, DumperMode} = prepare_ops(Tid, OrigOps, WaitFor, false, [], optional), InitBy = schema_prepare, - GoodRes = {Modified, + GoodRes = {Modified, Commit#commit{schema_ops = lists:reverse(Ops)}, DumperMode}, case DumperMode of @@ -1686,7 +1788,7 @@ prepare_commit(Tid, Commit, WaitFor) -> end end, case Ops of - [] -> + [] -> ignore; _ -> %% We need to grab a dumper lock here, the log may not @@ -1698,20 +1800,20 @@ prepare_commit(Tid, Commit, WaitFor) -> prepare_ops(Tid, [Op | Ops], WaitFor, Changed, Acc, DumperMode) -> case prepare_op(Tid, Op, WaitFor) of - {true, mandatory} -> + {true, mandatory} -> prepare_ops(Tid, Ops, WaitFor, Changed, [Op | Acc], mandatory); - {true, optional} -> + {true, optional} -> prepare_ops(Tid, Ops, WaitFor, Changed, [Op | Acc], DumperMode); - {true, Ops2, mandatory} -> + {true, Ops2, mandatory} -> prepare_ops(Tid, Ops, WaitFor, true, Ops2 ++ Acc, mandatory); - {true, Ops2, optional} -> + {true, Ops2, optional} -> prepare_ops(Tid, Ops, WaitFor, true, Ops2 ++ Acc, DumperMode); - {false, optional} -> + {false, optional} -> prepare_ops(Tid, Ops, WaitFor, true, Acc, DumperMode) end; prepare_ops(_Tid, [], _WaitFor, Changed, Acc, DumperMode) -> {Changed, Acc, DumperMode}. - + %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% Prepare for commit %% returns true if Op should be included, i.e. unmodified @@ -1730,19 +1832,22 @@ prepare_op(_Tid, {op, rec, unknown, Rec}, _WaitFor) -> prepare_op(_Tid, {op, announce_im_running, Node, SchemaDef, Running, RemoteRunning}, _WaitFor) -> SchemaCs = list2cs(SchemaDef), - if - Node == node() -> %% Announce has already run on local node + if + Node == node() -> %% Announce has already run on local node ignore; %% from do_merge_schema true -> - NewNodes = mnesia_lib:uniq(Running++RemoteRunning) -- val({current,db_nodes}), + %% If a node has restarted it may still linger in db_nodes, + %% but have been removed from recover_nodes + Current = mnesia_lib:intersect(val({current,db_nodes}), [node()|val(recover_nodes)]), + NewNodes = mnesia_lib:uniq(Running++RemoteRunning) -- Current, mnesia_lib:set(prepare_op, {announce_im_running,NewNodes}), announce_im_running(NewNodes, SchemaCs) end, {false, optional}; -prepare_op(_Tid, {op, sync_trans}, {part, CoordPid}) -> +prepare_op(_Tid, {op, sync_trans}, {part, CoordPid}) -> CoordPid ! {sync_trans, self()}, - receive + receive {sync_trans, CoordPid} -> {false, optional}; {mnesia_down, _Node} = Else -> @@ -1753,7 +1858,7 @@ prepare_op(_Tid, {op, sync_trans}, {part, CoordPid}) -> mnesia:abort(Else) end; -prepare_op(_Tid, {op, sync_trans}, {coord, Nodes}) -> +prepare_op(_Tid, {op, sync_trans}, {coord, Nodes}) -> case receive_sync(Nodes, []) of {abort, Reason} -> mnesia_lib:verbose("sync_op terminated due to ~p~n", [Reason]), @@ -1784,7 +1889,7 @@ prepare_op(Tid, {op, create_table, TabDef}, _WaitFor) -> create_ram_table(Tab, Cs#cstruct.type), create_disc_table(Tab), insert_cstruct(Tid, Cs, false), - {true, optional}; + {true, optional}; disc_only_copies -> mnesia_lib:set({Tab, create_table},true), create_disc_only_table(Tab,Cs#cstruct.type), @@ -1803,15 +1908,15 @@ prepare_op(Tid, {op, add_table_copy, Storage, Node, TabDef}, _WaitFor) -> if Tab == schema -> {true, optional}; - + Node == node() -> - case mnesia_lib:val({schema, storage_type}) of - ram_copies when Storage /= ram_copies -> + case mnesia_lib:val({schema, storage_type}) of + ram_copies when Storage /= ram_copies -> Error = {combine_error, Tab, "has no disc", Node}, mnesia:abort(Error); _ -> ok - end, + end, %% Tables are created by mnesia_loader get_network code insert_cstruct(Tid, Cs, true), case mnesia_controller:get_network_copy(Tab, Cs) of @@ -1848,22 +1953,22 @@ prepare_op(Tid, {op, add_table_copy, Storage, Node, TabDef}, _WaitFor) -> prepare_op(Tid, {op, del_table_copy, _Storage, Node, TabDef}, _WaitFor) -> Cs = list2cs(TabDef), Tab = Cs#cstruct.name, - + if %% Schema table lock is always required to run a schema op. %% No need to look it. - node(Tid#tid.pid) == node(), Tab /= schema -> + node(Tid#tid.pid) == node(), Tab /= schema -> Self = self(), Pid = spawn_link(fun() -> lock_del_table(Tab, Node, Cs, Self) end), put(mnesia_lock, Pid), - receive - {Pid, updated} -> + receive + {Pid, updated} -> {true, optional}; {Pid, FailReason} -> mnesia:abort(FailReason); {'EXIT', Pid, Reason} -> mnesia:abort(Reason) - end; + end; true -> {true, optional} end; @@ -1874,12 +1979,12 @@ prepare_op(_Tid, {op, change_table_copy_type, N, FromS, ToS, TabDef}, _WaitFor) Tab = Cs#cstruct.name, NotActive = mnesia_lib:not_active_here(Tab), - - if + + if NotActive == true -> mnesia:abort({not_active, Tab, node()}); - - Tab == schema -> + + Tab == schema -> case {FromS, ToS} of {ram_copies, disc_copies} -> case mnesia:system_info(schema_location) of @@ -1889,7 +1994,7 @@ prepare_op(_Tid, {op, change_table_copy_type, N, FromS, ToS, TabDef}, _WaitFor) mnesia:abort({combine_error, Tab, node(), "schema_location must be opt_disc"}) end, - Dir = mnesia_lib:dir(), + Dir = mnesia_lib:dir(), case opt_create_dir(true, Dir) of ok -> purge_dir(Dir, []), @@ -1913,18 +2018,18 @@ prepare_op(_Tid, {op, change_table_copy_type, N, FromS, ToS, TabDef}, _WaitFor) _ -> mnesia:abort({combine_error, Tab, ToS}) end; - - FromS == ram_copies -> + + FromS == ram_copies -> case mnesia_monitor:use_dir() of - true -> + true -> Dat = mnesia_lib:tab2dcd(Tab), case mnesia_lib:exists(Dat) of true -> mnesia:abort({combine_error, Tab, node(), "Table dump exists"}); false -> - case ToS of - disc_copies -> + case ToS of + disc_copies -> mnesia_log:ets2dcd(Tab, dmp); disc_only_copies -> mnesia_dumper:raw_named_dump_table(Tab, dmp) @@ -1934,7 +2039,7 @@ prepare_op(_Tid, {op, change_table_copy_type, N, FromS, ToS, TabDef}, _WaitFor) false -> mnesia:abort({has_no_disc, node()}) end; - + FromS == disc_copies, ToS == disc_only_copies -> mnesia_dumper:raw_named_dump_table(Tab, dmp); FromS == disc_only_copies -> @@ -1966,7 +2071,7 @@ prepare_op(_Tid, {op, dump_table, unknown, TabDef}, _WaitFor) -> case lists:member(node(), Cs#cstruct.ram_copies) of true -> case mnesia_monitor:use_dir() of - true -> + true -> mnesia_log:ets2dcd(Tab, dmp), Size = mnesia:table_info(Tab, size), {true, [{op, dump_table, Size, TabDef}], optional}; @@ -2004,7 +2109,7 @@ prepare_op(_Tid, {op, transform, Fun, TabDef}, _WaitFor) -> mnesia_lib:db_fixtable(Storage, Tab, true), Key = mnesia_lib:db_first(Tab), Op = {op, transform, Fun, TabDef}, - case catch transform_objs(Fun, Tab, RecName, + case catch transform_objs(Fun, Tab, RecName, Key, NewArity, Storage, Type, [Op]) of {'EXIT', Reason} -> mnesia_lib:db_fixtable(Storage, Tab, false), @@ -2018,7 +2123,7 @@ prepare_op(_Tid, {op, transform, Fun, TabDef}, _WaitFor) -> prepare_op(_Tid, {op, merge_schema, TabDef}, _WaitFor) -> Cs = list2cs(TabDef), case verify_merge(Cs) of - ok -> + ok -> {true, optional}; Error -> verbose("Merge_Schema ~p failed on ~p: ~p~n", [_Tid,node(),Error]), @@ -2039,7 +2144,7 @@ create_ram_table(Tab, Type) -> create_disc_table(Tab) -> File = mnesia_lib:tab2dcd(Tab), file:delete(File), - FArg = [{file, File}, {name, {mnesia,create}}, + FArg = [{file, File}, {name, {mnesia,create}}, {repair, false}, {mode, read_write}], case mnesia_monitor:open_log(FArg) of {ok,Log} -> @@ -2070,32 +2175,37 @@ receive_sync([], Pids) -> receive_sync(Nodes, Pids) -> receive {sync_trans, Pid} -> - Node = node(Pid), + Node = node(Pid), receive_sync(lists:delete(Node, Nodes), [Pid | Pids]); Else -> {abort, Else} end. -lock_del_table(Tab, Node, Cs, Father) -> +lock_del_table(Tab, NewNode, Cs0, Father) -> Ns = val({schema, active_replicas}), process_flag(trap_exit,true), Lock = fun() -> mnesia:write_lock_table(Tab), - {Res, []} = rpc:multicall(Ns, ?MODULE, set_where_to_read, [Tab, Node, Cs]), + %% Sigh using cs record + Set = fun(Node) -> + [Cs] = normalize_cs([Cs0], Node), + rpc:call(Node, ?MODULE, set_where_to_read, [Tab, NewNode, Cs]) + end, + Res = [Set(Node) || Node <- Ns], Filter = fun(ok) -> false; ({badrpc, {'EXIT', {undef, _}}}) -> %% This will be the case we talks with elder nodes - %% than 3.8.2, they will set where_to_read without - %% getting a lock. + %% than 3.8.2, they will set where_to_read without + %% getting a lock. false; (_) -> true end, case lists:filter(Filter, Res) of - [] -> + [] -> Father ! {self(), updated}, - %% When transaction is commited the process dies + %% When transaction is commited the process dies %% and the lock is released. receive _ -> ok end; Err -> @@ -2112,7 +2222,7 @@ lock_del_table(Tab, Node, Cs, Father) -> exit(normal). set_where_to_read(Tab, Node, Cs) -> - case mnesia_lib:val({Tab, where_to_read}) of + case mnesia_lib:val({Tab, where_to_read}) of Node -> case Cs#cstruct.local_content of true -> @@ -2131,16 +2241,16 @@ transform_objs(_Fun, _Tab, _RT, '$end_of_table', _NewArity, _Storage, _Type, Acc transform_objs(Fun, Tab, RecName, Key, A, Storage, Type, Acc) -> Objs = mnesia_lib:db_get(Tab, Key), NextKey = mnesia_lib:db_next_key(Tab, Key), - Oid = {Tab, Key}, + Oid = {Tab, Key}, NewObjs = {Ws, Ds} = transform_obj(Tab, RecName, Key, Fun, Objs, A, Type, [], []), - if - NewObjs == {[], []} -> + if + NewObjs == {[], []} -> transform_objs(Fun, Tab, RecName, NextKey, A, Storage, Type, Acc); - Type == bag -> + Type == bag -> transform_objs(Fun, Tab, RecName, NextKey, A, Storage, Type, [{op, rec, Storage, {Oid, Ws, write}}, {op, rec, Storage, {Oid, [Oid], delete}} | Acc]); - Ds == [] -> + Ds == [] -> %% Type is set or ordered_set, no need to delete the record first transform_objs(Fun, Tab, RecName, NextKey, A, Storage, Type, [{op, rec, Storage, {Oid, Ws, write}} | Acc]); @@ -2161,15 +2271,15 @@ transform_obj(Tab, RecName, Key, Fun, [Obj|Rest], NewArity, Type, Ws, Ds) -> NewObj == Obj -> transform_obj(Tab, RecName, Key, Fun, Rest, NewArity, Type, Ws, Ds); RecName == element(1, NewObj), Key == element(2, NewObj) -> - transform_obj(Tab, RecName, Key, Fun, Rest, NewArity, + transform_obj(Tab, RecName, Key, Fun, Rest, NewArity, Type, [NewObj | Ws], Ds); - NewObj == delete -> - case Type of + NewObj == delete -> + case Type of bag -> %% Just don't write that object - transform_obj(Tab, RecName, Key, Fun, Rest, - NewArity, Type, Ws, Ds); + transform_obj(Tab, RecName, Key, Fun, Rest, + NewArity, Type, Ws, Ds); _ -> - transform_obj(Tab, RecName, Key, Fun, Rest, NewArity, + transform_obj(Tab, RecName, Key, Fun, Rest, NewArity, Type, Ws, [NewObj | Ds]) end; true -> @@ -2193,7 +2303,7 @@ undo_prepare_commit(Tid, Commit) -> %% Undo in reverse order undo_prepare_ops(Tid, [Op | Ops]) -> - case element(1, Op) of + case element(1, Op) of TheOp when TheOp /= op, TheOp /= restore_op -> undo_prepare_ops(Tid, Ops); _ -> @@ -2220,7 +2330,7 @@ undo_prepare_op(Tid, {op, create_table, TabDef}) -> mnesia_lib:unset({Tab, create_table}), delete_cstruct(Tid, Cs), case mnesia_lib:cs_to_storage_type(node(), Cs) of - unknown -> + unknown -> ok; ram_copies -> ram_delete_table(Tab, ram_copies); @@ -2235,7 +2345,7 @@ undo_prepare_op(Tid, {op, create_table, TabDef}) -> %% disc_delete_table(Tab, Storage), file:delete(Dat) end; - + undo_prepare_op(Tid, {op, add_table_copy, Storage, Node, TabDef}) -> Cs = list2cs(TabDef), Tab = Cs#cstruct.name, @@ -2260,21 +2370,22 @@ undo_prepare_op(Tid, {op, add_table_copy, Storage, Node, TabDef}) -> Cs2 = new_cs(Cs, Node, Storage, del), insert_cstruct(Tid, Cs2, true) % Don't care about the version end; - -undo_prepare_op(_Tid, {op, del_table_copy, _, Node, TabDef}) + +undo_prepare_op(_Tid, {op, del_table_copy, _, Node, TabDef}) when Node == node() -> + WriteLocker = get(mnesia_lock), + WriteLocker =/= undefined andalso (WriteLocker ! die), Cs = list2cs(TabDef), Tab = Cs#cstruct.name, mnesia_lib:set({Tab, where_to_read}, Node); - -undo_prepare_op(_Tid, {op, change_table_copy_type, N, FromS, ToS, TabDef}) +undo_prepare_op(_Tid, {op, change_table_copy_type, N, FromS, ToS, TabDef}) when N == node() -> Cs = list2cs(TabDef), Tab = Cs#cstruct.name, mnesia_checkpoint:tm_change_table_copy_type(Tab, ToS, FromS), Dmp = mnesia_lib:tab2dmp(Tab), - + case {FromS, ToS} of {ram_copies, disc_copies} when Tab == schema -> file:delete(Dmp), @@ -2328,9 +2439,9 @@ ram_delete_table(Tab, Storage) -> ignore; disc_only_copies -> ignore; - _Else -> + _Else -> %% delete possible index files and data ..... - %% Got to catch this since if no info has been set in the + %% Got to catch this since if no info has been set in the %% mnesia_gvar it will crash catch mnesia_index:del_transient(Tab, Storage), case ?catch_val({Tab, {index, snmp}}) of @@ -2400,7 +2511,7 @@ has_known_suffix(File, [Suffix | Tail], false) -> has_known_suffix(File, Tail, lists:suffix(Suffix, File)); has_known_suffix(_File, [], Bool) -> Bool. - + known_suffixes() -> real_suffixes() ++ tmp_suffixes(). real_suffixes() -> [".DAT", ".LOG", ".BUP", ".DCL", ".DCD"]. @@ -2423,11 +2534,11 @@ info2(Tab, [{frag_hash, _V} | Tail]) -> % Ignore frag_hash info2(Tab, [{P, V} | Tail]) -> io:format("~-20w -> ~p~n",[P,V]), info2(Tab, Tail); -info2(_, []) -> +info2(_, []) -> io:format("~n", []). get_table_properties(Tab) -> - case catch mnesia_lib:db_match_object(ram_copies, + case catch mnesia_lib:db_match_object(ram_copies, mnesia_gvar, {{Tab, '_'}, '_'}) of {'EXIT', _} -> mnesia:abort({no_exists, Tab, all}); @@ -2455,9 +2566,9 @@ get_table_properties(Tab) -> recs = error_recs }). -restore(Opaque) -> +restore(Opaque) -> restore(Opaque, [], mnesia_monitor:get_env(backup_module)). -restore(Opaque, Args) when is_list(Args) -> +restore(Opaque, Args) when is_list(Args) -> restore(Opaque, Args, mnesia_monitor:get_env(backup_module)); restore(_Opaque, BadArg) -> {aborted, {badarg, BadArg}}. @@ -2468,7 +2579,7 @@ restore(Opaque, Args, Module) when is_list(Args), is_atom(Module) -> case mnesia_bup:read_schema(R#r.module, Opaque) of {error, Reason} -> {aborted, Reason}; - BupSchema -> + BupSchema -> schema_transaction(fun() -> do_restore(R, BupSchema) end) end; {'EXIT', Reason} -> @@ -2502,8 +2613,8 @@ check_restore_arg({keep_tables, List}, R) when is_list(List) -> check_restore_arg({skip_tables, List}, R) when is_list(List) -> TableList = [{Tab, skip_tables} || Tab <- List], R#r{table_options = R#r.table_options ++ TableList}; -check_restore_arg({default_op, Op}, R) -> - case Op of +check_restore_arg({default_op, Op}, R) -> + case Op of clear_tables -> ok; recreate_tables -> ok; keep_tables -> ok; @@ -2534,12 +2645,12 @@ restore_items([Rec | Recs], Header, Schema, R) -> case lists:keysearch(Tab, 1, R#r.tables) of {value, {Tab, Where0, Snmp, RecName}} -> Where = case Where0 of - undefined -> + undefined -> val({Tab, where_to_commit}); _ -> Where0 end, - {Rest, NRecs} = restore_tab_items([Rec | Recs], Tab, + {Rest, NRecs} = restore_tab_items([Rec | Recs], Tab, RecName, Where, Snmp, R#r.recs, R#r.insert_op), restore_items(Rest, Header, Schema, R#r{recs = NRecs}); @@ -2547,12 +2658,12 @@ restore_items([Rec | Recs], Header, Schema, R) -> Rest = skip_tab_items(Recs, Tab), restore_items(Rest, Header, Schema, R) end; - + restore_items([], _Header, _Schema, R) -> R. restore_func(Tab, R) -> - case lists:keysearch(Tab, 1, R#r.table_options) of + case lists:keysearch(Tab, 1, R#r.table_options) of {value, {Tab, OP}} -> OP; false -> @@ -2564,24 +2675,24 @@ where_to_commit(Tab, CsList) -> Disc = [{N, disc_copies} || N <- pick(Tab, disc_copies, CsList, [])], DiscO = [{N, disc_only_copies} || N <- pick(Tab, disc_only_copies, CsList, [])], Ram ++ Disc ++ DiscO. - + %% Changes of the Meta info of schema itself is not allowed restore_schema([{schema, schema, _List} | Schema], R) -> restore_schema(Schema, R); restore_schema([{schema, Tab, List} | Schema], R) -> case restore_func(Tab, R) of - clear_tables -> + clear_tables -> do_clear_table(Tab), - Snmp = val({Tab, snmp}), - RecName = val({Tab, record_name}), + Snmp = val({Tab, snmp}), + RecName = val({Tab, record_name}), R2 = R#r{tables = [{Tab, undefined, Snmp, RecName} | R#r.tables]}, restore_schema(Schema, R2); - recreate_tables -> + recreate_tables -> case ?catch_val({Tab, cstruct}) of - {'EXIT', _} -> + {'EXIT', _} -> TidTs = {_Mod, Tid, Ts} = get(mnesia_activity_state), RunningNodes = val({current, db_nodes}), - Nodes = mnesia_lib:intersect(mnesia_lib:cs_to_nodes(list2cs(List)), + Nodes = mnesia_lib:intersect(mnesia_lib:cs_to_nodes(list2cs(List)), RunningNodes), mnesia_locker:wlock_no_exist(Tid, Ts#tidstore.store, Tab, Nodes), TidTs; @@ -2589,20 +2700,20 @@ restore_schema([{schema, Tab, List} | Schema], R) -> TidTs = get_tid_ts_and_lock(Tab, write) end, NC = {cookie, ?unique_cookie}, - List2 = lists:keyreplace(cookie, 1, List, NC), + List2 = lists:keyreplace(cookie, 1, List, NC), Where = where_to_commit(Tab, List2), Snmp = pick(Tab, snmp, List2, []), RecName = pick(Tab, record_name, List2, Tab), insert_schema_ops(TidTs, [{op, restore_recreate, List2}]), R2 = R#r{tables = [{Tab, Where, Snmp, RecName} | R#r.tables]}, restore_schema(Schema, R2); - keep_tables -> + keep_tables -> get_tid_ts_and_lock(Tab, write), Snmp = val({Tab, snmp}), - RecName = val({Tab, record_name}), + RecName = val({Tab, record_name}), R2 = R#r{tables = [{Tab, undefined, Snmp, RecName} | R#r.tables]}, restore_schema(Schema, R2); - skip_tables -> + skip_tables -> restore_schema(Schema, R) end; @@ -2613,7 +2724,7 @@ restore_schema([{schema, Tab} | Schema], R) -> restore_schema([], R) -> R. -restore_tab_items([Rec | Rest], Tab, RecName, Where, Snmp, Recs, Op) +restore_tab_items([Rec | Rest], Tab, RecName, Where, Snmp, Recs, Op) when element(1, Rec) == Tab -> NewRecs = Op(Rec, Recs, RecName, Where, Snmp), restore_tab_items(Rest, Tab, RecName, Where, Snmp, NewRecs, Op); @@ -2621,7 +2732,7 @@ restore_tab_items([Rec | Rest], Tab, RecName, Where, Snmp, Recs, Op) restore_tab_items(Rest, _Tab, _RecName, _Where, _Snmp, Recs, _Op) -> {Rest, Recs}. -skip_tab_items([Rec| Rest], Tab) +skip_tab_items([Rec| Rest], Tab) when element(1, Rec) == Tab -> skip_tab_items(Rest, Tab); skip_tab_items(Recs, _) -> @@ -2656,7 +2767,6 @@ merge_schema() -> merge_schema(UserFun) -> schema_transaction(fun() -> UserFun(fun(Arg) -> do_merge_schema(Arg) end) end). - do_merge_schema(LockTabs0) -> {_Mod, Tid, Ts} = get_tid_ts_and_lock(schema, write), LockTabs = [{T, tab_to_nodes(T)} || T <- LockTabs0], @@ -2678,14 +2788,14 @@ do_merge_schema(LockTabs0) -> [mnesia_locker:wlock_no_exist( Tid, Store, T, mnesia_lib:intersect(Ns, OtherNodes)) || {T,Ns} <- LockTabs], - case rpc:call(Node, mnesia_controller, get_cstructs, []) of + case fetch_cstructs(Node) of {cstructs, Cstructs, RemoteRunning1} -> LockedAlready = Running ++ [Node], {New, Old} = mnesia_recover:connect_nodes(RemoteRunning1), RemoteRunning = mnesia_lib:intersect(New ++ Old, RemoteRunning1), - if + if RemoteRunning /= RemoteRunning1 -> - mnesia_lib:error("Mnesia on ~p could not connect to node(s) ~p~n", + mnesia_lib:error("Mnesia on ~p could not connect to node(s) ~p~n", [node(), RemoteRunning1 -- RemoteRunning]), mnesia:abort({node_not_running, RemoteRunning1 -- RemoteRunning}); true -> ok @@ -2695,24 +2805,24 @@ do_merge_schema(LockTabs0) -> [mnesia_locker:wlock_no_exist(Tid, Store, T, mnesia_lib:intersect(Ns,NeedsLock)) || {T,Ns} <- LockTabs], - {value, SchemaCs} = - lists:keysearch(schema, #cstruct.name, Cstructs), + NeedsConversion = need_old_cstructs(NeedsLock ++ LockedAlready), + {value, SchemaCs} = lists:keysearch(schema, #cstruct.name, Cstructs), + SchemaDef = cs2list(NeedsConversion, SchemaCs), %% Announce that Node is running - A = [{op, announce_im_running, node(), - cs2list(SchemaCs), Running, RemoteRunning}], + A = [{op, announce_im_running, node(), SchemaDef, Running, RemoteRunning}], do_insert_schema_ops(Store, A), - + %% Introduce remote tables to local node - do_insert_schema_ops(Store, make_merge_schema(Node, Cstructs)), - + do_insert_schema_ops(Store, make_merge_schema(Node, NeedsConversion, Cstructs)), + %% Introduce local tables to remote nodes Tabs = val({schema, tables}), Ops = [{op, merge_schema, get_create_list(T)} || T <- Tabs, not lists:keymember(T, #cstruct.name, Cstructs)], do_insert_schema_ops(Store, Ops), - + %% Ensure that the txn will be committed on all nodes NewNodes = RemoteRunning -- Running, mnesia_lib:set(prepare_op, {announce_im_running,NewNodes}), @@ -2728,19 +2838,52 @@ do_merge_schema(LockTabs0) -> not_merged end. +fetch_cstructs(Node) -> + case mnesia_monitor:needs_protocol_conversion(Node) of + true -> + case rpc:call(Node, mnesia_controller, get_cstructs, []) of + {cstructs, Cs0, RR} -> + {cstructs, [list2cs(cs2list(Cs)) || Cs <- Cs0], RR}; + Err -> Err + end; + false -> + rpc:call(Node, mnesia_controller, get_remote_cstructs, []) + end. + +need_old_cstructs() -> + need_old_cstructs(val({schema, where_to_write})). + +need_old_cstructs(Nodes) -> + Filter = fun(Node) -> not mnesia_monitor:needs_protocol_conversion(Node) end, + case lists:dropwhile(Filter, Nodes) of + [] -> false; + [Node|_] -> + case rpc:call(Node, mnesia_lib, val, [{schema,cstruct}]) of + #cstruct{} -> + %% mnesia_lib:warning("Mnesia on ~p do not need to convert cstruct (~p)~n", + %% [node(), Node]), + false; + {badrpc, _} -> + need_old_cstructs(lists:delete(Node,Nodes)); + Cs when element(1, Cs) == cstruct, tuple_size(Cs) == 17 -> + ver4_4_18; % Without majority + Cs when element(1, Cs) == cstruct, tuple_size(Cs) == 18 -> + ver4_4_19 % With majority + end + end. + tab_to_nodes(Tab) when is_atom(Tab) -> Cs = val({Tab, cstruct}), mnesia_lib:cs_to_nodes(Cs). -make_merge_schema(Node, [Cs | Cstructs]) -> - Ops = do_make_merge_schema(Node, Cs), - Ops ++ make_merge_schema(Node, Cstructs); -make_merge_schema(_Node, []) -> +make_merge_schema(Node, NeedsConv, [Cs | Cstructs]) -> + Ops = do_make_merge_schema(Node, NeedsConv, Cs), + Ops ++ make_merge_schema(Node, NeedsConv, Cstructs); +make_merge_schema(_Node, _, []) -> []. %% Merge definitions of schema table -do_make_merge_schema(Node, RemoteCs) - when RemoteCs#cstruct.name == schema -> +do_make_merge_schema(Node, NeedsConv, RemoteCs = #cstruct{name = schema}) -> Cs = val({schema, cstruct}), Masters = mnesia_recover:get_master_nodes(schema), HasRemoteMaster = lists:member(Node, Masters), @@ -2750,15 +2893,15 @@ do_make_merge_schema(Node, RemoteCs) StCsLocal = mnesia_lib:cs_to_storage_type(node(), Cs), StRcsLocal = mnesia_lib:cs_to_storage_type(node(), RemoteCs), StCsRemote = mnesia_lib:cs_to_storage_type(Node, Cs), - StRcsRemote = mnesia_lib:cs_to_storage_type(Node, RemoteCs), - + StRcsRemote = mnesia_lib:cs_to_storage_type(Node, RemoteCs), + if Cs#cstruct.cookie == RemoteCs#cstruct.cookie, Cs#cstruct.version == RemoteCs#cstruct.version -> %% Great, we have the same cookie and version %% and do not need to merge cstructs []; - + Cs#cstruct.cookie /= RemoteCs#cstruct.cookie, Cs#cstruct.disc_copies /= [], RemoteCs#cstruct.disc_copies /= [] -> @@ -2769,14 +2912,14 @@ do_make_merge_schema(Node, RemoteCs) HasRemoteMaster == false -> %% Choose local cstruct, %% since it's the master - [{op, merge_schema, cs2list(Cs)}]; + [{op, merge_schema, cs2list(NeedsConv, Cs)}]; HasRemoteMaster == true, HasLocalMaster == false -> %% Choose remote cstruct, %% since it's the master - [{op, merge_schema, cs2list(RemoteCs)}]; - + [{op, merge_schema, cs2list(NeedsConv, RemoteCs)}]; + true -> Str = io_lib:format("Incompatible schema cookies. " "Please, restart from old backup." @@ -2784,12 +2927,12 @@ do_make_merge_schema(Node, RemoteCs) [Node, cs2list(RemoteCs), node(), cs2list(Cs)]), throw(Str) end; - + StCsLocal /= StRcsLocal, StRcsLocal /= unknown, StCsLocal /= ram_copies -> Str = io_lib:format("Incompatible schema storage types (local). " "on ~w storage ~w, on ~w storage ~w~n", [node(), StCsLocal, Node, StRcsLocal]), - throw(Str); + throw(Str); StCsRemote /= StRcsRemote, StCsRemote /= unknown, StRcsRemote /= ram_copies -> Str = io_lib:format("Incompatible schema storage types (remote). " "on ~w cs ~w, on ~w rcs ~w~n", @@ -2800,27 +2943,27 @@ do_make_merge_schema(Node, RemoteCs) %% Choose local cstruct, %% since it involves disc nodes MergedCs = merge_cstructs(Cs, RemoteCs, Force), - [{op, merge_schema, cs2list(MergedCs)}]; - + [{op, merge_schema, cs2list(NeedsConv, MergedCs)}]; + RemoteCs#cstruct.disc_copies /= [] -> %% Choose remote cstruct, %% since it involves disc nodes MergedCs = merge_cstructs(RemoteCs, Cs, Force), - [{op, merge_schema, cs2list(MergedCs)}]; + [{op, merge_schema, cs2list(NeedsConv, MergedCs)}]; Cs > RemoteCs -> %% Choose remote cstruct MergedCs = merge_cstructs(RemoteCs, Cs, Force), - [{op, merge_schema, cs2list(MergedCs)}]; - + [{op, merge_schema, cs2list(NeedsConv, MergedCs)}]; + true -> %% Choose local cstruct MergedCs = merge_cstructs(Cs, RemoteCs, Force), - [{op, merge_schema, cs2list(MergedCs)}] + [{op, merge_schema, cs2list(NeedsConv, MergedCs)}] end; %% Merge definitions of normal table -do_make_merge_schema(Node, RemoteCs) -> +do_make_merge_schema(Node, NeedsConv, RemoteCs = #cstruct{}) -> Tab = RemoteCs#cstruct.name, Masters = mnesia_recover:get_master_nodes(schema), HasRemoteMaster = lists:member(Node, Masters), @@ -2829,27 +2972,27 @@ do_make_merge_schema(Node, RemoteCs) -> case ?catch_val({Tab, cstruct}) of {'EXIT', _} -> %% A completely new table, created while Node was down - [{op, merge_schema, cs2list(RemoteCs)}]; + [{op, merge_schema, cs2list(NeedsConv, RemoteCs)}]; Cs when Cs#cstruct.cookie == RemoteCs#cstruct.cookie -> if Cs#cstruct.version == RemoteCs#cstruct.version -> %% We have exactly the same version of the %% table def []; - + Cs#cstruct.version > RemoteCs#cstruct.version -> %% Oops, we have different versions %% of the table def, lets merge them. %% The only changes that may have occurred %% is that new replicas may have been added. MergedCs = merge_cstructs(Cs, RemoteCs, Force), - [{op, merge_schema, cs2list(MergedCs)}]; - + [{op, merge_schema, cs2list(NeedsConv, MergedCs)}]; + Cs#cstruct.version < RemoteCs#cstruct.version -> %% Oops, we have different versions %% of the table def, lets merge them MergedCs = merge_cstructs(RemoteCs, Cs, Force), - [{op, merge_schema, cs2list(MergedCs)}] + [{op, merge_schema, cs2list(NeedsConv, MergedCs)}] end; Cs -> %% Different cookies, not possible to merge @@ -2858,14 +3001,14 @@ do_make_merge_schema(Node, RemoteCs) -> HasRemoteMaster == false -> %% Choose local cstruct, %% since it's the master - [{op, merge_schema, cs2list(Cs)}]; + [{op, merge_schema, cs2list(NeedsConv, Cs)}]; HasRemoteMaster == true, HasLocalMaster == false -> %% Choose remote cstruct, %% since it's the master - [{op, merge_schema, cs2list(RemoteCs)}]; - + [{op, merge_schema, cs2list(NeedsConv, RemoteCs)}]; + true -> Str = io_lib:format("Bad cookie in table definition" " ~w: ~w = ~w, ~w = ~w~n", @@ -2935,7 +3078,7 @@ compare_storage_type(true, One, Another) -> compare_storage_type(false, Another, One); compare_storage_type(false, _One, _Another) -> incompatible. - + change_storage_type(N, ram_copies, Cs) -> Nodes = [N | Cs#cstruct.ram_copies], Cs#cstruct{ram_copies = mnesia_lib:uniq(Nodes)}; @@ -2971,6 +3114,7 @@ merge_versions(AnythingNew, Cs, RemoteCs, Force) -> Cs#cstruct.index == RemoteCs#cstruct.index, Cs#cstruct.snmp == RemoteCs#cstruct.snmp, Cs#cstruct.access_mode == RemoteCs#cstruct.access_mode, + Cs#cstruct.majority == RemoteCs#cstruct.majority, Cs#cstruct.load_order == RemoteCs#cstruct.load_order, Cs#cstruct.user_properties == RemoteCs#cstruct.user_properties -> do_merge_versions(AnythingNew, Cs, RemoteCs); @@ -3016,14 +3160,14 @@ verify_merge(RemoteCs) -> if StCsLocal == StRcsLocal -> ok; StCsLocal == unknown -> ok; - (StRcsLocal == unknown), (HasRemoteMaster == false) -> + (StRcsLocal == unknown), (HasRemoteMaster == false) -> {merge_error, Cs, RemoteCs}; %% Trust the merger true -> ok end end. -announce_im_running([N | Ns], SchemaCs) -> +announce_im_running([N | Ns], SchemaCs) -> {L1, L2} = mnesia_recover:connect_nodes([N]), case lists:member(N, L1) or lists:member(N, L2) of true -> @@ -3040,7 +3184,7 @@ announce_im_running([], _) -> unannounce_im_running([N | Ns]) -> mnesia_lib:del({current, db_nodes}, N), - mnesia_controller:del_active_replica(schema, N), + mnesia_controller:del_active_replica(schema, N), unannounce_im_running(Ns); unannounce_im_running([]) -> ok. diff --git a/lib/mnesia/src/mnesia_text.erl b/lib/mnesia/src/mnesia_text.erl index ab1362f6b6..0906d18da9 100644 --- a/lib/mnesia/src/mnesia_text.erl +++ b/lib/mnesia/src/mnesia_text.erl @@ -1,7 +1,7 @@ %% %% %CopyrightBegin% %% -%% Copyright Ericsson AB 1996-2010. All Rights Reserved. +%% Copyright Ericsson AB 1996-2011. All Rights Reserved. %% %% The contents of this file are subject to the Erlang Public License, %% Version 1.1, (the "License"); you may not use this file except in @@ -181,9 +181,6 @@ read_term_from_stream(Stream, File, Line) -> Str = Mod:format_error(What), io:format("Error in line:~p of:~p ~s\n", [NewLine, File, Str]), - error; - T -> - io:format("Error2 **~p~n",[T]), error end; {eof,_EndLine} -> diff --git a/lib/mnesia/src/mnesia_tm.erl b/lib/mnesia/src/mnesia_tm.erl index bb8e788b40..f62f7cb7c8 100644 --- a/lib/mnesia/src/mnesia_tm.erl +++ b/lib/mnesia/src/mnesia_tm.erl @@ -64,7 +64,8 @@ prev_tab = [], % initiate to a non valid table name prev_types, prev_snmp, - types + types, + majority = [] }). -record(participant, {tid, pid, commit, disc_nodes = [], @@ -1100,9 +1101,12 @@ t_commit(Type) -> 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, Tid, Prep#prep.records, Store) + multi_commit(read_only, + majority_attr(Prep), + Tid, Prep#prep.records, Store) end; true -> %% nested commit @@ -1117,6 +1121,10 @@ t_commit(Type) -> 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. @@ -1222,11 +1230,13 @@ prepare_items(Tid, Tab, Key, Items, Prep) -> {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. @@ -1235,6 +1245,33 @@ 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 [] -> @@ -1261,10 +1298,15 @@ prepare_snmp(Tid, Tab, Key, Types, Us, Items, Recs) -> prepare_nodes(Tid, Types, [{clear_table, Tab}], Recs, snmp) end. -check_prep(Prep, Types) when Prep#prep.types == Types -> +check_prep(#prep{majority = [], types = Types} = Prep, Types) -> Prep; -check_prep(Prep, Types) when Prep#prep.types == undefined -> - Prep#prep{types = Types}; +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}. @@ -1311,7 +1353,7 @@ prepare_node(_Node, _Storage, [], Rec, _Kind) -> %% multi_commit((Protocol, Tid, CommitRecords, Store) %% Local work is always performed in users process -multi_commit(read_only, Tid, CR, _Store) -> +multi_commit(read_only, _Maj = [], Tid, CR, _Store) -> %% This featherweight commit protocol is used when no %% updates has been performed in the transaction. @@ -1324,7 +1366,7 @@ multi_commit(read_only, Tid, CR, _Store) -> ?MODULE ! {delete_transaction, Tid}, do_commit; -multi_commit(sym_trans, Tid, CR, Store) -> +multi_commit(sym_trans, _Maj = [], Tid, CR, Store) -> %% This lightweight commit protocol is used when all %% the involved tables are replicated symetrically. %% Their storage types must match on each node. @@ -1376,7 +1418,7 @@ multi_commit(sym_trans, Tid, CR, Store) -> [{tid, Tid}, {outcome, Outcome}]), Outcome; -multi_commit(sync_sym_trans, Tid, CR, Store) -> +multi_commit(sync_sym_trans, _Maj = [], Tid, CR, Store) -> %% This protocol is the same as sym_trans except that it %% uses syncronized calls to disk_log and syncronized commits %% when several nodes are involved. @@ -1408,7 +1450,7 @@ multi_commit(sync_sym_trans, Tid, CR, Store) -> [{tid, Tid}, {outcome, Outcome}]), Outcome; -multi_commit(asym_trans, Tid, CR, Store) -> +multi_commit(asym_trans, Majority, Tid, CR, Store) -> %% This more expensive commit protocol is used when %% table definitions are changed (schema transactions). %% It is also used when the involved tables are @@ -1469,6 +1511,10 @@ multi_commit(asym_trans, Tid, CR, Store) -> {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(asym_trans, Tid, CR2, DiscNs, RamNs), |